| 1 |
/// Auto-generated prime length butterflies |
| 2 |
/// The code here is mostly autogenerated by the python script tools/gen_sse_butterflies.py, and then translated from SSE to WASM SIMD. |
| 3 |
/// |
| 4 |
/// The algorithm is derived directly from the definition of the DFT, by eliminating any repeated calculations. |
| 5 |
/// See the comments in src/algorithm/butterflies.rs for a detailed description. |
| 6 |
/// |
| 7 |
/// The script generates the code for performing a single f64 fft, as well as dual f32 fft. |
| 8 |
/// It also generates the code for reading and writing the input and output. |
| 9 |
/// The single 32-bit ffts reuse the dual ffts. |
| 10 |
use core::arch::wasm32::*; |
| 11 |
use num_complex::Complex; |
| 12 |
|
| 13 |
use crate::{common::FftNum, FftDirection}; |
| 14 |
|
| 15 |
use crate::array_utils; |
| 16 |
use crate::array_utils::workaround_transmute_mut; |
| 17 |
use crate::array_utils::DoubleBuf; |
| 18 |
use crate::common::{fft_error_inplace, fft_error_outofplace}; |
| 19 |
use crate::twiddles; |
| 20 |
use crate::{Direction, Fft, Length}; |
| 21 |
|
| 22 |
use super::wasm_simd_butterflies::{parallel_fft2_interleaved_f32, solo_fft2_f64}; |
| 23 |
use super::wasm_simd_common::{assert_f32, assert_f64}; |
| 24 |
use super::wasm_simd_utils::*; |
| 25 |
use super::wasm_simd_vector::WasmSimdArrayMut; |
| 26 |
|
| 27 |
// _____ _________ _ _ _ |
| 28 |
// |___ | |___ /___ \| |__ (_) |_ |
| 29 |
// / / _____ |_ \ __) | '_ \| | __| |
| 30 |
// / / |_____| ___) / __/| |_) | | |_ |
| 31 |
// /_/ |____/_____|_.__/|_|\__| |
| 32 |
// |
| 33 |
|
| 34 |
pub struct WasmSimdF32Butterfly7<T> { |
| 35 |
direction: FftDirection, |
| 36 |
_phantom: std::marker::PhantomData<T>, |
| 37 |
rotate: Rotate90F32, |
| 38 |
twiddle1re: v128, |
| 39 |
twiddle1im: v128, |
| 40 |
twiddle2re: v128, |
| 41 |
twiddle2im: v128, |
| 42 |
twiddle3re: v128, |
| 43 |
twiddle3im: v128, |
| 44 |
} |
| 45 |
|
| 46 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 47 |
WasmSimdF32Butterfly7, |
| 48 |
7, |
| 49 |
|this: &WasmSimdF32Butterfly7<_>| this.direction |
| 50 |
); |
| 51 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 52 |
WasmSimdF32Butterfly7, |
| 53 |
7, |
| 54 |
|this: &WasmSimdF32Butterfly7<_>| { this.direction } |
| 55 |
); |
| 56 |
impl<T: FftNum> WasmSimdF32Butterfly7<T> { |
| 57 |
#[inline(always)] |
| 58 |
pub fn new(direction: FftDirection) -> Self { |
| 59 |
assert_f32::<T>(); |
| 60 |
let rotate = Rotate90F32::new(true); |
| 61 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 7, direction); |
| 62 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 7, direction); |
| 63 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 7, direction); |
| 64 |
let twiddle1re = f32x4_splat(tw1.re); |
| 65 |
let twiddle1im = f32x4_splat(tw1.im); |
| 66 |
let twiddle2re = f32x4_splat(tw2.re); |
| 67 |
let twiddle2im = f32x4_splat(tw2.im); |
| 68 |
let twiddle3re = f32x4_splat(tw3.re); |
| 69 |
let twiddle3im = f32x4_splat(tw3.im); |
| 70 |
|
| 71 |
Self { |
| 72 |
direction, |
| 73 |
_phantom: std::marker::PhantomData, |
| 74 |
rotate, |
| 75 |
twiddle1re, |
| 76 |
twiddle1im, |
| 77 |
twiddle2re, |
| 78 |
twiddle2im, |
| 79 |
twiddle3re, |
| 80 |
twiddle3im, |
| 81 |
} |
| 82 |
} |
| 83 |
#[inline(always)] |
| 84 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 85 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6}); |
| 86 |
|
| 87 |
let out = self.perform_parallel_fft_direct(values); |
| 88 |
|
| 89 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6}); |
| 90 |
} |
| 91 |
|
| 92 |
#[inline(always)] |
| 93 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 94 |
&self, |
| 95 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 96 |
) { |
| 97 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12}); |
| 98 |
|
| 99 |
let values = [ |
| 100 |
extract_lo_hi_f32(input_packed[0], input_packed[3]), |
| 101 |
extract_hi_lo_f32(input_packed[0], input_packed[4]), |
| 102 |
extract_lo_hi_f32(input_packed[1], input_packed[4]), |
| 103 |
extract_hi_lo_f32(input_packed[1], input_packed[5]), |
| 104 |
extract_lo_hi_f32(input_packed[2], input_packed[5]), |
| 105 |
extract_hi_lo_f32(input_packed[2], input_packed[6]), |
| 106 |
extract_lo_hi_f32(input_packed[3], input_packed[6]), |
| 107 |
]; |
| 108 |
|
| 109 |
let out = self.perform_parallel_fft_direct(values); |
| 110 |
|
| 111 |
let out_packed = [ |
| 112 |
extract_lo_lo_f32(out[0], out[1]), |
| 113 |
extract_lo_lo_f32(out[2], out[3]), |
| 114 |
extract_lo_lo_f32(out[4], out[5]), |
| 115 |
extract_lo_hi_f32(out[6], out[0]), |
| 116 |
extract_hi_hi_f32(out[1], out[2]), |
| 117 |
extract_hi_hi_f32(out[3], out[4]), |
| 118 |
extract_hi_hi_f32(out[5], out[6]), |
| 119 |
]; |
| 120 |
|
| 121 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0,1,2,3,4,5,6}); |
| 122 |
} |
| 123 |
|
| 124 |
#[inline(always)] |
| 125 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 7]) -> [v128; 7] { |
| 126 |
let [x1p6, x1m6] = parallel_fft2_interleaved_f32(values[1], values[6]); |
| 127 |
let [x2p5, x2m5] = parallel_fft2_interleaved_f32(values[2], values[5]); |
| 128 |
let [x3p4, x3m4] = parallel_fft2_interleaved_f32(values[3], values[4]); |
| 129 |
|
| 130 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p6); |
| 131 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p5); |
| 132 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p4); |
| 133 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p6); |
| 134 |
let t_a2_2 = f32x4_mul(self.twiddle3re, x2p5); |
| 135 |
let t_a2_3 = f32x4_mul(self.twiddle1re, x3p4); |
| 136 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p6); |
| 137 |
let t_a3_2 = f32x4_mul(self.twiddle1re, x2p5); |
| 138 |
let t_a3_3 = f32x4_mul(self.twiddle2re, x3p4); |
| 139 |
|
| 140 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m6); |
| 141 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m5); |
| 142 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m4); |
| 143 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m6); |
| 144 |
let t_b2_2 = f32x4_mul(self.twiddle3im, x2m5); |
| 145 |
let t_b2_3 = f32x4_mul(self.twiddle1im, x3m4); |
| 146 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m6); |
| 147 |
let t_b3_2 = f32x4_mul(self.twiddle1im, x2m5); |
| 148 |
let t_b3_3 = f32x4_mul(self.twiddle2im, x3m4); |
| 149 |
|
| 150 |
let x0 = values[0]; |
| 151 |
let t_a1 = calc_f32!(x0 + t_a1_1 + t_a1_2 + t_a1_3); |
| 152 |
let t_a2 = calc_f32!(x0 + t_a2_1 + t_a2_2 + t_a2_3); |
| 153 |
let t_a3 = calc_f32!(x0 + t_a3_1 + t_a3_2 + t_a3_3); |
| 154 |
|
| 155 |
let t_b1 = calc_f32!(t_b1_1 + t_b1_2 + t_b1_3); |
| 156 |
let t_b2 = calc_f32!(t_b2_1 - t_b2_2 - t_b2_3); |
| 157 |
let t_b3 = calc_f32!(t_b3_1 - t_b3_2 + t_b3_3); |
| 158 |
|
| 159 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 160 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 161 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 162 |
|
| 163 |
let y0 = calc_f32!(x0 + x1p6 + x2p5 + x3p4); |
| 164 |
let [y1, y6] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 165 |
let [y2, y5] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 166 |
let [y3, y4] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 167 |
[y0, y1, y2, y3, y4, y5, y6] |
| 168 |
} |
| 169 |
} |
| 170 |
|
| 171 |
// _____ __ _ _ _ _ _ |
| 172 |
// |___ | / /_ | || | | |__ (_) |_ |
| 173 |
// / / _____ | '_ \| || |_| '_ \| | __| |
| 174 |
// / / |_____| | (_) |__ _| |_) | | |_ |
| 175 |
// /_/ \___/ |_| |_.__/|_|\__| |
| 176 |
// |
| 177 |
|
| 178 |
pub struct WasmSimdF64Butterfly7<T> { |
| 179 |
direction: FftDirection, |
| 180 |
_phantom: std::marker::PhantomData<T>, |
| 181 |
rotate: Rotate90F64, |
| 182 |
twiddle1re: v128, |
| 183 |
twiddle1im: v128, |
| 184 |
twiddle2re: v128, |
| 185 |
twiddle2im: v128, |
| 186 |
twiddle3re: v128, |
| 187 |
twiddle3im: v128, |
| 188 |
} |
| 189 |
|
| 190 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 191 |
WasmSimdF64Butterfly7, |
| 192 |
7, |
| 193 |
|this: &WasmSimdF64Butterfly7<_>| this.direction |
| 194 |
); |
| 195 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 196 |
WasmSimdF64Butterfly7, |
| 197 |
7, |
| 198 |
|this: &WasmSimdF64Butterfly7<_>| { this.direction } |
| 199 |
); |
| 200 |
impl<T: FftNum> WasmSimdF64Butterfly7<T> { |
| 201 |
#[inline(always)] |
| 202 |
pub fn new(direction: FftDirection) -> Self { |
| 203 |
assert_f64::<T>(); |
| 204 |
let rotate = Rotate90F64::new(true); |
| 205 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 7, direction); |
| 206 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 7, direction); |
| 207 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 7, direction); |
| 208 |
let twiddle1re = f64x2_splat(tw1.re); |
| 209 |
let twiddle1im = f64x2_splat(tw1.im); |
| 210 |
let twiddle2re = f64x2_splat(tw2.re); |
| 211 |
let twiddle2im = f64x2_splat(tw2.im); |
| 212 |
let twiddle3re = f64x2_splat(tw3.re); |
| 213 |
let twiddle3im = f64x2_splat(tw3.im); |
| 214 |
|
| 215 |
Self { |
| 216 |
direction, |
| 217 |
_phantom: std::marker::PhantomData, |
| 218 |
rotate, |
| 219 |
twiddle1re, |
| 220 |
twiddle1im, |
| 221 |
twiddle2re, |
| 222 |
twiddle2im, |
| 223 |
twiddle3re, |
| 224 |
twiddle3im, |
| 225 |
} |
| 226 |
} |
| 227 |
|
| 228 |
#[inline(always)] |
| 229 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 230 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6}); |
| 231 |
|
| 232 |
let out = self.perform_fft_direct(values); |
| 233 |
|
| 234 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6}); |
| 235 |
} |
| 236 |
|
| 237 |
#[inline(always)] |
| 238 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 7]) -> [v128; 7] { |
| 239 |
let [x1p6, x1m6] = solo_fft2_f64(values[1], values[6]); |
| 240 |
let [x2p5, x2m5] = solo_fft2_f64(values[2], values[5]); |
| 241 |
let [x3p4, x3m4] = solo_fft2_f64(values[3], values[4]); |
| 242 |
|
| 243 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p6); |
| 244 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p5); |
| 245 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p4); |
| 246 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p6); |
| 247 |
let t_a2_2 = f64x2_mul(self.twiddle3re, x2p5); |
| 248 |
let t_a2_3 = f64x2_mul(self.twiddle1re, x3p4); |
| 249 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p6); |
| 250 |
let t_a3_2 = f64x2_mul(self.twiddle1re, x2p5); |
| 251 |
let t_a3_3 = f64x2_mul(self.twiddle2re, x3p4); |
| 252 |
|
| 253 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m6); |
| 254 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m5); |
| 255 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m4); |
| 256 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m6); |
| 257 |
let t_b2_2 = f64x2_mul(self.twiddle3im, x2m5); |
| 258 |
let t_b2_3 = f64x2_mul(self.twiddle1im, x3m4); |
| 259 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m6); |
| 260 |
let t_b3_2 = f64x2_mul(self.twiddle1im, x2m5); |
| 261 |
let t_b3_3 = f64x2_mul(self.twiddle2im, x3m4); |
| 262 |
|
| 263 |
let x0 = values[0]; |
| 264 |
let t_a1 = calc_f64!(x0 + t_a1_1 + t_a1_2 + t_a1_3); |
| 265 |
let t_a2 = calc_f64!(x0 + t_a2_1 + t_a2_2 + t_a2_3); |
| 266 |
let t_a3 = calc_f64!(x0 + t_a3_1 + t_a3_2 + t_a3_3); |
| 267 |
|
| 268 |
let t_b1 = calc_f64!(t_b1_1 + t_b1_2 + t_b1_3); |
| 269 |
let t_b2 = calc_f64!(t_b2_1 - t_b2_2 - t_b2_3); |
| 270 |
let t_b3 = calc_f64!(t_b3_1 - t_b3_2 + t_b3_3); |
| 271 |
|
| 272 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 273 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 274 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 275 |
|
| 276 |
let y0 = calc_f64!(x0 + x1p6 + x2p5 + x3p4); |
| 277 |
let [y1, y6] = solo_fft2_f64(t_a1, t_b1_rot); |
| 278 |
let [y2, y5] = solo_fft2_f64(t_a2, t_b2_rot); |
| 279 |
let [y3, y4] = solo_fft2_f64(t_a3, t_b3_rot); |
| 280 |
[y0, y1, y2, y3, y4, y5, y6] |
| 281 |
} |
| 282 |
} |
| 283 |
|
| 284 |
// _ _ _________ _ _ _ |
| 285 |
// / / | |___ /___ \| |__ (_) |_ |
| 286 |
// | | | _____ |_ \ __) | '_ \| | __| |
| 287 |
// | | | |_____| ___) / __/| |_) | | |_ |
| 288 |
// |_|_| |____/_____|_.__/|_|\__| |
| 289 |
// |
| 290 |
|
| 291 |
pub struct WasmSimdF32Butterfly11<T> { |
| 292 |
direction: FftDirection, |
| 293 |
_phantom: std::marker::PhantomData<T>, |
| 294 |
rotate: Rotate90F32, |
| 295 |
twiddle1re: v128, |
| 296 |
twiddle1im: v128, |
| 297 |
twiddle2re: v128, |
| 298 |
twiddle2im: v128, |
| 299 |
twiddle3re: v128, |
| 300 |
twiddle3im: v128, |
| 301 |
twiddle4re: v128, |
| 302 |
twiddle4im: v128, |
| 303 |
twiddle5re: v128, |
| 304 |
twiddle5im: v128, |
| 305 |
} |
| 306 |
|
| 307 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 308 |
WasmSimdF32Butterfly11, |
| 309 |
11, |
| 310 |
|this: &WasmSimdF32Butterfly11<_>| this.direction |
| 311 |
); |
| 312 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 313 |
WasmSimdF32Butterfly11, |
| 314 |
11, |
| 315 |
|this: &WasmSimdF32Butterfly11<_>| this.direction |
| 316 |
); |
| 317 |
impl<T: FftNum> WasmSimdF32Butterfly11<T> { |
| 318 |
#[inline(always)] |
| 319 |
pub fn new(direction: FftDirection) -> Self { |
| 320 |
assert_f32::<T>(); |
| 321 |
let rotate = Rotate90F32::new(true); |
| 322 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 11, direction); |
| 323 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 11, direction); |
| 324 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 11, direction); |
| 325 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 11, direction); |
| 326 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 11, direction); |
| 327 |
let twiddle1re = f32x4_splat(tw1.re); |
| 328 |
let twiddle1im = f32x4_splat(tw1.im); |
| 329 |
let twiddle2re = f32x4_splat(tw2.re); |
| 330 |
let twiddle2im = f32x4_splat(tw2.im); |
| 331 |
let twiddle3re = f32x4_splat(tw3.re); |
| 332 |
let twiddle3im = f32x4_splat(tw3.im); |
| 333 |
let twiddle4re = f32x4_splat(tw4.re); |
| 334 |
let twiddle4im = f32x4_splat(tw4.im); |
| 335 |
let twiddle5re = f32x4_splat(tw5.re); |
| 336 |
let twiddle5im = f32x4_splat(tw5.im); |
| 337 |
|
| 338 |
Self { |
| 339 |
direction, |
| 340 |
_phantom: std::marker::PhantomData, |
| 341 |
rotate, |
| 342 |
twiddle1re, |
| 343 |
twiddle1im, |
| 344 |
twiddle2re, |
| 345 |
twiddle2im, |
| 346 |
twiddle3re, |
| 347 |
twiddle3im, |
| 348 |
twiddle4re, |
| 349 |
twiddle4im, |
| 350 |
twiddle5re, |
| 351 |
twiddle5im, |
| 352 |
} |
| 353 |
} |
| 354 |
|
| 355 |
#[inline(always)] |
| 356 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 357 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); |
| 358 |
|
| 359 |
let out = self.perform_parallel_fft_direct(values); |
| 360 |
|
| 361 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); |
| 362 |
} |
| 363 |
|
| 364 |
#[inline(always)] |
| 365 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 366 |
&self, |
| 367 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 368 |
) { |
| 369 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20}); |
| 370 |
|
| 371 |
let values = [ |
| 372 |
extract_lo_hi_f32(input_packed[0], input_packed[5]), |
| 373 |
extract_hi_lo_f32(input_packed[0], input_packed[6]), |
| 374 |
extract_lo_hi_f32(input_packed[1], input_packed[6]), |
| 375 |
extract_hi_lo_f32(input_packed[1], input_packed[7]), |
| 376 |
extract_lo_hi_f32(input_packed[2], input_packed[7]), |
| 377 |
extract_hi_lo_f32(input_packed[2], input_packed[8]), |
| 378 |
extract_lo_hi_f32(input_packed[3], input_packed[8]), |
| 379 |
extract_hi_lo_f32(input_packed[3], input_packed[9]), |
| 380 |
extract_lo_hi_f32(input_packed[4], input_packed[9]), |
| 381 |
extract_hi_lo_f32(input_packed[4], input_packed[10]), |
| 382 |
extract_lo_hi_f32(input_packed[5], input_packed[10]), |
| 383 |
]; |
| 384 |
|
| 385 |
let out = self.perform_parallel_fft_direct(values); |
| 386 |
|
| 387 |
let out_packed = [ |
| 388 |
extract_lo_lo_f32(out[0], out[1]), |
| 389 |
extract_lo_lo_f32(out[2], out[3]), |
| 390 |
extract_lo_lo_f32(out[4], out[5]), |
| 391 |
extract_lo_lo_f32(out[6], out[7]), |
| 392 |
extract_lo_lo_f32(out[8], out[9]), |
| 393 |
extract_lo_hi_f32(out[10], out[0]), |
| 394 |
extract_hi_hi_f32(out[1], out[2]), |
| 395 |
extract_hi_hi_f32(out[3], out[4]), |
| 396 |
extract_hi_hi_f32(out[5], out[6]), |
| 397 |
extract_hi_hi_f32(out[7], out[8]), |
| 398 |
extract_hi_hi_f32(out[9], out[10]), |
| 399 |
]; |
| 400 |
|
| 401 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); |
| 402 |
} |
| 403 |
|
| 404 |
#[inline(always)] |
| 405 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 11]) -> [v128; 11] { |
| 406 |
let [x1p10, x1m10] = parallel_fft2_interleaved_f32(values[1], values[10]); |
| 407 |
let [x2p9, x2m9] = parallel_fft2_interleaved_f32(values[2], values[9]); |
| 408 |
let [x3p8, x3m8] = parallel_fft2_interleaved_f32(values[3], values[8]); |
| 409 |
let [x4p7, x4m7] = parallel_fft2_interleaved_f32(values[4], values[7]); |
| 410 |
let [x5p6, x5m6] = parallel_fft2_interleaved_f32(values[5], values[6]); |
| 411 |
|
| 412 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p10); |
| 413 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p9); |
| 414 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p8); |
| 415 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p7); |
| 416 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p6); |
| 417 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p10); |
| 418 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p9); |
| 419 |
let t_a2_3 = f32x4_mul(self.twiddle5re, x3p8); |
| 420 |
let t_a2_4 = f32x4_mul(self.twiddle3re, x4p7); |
| 421 |
let t_a2_5 = f32x4_mul(self.twiddle1re, x5p6); |
| 422 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p10); |
| 423 |
let t_a3_2 = f32x4_mul(self.twiddle5re, x2p9); |
| 424 |
let t_a3_3 = f32x4_mul(self.twiddle2re, x3p8); |
| 425 |
let t_a3_4 = f32x4_mul(self.twiddle1re, x4p7); |
| 426 |
let t_a3_5 = f32x4_mul(self.twiddle4re, x5p6); |
| 427 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p10); |
| 428 |
let t_a4_2 = f32x4_mul(self.twiddle3re, x2p9); |
| 429 |
let t_a4_3 = f32x4_mul(self.twiddle1re, x3p8); |
| 430 |
let t_a4_4 = f32x4_mul(self.twiddle5re, x4p7); |
| 431 |
let t_a4_5 = f32x4_mul(self.twiddle2re, x5p6); |
| 432 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p10); |
| 433 |
let t_a5_2 = f32x4_mul(self.twiddle1re, x2p9); |
| 434 |
let t_a5_3 = f32x4_mul(self.twiddle4re, x3p8); |
| 435 |
let t_a5_4 = f32x4_mul(self.twiddle2re, x4p7); |
| 436 |
let t_a5_5 = f32x4_mul(self.twiddle3re, x5p6); |
| 437 |
|
| 438 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m10); |
| 439 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m9); |
| 440 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m8); |
| 441 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m7); |
| 442 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m6); |
| 443 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m10); |
| 444 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m9); |
| 445 |
let t_b2_3 = f32x4_mul(self.twiddle5im, x3m8); |
| 446 |
let t_b2_4 = f32x4_mul(self.twiddle3im, x4m7); |
| 447 |
let t_b2_5 = f32x4_mul(self.twiddle1im, x5m6); |
| 448 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m10); |
| 449 |
let t_b3_2 = f32x4_mul(self.twiddle5im, x2m9); |
| 450 |
let t_b3_3 = f32x4_mul(self.twiddle2im, x3m8); |
| 451 |
let t_b3_4 = f32x4_mul(self.twiddle1im, x4m7); |
| 452 |
let t_b3_5 = f32x4_mul(self.twiddle4im, x5m6); |
| 453 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m10); |
| 454 |
let t_b4_2 = f32x4_mul(self.twiddle3im, x2m9); |
| 455 |
let t_b4_3 = f32x4_mul(self.twiddle1im, x3m8); |
| 456 |
let t_b4_4 = f32x4_mul(self.twiddle5im, x4m7); |
| 457 |
let t_b4_5 = f32x4_mul(self.twiddle2im, x5m6); |
| 458 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m10); |
| 459 |
let t_b5_2 = f32x4_mul(self.twiddle1im, x2m9); |
| 460 |
let t_b5_3 = f32x4_mul(self.twiddle4im, x3m8); |
| 461 |
let t_b5_4 = f32x4_mul(self.twiddle2im, x4m7); |
| 462 |
let t_b5_5 = f32x4_mul(self.twiddle3im, x5m6); |
| 463 |
|
| 464 |
let x0 = values[0]; |
| 465 |
let t_a1 = calc_f32!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5); |
| 466 |
let t_a2 = calc_f32!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5); |
| 467 |
let t_a3 = calc_f32!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5); |
| 468 |
let t_a4 = calc_f32!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5); |
| 469 |
let t_a5 = calc_f32!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5); |
| 470 |
|
| 471 |
let t_b1 = calc_f32!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5); |
| 472 |
let t_b2 = calc_f32!(t_b2_1 + t_b2_2 - t_b2_3 - t_b2_4 - t_b2_5); |
| 473 |
let t_b3 = calc_f32!(t_b3_1 - t_b3_2 - t_b3_3 + t_b3_4 + t_b3_5); |
| 474 |
let t_b4 = calc_f32!(t_b4_1 - t_b4_2 + t_b4_3 + t_b4_4 - t_b4_5); |
| 475 |
let t_b5 = calc_f32!(t_b5_1 - t_b5_2 + t_b5_3 - t_b5_4 + t_b5_5); |
| 476 |
|
| 477 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 478 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 479 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 480 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 481 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 482 |
|
| 483 |
let y0 = calc_f32!(x0 + x1p10 + x2p9 + x3p8 + x4p7 + x5p6); |
| 484 |
let [y1, y10] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 485 |
let [y2, y9] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 486 |
let [y3, y8] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 487 |
let [y4, y7] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 488 |
let [y5, y6] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 489 |
[y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10] |
| 490 |
} |
| 491 |
} |
| 492 |
|
| 493 |
// _ _ __ _ _ _ _ _ |
| 494 |
// / / | / /_ | || | | |__ (_) |_ |
| 495 |
// | | | _____ | '_ \| || |_| '_ \| | __| |
| 496 |
// | | | |_____| | (_) |__ _| |_) | | |_ |
| 497 |
// |_|_| \___/ |_| |_.__/|_|\__| |
| 498 |
// |
| 499 |
|
| 500 |
pub struct WasmSimdF64Butterfly11<T> { |
| 501 |
direction: FftDirection, |
| 502 |
_phantom: std::marker::PhantomData<T>, |
| 503 |
rotate: Rotate90F64, |
| 504 |
twiddle1re: v128, |
| 505 |
twiddle1im: v128, |
| 506 |
twiddle2re: v128, |
| 507 |
twiddle2im: v128, |
| 508 |
twiddle3re: v128, |
| 509 |
twiddle3im: v128, |
| 510 |
twiddle4re: v128, |
| 511 |
twiddle4im: v128, |
| 512 |
twiddle5re: v128, |
| 513 |
twiddle5im: v128, |
| 514 |
} |
| 515 |
|
| 516 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 517 |
WasmSimdF64Butterfly11, |
| 518 |
11, |
| 519 |
|this: &WasmSimdF64Butterfly11<_>| this.direction |
| 520 |
); |
| 521 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 522 |
WasmSimdF64Butterfly11, |
| 523 |
11, |
| 524 |
|this: &WasmSimdF64Butterfly11<_>| this.direction |
| 525 |
); |
| 526 |
impl<T: FftNum> WasmSimdF64Butterfly11<T> { |
| 527 |
#[inline(always)] |
| 528 |
pub fn new(direction: FftDirection) -> Self { |
| 529 |
assert_f64::<T>(); |
| 530 |
let rotate = Rotate90F64::new(true); |
| 531 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 11, direction); |
| 532 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 11, direction); |
| 533 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 11, direction); |
| 534 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 11, direction); |
| 535 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 11, direction); |
| 536 |
let twiddle1re = f64x2_splat(tw1.re); |
| 537 |
let twiddle1im = f64x2_splat(tw1.im); |
| 538 |
let twiddle2re = f64x2_splat(tw2.re); |
| 539 |
let twiddle2im = f64x2_splat(tw2.im); |
| 540 |
let twiddle3re = f64x2_splat(tw3.re); |
| 541 |
let twiddle3im = f64x2_splat(tw3.im); |
| 542 |
let twiddle4re = f64x2_splat(tw4.re); |
| 543 |
let twiddle4im = f64x2_splat(tw4.im); |
| 544 |
let twiddle5re = f64x2_splat(tw5.re); |
| 545 |
let twiddle5im = f64x2_splat(tw5.im); |
| 546 |
|
| 547 |
Self { |
| 548 |
direction, |
| 549 |
_phantom: std::marker::PhantomData, |
| 550 |
rotate, |
| 551 |
twiddle1re, |
| 552 |
twiddle1im, |
| 553 |
twiddle2re, |
| 554 |
twiddle2im, |
| 555 |
twiddle3re, |
| 556 |
twiddle3im, |
| 557 |
twiddle4re, |
| 558 |
twiddle4im, |
| 559 |
twiddle5re, |
| 560 |
twiddle5im, |
| 561 |
} |
| 562 |
} |
| 563 |
|
| 564 |
#[inline(always)] |
| 565 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 566 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); |
| 567 |
|
| 568 |
let out = self.perform_fft_direct(values); |
| 569 |
|
| 570 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10}); |
| 571 |
} |
| 572 |
|
| 573 |
#[inline(always)] |
| 574 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 11]) -> [v128; 11] { |
| 575 |
let [x1p10, x1m10] = solo_fft2_f64(values[1], values[10]); |
| 576 |
let [x2p9, x2m9] = solo_fft2_f64(values[2], values[9]); |
| 577 |
let [x3p8, x3m8] = solo_fft2_f64(values[3], values[8]); |
| 578 |
let [x4p7, x4m7] = solo_fft2_f64(values[4], values[7]); |
| 579 |
let [x5p6, x5m6] = solo_fft2_f64(values[5], values[6]); |
| 580 |
|
| 581 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p10); |
| 582 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p9); |
| 583 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p8); |
| 584 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p7); |
| 585 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p6); |
| 586 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p10); |
| 587 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p9); |
| 588 |
let t_a2_3 = f64x2_mul(self.twiddle5re, x3p8); |
| 589 |
let t_a2_4 = f64x2_mul(self.twiddle3re, x4p7); |
| 590 |
let t_a2_5 = f64x2_mul(self.twiddle1re, x5p6); |
| 591 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p10); |
| 592 |
let t_a3_2 = f64x2_mul(self.twiddle5re, x2p9); |
| 593 |
let t_a3_3 = f64x2_mul(self.twiddle2re, x3p8); |
| 594 |
let t_a3_4 = f64x2_mul(self.twiddle1re, x4p7); |
| 595 |
let t_a3_5 = f64x2_mul(self.twiddle4re, x5p6); |
| 596 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p10); |
| 597 |
let t_a4_2 = f64x2_mul(self.twiddle3re, x2p9); |
| 598 |
let t_a4_3 = f64x2_mul(self.twiddle1re, x3p8); |
| 599 |
let t_a4_4 = f64x2_mul(self.twiddle5re, x4p7); |
| 600 |
let t_a4_5 = f64x2_mul(self.twiddle2re, x5p6); |
| 601 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p10); |
| 602 |
let t_a5_2 = f64x2_mul(self.twiddle1re, x2p9); |
| 603 |
let t_a5_3 = f64x2_mul(self.twiddle4re, x3p8); |
| 604 |
let t_a5_4 = f64x2_mul(self.twiddle2re, x4p7); |
| 605 |
let t_a5_5 = f64x2_mul(self.twiddle3re, x5p6); |
| 606 |
|
| 607 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m10); |
| 608 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m9); |
| 609 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m8); |
| 610 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m7); |
| 611 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m6); |
| 612 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m10); |
| 613 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m9); |
| 614 |
let t_b2_3 = f64x2_mul(self.twiddle5im, x3m8); |
| 615 |
let t_b2_4 = f64x2_mul(self.twiddle3im, x4m7); |
| 616 |
let t_b2_5 = f64x2_mul(self.twiddle1im, x5m6); |
| 617 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m10); |
| 618 |
let t_b3_2 = f64x2_mul(self.twiddle5im, x2m9); |
| 619 |
let t_b3_3 = f64x2_mul(self.twiddle2im, x3m8); |
| 620 |
let t_b3_4 = f64x2_mul(self.twiddle1im, x4m7); |
| 621 |
let t_b3_5 = f64x2_mul(self.twiddle4im, x5m6); |
| 622 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m10); |
| 623 |
let t_b4_2 = f64x2_mul(self.twiddle3im, x2m9); |
| 624 |
let t_b4_3 = f64x2_mul(self.twiddle1im, x3m8); |
| 625 |
let t_b4_4 = f64x2_mul(self.twiddle5im, x4m7); |
| 626 |
let t_b4_5 = f64x2_mul(self.twiddle2im, x5m6); |
| 627 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m10); |
| 628 |
let t_b5_2 = f64x2_mul(self.twiddle1im, x2m9); |
| 629 |
let t_b5_3 = f64x2_mul(self.twiddle4im, x3m8); |
| 630 |
let t_b5_4 = f64x2_mul(self.twiddle2im, x4m7); |
| 631 |
let t_b5_5 = f64x2_mul(self.twiddle3im, x5m6); |
| 632 |
|
| 633 |
let x0 = values[0]; |
| 634 |
let t_a1 = calc_f64!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5); |
| 635 |
let t_a2 = calc_f64!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5); |
| 636 |
let t_a3 = calc_f64!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5); |
| 637 |
let t_a4 = calc_f64!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5); |
| 638 |
let t_a5 = calc_f64!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5); |
| 639 |
|
| 640 |
let t_b1 = calc_f64!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5); |
| 641 |
let t_b2 = calc_f64!(t_b2_1 + t_b2_2 - t_b2_3 - t_b2_4 - t_b2_5); |
| 642 |
let t_b3 = calc_f64!(t_b3_1 - t_b3_2 - t_b3_3 + t_b3_4 + t_b3_5); |
| 643 |
let t_b4 = calc_f64!(t_b4_1 - t_b4_2 + t_b4_3 + t_b4_4 - t_b4_5); |
| 644 |
let t_b5 = calc_f64!(t_b5_1 - t_b5_2 + t_b5_3 - t_b5_4 + t_b5_5); |
| 645 |
|
| 646 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 647 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 648 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 649 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 650 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 651 |
|
| 652 |
let y0 = calc_f64!(x0 + x1p10 + x2p9 + x3p8 + x4p7 + x5p6); |
| 653 |
let [y1, y10] = solo_fft2_f64(t_a1, t_b1_rot); |
| 654 |
let [y2, y9] = solo_fft2_f64(t_a2, t_b2_rot); |
| 655 |
let [y3, y8] = solo_fft2_f64(t_a3, t_b3_rot); |
| 656 |
let [y4, y7] = solo_fft2_f64(t_a4, t_b4_rot); |
| 657 |
let [y5, y6] = solo_fft2_f64(t_a5, t_b5_rot); |
| 658 |
[y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10] |
| 659 |
} |
| 660 |
} |
| 661 |
|
| 662 |
// _ _____ _________ _ _ _ |
| 663 |
// / |___ / |___ /___ \| |__ (_) |_ |
| 664 |
// | | |_ \ _____ |_ \ __) | '_ \| | __| |
| 665 |
// | |___) | |_____| ___) / __/| |_) | | |_ |
| 666 |
// |_|____/ |____/_____|_.__/|_|\__| |
| 667 |
// |
| 668 |
|
| 669 |
pub struct WasmSimdF32Butterfly13<T> { |
| 670 |
direction: FftDirection, |
| 671 |
_phantom: std::marker::PhantomData<T>, |
| 672 |
rotate: Rotate90F32, |
| 673 |
twiddle1re: v128, |
| 674 |
twiddle1im: v128, |
| 675 |
twiddle2re: v128, |
| 676 |
twiddle2im: v128, |
| 677 |
twiddle3re: v128, |
| 678 |
twiddle3im: v128, |
| 679 |
twiddle4re: v128, |
| 680 |
twiddle4im: v128, |
| 681 |
twiddle5re: v128, |
| 682 |
twiddle5im: v128, |
| 683 |
twiddle6re: v128, |
| 684 |
twiddle6im: v128, |
| 685 |
} |
| 686 |
|
| 687 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 688 |
WasmSimdF32Butterfly13, |
| 689 |
13, |
| 690 |
|this: &WasmSimdF32Butterfly13<_>| this.direction |
| 691 |
); |
| 692 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 693 |
WasmSimdF32Butterfly13, |
| 694 |
13, |
| 695 |
|this: &WasmSimdF32Butterfly13<_>| this.direction |
| 696 |
); |
| 697 |
impl<T: FftNum> WasmSimdF32Butterfly13<T> { |
| 698 |
#[inline(always)] |
| 699 |
pub fn new(direction: FftDirection) -> Self { |
| 700 |
assert_f32::<T>(); |
| 701 |
let rotate = Rotate90F32::new(true); |
| 702 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 13, direction); |
| 703 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 13, direction); |
| 704 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 13, direction); |
| 705 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 13, direction); |
| 706 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 13, direction); |
| 707 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 13, direction); |
| 708 |
let twiddle1re = f32x4_splat(tw1.re); |
| 709 |
let twiddle1im = f32x4_splat(tw1.im); |
| 710 |
let twiddle2re = f32x4_splat(tw2.re); |
| 711 |
let twiddle2im = f32x4_splat(tw2.im); |
| 712 |
let twiddle3re = f32x4_splat(tw3.re); |
| 713 |
let twiddle3im = f32x4_splat(tw3.im); |
| 714 |
let twiddle4re = f32x4_splat(tw4.re); |
| 715 |
let twiddle4im = f32x4_splat(tw4.im); |
| 716 |
let twiddle5re = f32x4_splat(tw5.re); |
| 717 |
let twiddle5im = f32x4_splat(tw5.im); |
| 718 |
let twiddle6re = f32x4_splat(tw6.re); |
| 719 |
let twiddle6im = f32x4_splat(tw6.im); |
| 720 |
|
| 721 |
Self { |
| 722 |
direction, |
| 723 |
_phantom: std::marker::PhantomData, |
| 724 |
rotate, |
| 725 |
twiddle1re, |
| 726 |
twiddle1im, |
| 727 |
twiddle2re, |
| 728 |
twiddle2im, |
| 729 |
twiddle3re, |
| 730 |
twiddle3im, |
| 731 |
twiddle4re, |
| 732 |
twiddle4im, |
| 733 |
twiddle5re, |
| 734 |
twiddle5im, |
| 735 |
twiddle6re, |
| 736 |
twiddle6im, |
| 737 |
} |
| 738 |
} |
| 739 |
|
| 740 |
#[inline(always)] |
| 741 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 742 |
let values = |
| 743 |
read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}); |
| 744 |
|
| 745 |
let out = self.perform_parallel_fft_direct(values); |
| 746 |
|
| 747 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}); |
| 748 |
} |
| 749 |
|
| 750 |
#[inline(always)] |
| 751 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 752 |
&self, |
| 753 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 754 |
) { |
| 755 |
let input_packed = |
| 756 |
read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24}); |
| 757 |
|
| 758 |
let values = [ |
| 759 |
extract_lo_hi_f32(input_packed[0], input_packed[6]), |
| 760 |
extract_hi_lo_f32(input_packed[0], input_packed[7]), |
| 761 |
extract_lo_hi_f32(input_packed[1], input_packed[7]), |
| 762 |
extract_hi_lo_f32(input_packed[1], input_packed[8]), |
| 763 |
extract_lo_hi_f32(input_packed[2], input_packed[8]), |
| 764 |
extract_hi_lo_f32(input_packed[2], input_packed[9]), |
| 765 |
extract_lo_hi_f32(input_packed[3], input_packed[9]), |
| 766 |
extract_hi_lo_f32(input_packed[3], input_packed[10]), |
| 767 |
extract_lo_hi_f32(input_packed[4], input_packed[10]), |
| 768 |
extract_hi_lo_f32(input_packed[4], input_packed[11]), |
| 769 |
extract_lo_hi_f32(input_packed[5], input_packed[11]), |
| 770 |
extract_hi_lo_f32(input_packed[5], input_packed[12]), |
| 771 |
extract_lo_hi_f32(input_packed[6], input_packed[12]), |
| 772 |
]; |
| 773 |
|
| 774 |
let out = self.perform_parallel_fft_direct(values); |
| 775 |
|
| 776 |
let out_packed = [ |
| 777 |
extract_lo_lo_f32(out[0], out[1]), |
| 778 |
extract_lo_lo_f32(out[2], out[3]), |
| 779 |
extract_lo_lo_f32(out[4], out[5]), |
| 780 |
extract_lo_lo_f32(out[6], out[7]), |
| 781 |
extract_lo_lo_f32(out[8], out[9]), |
| 782 |
extract_lo_lo_f32(out[10], out[11]), |
| 783 |
extract_lo_hi_f32(out[12], out[0]), |
| 784 |
extract_hi_hi_f32(out[1], out[2]), |
| 785 |
extract_hi_hi_f32(out[3], out[4]), |
| 786 |
extract_hi_hi_f32(out[5], out[6]), |
| 787 |
extract_hi_hi_f32(out[7], out[8]), |
| 788 |
extract_hi_hi_f32(out[9], out[10]), |
| 789 |
extract_hi_hi_f32(out[11], out[12]), |
| 790 |
]; |
| 791 |
|
| 792 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}); |
| 793 |
} |
| 794 |
|
| 795 |
#[inline(always)] |
| 796 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 13]) -> [v128; 13] { |
| 797 |
let [x1p12, x1m12] = parallel_fft2_interleaved_f32(values[1], values[12]); |
| 798 |
let [x2p11, x2m11] = parallel_fft2_interleaved_f32(values[2], values[11]); |
| 799 |
let [x3p10, x3m10] = parallel_fft2_interleaved_f32(values[3], values[10]); |
| 800 |
let [x4p9, x4m9] = parallel_fft2_interleaved_f32(values[4], values[9]); |
| 801 |
let [x5p8, x5m8] = parallel_fft2_interleaved_f32(values[5], values[8]); |
| 802 |
let [x6p7, x6m7] = parallel_fft2_interleaved_f32(values[6], values[7]); |
| 803 |
|
| 804 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p12); |
| 805 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p11); |
| 806 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p10); |
| 807 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p9); |
| 808 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p8); |
| 809 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p7); |
| 810 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p12); |
| 811 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p11); |
| 812 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p10); |
| 813 |
let t_a2_4 = f32x4_mul(self.twiddle5re, x4p9); |
| 814 |
let t_a2_5 = f32x4_mul(self.twiddle3re, x5p8); |
| 815 |
let t_a2_6 = f32x4_mul(self.twiddle1re, x6p7); |
| 816 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p12); |
| 817 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p11); |
| 818 |
let t_a3_3 = f32x4_mul(self.twiddle4re, x3p10); |
| 819 |
let t_a3_4 = f32x4_mul(self.twiddle1re, x4p9); |
| 820 |
let t_a3_5 = f32x4_mul(self.twiddle2re, x5p8); |
| 821 |
let t_a3_6 = f32x4_mul(self.twiddle5re, x6p7); |
| 822 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p12); |
| 823 |
let t_a4_2 = f32x4_mul(self.twiddle5re, x2p11); |
| 824 |
let t_a4_3 = f32x4_mul(self.twiddle1re, x3p10); |
| 825 |
let t_a4_4 = f32x4_mul(self.twiddle3re, x4p9); |
| 826 |
let t_a4_5 = f32x4_mul(self.twiddle6re, x5p8); |
| 827 |
let t_a4_6 = f32x4_mul(self.twiddle2re, x6p7); |
| 828 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p12); |
| 829 |
let t_a5_2 = f32x4_mul(self.twiddle3re, x2p11); |
| 830 |
let t_a5_3 = f32x4_mul(self.twiddle2re, x3p10); |
| 831 |
let t_a5_4 = f32x4_mul(self.twiddle6re, x4p9); |
| 832 |
let t_a5_5 = f32x4_mul(self.twiddle1re, x5p8); |
| 833 |
let t_a5_6 = f32x4_mul(self.twiddle4re, x6p7); |
| 834 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p12); |
| 835 |
let t_a6_2 = f32x4_mul(self.twiddle1re, x2p11); |
| 836 |
let t_a6_3 = f32x4_mul(self.twiddle5re, x3p10); |
| 837 |
let t_a6_4 = f32x4_mul(self.twiddle2re, x4p9); |
| 838 |
let t_a6_5 = f32x4_mul(self.twiddle4re, x5p8); |
| 839 |
let t_a6_6 = f32x4_mul(self.twiddle3re, x6p7); |
| 840 |
|
| 841 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m12); |
| 842 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m11); |
| 843 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m10); |
| 844 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m9); |
| 845 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m8); |
| 846 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m7); |
| 847 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m12); |
| 848 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m11); |
| 849 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m10); |
| 850 |
let t_b2_4 = f32x4_mul(self.twiddle5im, x4m9); |
| 851 |
let t_b2_5 = f32x4_mul(self.twiddle3im, x5m8); |
| 852 |
let t_b2_6 = f32x4_mul(self.twiddle1im, x6m7); |
| 853 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m12); |
| 854 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m11); |
| 855 |
let t_b3_3 = f32x4_mul(self.twiddle4im, x3m10); |
| 856 |
let t_b3_4 = f32x4_mul(self.twiddle1im, x4m9); |
| 857 |
let t_b3_5 = f32x4_mul(self.twiddle2im, x5m8); |
| 858 |
let t_b3_6 = f32x4_mul(self.twiddle5im, x6m7); |
| 859 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m12); |
| 860 |
let t_b4_2 = f32x4_mul(self.twiddle5im, x2m11); |
| 861 |
let t_b4_3 = f32x4_mul(self.twiddle1im, x3m10); |
| 862 |
let t_b4_4 = f32x4_mul(self.twiddle3im, x4m9); |
| 863 |
let t_b4_5 = f32x4_mul(self.twiddle6im, x5m8); |
| 864 |
let t_b4_6 = f32x4_mul(self.twiddle2im, x6m7); |
| 865 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m12); |
| 866 |
let t_b5_2 = f32x4_mul(self.twiddle3im, x2m11); |
| 867 |
let t_b5_3 = f32x4_mul(self.twiddle2im, x3m10); |
| 868 |
let t_b5_4 = f32x4_mul(self.twiddle6im, x4m9); |
| 869 |
let t_b5_5 = f32x4_mul(self.twiddle1im, x5m8); |
| 870 |
let t_b5_6 = f32x4_mul(self.twiddle4im, x6m7); |
| 871 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m12); |
| 872 |
let t_b6_2 = f32x4_mul(self.twiddle1im, x2m11); |
| 873 |
let t_b6_3 = f32x4_mul(self.twiddle5im, x3m10); |
| 874 |
let t_b6_4 = f32x4_mul(self.twiddle2im, x4m9); |
| 875 |
let t_b6_5 = f32x4_mul(self.twiddle4im, x5m8); |
| 876 |
let t_b6_6 = f32x4_mul(self.twiddle3im, x6m7); |
| 877 |
|
| 878 |
let x0 = values[0]; |
| 879 |
let t_a1 = calc_f32!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6); |
| 880 |
let t_a2 = calc_f32!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6); |
| 881 |
let t_a3 = calc_f32!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6); |
| 882 |
let t_a4 = calc_f32!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6); |
| 883 |
let t_a5 = calc_f32!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6); |
| 884 |
let t_a6 = calc_f32!(x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6); |
| 885 |
|
| 886 |
let t_b1 = calc_f32!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6); |
| 887 |
let t_b2 = calc_f32!(t_b2_1 + t_b2_2 + t_b2_3 - t_b2_4 - t_b2_5 - t_b2_6); |
| 888 |
let t_b3 = calc_f32!(t_b3_1 + t_b3_2 - t_b3_3 - t_b3_4 + t_b3_5 + t_b3_6); |
| 889 |
let t_b4 = calc_f32!(t_b4_1 - t_b4_2 - t_b4_3 + t_b4_4 - t_b4_5 - t_b4_6); |
| 890 |
let t_b5 = calc_f32!(t_b5_1 - t_b5_2 + t_b5_3 - t_b5_4 - t_b5_5 + t_b5_6); |
| 891 |
let t_b6 = calc_f32!(t_b6_1 - t_b6_2 + t_b6_3 - t_b6_4 + t_b6_5 - t_b6_6); |
| 892 |
|
| 893 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 894 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 895 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 896 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 897 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 898 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 899 |
|
| 900 |
let y0 = calc_f32!(x0 + x1p12 + x2p11 + x3p10 + x4p9 + x5p8 + x6p7); |
| 901 |
let [y1, y12] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 902 |
let [y2, y11] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 903 |
let [y3, y10] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 904 |
let [y4, y9] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 905 |
let [y5, y8] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 906 |
let [y6, y7] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 907 |
[y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12] |
| 908 |
} |
| 909 |
} |
| 910 |
|
| 911 |
// _ _____ __ _ _ _ _ _ |
| 912 |
// / |___ / / /_ | || | | |__ (_) |_ |
| 913 |
// | | |_ \ _____ | '_ \| || |_| '_ \| | __| |
| 914 |
// | |___) | |_____| | (_) |__ _| |_) | | |_ |
| 915 |
// |_|____/ \___/ |_| |_.__/|_|\__| |
| 916 |
// |
| 917 |
|
| 918 |
pub struct WasmSimdF64Butterfly13<T> { |
| 919 |
direction: FftDirection, |
| 920 |
_phantom: std::marker::PhantomData<T>, |
| 921 |
rotate: Rotate90F64, |
| 922 |
twiddle1re: v128, |
| 923 |
twiddle1im: v128, |
| 924 |
twiddle2re: v128, |
| 925 |
twiddle2im: v128, |
| 926 |
twiddle3re: v128, |
| 927 |
twiddle3im: v128, |
| 928 |
twiddle4re: v128, |
| 929 |
twiddle4im: v128, |
| 930 |
twiddle5re: v128, |
| 931 |
twiddle5im: v128, |
| 932 |
twiddle6re: v128, |
| 933 |
twiddle6im: v128, |
| 934 |
} |
| 935 |
|
| 936 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 937 |
WasmSimdF64Butterfly13, |
| 938 |
13, |
| 939 |
|this: &WasmSimdF64Butterfly13<_>| this.direction |
| 940 |
); |
| 941 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 942 |
WasmSimdF64Butterfly13, |
| 943 |
13, |
| 944 |
|this: &WasmSimdF64Butterfly13<_>| this.direction |
| 945 |
); |
| 946 |
impl<T: FftNum> WasmSimdF64Butterfly13<T> { |
| 947 |
#[inline(always)] |
| 948 |
pub fn new(direction: FftDirection) -> Self { |
| 949 |
assert_f64::<T>(); |
| 950 |
let rotate = Rotate90F64::new(true); |
| 951 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 13, direction); |
| 952 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 13, direction); |
| 953 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 13, direction); |
| 954 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 13, direction); |
| 955 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 13, direction); |
| 956 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 13, direction); |
| 957 |
let twiddle1re = f64x2_splat(tw1.re); |
| 958 |
let twiddle1im = f64x2_splat(tw1.im); |
| 959 |
let twiddle2re = f64x2_splat(tw2.re); |
| 960 |
let twiddle2im = f64x2_splat(tw2.im); |
| 961 |
let twiddle3re = f64x2_splat(tw3.re); |
| 962 |
let twiddle3im = f64x2_splat(tw3.im); |
| 963 |
let twiddle4re = f64x2_splat(tw4.re); |
| 964 |
let twiddle4im = f64x2_splat(tw4.im); |
| 965 |
let twiddle5re = f64x2_splat(tw5.re); |
| 966 |
let twiddle5im = f64x2_splat(tw5.im); |
| 967 |
let twiddle6re = f64x2_splat(tw6.re); |
| 968 |
let twiddle6im = f64x2_splat(tw6.im); |
| 969 |
|
| 970 |
Self { |
| 971 |
direction, |
| 972 |
_phantom: std::marker::PhantomData, |
| 973 |
rotate, |
| 974 |
twiddle1re, |
| 975 |
twiddle1im, |
| 976 |
twiddle2re, |
| 977 |
twiddle2im, |
| 978 |
twiddle3re, |
| 979 |
twiddle3im, |
| 980 |
twiddle4re, |
| 981 |
twiddle4im, |
| 982 |
twiddle5re, |
| 983 |
twiddle5im, |
| 984 |
twiddle6re, |
| 985 |
twiddle6im, |
| 986 |
} |
| 987 |
} |
| 988 |
|
| 989 |
#[inline(always)] |
| 990 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 991 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}); |
| 992 |
|
| 993 |
let out = self.perform_fft_direct(values); |
| 994 |
|
| 995 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12}); |
| 996 |
} |
| 997 |
|
| 998 |
#[inline(always)] |
| 999 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 13]) -> [v128; 13] { |
| 1000 |
let [x1p12, x1m12] = solo_fft2_f64(values[1], values[12]); |
| 1001 |
let [x2p11, x2m11] = solo_fft2_f64(values[2], values[11]); |
| 1002 |
let [x3p10, x3m10] = solo_fft2_f64(values[3], values[10]); |
| 1003 |
let [x4p9, x4m9] = solo_fft2_f64(values[4], values[9]); |
| 1004 |
let [x5p8, x5m8] = solo_fft2_f64(values[5], values[8]); |
| 1005 |
let [x6p7, x6m7] = solo_fft2_f64(values[6], values[7]); |
| 1006 |
|
| 1007 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p12); |
| 1008 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p11); |
| 1009 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p10); |
| 1010 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p9); |
| 1011 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p8); |
| 1012 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p7); |
| 1013 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p12); |
| 1014 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p11); |
| 1015 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p10); |
| 1016 |
let t_a2_4 = f64x2_mul(self.twiddle5re, x4p9); |
| 1017 |
let t_a2_5 = f64x2_mul(self.twiddle3re, x5p8); |
| 1018 |
let t_a2_6 = f64x2_mul(self.twiddle1re, x6p7); |
| 1019 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p12); |
| 1020 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p11); |
| 1021 |
let t_a3_3 = f64x2_mul(self.twiddle4re, x3p10); |
| 1022 |
let t_a3_4 = f64x2_mul(self.twiddle1re, x4p9); |
| 1023 |
let t_a3_5 = f64x2_mul(self.twiddle2re, x5p8); |
| 1024 |
let t_a3_6 = f64x2_mul(self.twiddle5re, x6p7); |
| 1025 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p12); |
| 1026 |
let t_a4_2 = f64x2_mul(self.twiddle5re, x2p11); |
| 1027 |
let t_a4_3 = f64x2_mul(self.twiddle1re, x3p10); |
| 1028 |
let t_a4_4 = f64x2_mul(self.twiddle3re, x4p9); |
| 1029 |
let t_a4_5 = f64x2_mul(self.twiddle6re, x5p8); |
| 1030 |
let t_a4_6 = f64x2_mul(self.twiddle2re, x6p7); |
| 1031 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p12); |
| 1032 |
let t_a5_2 = f64x2_mul(self.twiddle3re, x2p11); |
| 1033 |
let t_a5_3 = f64x2_mul(self.twiddle2re, x3p10); |
| 1034 |
let t_a5_4 = f64x2_mul(self.twiddle6re, x4p9); |
| 1035 |
let t_a5_5 = f64x2_mul(self.twiddle1re, x5p8); |
| 1036 |
let t_a5_6 = f64x2_mul(self.twiddle4re, x6p7); |
| 1037 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p12); |
| 1038 |
let t_a6_2 = f64x2_mul(self.twiddle1re, x2p11); |
| 1039 |
let t_a6_3 = f64x2_mul(self.twiddle5re, x3p10); |
| 1040 |
let t_a6_4 = f64x2_mul(self.twiddle2re, x4p9); |
| 1041 |
let t_a6_5 = f64x2_mul(self.twiddle4re, x5p8); |
| 1042 |
let t_a6_6 = f64x2_mul(self.twiddle3re, x6p7); |
| 1043 |
|
| 1044 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m12); |
| 1045 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m11); |
| 1046 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m10); |
| 1047 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m9); |
| 1048 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m8); |
| 1049 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m7); |
| 1050 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m12); |
| 1051 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m11); |
| 1052 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m10); |
| 1053 |
let t_b2_4 = f64x2_mul(self.twiddle5im, x4m9); |
| 1054 |
let t_b2_5 = f64x2_mul(self.twiddle3im, x5m8); |
| 1055 |
let t_b2_6 = f64x2_mul(self.twiddle1im, x6m7); |
| 1056 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m12); |
| 1057 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m11); |
| 1058 |
let t_b3_3 = f64x2_mul(self.twiddle4im, x3m10); |
| 1059 |
let t_b3_4 = f64x2_mul(self.twiddle1im, x4m9); |
| 1060 |
let t_b3_5 = f64x2_mul(self.twiddle2im, x5m8); |
| 1061 |
let t_b3_6 = f64x2_mul(self.twiddle5im, x6m7); |
| 1062 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m12); |
| 1063 |
let t_b4_2 = f64x2_mul(self.twiddle5im, x2m11); |
| 1064 |
let t_b4_3 = f64x2_mul(self.twiddle1im, x3m10); |
| 1065 |
let t_b4_4 = f64x2_mul(self.twiddle3im, x4m9); |
| 1066 |
let t_b4_5 = f64x2_mul(self.twiddle6im, x5m8); |
| 1067 |
let t_b4_6 = f64x2_mul(self.twiddle2im, x6m7); |
| 1068 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m12); |
| 1069 |
let t_b5_2 = f64x2_mul(self.twiddle3im, x2m11); |
| 1070 |
let t_b5_3 = f64x2_mul(self.twiddle2im, x3m10); |
| 1071 |
let t_b5_4 = f64x2_mul(self.twiddle6im, x4m9); |
| 1072 |
let t_b5_5 = f64x2_mul(self.twiddle1im, x5m8); |
| 1073 |
let t_b5_6 = f64x2_mul(self.twiddle4im, x6m7); |
| 1074 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m12); |
| 1075 |
let t_b6_2 = f64x2_mul(self.twiddle1im, x2m11); |
| 1076 |
let t_b6_3 = f64x2_mul(self.twiddle5im, x3m10); |
| 1077 |
let t_b6_4 = f64x2_mul(self.twiddle2im, x4m9); |
| 1078 |
let t_b6_5 = f64x2_mul(self.twiddle4im, x5m8); |
| 1079 |
let t_b6_6 = f64x2_mul(self.twiddle3im, x6m7); |
| 1080 |
|
| 1081 |
let x0 = values[0]; |
| 1082 |
let t_a1 = calc_f64!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6); |
| 1083 |
let t_a2 = calc_f64!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6); |
| 1084 |
let t_a3 = calc_f64!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6); |
| 1085 |
let t_a4 = calc_f64!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6); |
| 1086 |
let t_a5 = calc_f64!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6); |
| 1087 |
let t_a6 = calc_f64!(x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6); |
| 1088 |
|
| 1089 |
let t_b1 = calc_f64!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6); |
| 1090 |
let t_b2 = calc_f64!(t_b2_1 + t_b2_2 + t_b2_3 - t_b2_4 - t_b2_5 - t_b2_6); |
| 1091 |
let t_b3 = calc_f64!(t_b3_1 + t_b3_2 - t_b3_3 - t_b3_4 + t_b3_5 + t_b3_6); |
| 1092 |
let t_b4 = calc_f64!(t_b4_1 - t_b4_2 - t_b4_3 + t_b4_4 - t_b4_5 - t_b4_6); |
| 1093 |
let t_b5 = calc_f64!(t_b5_1 - t_b5_2 + t_b5_3 - t_b5_4 - t_b5_5 + t_b5_6); |
| 1094 |
let t_b6 = calc_f64!(t_b6_1 - t_b6_2 + t_b6_3 - t_b6_4 + t_b6_5 - t_b6_6); |
| 1095 |
|
| 1096 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 1097 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 1098 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 1099 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 1100 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 1101 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 1102 |
|
| 1103 |
let y0 = calc_f64!(x0 + x1p12 + x2p11 + x3p10 + x4p9 + x5p8 + x6p7); |
| 1104 |
let [y1, y12] = solo_fft2_f64(t_a1, t_b1_rot); |
| 1105 |
let [y2, y11] = solo_fft2_f64(t_a2, t_b2_rot); |
| 1106 |
let [y3, y10] = solo_fft2_f64(t_a3, t_b3_rot); |
| 1107 |
let [y4, y9] = solo_fft2_f64(t_a4, t_b4_rot); |
| 1108 |
let [y5, y8] = solo_fft2_f64(t_a5, t_b5_rot); |
| 1109 |
let [y6, y7] = solo_fft2_f64(t_a6, t_b6_rot); |
| 1110 |
[y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12] |
| 1111 |
} |
| 1112 |
} |
| 1113 |
|
| 1114 |
// _ _____ _________ _ _ _ |
| 1115 |
// / |___ | |___ /___ \| |__ (_) |_ |
| 1116 |
// | | / / _____ |_ \ __) | '_ \| | __| |
| 1117 |
// | | / / |_____| ___) / __/| |_) | | |_ |
| 1118 |
// |_|/_/ |____/_____|_.__/|_|\__| |
| 1119 |
// |
| 1120 |
|
| 1121 |
pub struct WasmSimdF32Butterfly17<T> { |
| 1122 |
direction: FftDirection, |
| 1123 |
_phantom: std::marker::PhantomData<T>, |
| 1124 |
rotate: Rotate90F32, |
| 1125 |
twiddle1re: v128, |
| 1126 |
twiddle1im: v128, |
| 1127 |
twiddle2re: v128, |
| 1128 |
twiddle2im: v128, |
| 1129 |
twiddle3re: v128, |
| 1130 |
twiddle3im: v128, |
| 1131 |
twiddle4re: v128, |
| 1132 |
twiddle4im: v128, |
| 1133 |
twiddle5re: v128, |
| 1134 |
twiddle5im: v128, |
| 1135 |
twiddle6re: v128, |
| 1136 |
twiddle6im: v128, |
| 1137 |
twiddle7re: v128, |
| 1138 |
twiddle7im: v128, |
| 1139 |
twiddle8re: v128, |
| 1140 |
twiddle8im: v128, |
| 1141 |
} |
| 1142 |
|
| 1143 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 1144 |
WasmSimdF32Butterfly17, |
| 1145 |
17, |
| 1146 |
|this: &WasmSimdF32Butterfly17<_>| this.direction |
| 1147 |
); |
| 1148 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 1149 |
WasmSimdF32Butterfly17, |
| 1150 |
17, |
| 1151 |
|this: &WasmSimdF32Butterfly17<_>| this.direction |
| 1152 |
); |
| 1153 |
impl<T: FftNum> WasmSimdF32Butterfly17<T> { |
| 1154 |
#[inline(always)] |
| 1155 |
pub fn new(direction: FftDirection) -> Self { |
| 1156 |
assert_f32::<T>(); |
| 1157 |
let rotate = Rotate90F32::new(true); |
| 1158 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 17, direction); |
| 1159 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 17, direction); |
| 1160 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 17, direction); |
| 1161 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 17, direction); |
| 1162 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 17, direction); |
| 1163 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 17, direction); |
| 1164 |
let tw7: Complex<f32> = twiddles::compute_twiddle(7, 17, direction); |
| 1165 |
let tw8: Complex<f32> = twiddles::compute_twiddle(8, 17, direction); |
| 1166 |
let twiddle1re = f32x4_splat(tw1.re); |
| 1167 |
let twiddle1im = f32x4_splat(tw1.im); |
| 1168 |
let twiddle2re = f32x4_splat(tw2.re); |
| 1169 |
let twiddle2im = f32x4_splat(tw2.im); |
| 1170 |
let twiddle3re = f32x4_splat(tw3.re); |
| 1171 |
let twiddle3im = f32x4_splat(tw3.im); |
| 1172 |
let twiddle4re = f32x4_splat(tw4.re); |
| 1173 |
let twiddle4im = f32x4_splat(tw4.im); |
| 1174 |
let twiddle5re = f32x4_splat(tw5.re); |
| 1175 |
let twiddle5im = f32x4_splat(tw5.im); |
| 1176 |
let twiddle6re = f32x4_splat(tw6.re); |
| 1177 |
let twiddle6im = f32x4_splat(tw6.im); |
| 1178 |
let twiddle7re = f32x4_splat(tw7.re); |
| 1179 |
let twiddle7im = f32x4_splat(tw7.im); |
| 1180 |
let twiddle8re = f32x4_splat(tw8.re); |
| 1181 |
let twiddle8im = f32x4_splat(tw8.im); |
| 1182 |
|
| 1183 |
Self { |
| 1184 |
direction, |
| 1185 |
_phantom: std::marker::PhantomData, |
| 1186 |
rotate, |
| 1187 |
twiddle1re, |
| 1188 |
twiddle1im, |
| 1189 |
twiddle2re, |
| 1190 |
twiddle2im, |
| 1191 |
twiddle3re, |
| 1192 |
twiddle3im, |
| 1193 |
twiddle4re, |
| 1194 |
twiddle4im, |
| 1195 |
twiddle5re, |
| 1196 |
twiddle5im, |
| 1197 |
twiddle6re, |
| 1198 |
twiddle6im, |
| 1199 |
twiddle7re, |
| 1200 |
twiddle7im, |
| 1201 |
twiddle8re, |
| 1202 |
twiddle8im, |
| 1203 |
} |
| 1204 |
} |
| 1205 |
|
| 1206 |
#[inline(always)] |
| 1207 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 1208 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}); |
| 1209 |
|
| 1210 |
let out = self.perform_parallel_fft_direct(values); |
| 1211 |
|
| 1212 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}); |
| 1213 |
} |
| 1214 |
|
| 1215 |
#[inline(always)] |
| 1216 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 1217 |
&self, |
| 1218 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 1219 |
) { |
| 1220 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32}); |
| 1221 |
|
| 1222 |
let values = [ |
| 1223 |
extract_lo_hi_f32(input_packed[0], input_packed[8]), |
| 1224 |
extract_hi_lo_f32(input_packed[0], input_packed[9]), |
| 1225 |
extract_lo_hi_f32(input_packed[1], input_packed[9]), |
| 1226 |
extract_hi_lo_f32(input_packed[1], input_packed[10]), |
| 1227 |
extract_lo_hi_f32(input_packed[2], input_packed[10]), |
| 1228 |
extract_hi_lo_f32(input_packed[2], input_packed[11]), |
| 1229 |
extract_lo_hi_f32(input_packed[3], input_packed[11]), |
| 1230 |
extract_hi_lo_f32(input_packed[3], input_packed[12]), |
| 1231 |
extract_lo_hi_f32(input_packed[4], input_packed[12]), |
| 1232 |
extract_hi_lo_f32(input_packed[4], input_packed[13]), |
| 1233 |
extract_lo_hi_f32(input_packed[5], input_packed[13]), |
| 1234 |
extract_hi_lo_f32(input_packed[5], input_packed[14]), |
| 1235 |
extract_lo_hi_f32(input_packed[6], input_packed[14]), |
| 1236 |
extract_hi_lo_f32(input_packed[6], input_packed[15]), |
| 1237 |
extract_lo_hi_f32(input_packed[7], input_packed[15]), |
| 1238 |
extract_hi_lo_f32(input_packed[7], input_packed[16]), |
| 1239 |
extract_lo_hi_f32(input_packed[8], input_packed[16]), |
| 1240 |
]; |
| 1241 |
|
| 1242 |
let out = self.perform_parallel_fft_direct(values); |
| 1243 |
|
| 1244 |
let out_packed = [ |
| 1245 |
extract_lo_lo_f32(out[0], out[1]), |
| 1246 |
extract_lo_lo_f32(out[2], out[3]), |
| 1247 |
extract_lo_lo_f32(out[4], out[5]), |
| 1248 |
extract_lo_lo_f32(out[6], out[7]), |
| 1249 |
extract_lo_lo_f32(out[8], out[9]), |
| 1250 |
extract_lo_lo_f32(out[10], out[11]), |
| 1251 |
extract_lo_lo_f32(out[12], out[13]), |
| 1252 |
extract_lo_lo_f32(out[14], out[15]), |
| 1253 |
extract_lo_hi_f32(out[16], out[0]), |
| 1254 |
extract_hi_hi_f32(out[1], out[2]), |
| 1255 |
extract_hi_hi_f32(out[3], out[4]), |
| 1256 |
extract_hi_hi_f32(out[5], out[6]), |
| 1257 |
extract_hi_hi_f32(out[7], out[8]), |
| 1258 |
extract_hi_hi_f32(out[9], out[10]), |
| 1259 |
extract_hi_hi_f32(out[11], out[12]), |
| 1260 |
extract_hi_hi_f32(out[13], out[14]), |
| 1261 |
extract_hi_hi_f32(out[15], out[16]), |
| 1262 |
]; |
| 1263 |
|
| 1264 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}); |
| 1265 |
} |
| 1266 |
|
| 1267 |
#[inline(always)] |
| 1268 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 17]) -> [v128; 17] { |
| 1269 |
let [x1p16, x1m16] = parallel_fft2_interleaved_f32(values[1], values[16]); |
| 1270 |
let [x2p15, x2m15] = parallel_fft2_interleaved_f32(values[2], values[15]); |
| 1271 |
let [x3p14, x3m14] = parallel_fft2_interleaved_f32(values[3], values[14]); |
| 1272 |
let [x4p13, x4m13] = parallel_fft2_interleaved_f32(values[4], values[13]); |
| 1273 |
let [x5p12, x5m12] = parallel_fft2_interleaved_f32(values[5], values[12]); |
| 1274 |
let [x6p11, x6m11] = parallel_fft2_interleaved_f32(values[6], values[11]); |
| 1275 |
let [x7p10, x7m10] = parallel_fft2_interleaved_f32(values[7], values[10]); |
| 1276 |
let [x8p9, x8m9] = parallel_fft2_interleaved_f32(values[8], values[9]); |
| 1277 |
|
| 1278 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p16); |
| 1279 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p15); |
| 1280 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p14); |
| 1281 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p13); |
| 1282 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p12); |
| 1283 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p11); |
| 1284 |
let t_a1_7 = f32x4_mul(self.twiddle7re, x7p10); |
| 1285 |
let t_a1_8 = f32x4_mul(self.twiddle8re, x8p9); |
| 1286 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p16); |
| 1287 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p15); |
| 1288 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p14); |
| 1289 |
let t_a2_4 = f32x4_mul(self.twiddle8re, x4p13); |
| 1290 |
let t_a2_5 = f32x4_mul(self.twiddle7re, x5p12); |
| 1291 |
let t_a2_6 = f32x4_mul(self.twiddle5re, x6p11); |
| 1292 |
let t_a2_7 = f32x4_mul(self.twiddle3re, x7p10); |
| 1293 |
let t_a2_8 = f32x4_mul(self.twiddle1re, x8p9); |
| 1294 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p16); |
| 1295 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p15); |
| 1296 |
let t_a3_3 = f32x4_mul(self.twiddle8re, x3p14); |
| 1297 |
let t_a3_4 = f32x4_mul(self.twiddle5re, x4p13); |
| 1298 |
let t_a3_5 = f32x4_mul(self.twiddle2re, x5p12); |
| 1299 |
let t_a3_6 = f32x4_mul(self.twiddle1re, x6p11); |
| 1300 |
let t_a3_7 = f32x4_mul(self.twiddle4re, x7p10); |
| 1301 |
let t_a3_8 = f32x4_mul(self.twiddle7re, x8p9); |
| 1302 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p16); |
| 1303 |
let t_a4_2 = f32x4_mul(self.twiddle8re, x2p15); |
| 1304 |
let t_a4_3 = f32x4_mul(self.twiddle5re, x3p14); |
| 1305 |
let t_a4_4 = f32x4_mul(self.twiddle1re, x4p13); |
| 1306 |
let t_a4_5 = f32x4_mul(self.twiddle3re, x5p12); |
| 1307 |
let t_a4_6 = f32x4_mul(self.twiddle7re, x6p11); |
| 1308 |
let t_a4_7 = f32x4_mul(self.twiddle6re, x7p10); |
| 1309 |
let t_a4_8 = f32x4_mul(self.twiddle2re, x8p9); |
| 1310 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p16); |
| 1311 |
let t_a5_2 = f32x4_mul(self.twiddle7re, x2p15); |
| 1312 |
let t_a5_3 = f32x4_mul(self.twiddle2re, x3p14); |
| 1313 |
let t_a5_4 = f32x4_mul(self.twiddle3re, x4p13); |
| 1314 |
let t_a5_5 = f32x4_mul(self.twiddle8re, x5p12); |
| 1315 |
let t_a5_6 = f32x4_mul(self.twiddle4re, x6p11); |
| 1316 |
let t_a5_7 = f32x4_mul(self.twiddle1re, x7p10); |
| 1317 |
let t_a5_8 = f32x4_mul(self.twiddle6re, x8p9); |
| 1318 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p16); |
| 1319 |
let t_a6_2 = f32x4_mul(self.twiddle5re, x2p15); |
| 1320 |
let t_a6_3 = f32x4_mul(self.twiddle1re, x3p14); |
| 1321 |
let t_a6_4 = f32x4_mul(self.twiddle7re, x4p13); |
| 1322 |
let t_a6_5 = f32x4_mul(self.twiddle4re, x5p12); |
| 1323 |
let t_a6_6 = f32x4_mul(self.twiddle2re, x6p11); |
| 1324 |
let t_a6_7 = f32x4_mul(self.twiddle8re, x7p10); |
| 1325 |
let t_a6_8 = f32x4_mul(self.twiddle3re, x8p9); |
| 1326 |
let t_a7_1 = f32x4_mul(self.twiddle7re, x1p16); |
| 1327 |
let t_a7_2 = f32x4_mul(self.twiddle3re, x2p15); |
| 1328 |
let t_a7_3 = f32x4_mul(self.twiddle4re, x3p14); |
| 1329 |
let t_a7_4 = f32x4_mul(self.twiddle6re, x4p13); |
| 1330 |
let t_a7_5 = f32x4_mul(self.twiddle1re, x5p12); |
| 1331 |
let t_a7_6 = f32x4_mul(self.twiddle8re, x6p11); |
| 1332 |
let t_a7_7 = f32x4_mul(self.twiddle2re, x7p10); |
| 1333 |
let t_a7_8 = f32x4_mul(self.twiddle5re, x8p9); |
| 1334 |
let t_a8_1 = f32x4_mul(self.twiddle8re, x1p16); |
| 1335 |
let t_a8_2 = f32x4_mul(self.twiddle1re, x2p15); |
| 1336 |
let t_a8_3 = f32x4_mul(self.twiddle7re, x3p14); |
| 1337 |
let t_a8_4 = f32x4_mul(self.twiddle2re, x4p13); |
| 1338 |
let t_a8_5 = f32x4_mul(self.twiddle6re, x5p12); |
| 1339 |
let t_a8_6 = f32x4_mul(self.twiddle3re, x6p11); |
| 1340 |
let t_a8_7 = f32x4_mul(self.twiddle5re, x7p10); |
| 1341 |
let t_a8_8 = f32x4_mul(self.twiddle4re, x8p9); |
| 1342 |
|
| 1343 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m16); |
| 1344 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m15); |
| 1345 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m14); |
| 1346 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m13); |
| 1347 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m12); |
| 1348 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m11); |
| 1349 |
let t_b1_7 = f32x4_mul(self.twiddle7im, x7m10); |
| 1350 |
let t_b1_8 = f32x4_mul(self.twiddle8im, x8m9); |
| 1351 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m16); |
| 1352 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m15); |
| 1353 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m14); |
| 1354 |
let t_b2_4 = f32x4_mul(self.twiddle8im, x4m13); |
| 1355 |
let t_b2_5 = f32x4_mul(self.twiddle7im, x5m12); |
| 1356 |
let t_b2_6 = f32x4_mul(self.twiddle5im, x6m11); |
| 1357 |
let t_b2_7 = f32x4_mul(self.twiddle3im, x7m10); |
| 1358 |
let t_b2_8 = f32x4_mul(self.twiddle1im, x8m9); |
| 1359 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m16); |
| 1360 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m15); |
| 1361 |
let t_b3_3 = f32x4_mul(self.twiddle8im, x3m14); |
| 1362 |
let t_b3_4 = f32x4_mul(self.twiddle5im, x4m13); |
| 1363 |
let t_b3_5 = f32x4_mul(self.twiddle2im, x5m12); |
| 1364 |
let t_b3_6 = f32x4_mul(self.twiddle1im, x6m11); |
| 1365 |
let t_b3_7 = f32x4_mul(self.twiddle4im, x7m10); |
| 1366 |
let t_b3_8 = f32x4_mul(self.twiddle7im, x8m9); |
| 1367 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m16); |
| 1368 |
let t_b4_2 = f32x4_mul(self.twiddle8im, x2m15); |
| 1369 |
let t_b4_3 = f32x4_mul(self.twiddle5im, x3m14); |
| 1370 |
let t_b4_4 = f32x4_mul(self.twiddle1im, x4m13); |
| 1371 |
let t_b4_5 = f32x4_mul(self.twiddle3im, x5m12); |
| 1372 |
let t_b4_6 = f32x4_mul(self.twiddle7im, x6m11); |
| 1373 |
let t_b4_7 = f32x4_mul(self.twiddle6im, x7m10); |
| 1374 |
let t_b4_8 = f32x4_mul(self.twiddle2im, x8m9); |
| 1375 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m16); |
| 1376 |
let t_b5_2 = f32x4_mul(self.twiddle7im, x2m15); |
| 1377 |
let t_b5_3 = f32x4_mul(self.twiddle2im, x3m14); |
| 1378 |
let t_b5_4 = f32x4_mul(self.twiddle3im, x4m13); |
| 1379 |
let t_b5_5 = f32x4_mul(self.twiddle8im, x5m12); |
| 1380 |
let t_b5_6 = f32x4_mul(self.twiddle4im, x6m11); |
| 1381 |
let t_b5_7 = f32x4_mul(self.twiddle1im, x7m10); |
| 1382 |
let t_b5_8 = f32x4_mul(self.twiddle6im, x8m9); |
| 1383 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m16); |
| 1384 |
let t_b6_2 = f32x4_mul(self.twiddle5im, x2m15); |
| 1385 |
let t_b6_3 = f32x4_mul(self.twiddle1im, x3m14); |
| 1386 |
let t_b6_4 = f32x4_mul(self.twiddle7im, x4m13); |
| 1387 |
let t_b6_5 = f32x4_mul(self.twiddle4im, x5m12); |
| 1388 |
let t_b6_6 = f32x4_mul(self.twiddle2im, x6m11); |
| 1389 |
let t_b6_7 = f32x4_mul(self.twiddle8im, x7m10); |
| 1390 |
let t_b6_8 = f32x4_mul(self.twiddle3im, x8m9); |
| 1391 |
let t_b7_1 = f32x4_mul(self.twiddle7im, x1m16); |
| 1392 |
let t_b7_2 = f32x4_mul(self.twiddle3im, x2m15); |
| 1393 |
let t_b7_3 = f32x4_mul(self.twiddle4im, x3m14); |
| 1394 |
let t_b7_4 = f32x4_mul(self.twiddle6im, x4m13); |
| 1395 |
let t_b7_5 = f32x4_mul(self.twiddle1im, x5m12); |
| 1396 |
let t_b7_6 = f32x4_mul(self.twiddle8im, x6m11); |
| 1397 |
let t_b7_7 = f32x4_mul(self.twiddle2im, x7m10); |
| 1398 |
let t_b7_8 = f32x4_mul(self.twiddle5im, x8m9); |
| 1399 |
let t_b8_1 = f32x4_mul(self.twiddle8im, x1m16); |
| 1400 |
let t_b8_2 = f32x4_mul(self.twiddle1im, x2m15); |
| 1401 |
let t_b8_3 = f32x4_mul(self.twiddle7im, x3m14); |
| 1402 |
let t_b8_4 = f32x4_mul(self.twiddle2im, x4m13); |
| 1403 |
let t_b8_5 = f32x4_mul(self.twiddle6im, x5m12); |
| 1404 |
let t_b8_6 = f32x4_mul(self.twiddle3im, x6m11); |
| 1405 |
let t_b8_7 = f32x4_mul(self.twiddle5im, x7m10); |
| 1406 |
let t_b8_8 = f32x4_mul(self.twiddle4im, x8m9); |
| 1407 |
|
| 1408 |
let x0 = values[0]; |
| 1409 |
let t_a1 = |
| 1410 |
calc_f32!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6 + t_a1_7 + t_a1_8); |
| 1411 |
let t_a2 = |
| 1412 |
calc_f32!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6 + t_a2_7 + t_a2_8); |
| 1413 |
let t_a3 = |
| 1414 |
calc_f32!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6 + t_a3_7 + t_a3_8); |
| 1415 |
let t_a4 = |
| 1416 |
calc_f32!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6 + t_a4_7 + t_a4_8); |
| 1417 |
let t_a5 = |
| 1418 |
calc_f32!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6 + t_a5_7 + t_a5_8); |
| 1419 |
let t_a6 = |
| 1420 |
calc_f32!(x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6 + t_a6_7 + t_a6_8); |
| 1421 |
let t_a7 = |
| 1422 |
calc_f32!(x0 + t_a7_1 + t_a7_2 + t_a7_3 + t_a7_4 + t_a7_5 + t_a7_6 + t_a7_7 + t_a7_8); |
| 1423 |
let t_a8 = |
| 1424 |
calc_f32!(x0 + t_a8_1 + t_a8_2 + t_a8_3 + t_a8_4 + t_a8_5 + t_a8_6 + t_a8_7 + t_a8_8); |
| 1425 |
|
| 1426 |
let t_b1 = calc_f32!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6 + t_b1_7 + t_b1_8); |
| 1427 |
let t_b2 = calc_f32!(t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 - t_b2_5 - t_b2_6 - t_b2_7 - t_b2_8); |
| 1428 |
let t_b3 = calc_f32!(t_b3_1 + t_b3_2 - t_b3_3 - t_b3_4 - t_b3_5 + t_b3_6 + t_b3_7 + t_b3_8); |
| 1429 |
let t_b4 = calc_f32!(t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 + t_b4_5 + t_b4_6 - t_b4_7 - t_b4_8); |
| 1430 |
let t_b5 = calc_f32!(t_b5_1 - t_b5_2 - t_b5_3 + t_b5_4 + t_b5_5 - t_b5_6 + t_b5_7 + t_b5_8); |
| 1431 |
let t_b6 = calc_f32!(t_b6_1 - t_b6_2 + t_b6_3 + t_b6_4 - t_b6_5 + t_b6_6 + t_b6_7 - t_b6_8); |
| 1432 |
let t_b7 = calc_f32!(t_b7_1 - t_b7_2 + t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 + t_b7_8); |
| 1433 |
let t_b8 = calc_f32!(t_b8_1 - t_b8_2 + t_b8_3 - t_b8_4 + t_b8_5 - t_b8_6 + t_b8_7 - t_b8_8); |
| 1434 |
|
| 1435 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 1436 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 1437 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 1438 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 1439 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 1440 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 1441 |
let t_b7_rot = self.rotate.rotate_both(t_b7); |
| 1442 |
let t_b8_rot = self.rotate.rotate_both(t_b8); |
| 1443 |
|
| 1444 |
let y0 = calc_f32!(x0 + x1p16 + x2p15 + x3p14 + x4p13 + x5p12 + x6p11 + x7p10 + x8p9); |
| 1445 |
let [y1, y16] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 1446 |
let [y2, y15] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 1447 |
let [y3, y14] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 1448 |
let [y4, y13] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 1449 |
let [y5, y12] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 1450 |
let [y6, y11] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 1451 |
let [y7, y10] = parallel_fft2_interleaved_f32(t_a7, t_b7_rot); |
| 1452 |
let [y8, y9] = parallel_fft2_interleaved_f32(t_a8, t_b8_rot); |
| 1453 |
[ |
| 1454 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, |
| 1455 |
] |
| 1456 |
} |
| 1457 |
} |
| 1458 |
|
| 1459 |
// _ _____ __ _ _ _ _ _ |
| 1460 |
// / |___ | / /_ | || | | |__ (_) |_ |
| 1461 |
// | | / / _____ | '_ \| || |_| '_ \| | __| |
| 1462 |
// | | / / |_____| | (_) |__ _| |_) | | |_ |
| 1463 |
// |_|/_/ \___/ |_| |_.__/|_|\__| |
| 1464 |
// |
| 1465 |
|
| 1466 |
pub struct WasmSimdF64Butterfly17<T> { |
| 1467 |
direction: FftDirection, |
| 1468 |
_phantom: std::marker::PhantomData<T>, |
| 1469 |
rotate: Rotate90F64, |
| 1470 |
twiddle1re: v128, |
| 1471 |
twiddle1im: v128, |
| 1472 |
twiddle2re: v128, |
| 1473 |
twiddle2im: v128, |
| 1474 |
twiddle3re: v128, |
| 1475 |
twiddle3im: v128, |
| 1476 |
twiddle4re: v128, |
| 1477 |
twiddle4im: v128, |
| 1478 |
twiddle5re: v128, |
| 1479 |
twiddle5im: v128, |
| 1480 |
twiddle6re: v128, |
| 1481 |
twiddle6im: v128, |
| 1482 |
twiddle7re: v128, |
| 1483 |
twiddle7im: v128, |
| 1484 |
twiddle8re: v128, |
| 1485 |
twiddle8im: v128, |
| 1486 |
} |
| 1487 |
|
| 1488 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 1489 |
WasmSimdF64Butterfly17, |
| 1490 |
17, |
| 1491 |
|this: &WasmSimdF64Butterfly17<_>| this.direction |
| 1492 |
); |
| 1493 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 1494 |
WasmSimdF64Butterfly17, |
| 1495 |
17, |
| 1496 |
|this: &WasmSimdF64Butterfly17<_>| this.direction |
| 1497 |
); |
| 1498 |
impl<T: FftNum> WasmSimdF64Butterfly17<T> { |
| 1499 |
#[inline(always)] |
| 1500 |
pub fn new(direction: FftDirection) -> Self { |
| 1501 |
assert_f64::<T>(); |
| 1502 |
let rotate = Rotate90F64::new(true); |
| 1503 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 17, direction); |
| 1504 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 17, direction); |
| 1505 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 17, direction); |
| 1506 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 17, direction); |
| 1507 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 17, direction); |
| 1508 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 17, direction); |
| 1509 |
let tw7: Complex<f64> = twiddles::compute_twiddle(7, 17, direction); |
| 1510 |
let tw8: Complex<f64> = twiddles::compute_twiddle(8, 17, direction); |
| 1511 |
let twiddle1re = f64x2_splat(tw1.re); |
| 1512 |
let twiddle1im = f64x2_splat(tw1.im); |
| 1513 |
let twiddle2re = f64x2_splat(tw2.re); |
| 1514 |
let twiddle2im = f64x2_splat(tw2.im); |
| 1515 |
let twiddle3re = f64x2_splat(tw3.re); |
| 1516 |
let twiddle3im = f64x2_splat(tw3.im); |
| 1517 |
let twiddle4re = f64x2_splat(tw4.re); |
| 1518 |
let twiddle4im = f64x2_splat(tw4.im); |
| 1519 |
let twiddle5re = f64x2_splat(tw5.re); |
| 1520 |
let twiddle5im = f64x2_splat(tw5.im); |
| 1521 |
let twiddle6re = f64x2_splat(tw6.re); |
| 1522 |
let twiddle6im = f64x2_splat(tw6.im); |
| 1523 |
let twiddle7re = f64x2_splat(tw7.re); |
| 1524 |
let twiddle7im = f64x2_splat(tw7.im); |
| 1525 |
let twiddle8re = f64x2_splat(tw8.re); |
| 1526 |
let twiddle8im = f64x2_splat(tw8.im); |
| 1527 |
|
| 1528 |
Self { |
| 1529 |
direction, |
| 1530 |
_phantom: std::marker::PhantomData, |
| 1531 |
rotate, |
| 1532 |
twiddle1re, |
| 1533 |
twiddle1im, |
| 1534 |
twiddle2re, |
| 1535 |
twiddle2im, |
| 1536 |
twiddle3re, |
| 1537 |
twiddle3im, |
| 1538 |
twiddle4re, |
| 1539 |
twiddle4im, |
| 1540 |
twiddle5re, |
| 1541 |
twiddle5im, |
| 1542 |
twiddle6re, |
| 1543 |
twiddle6im, |
| 1544 |
twiddle7re, |
| 1545 |
twiddle7im, |
| 1546 |
twiddle8re, |
| 1547 |
twiddle8im, |
| 1548 |
} |
| 1549 |
} |
| 1550 |
|
| 1551 |
#[inline(always)] |
| 1552 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 1553 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}); |
| 1554 |
|
| 1555 |
let out = self.perform_fft_direct(values); |
| 1556 |
|
| 1557 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16}); |
| 1558 |
} |
| 1559 |
|
| 1560 |
#[inline(always)] |
| 1561 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 17]) -> [v128; 17] { |
| 1562 |
let [x1p16, x1m16] = solo_fft2_f64(values[1], values[16]); |
| 1563 |
let [x2p15, x2m15] = solo_fft2_f64(values[2], values[15]); |
| 1564 |
let [x3p14, x3m14] = solo_fft2_f64(values[3], values[14]); |
| 1565 |
let [x4p13, x4m13] = solo_fft2_f64(values[4], values[13]); |
| 1566 |
let [x5p12, x5m12] = solo_fft2_f64(values[5], values[12]); |
| 1567 |
let [x6p11, x6m11] = solo_fft2_f64(values[6], values[11]); |
| 1568 |
let [x7p10, x7m10] = solo_fft2_f64(values[7], values[10]); |
| 1569 |
let [x8p9, x8m9] = solo_fft2_f64(values[8], values[9]); |
| 1570 |
|
| 1571 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p16); |
| 1572 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p15); |
| 1573 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p14); |
| 1574 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p13); |
| 1575 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p12); |
| 1576 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p11); |
| 1577 |
let t_a1_7 = f64x2_mul(self.twiddle7re, x7p10); |
| 1578 |
let t_a1_8 = f64x2_mul(self.twiddle8re, x8p9); |
| 1579 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p16); |
| 1580 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p15); |
| 1581 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p14); |
| 1582 |
let t_a2_4 = f64x2_mul(self.twiddle8re, x4p13); |
| 1583 |
let t_a2_5 = f64x2_mul(self.twiddle7re, x5p12); |
| 1584 |
let t_a2_6 = f64x2_mul(self.twiddle5re, x6p11); |
| 1585 |
let t_a2_7 = f64x2_mul(self.twiddle3re, x7p10); |
| 1586 |
let t_a2_8 = f64x2_mul(self.twiddle1re, x8p9); |
| 1587 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p16); |
| 1588 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p15); |
| 1589 |
let t_a3_3 = f64x2_mul(self.twiddle8re, x3p14); |
| 1590 |
let t_a3_4 = f64x2_mul(self.twiddle5re, x4p13); |
| 1591 |
let t_a3_5 = f64x2_mul(self.twiddle2re, x5p12); |
| 1592 |
let t_a3_6 = f64x2_mul(self.twiddle1re, x6p11); |
| 1593 |
let t_a3_7 = f64x2_mul(self.twiddle4re, x7p10); |
| 1594 |
let t_a3_8 = f64x2_mul(self.twiddle7re, x8p9); |
| 1595 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p16); |
| 1596 |
let t_a4_2 = f64x2_mul(self.twiddle8re, x2p15); |
| 1597 |
let t_a4_3 = f64x2_mul(self.twiddle5re, x3p14); |
| 1598 |
let t_a4_4 = f64x2_mul(self.twiddle1re, x4p13); |
| 1599 |
let t_a4_5 = f64x2_mul(self.twiddle3re, x5p12); |
| 1600 |
let t_a4_6 = f64x2_mul(self.twiddle7re, x6p11); |
| 1601 |
let t_a4_7 = f64x2_mul(self.twiddle6re, x7p10); |
| 1602 |
let t_a4_8 = f64x2_mul(self.twiddle2re, x8p9); |
| 1603 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p16); |
| 1604 |
let t_a5_2 = f64x2_mul(self.twiddle7re, x2p15); |
| 1605 |
let t_a5_3 = f64x2_mul(self.twiddle2re, x3p14); |
| 1606 |
let t_a5_4 = f64x2_mul(self.twiddle3re, x4p13); |
| 1607 |
let t_a5_5 = f64x2_mul(self.twiddle8re, x5p12); |
| 1608 |
let t_a5_6 = f64x2_mul(self.twiddle4re, x6p11); |
| 1609 |
let t_a5_7 = f64x2_mul(self.twiddle1re, x7p10); |
| 1610 |
let t_a5_8 = f64x2_mul(self.twiddle6re, x8p9); |
| 1611 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p16); |
| 1612 |
let t_a6_2 = f64x2_mul(self.twiddle5re, x2p15); |
| 1613 |
let t_a6_3 = f64x2_mul(self.twiddle1re, x3p14); |
| 1614 |
let t_a6_4 = f64x2_mul(self.twiddle7re, x4p13); |
| 1615 |
let t_a6_5 = f64x2_mul(self.twiddle4re, x5p12); |
| 1616 |
let t_a6_6 = f64x2_mul(self.twiddle2re, x6p11); |
| 1617 |
let t_a6_7 = f64x2_mul(self.twiddle8re, x7p10); |
| 1618 |
let t_a6_8 = f64x2_mul(self.twiddle3re, x8p9); |
| 1619 |
let t_a7_1 = f64x2_mul(self.twiddle7re, x1p16); |
| 1620 |
let t_a7_2 = f64x2_mul(self.twiddle3re, x2p15); |
| 1621 |
let t_a7_3 = f64x2_mul(self.twiddle4re, x3p14); |
| 1622 |
let t_a7_4 = f64x2_mul(self.twiddle6re, x4p13); |
| 1623 |
let t_a7_5 = f64x2_mul(self.twiddle1re, x5p12); |
| 1624 |
let t_a7_6 = f64x2_mul(self.twiddle8re, x6p11); |
| 1625 |
let t_a7_7 = f64x2_mul(self.twiddle2re, x7p10); |
| 1626 |
let t_a7_8 = f64x2_mul(self.twiddle5re, x8p9); |
| 1627 |
let t_a8_1 = f64x2_mul(self.twiddle8re, x1p16); |
| 1628 |
let t_a8_2 = f64x2_mul(self.twiddle1re, x2p15); |
| 1629 |
let t_a8_3 = f64x2_mul(self.twiddle7re, x3p14); |
| 1630 |
let t_a8_4 = f64x2_mul(self.twiddle2re, x4p13); |
| 1631 |
let t_a8_5 = f64x2_mul(self.twiddle6re, x5p12); |
| 1632 |
let t_a8_6 = f64x2_mul(self.twiddle3re, x6p11); |
| 1633 |
let t_a8_7 = f64x2_mul(self.twiddle5re, x7p10); |
| 1634 |
let t_a8_8 = f64x2_mul(self.twiddle4re, x8p9); |
| 1635 |
|
| 1636 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m16); |
| 1637 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m15); |
| 1638 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m14); |
| 1639 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m13); |
| 1640 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m12); |
| 1641 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m11); |
| 1642 |
let t_b1_7 = f64x2_mul(self.twiddle7im, x7m10); |
| 1643 |
let t_b1_8 = f64x2_mul(self.twiddle8im, x8m9); |
| 1644 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m16); |
| 1645 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m15); |
| 1646 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m14); |
| 1647 |
let t_b2_4 = f64x2_mul(self.twiddle8im, x4m13); |
| 1648 |
let t_b2_5 = f64x2_mul(self.twiddle7im, x5m12); |
| 1649 |
let t_b2_6 = f64x2_mul(self.twiddle5im, x6m11); |
| 1650 |
let t_b2_7 = f64x2_mul(self.twiddle3im, x7m10); |
| 1651 |
let t_b2_8 = f64x2_mul(self.twiddle1im, x8m9); |
| 1652 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m16); |
| 1653 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m15); |
| 1654 |
let t_b3_3 = f64x2_mul(self.twiddle8im, x3m14); |
| 1655 |
let t_b3_4 = f64x2_mul(self.twiddle5im, x4m13); |
| 1656 |
let t_b3_5 = f64x2_mul(self.twiddle2im, x5m12); |
| 1657 |
let t_b3_6 = f64x2_mul(self.twiddle1im, x6m11); |
| 1658 |
let t_b3_7 = f64x2_mul(self.twiddle4im, x7m10); |
| 1659 |
let t_b3_8 = f64x2_mul(self.twiddle7im, x8m9); |
| 1660 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m16); |
| 1661 |
let t_b4_2 = f64x2_mul(self.twiddle8im, x2m15); |
| 1662 |
let t_b4_3 = f64x2_mul(self.twiddle5im, x3m14); |
| 1663 |
let t_b4_4 = f64x2_mul(self.twiddle1im, x4m13); |
| 1664 |
let t_b4_5 = f64x2_mul(self.twiddle3im, x5m12); |
| 1665 |
let t_b4_6 = f64x2_mul(self.twiddle7im, x6m11); |
| 1666 |
let t_b4_7 = f64x2_mul(self.twiddle6im, x7m10); |
| 1667 |
let t_b4_8 = f64x2_mul(self.twiddle2im, x8m9); |
| 1668 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m16); |
| 1669 |
let t_b5_2 = f64x2_mul(self.twiddle7im, x2m15); |
| 1670 |
let t_b5_3 = f64x2_mul(self.twiddle2im, x3m14); |
| 1671 |
let t_b5_4 = f64x2_mul(self.twiddle3im, x4m13); |
| 1672 |
let t_b5_5 = f64x2_mul(self.twiddle8im, x5m12); |
| 1673 |
let t_b5_6 = f64x2_mul(self.twiddle4im, x6m11); |
| 1674 |
let t_b5_7 = f64x2_mul(self.twiddle1im, x7m10); |
| 1675 |
let t_b5_8 = f64x2_mul(self.twiddle6im, x8m9); |
| 1676 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m16); |
| 1677 |
let t_b6_2 = f64x2_mul(self.twiddle5im, x2m15); |
| 1678 |
let t_b6_3 = f64x2_mul(self.twiddle1im, x3m14); |
| 1679 |
let t_b6_4 = f64x2_mul(self.twiddle7im, x4m13); |
| 1680 |
let t_b6_5 = f64x2_mul(self.twiddle4im, x5m12); |
| 1681 |
let t_b6_6 = f64x2_mul(self.twiddle2im, x6m11); |
| 1682 |
let t_b6_7 = f64x2_mul(self.twiddle8im, x7m10); |
| 1683 |
let t_b6_8 = f64x2_mul(self.twiddle3im, x8m9); |
| 1684 |
let t_b7_1 = f64x2_mul(self.twiddle7im, x1m16); |
| 1685 |
let t_b7_2 = f64x2_mul(self.twiddle3im, x2m15); |
| 1686 |
let t_b7_3 = f64x2_mul(self.twiddle4im, x3m14); |
| 1687 |
let t_b7_4 = f64x2_mul(self.twiddle6im, x4m13); |
| 1688 |
let t_b7_5 = f64x2_mul(self.twiddle1im, x5m12); |
| 1689 |
let t_b7_6 = f64x2_mul(self.twiddle8im, x6m11); |
| 1690 |
let t_b7_7 = f64x2_mul(self.twiddle2im, x7m10); |
| 1691 |
let t_b7_8 = f64x2_mul(self.twiddle5im, x8m9); |
| 1692 |
let t_b8_1 = f64x2_mul(self.twiddle8im, x1m16); |
| 1693 |
let t_b8_2 = f64x2_mul(self.twiddle1im, x2m15); |
| 1694 |
let t_b8_3 = f64x2_mul(self.twiddle7im, x3m14); |
| 1695 |
let t_b8_4 = f64x2_mul(self.twiddle2im, x4m13); |
| 1696 |
let t_b8_5 = f64x2_mul(self.twiddle6im, x5m12); |
| 1697 |
let t_b8_6 = f64x2_mul(self.twiddle3im, x6m11); |
| 1698 |
let t_b8_7 = f64x2_mul(self.twiddle5im, x7m10); |
| 1699 |
let t_b8_8 = f64x2_mul(self.twiddle4im, x8m9); |
| 1700 |
|
| 1701 |
let x0 = values[0]; |
| 1702 |
let t_a1 = |
| 1703 |
calc_f64!(x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6 + t_a1_7 + t_a1_8); |
| 1704 |
let t_a2 = |
| 1705 |
calc_f64!(x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6 + t_a2_7 + t_a2_8); |
| 1706 |
let t_a3 = |
| 1707 |
calc_f64!(x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6 + t_a3_7 + t_a3_8); |
| 1708 |
let t_a4 = |
| 1709 |
calc_f64!(x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6 + t_a4_7 + t_a4_8); |
| 1710 |
let t_a5 = |
| 1711 |
calc_f64!(x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6 + t_a5_7 + t_a5_8); |
| 1712 |
let t_a6 = |
| 1713 |
calc_f64!(x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6 + t_a6_7 + t_a6_8); |
| 1714 |
let t_a7 = |
| 1715 |
calc_f64!(x0 + t_a7_1 + t_a7_2 + t_a7_3 + t_a7_4 + t_a7_5 + t_a7_6 + t_a7_7 + t_a7_8); |
| 1716 |
let t_a8 = |
| 1717 |
calc_f64!(x0 + t_a8_1 + t_a8_2 + t_a8_3 + t_a8_4 + t_a8_5 + t_a8_6 + t_a8_7 + t_a8_8); |
| 1718 |
|
| 1719 |
let t_b1 = calc_f64!(t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6 + t_b1_7 + t_b1_8); |
| 1720 |
let t_b2 = calc_f64!(t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 - t_b2_5 - t_b2_6 - t_b2_7 - t_b2_8); |
| 1721 |
let t_b3 = calc_f64!(t_b3_1 + t_b3_2 - t_b3_3 - t_b3_4 - t_b3_5 + t_b3_6 + t_b3_7 + t_b3_8); |
| 1722 |
let t_b4 = calc_f64!(t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 + t_b4_5 + t_b4_6 - t_b4_7 - t_b4_8); |
| 1723 |
let t_b5 = calc_f64!(t_b5_1 - t_b5_2 - t_b5_3 + t_b5_4 + t_b5_5 - t_b5_6 + t_b5_7 + t_b5_8); |
| 1724 |
let t_b6 = calc_f64!(t_b6_1 - t_b6_2 + t_b6_3 + t_b6_4 - t_b6_5 + t_b6_6 + t_b6_7 - t_b6_8); |
| 1725 |
let t_b7 = calc_f64!(t_b7_1 - t_b7_2 + t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 + t_b7_8); |
| 1726 |
let t_b8 = calc_f64!(t_b8_1 - t_b8_2 + t_b8_3 - t_b8_4 + t_b8_5 - t_b8_6 + t_b8_7 - t_b8_8); |
| 1727 |
|
| 1728 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 1729 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 1730 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 1731 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 1732 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 1733 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 1734 |
let t_b7_rot = self.rotate.rotate(t_b7); |
| 1735 |
let t_b8_rot = self.rotate.rotate(t_b8); |
| 1736 |
|
| 1737 |
let y0 = calc_f64!(x0 + x1p16 + x2p15 + x3p14 + x4p13 + x5p12 + x6p11 + x7p10 + x8p9); |
| 1738 |
let [y1, y16] = solo_fft2_f64(t_a1, t_b1_rot); |
| 1739 |
let [y2, y15] = solo_fft2_f64(t_a2, t_b2_rot); |
| 1740 |
let [y3, y14] = solo_fft2_f64(t_a3, t_b3_rot); |
| 1741 |
let [y4, y13] = solo_fft2_f64(t_a4, t_b4_rot); |
| 1742 |
let [y5, y12] = solo_fft2_f64(t_a5, t_b5_rot); |
| 1743 |
let [y6, y11] = solo_fft2_f64(t_a6, t_b6_rot); |
| 1744 |
let [y7, y10] = solo_fft2_f64(t_a7, t_b7_rot); |
| 1745 |
let [y8, y9] = solo_fft2_f64(t_a8, t_b8_rot); |
| 1746 |
[ |
| 1747 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, |
| 1748 |
] |
| 1749 |
} |
| 1750 |
} |
| 1751 |
|
| 1752 |
// _ ___ _________ _ _ _ |
| 1753 |
// / |/ _ \ |___ /___ \| |__ (_) |_ |
| 1754 |
// | | (_) | _____ |_ \ __) | '_ \| | __| |
| 1755 |
// | |\__, | |_____| ___) / __/| |_) | | |_ |
| 1756 |
// |_| /_/ |____/_____|_.__/|_|\__| |
| 1757 |
// |
| 1758 |
pub struct WasmSimdF32Butterfly19<T> { |
| 1759 |
direction: FftDirection, |
| 1760 |
_phantom: std::marker::PhantomData<T>, |
| 1761 |
rotate: Rotate90F32, |
| 1762 |
twiddle1re: v128, |
| 1763 |
twiddle1im: v128, |
| 1764 |
twiddle2re: v128, |
| 1765 |
twiddle2im: v128, |
| 1766 |
twiddle3re: v128, |
| 1767 |
twiddle3im: v128, |
| 1768 |
twiddle4re: v128, |
| 1769 |
twiddle4im: v128, |
| 1770 |
twiddle5re: v128, |
| 1771 |
twiddle5im: v128, |
| 1772 |
twiddle6re: v128, |
| 1773 |
twiddle6im: v128, |
| 1774 |
twiddle7re: v128, |
| 1775 |
twiddle7im: v128, |
| 1776 |
twiddle8re: v128, |
| 1777 |
twiddle8im: v128, |
| 1778 |
twiddle9re: v128, |
| 1779 |
twiddle9im: v128, |
| 1780 |
} |
| 1781 |
|
| 1782 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 1783 |
WasmSimdF32Butterfly19, |
| 1784 |
19, |
| 1785 |
|this: &WasmSimdF32Butterfly19<_>| this.direction |
| 1786 |
); |
| 1787 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 1788 |
WasmSimdF32Butterfly19, |
| 1789 |
19, |
| 1790 |
|this: &WasmSimdF32Butterfly19<_>| this.direction |
| 1791 |
); |
| 1792 |
impl<T: FftNum> WasmSimdF32Butterfly19<T> { |
| 1793 |
#[inline(always)] |
| 1794 |
pub fn new(direction: FftDirection) -> Self { |
| 1795 |
assert_f32::<T>(); |
| 1796 |
let rotate = Rotate90F32::new(true); |
| 1797 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 19, direction); |
| 1798 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 19, direction); |
| 1799 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 19, direction); |
| 1800 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 19, direction); |
| 1801 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 19, direction); |
| 1802 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 19, direction); |
| 1803 |
let tw7: Complex<f32> = twiddles::compute_twiddle(7, 19, direction); |
| 1804 |
let tw8: Complex<f32> = twiddles::compute_twiddle(8, 19, direction); |
| 1805 |
let tw9: Complex<f32> = twiddles::compute_twiddle(9, 19, direction); |
| 1806 |
let twiddle1re = f32x4_splat(tw1.re); |
| 1807 |
let twiddle1im = f32x4_splat(tw1.im); |
| 1808 |
let twiddle2re = f32x4_splat(tw2.re); |
| 1809 |
let twiddle2im = f32x4_splat(tw2.im); |
| 1810 |
let twiddle3re = f32x4_splat(tw3.re); |
| 1811 |
let twiddle3im = f32x4_splat(tw3.im); |
| 1812 |
let twiddle4re = f32x4_splat(tw4.re); |
| 1813 |
let twiddle4im = f32x4_splat(tw4.im); |
| 1814 |
let twiddle5re = f32x4_splat(tw5.re); |
| 1815 |
let twiddle5im = f32x4_splat(tw5.im); |
| 1816 |
let twiddle6re = f32x4_splat(tw6.re); |
| 1817 |
let twiddle6im = f32x4_splat(tw6.im); |
| 1818 |
let twiddle7re = f32x4_splat(tw7.re); |
| 1819 |
let twiddle7im = f32x4_splat(tw7.im); |
| 1820 |
let twiddle8re = f32x4_splat(tw8.re); |
| 1821 |
let twiddle8im = f32x4_splat(tw8.im); |
| 1822 |
let twiddle9re = f32x4_splat(tw9.re); |
| 1823 |
let twiddle9im = f32x4_splat(tw9.im); |
| 1824 |
|
| 1825 |
Self { |
| 1826 |
direction, |
| 1827 |
_phantom: std::marker::PhantomData, |
| 1828 |
rotate, |
| 1829 |
twiddle1re, |
| 1830 |
twiddle1im, |
| 1831 |
twiddle2re, |
| 1832 |
twiddle2im, |
| 1833 |
twiddle3re, |
| 1834 |
twiddle3im, |
| 1835 |
twiddle4re, |
| 1836 |
twiddle4im, |
| 1837 |
twiddle5re, |
| 1838 |
twiddle5im, |
| 1839 |
twiddle6re, |
| 1840 |
twiddle6im, |
| 1841 |
twiddle7re, |
| 1842 |
twiddle7im, |
| 1843 |
twiddle8re, |
| 1844 |
twiddle8im, |
| 1845 |
twiddle9re, |
| 1846 |
twiddle9im, |
| 1847 |
} |
| 1848 |
} |
| 1849 |
|
| 1850 |
#[inline(always)] |
| 1851 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 1852 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}); |
| 1853 |
|
| 1854 |
let out = self.perform_parallel_fft_direct(values); |
| 1855 |
|
| 1856 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}); |
| 1857 |
} |
| 1858 |
|
| 1859 |
#[inline(always)] |
| 1860 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 1861 |
&self, |
| 1862 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 1863 |
) { |
| 1864 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36}); |
| 1865 |
|
| 1866 |
let values = [ |
| 1867 |
extract_lo_hi_f32(input_packed[0], input_packed[9]), |
| 1868 |
extract_hi_lo_f32(input_packed[0], input_packed[10]), |
| 1869 |
extract_lo_hi_f32(input_packed[1], input_packed[10]), |
| 1870 |
extract_hi_lo_f32(input_packed[1], input_packed[11]), |
| 1871 |
extract_lo_hi_f32(input_packed[2], input_packed[11]), |
| 1872 |
extract_hi_lo_f32(input_packed[2], input_packed[12]), |
| 1873 |
extract_lo_hi_f32(input_packed[3], input_packed[12]), |
| 1874 |
extract_hi_lo_f32(input_packed[3], input_packed[13]), |
| 1875 |
extract_lo_hi_f32(input_packed[4], input_packed[13]), |
| 1876 |
extract_hi_lo_f32(input_packed[4], input_packed[14]), |
| 1877 |
extract_lo_hi_f32(input_packed[5], input_packed[14]), |
| 1878 |
extract_hi_lo_f32(input_packed[5], input_packed[15]), |
| 1879 |
extract_lo_hi_f32(input_packed[6], input_packed[15]), |
| 1880 |
extract_hi_lo_f32(input_packed[6], input_packed[16]), |
| 1881 |
extract_lo_hi_f32(input_packed[7], input_packed[16]), |
| 1882 |
extract_hi_lo_f32(input_packed[7], input_packed[17]), |
| 1883 |
extract_lo_hi_f32(input_packed[8], input_packed[17]), |
| 1884 |
extract_hi_lo_f32(input_packed[8], input_packed[18]), |
| 1885 |
extract_lo_hi_f32(input_packed[9], input_packed[18]), |
| 1886 |
]; |
| 1887 |
|
| 1888 |
let out = self.perform_parallel_fft_direct(values); |
| 1889 |
|
| 1890 |
let out_packed = [ |
| 1891 |
extract_lo_lo_f32(out[0], out[1]), |
| 1892 |
extract_lo_lo_f32(out[2], out[3]), |
| 1893 |
extract_lo_lo_f32(out[4], out[5]), |
| 1894 |
extract_lo_lo_f32(out[6], out[7]), |
| 1895 |
extract_lo_lo_f32(out[8], out[9]), |
| 1896 |
extract_lo_lo_f32(out[10], out[11]), |
| 1897 |
extract_lo_lo_f32(out[12], out[13]), |
| 1898 |
extract_lo_lo_f32(out[14], out[15]), |
| 1899 |
extract_lo_lo_f32(out[16], out[17]), |
| 1900 |
extract_lo_hi_f32(out[18], out[0]), |
| 1901 |
extract_hi_hi_f32(out[1], out[2]), |
| 1902 |
extract_hi_hi_f32(out[3], out[4]), |
| 1903 |
extract_hi_hi_f32(out[5], out[6]), |
| 1904 |
extract_hi_hi_f32(out[7], out[8]), |
| 1905 |
extract_hi_hi_f32(out[9], out[10]), |
| 1906 |
extract_hi_hi_f32(out[11], out[12]), |
| 1907 |
extract_hi_hi_f32(out[13], out[14]), |
| 1908 |
extract_hi_hi_f32(out[15], out[16]), |
| 1909 |
extract_hi_hi_f32(out[17], out[18]), |
| 1910 |
]; |
| 1911 |
|
| 1912 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}); |
| 1913 |
} |
| 1914 |
|
| 1915 |
#[inline(always)] |
| 1916 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 19]) -> [v128; 19] { |
| 1917 |
let [x1p18, x1m18] = parallel_fft2_interleaved_f32(values[1], values[18]); |
| 1918 |
let [x2p17, x2m17] = parallel_fft2_interleaved_f32(values[2], values[17]); |
| 1919 |
let [x3p16, x3m16] = parallel_fft2_interleaved_f32(values[3], values[16]); |
| 1920 |
let [x4p15, x4m15] = parallel_fft2_interleaved_f32(values[4], values[15]); |
| 1921 |
let [x5p14, x5m14] = parallel_fft2_interleaved_f32(values[5], values[14]); |
| 1922 |
let [x6p13, x6m13] = parallel_fft2_interleaved_f32(values[6], values[13]); |
| 1923 |
let [x7p12, x7m12] = parallel_fft2_interleaved_f32(values[7], values[12]); |
| 1924 |
let [x8p11, x8m11] = parallel_fft2_interleaved_f32(values[8], values[11]); |
| 1925 |
let [x9p10, x9m10] = parallel_fft2_interleaved_f32(values[9], values[10]); |
| 1926 |
|
| 1927 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p18); |
| 1928 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p17); |
| 1929 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p16); |
| 1930 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p15); |
| 1931 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p14); |
| 1932 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p13); |
| 1933 |
let t_a1_7 = f32x4_mul(self.twiddle7re, x7p12); |
| 1934 |
let t_a1_8 = f32x4_mul(self.twiddle8re, x8p11); |
| 1935 |
let t_a1_9 = f32x4_mul(self.twiddle9re, x9p10); |
| 1936 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p18); |
| 1937 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p17); |
| 1938 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p16); |
| 1939 |
let t_a2_4 = f32x4_mul(self.twiddle8re, x4p15); |
| 1940 |
let t_a2_5 = f32x4_mul(self.twiddle9re, x5p14); |
| 1941 |
let t_a2_6 = f32x4_mul(self.twiddle7re, x6p13); |
| 1942 |
let t_a2_7 = f32x4_mul(self.twiddle5re, x7p12); |
| 1943 |
let t_a2_8 = f32x4_mul(self.twiddle3re, x8p11); |
| 1944 |
let t_a2_9 = f32x4_mul(self.twiddle1re, x9p10); |
| 1945 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p18); |
| 1946 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p17); |
| 1947 |
let t_a3_3 = f32x4_mul(self.twiddle9re, x3p16); |
| 1948 |
let t_a3_4 = f32x4_mul(self.twiddle7re, x4p15); |
| 1949 |
let t_a3_5 = f32x4_mul(self.twiddle4re, x5p14); |
| 1950 |
let t_a3_6 = f32x4_mul(self.twiddle1re, x6p13); |
| 1951 |
let t_a3_7 = f32x4_mul(self.twiddle2re, x7p12); |
| 1952 |
let t_a3_8 = f32x4_mul(self.twiddle5re, x8p11); |
| 1953 |
let t_a3_9 = f32x4_mul(self.twiddle8re, x9p10); |
| 1954 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p18); |
| 1955 |
let t_a4_2 = f32x4_mul(self.twiddle8re, x2p17); |
| 1956 |
let t_a4_3 = f32x4_mul(self.twiddle7re, x3p16); |
| 1957 |
let t_a4_4 = f32x4_mul(self.twiddle3re, x4p15); |
| 1958 |
let t_a4_5 = f32x4_mul(self.twiddle1re, x5p14); |
| 1959 |
let t_a4_6 = f32x4_mul(self.twiddle5re, x6p13); |
| 1960 |
let t_a4_7 = f32x4_mul(self.twiddle9re, x7p12); |
| 1961 |
let t_a4_8 = f32x4_mul(self.twiddle6re, x8p11); |
| 1962 |
let t_a4_9 = f32x4_mul(self.twiddle2re, x9p10); |
| 1963 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p18); |
| 1964 |
let t_a5_2 = f32x4_mul(self.twiddle9re, x2p17); |
| 1965 |
let t_a5_3 = f32x4_mul(self.twiddle4re, x3p16); |
| 1966 |
let t_a5_4 = f32x4_mul(self.twiddle1re, x4p15); |
| 1967 |
let t_a5_5 = f32x4_mul(self.twiddle6re, x5p14); |
| 1968 |
let t_a5_6 = f32x4_mul(self.twiddle8re, x6p13); |
| 1969 |
let t_a5_7 = f32x4_mul(self.twiddle3re, x7p12); |
| 1970 |
let t_a5_8 = f32x4_mul(self.twiddle2re, x8p11); |
| 1971 |
let t_a5_9 = f32x4_mul(self.twiddle7re, x9p10); |
| 1972 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p18); |
| 1973 |
let t_a6_2 = f32x4_mul(self.twiddle7re, x2p17); |
| 1974 |
let t_a6_3 = f32x4_mul(self.twiddle1re, x3p16); |
| 1975 |
let t_a6_4 = f32x4_mul(self.twiddle5re, x4p15); |
| 1976 |
let t_a6_5 = f32x4_mul(self.twiddle8re, x5p14); |
| 1977 |
let t_a6_6 = f32x4_mul(self.twiddle2re, x6p13); |
| 1978 |
let t_a6_7 = f32x4_mul(self.twiddle4re, x7p12); |
| 1979 |
let t_a6_8 = f32x4_mul(self.twiddle9re, x8p11); |
| 1980 |
let t_a6_9 = f32x4_mul(self.twiddle3re, x9p10); |
| 1981 |
let t_a7_1 = f32x4_mul(self.twiddle7re, x1p18); |
| 1982 |
let t_a7_2 = f32x4_mul(self.twiddle5re, x2p17); |
| 1983 |
let t_a7_3 = f32x4_mul(self.twiddle2re, x3p16); |
| 1984 |
let t_a7_4 = f32x4_mul(self.twiddle9re, x4p15); |
| 1985 |
let t_a7_5 = f32x4_mul(self.twiddle3re, x5p14); |
| 1986 |
let t_a7_6 = f32x4_mul(self.twiddle4re, x6p13); |
| 1987 |
let t_a7_7 = f32x4_mul(self.twiddle8re, x7p12); |
| 1988 |
let t_a7_8 = f32x4_mul(self.twiddle1re, x8p11); |
| 1989 |
let t_a7_9 = f32x4_mul(self.twiddle6re, x9p10); |
| 1990 |
let t_a8_1 = f32x4_mul(self.twiddle8re, x1p18); |
| 1991 |
let t_a8_2 = f32x4_mul(self.twiddle3re, x2p17); |
| 1992 |
let t_a8_3 = f32x4_mul(self.twiddle5re, x3p16); |
| 1993 |
let t_a8_4 = f32x4_mul(self.twiddle6re, x4p15); |
| 1994 |
let t_a8_5 = f32x4_mul(self.twiddle2re, x5p14); |
| 1995 |
let t_a8_6 = f32x4_mul(self.twiddle9re, x6p13); |
| 1996 |
let t_a8_7 = f32x4_mul(self.twiddle1re, x7p12); |
| 1997 |
let t_a8_8 = f32x4_mul(self.twiddle7re, x8p11); |
| 1998 |
let t_a8_9 = f32x4_mul(self.twiddle4re, x9p10); |
| 1999 |
let t_a9_1 = f32x4_mul(self.twiddle9re, x1p18); |
| 2000 |
let t_a9_2 = f32x4_mul(self.twiddle1re, x2p17); |
| 2001 |
let t_a9_3 = f32x4_mul(self.twiddle8re, x3p16); |
| 2002 |
let t_a9_4 = f32x4_mul(self.twiddle2re, x4p15); |
| 2003 |
let t_a9_5 = f32x4_mul(self.twiddle7re, x5p14); |
| 2004 |
let t_a9_6 = f32x4_mul(self.twiddle3re, x6p13); |
| 2005 |
let t_a9_7 = f32x4_mul(self.twiddle6re, x7p12); |
| 2006 |
let t_a9_8 = f32x4_mul(self.twiddle4re, x8p11); |
| 2007 |
let t_a9_9 = f32x4_mul(self.twiddle5re, x9p10); |
| 2008 |
|
| 2009 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m18); |
| 2010 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m17); |
| 2011 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m16); |
| 2012 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m15); |
| 2013 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m14); |
| 2014 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m13); |
| 2015 |
let t_b1_7 = f32x4_mul(self.twiddle7im, x7m12); |
| 2016 |
let t_b1_8 = f32x4_mul(self.twiddle8im, x8m11); |
| 2017 |
let t_b1_9 = f32x4_mul(self.twiddle9im, x9m10); |
| 2018 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m18); |
| 2019 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m17); |
| 2020 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m16); |
| 2021 |
let t_b2_4 = f32x4_mul(self.twiddle8im, x4m15); |
| 2022 |
let t_b2_5 = f32x4_mul(self.twiddle9im, x5m14); |
| 2023 |
let t_b2_6 = f32x4_mul(self.twiddle7im, x6m13); |
| 2024 |
let t_b2_7 = f32x4_mul(self.twiddle5im, x7m12); |
| 2025 |
let t_b2_8 = f32x4_mul(self.twiddle3im, x8m11); |
| 2026 |
let t_b2_9 = f32x4_mul(self.twiddle1im, x9m10); |
| 2027 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m18); |
| 2028 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m17); |
| 2029 |
let t_b3_3 = f32x4_mul(self.twiddle9im, x3m16); |
| 2030 |
let t_b3_4 = f32x4_mul(self.twiddle7im, x4m15); |
| 2031 |
let t_b3_5 = f32x4_mul(self.twiddle4im, x5m14); |
| 2032 |
let t_b3_6 = f32x4_mul(self.twiddle1im, x6m13); |
| 2033 |
let t_b3_7 = f32x4_mul(self.twiddle2im, x7m12); |
| 2034 |
let t_b3_8 = f32x4_mul(self.twiddle5im, x8m11); |
| 2035 |
let t_b3_9 = f32x4_mul(self.twiddle8im, x9m10); |
| 2036 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m18); |
| 2037 |
let t_b4_2 = f32x4_mul(self.twiddle8im, x2m17); |
| 2038 |
let t_b4_3 = f32x4_mul(self.twiddle7im, x3m16); |
| 2039 |
let t_b4_4 = f32x4_mul(self.twiddle3im, x4m15); |
| 2040 |
let t_b4_5 = f32x4_mul(self.twiddle1im, x5m14); |
| 2041 |
let t_b4_6 = f32x4_mul(self.twiddle5im, x6m13); |
| 2042 |
let t_b4_7 = f32x4_mul(self.twiddle9im, x7m12); |
| 2043 |
let t_b4_8 = f32x4_mul(self.twiddle6im, x8m11); |
| 2044 |
let t_b4_9 = f32x4_mul(self.twiddle2im, x9m10); |
| 2045 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m18); |
| 2046 |
let t_b5_2 = f32x4_mul(self.twiddle9im, x2m17); |
| 2047 |
let t_b5_3 = f32x4_mul(self.twiddle4im, x3m16); |
| 2048 |
let t_b5_4 = f32x4_mul(self.twiddle1im, x4m15); |
| 2049 |
let t_b5_5 = f32x4_mul(self.twiddle6im, x5m14); |
| 2050 |
let t_b5_6 = f32x4_mul(self.twiddle8im, x6m13); |
| 2051 |
let t_b5_7 = f32x4_mul(self.twiddle3im, x7m12); |
| 2052 |
let t_b5_8 = f32x4_mul(self.twiddle2im, x8m11); |
| 2053 |
let t_b5_9 = f32x4_mul(self.twiddle7im, x9m10); |
| 2054 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m18); |
| 2055 |
let t_b6_2 = f32x4_mul(self.twiddle7im, x2m17); |
| 2056 |
let t_b6_3 = f32x4_mul(self.twiddle1im, x3m16); |
| 2057 |
let t_b6_4 = f32x4_mul(self.twiddle5im, x4m15); |
| 2058 |
let t_b6_5 = f32x4_mul(self.twiddle8im, x5m14); |
| 2059 |
let t_b6_6 = f32x4_mul(self.twiddle2im, x6m13); |
| 2060 |
let t_b6_7 = f32x4_mul(self.twiddle4im, x7m12); |
| 2061 |
let t_b6_8 = f32x4_mul(self.twiddle9im, x8m11); |
| 2062 |
let t_b6_9 = f32x4_mul(self.twiddle3im, x9m10); |
| 2063 |
let t_b7_1 = f32x4_mul(self.twiddle7im, x1m18); |
| 2064 |
let t_b7_2 = f32x4_mul(self.twiddle5im, x2m17); |
| 2065 |
let t_b7_3 = f32x4_mul(self.twiddle2im, x3m16); |
| 2066 |
let t_b7_4 = f32x4_mul(self.twiddle9im, x4m15); |
| 2067 |
let t_b7_5 = f32x4_mul(self.twiddle3im, x5m14); |
| 2068 |
let t_b7_6 = f32x4_mul(self.twiddle4im, x6m13); |
| 2069 |
let t_b7_7 = f32x4_mul(self.twiddle8im, x7m12); |
| 2070 |
let t_b7_8 = f32x4_mul(self.twiddle1im, x8m11); |
| 2071 |
let t_b7_9 = f32x4_mul(self.twiddle6im, x9m10); |
| 2072 |
let t_b8_1 = f32x4_mul(self.twiddle8im, x1m18); |
| 2073 |
let t_b8_2 = f32x4_mul(self.twiddle3im, x2m17); |
| 2074 |
let t_b8_3 = f32x4_mul(self.twiddle5im, x3m16); |
| 2075 |
let t_b8_4 = f32x4_mul(self.twiddle6im, x4m15); |
| 2076 |
let t_b8_5 = f32x4_mul(self.twiddle2im, x5m14); |
| 2077 |
let t_b8_6 = f32x4_mul(self.twiddle9im, x6m13); |
| 2078 |
let t_b8_7 = f32x4_mul(self.twiddle1im, x7m12); |
| 2079 |
let t_b8_8 = f32x4_mul(self.twiddle7im, x8m11); |
| 2080 |
let t_b8_9 = f32x4_mul(self.twiddle4im, x9m10); |
| 2081 |
let t_b9_1 = f32x4_mul(self.twiddle9im, x1m18); |
| 2082 |
let t_b9_2 = f32x4_mul(self.twiddle1im, x2m17); |
| 2083 |
let t_b9_3 = f32x4_mul(self.twiddle8im, x3m16); |
| 2084 |
let t_b9_4 = f32x4_mul(self.twiddle2im, x4m15); |
| 2085 |
let t_b9_5 = f32x4_mul(self.twiddle7im, x5m14); |
| 2086 |
let t_b9_6 = f32x4_mul(self.twiddle3im, x6m13); |
| 2087 |
let t_b9_7 = f32x4_mul(self.twiddle6im, x7m12); |
| 2088 |
let t_b9_8 = f32x4_mul(self.twiddle4im, x8m11); |
| 2089 |
let t_b9_9 = f32x4_mul(self.twiddle5im, x9m10); |
| 2090 |
|
| 2091 |
let x0 = values[0]; |
| 2092 |
let t_a1 = calc_f32!( |
| 2093 |
x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6 + t_a1_7 + t_a1_8 + t_a1_9 |
| 2094 |
); |
| 2095 |
let t_a2 = calc_f32!( |
| 2096 |
x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6 + t_a2_7 + t_a2_8 + t_a2_9 |
| 2097 |
); |
| 2098 |
let t_a3 = calc_f32!( |
| 2099 |
x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6 + t_a3_7 + t_a3_8 + t_a3_9 |
| 2100 |
); |
| 2101 |
let t_a4 = calc_f32!( |
| 2102 |
x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6 + t_a4_7 + t_a4_8 + t_a4_9 |
| 2103 |
); |
| 2104 |
let t_a5 = calc_f32!( |
| 2105 |
x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6 + t_a5_7 + t_a5_8 + t_a5_9 |
| 2106 |
); |
| 2107 |
let t_a6 = calc_f32!( |
| 2108 |
x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6 + t_a6_7 + t_a6_8 + t_a6_9 |
| 2109 |
); |
| 2110 |
let t_a7 = calc_f32!( |
| 2111 |
x0 + t_a7_1 + t_a7_2 + t_a7_3 + t_a7_4 + t_a7_5 + t_a7_6 + t_a7_7 + t_a7_8 + t_a7_9 |
| 2112 |
); |
| 2113 |
let t_a8 = calc_f32!( |
| 2114 |
x0 + t_a8_1 + t_a8_2 + t_a8_3 + t_a8_4 + t_a8_5 + t_a8_6 + t_a8_7 + t_a8_8 + t_a8_9 |
| 2115 |
); |
| 2116 |
let t_a9 = calc_f32!( |
| 2117 |
x0 + t_a9_1 + t_a9_2 + t_a9_3 + t_a9_4 + t_a9_5 + t_a9_6 + t_a9_7 + t_a9_8 + t_a9_9 |
| 2118 |
); |
| 2119 |
|
| 2120 |
let t_b1 = calc_f32!( |
| 2121 |
t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6 + t_b1_7 + t_b1_8 + t_b1_9 |
| 2122 |
); |
| 2123 |
let t_b2 = calc_f32!( |
| 2124 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 - t_b2_5 - t_b2_6 - t_b2_7 - t_b2_8 - t_b2_9 |
| 2125 |
); |
| 2126 |
let t_b3 = calc_f32!( |
| 2127 |
t_b3_1 + t_b3_2 + t_b3_3 - t_b3_4 - t_b3_5 - t_b3_6 + t_b3_7 + t_b3_8 + t_b3_9 |
| 2128 |
); |
| 2129 |
let t_b4 = calc_f32!( |
| 2130 |
t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 + t_b4_5 + t_b4_6 + t_b4_7 - t_b4_8 - t_b4_9 |
| 2131 |
); |
| 2132 |
let t_b5 = calc_f32!( |
| 2133 |
t_b5_1 - t_b5_2 - t_b5_3 + t_b5_4 + t_b5_5 - t_b5_6 - t_b5_7 + t_b5_8 + t_b5_9 |
| 2134 |
); |
| 2135 |
let t_b6 = calc_f32!( |
| 2136 |
t_b6_1 - t_b6_2 - t_b6_3 + t_b6_4 - t_b6_5 - t_b6_6 + t_b6_7 - t_b6_8 - t_b6_9 |
| 2137 |
); |
| 2138 |
let t_b7 = calc_f32!( |
| 2139 |
t_b7_1 - t_b7_2 + t_b7_3 + t_b7_4 - t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 + t_b7_9 |
| 2140 |
); |
| 2141 |
let t_b8 = calc_f32!( |
| 2142 |
t_b8_1 - t_b8_2 + t_b8_3 - t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 - t_b8_9 |
| 2143 |
); |
| 2144 |
let t_b9 = calc_f32!( |
| 2145 |
t_b9_1 - t_b9_2 + t_b9_3 - t_b9_4 + t_b9_5 - t_b9_6 + t_b9_7 - t_b9_8 + t_b9_9 |
| 2146 |
); |
| 2147 |
|
| 2148 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 2149 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 2150 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 2151 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 2152 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 2153 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 2154 |
let t_b7_rot = self.rotate.rotate_both(t_b7); |
| 2155 |
let t_b8_rot = self.rotate.rotate_both(t_b8); |
| 2156 |
let t_b9_rot = self.rotate.rotate_both(t_b9); |
| 2157 |
|
| 2158 |
let y0 = |
| 2159 |
calc_f32!(x0 + x1p18 + x2p17 + x3p16 + x4p15 + x5p14 + x6p13 + x7p12 + x8p11 + x9p10); |
| 2160 |
let [y1, y18] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 2161 |
let [y2, y17] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 2162 |
let [y3, y16] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 2163 |
let [y4, y15] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 2164 |
let [y5, y14] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 2165 |
let [y6, y13] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 2166 |
let [y7, y12] = parallel_fft2_interleaved_f32(t_a7, t_b7_rot); |
| 2167 |
let [y8, y11] = parallel_fft2_interleaved_f32(t_a8, t_b8_rot); |
| 2168 |
let [y9, y10] = parallel_fft2_interleaved_f32(t_a9, t_b9_rot); |
| 2169 |
[ |
| 2170 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 2171 |
] |
| 2172 |
} |
| 2173 |
} |
| 2174 |
|
| 2175 |
// _ ___ __ _ _ _ _ _ |
| 2176 |
// / |/ _ \ / /_ | || | | |__ (_) |_ |
| 2177 |
// | | (_) | _____ | '_ \| || |_| '_ \| | __| |
| 2178 |
// | |\__, | |_____| | (_) |__ _| |_) | | |_ |
| 2179 |
// |_| /_/ \___/ |_| |_.__/|_|\__| |
| 2180 |
// |
| 2181 |
|
| 2182 |
pub struct WasmSimdF64Butterfly19<T> { |
| 2183 |
direction: FftDirection, |
| 2184 |
_phantom: std::marker::PhantomData<T>, |
| 2185 |
rotate: Rotate90F64, |
| 2186 |
twiddle1re: v128, |
| 2187 |
twiddle1im: v128, |
| 2188 |
twiddle2re: v128, |
| 2189 |
twiddle2im: v128, |
| 2190 |
twiddle3re: v128, |
| 2191 |
twiddle3im: v128, |
| 2192 |
twiddle4re: v128, |
| 2193 |
twiddle4im: v128, |
| 2194 |
twiddle5re: v128, |
| 2195 |
twiddle5im: v128, |
| 2196 |
twiddle6re: v128, |
| 2197 |
twiddle6im: v128, |
| 2198 |
twiddle7re: v128, |
| 2199 |
twiddle7im: v128, |
| 2200 |
twiddle8re: v128, |
| 2201 |
twiddle8im: v128, |
| 2202 |
twiddle9re: v128, |
| 2203 |
twiddle9im: v128, |
| 2204 |
} |
| 2205 |
|
| 2206 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 2207 |
WasmSimdF64Butterfly19, |
| 2208 |
19, |
| 2209 |
|this: &WasmSimdF64Butterfly19<_>| this.direction |
| 2210 |
); |
| 2211 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 2212 |
WasmSimdF64Butterfly19, |
| 2213 |
19, |
| 2214 |
|this: &WasmSimdF64Butterfly19<_>| this.direction |
| 2215 |
); |
| 2216 |
impl<T: FftNum> WasmSimdF64Butterfly19<T> { |
| 2217 |
#[inline(always)] |
| 2218 |
pub fn new(direction: FftDirection) -> Self { |
| 2219 |
assert_f64::<T>(); |
| 2220 |
let rotate = Rotate90F64::new(true); |
| 2221 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 19, direction); |
| 2222 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 19, direction); |
| 2223 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 19, direction); |
| 2224 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 19, direction); |
| 2225 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 19, direction); |
| 2226 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 19, direction); |
| 2227 |
let tw7: Complex<f64> = twiddles::compute_twiddle(7, 19, direction); |
| 2228 |
let tw8: Complex<f64> = twiddles::compute_twiddle(8, 19, direction); |
| 2229 |
let tw9: Complex<f64> = twiddles::compute_twiddle(9, 19, direction); |
| 2230 |
let twiddle1re = f64x2_splat(tw1.re); |
| 2231 |
let twiddle1im = f64x2_splat(tw1.im); |
| 2232 |
let twiddle2re = f64x2_splat(tw2.re); |
| 2233 |
let twiddle2im = f64x2_splat(tw2.im); |
| 2234 |
let twiddle3re = f64x2_splat(tw3.re); |
| 2235 |
let twiddle3im = f64x2_splat(tw3.im); |
| 2236 |
let twiddle4re = f64x2_splat(tw4.re); |
| 2237 |
let twiddle4im = f64x2_splat(tw4.im); |
| 2238 |
let twiddle5re = f64x2_splat(tw5.re); |
| 2239 |
let twiddle5im = f64x2_splat(tw5.im); |
| 2240 |
let twiddle6re = f64x2_splat(tw6.re); |
| 2241 |
let twiddle6im = f64x2_splat(tw6.im); |
| 2242 |
let twiddle7re = f64x2_splat(tw7.re); |
| 2243 |
let twiddle7im = f64x2_splat(tw7.im); |
| 2244 |
let twiddle8re = f64x2_splat(tw8.re); |
| 2245 |
let twiddle8im = f64x2_splat(tw8.im); |
| 2246 |
let twiddle9re = f64x2_splat(tw9.re); |
| 2247 |
let twiddle9im = f64x2_splat(tw9.im); |
| 2248 |
|
| 2249 |
Self { |
| 2250 |
direction, |
| 2251 |
_phantom: std::marker::PhantomData, |
| 2252 |
rotate, |
| 2253 |
twiddle1re, |
| 2254 |
twiddle1im, |
| 2255 |
twiddle2re, |
| 2256 |
twiddle2im, |
| 2257 |
twiddle3re, |
| 2258 |
twiddle3im, |
| 2259 |
twiddle4re, |
| 2260 |
twiddle4im, |
| 2261 |
twiddle5re, |
| 2262 |
twiddle5im, |
| 2263 |
twiddle6re, |
| 2264 |
twiddle6im, |
| 2265 |
twiddle7re, |
| 2266 |
twiddle7im, |
| 2267 |
twiddle8re, |
| 2268 |
twiddle8im, |
| 2269 |
twiddle9re, |
| 2270 |
twiddle9im, |
| 2271 |
} |
| 2272 |
} |
| 2273 |
|
| 2274 |
#[inline(always)] |
| 2275 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 2276 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}); |
| 2277 |
|
| 2278 |
let out = self.perform_fft_direct(values); |
| 2279 |
|
| 2280 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18}); |
| 2281 |
} |
| 2282 |
|
| 2283 |
#[inline(always)] |
| 2284 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 19]) -> [v128; 19] { |
| 2285 |
let [x1p18, x1m18] = solo_fft2_f64(values[1], values[18]); |
| 2286 |
let [x2p17, x2m17] = solo_fft2_f64(values[2], values[17]); |
| 2287 |
let [x3p16, x3m16] = solo_fft2_f64(values[3], values[16]); |
| 2288 |
let [x4p15, x4m15] = solo_fft2_f64(values[4], values[15]); |
| 2289 |
let [x5p14, x5m14] = solo_fft2_f64(values[5], values[14]); |
| 2290 |
let [x6p13, x6m13] = solo_fft2_f64(values[6], values[13]); |
| 2291 |
let [x7p12, x7m12] = solo_fft2_f64(values[7], values[12]); |
| 2292 |
let [x8p11, x8m11] = solo_fft2_f64(values[8], values[11]); |
| 2293 |
let [x9p10, x9m10] = solo_fft2_f64(values[9], values[10]); |
| 2294 |
|
| 2295 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p18); |
| 2296 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p17); |
| 2297 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p16); |
| 2298 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p15); |
| 2299 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p14); |
| 2300 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p13); |
| 2301 |
let t_a1_7 = f64x2_mul(self.twiddle7re, x7p12); |
| 2302 |
let t_a1_8 = f64x2_mul(self.twiddle8re, x8p11); |
| 2303 |
let t_a1_9 = f64x2_mul(self.twiddle9re, x9p10); |
| 2304 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p18); |
| 2305 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p17); |
| 2306 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p16); |
| 2307 |
let t_a2_4 = f64x2_mul(self.twiddle8re, x4p15); |
| 2308 |
let t_a2_5 = f64x2_mul(self.twiddle9re, x5p14); |
| 2309 |
let t_a2_6 = f64x2_mul(self.twiddle7re, x6p13); |
| 2310 |
let t_a2_7 = f64x2_mul(self.twiddle5re, x7p12); |
| 2311 |
let t_a2_8 = f64x2_mul(self.twiddle3re, x8p11); |
| 2312 |
let t_a2_9 = f64x2_mul(self.twiddle1re, x9p10); |
| 2313 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p18); |
| 2314 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p17); |
| 2315 |
let t_a3_3 = f64x2_mul(self.twiddle9re, x3p16); |
| 2316 |
let t_a3_4 = f64x2_mul(self.twiddle7re, x4p15); |
| 2317 |
let t_a3_5 = f64x2_mul(self.twiddle4re, x5p14); |
| 2318 |
let t_a3_6 = f64x2_mul(self.twiddle1re, x6p13); |
| 2319 |
let t_a3_7 = f64x2_mul(self.twiddle2re, x7p12); |
| 2320 |
let t_a3_8 = f64x2_mul(self.twiddle5re, x8p11); |
| 2321 |
let t_a3_9 = f64x2_mul(self.twiddle8re, x9p10); |
| 2322 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p18); |
| 2323 |
let t_a4_2 = f64x2_mul(self.twiddle8re, x2p17); |
| 2324 |
let t_a4_3 = f64x2_mul(self.twiddle7re, x3p16); |
| 2325 |
let t_a4_4 = f64x2_mul(self.twiddle3re, x4p15); |
| 2326 |
let t_a4_5 = f64x2_mul(self.twiddle1re, x5p14); |
| 2327 |
let t_a4_6 = f64x2_mul(self.twiddle5re, x6p13); |
| 2328 |
let t_a4_7 = f64x2_mul(self.twiddle9re, x7p12); |
| 2329 |
let t_a4_8 = f64x2_mul(self.twiddle6re, x8p11); |
| 2330 |
let t_a4_9 = f64x2_mul(self.twiddle2re, x9p10); |
| 2331 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p18); |
| 2332 |
let t_a5_2 = f64x2_mul(self.twiddle9re, x2p17); |
| 2333 |
let t_a5_3 = f64x2_mul(self.twiddle4re, x3p16); |
| 2334 |
let t_a5_4 = f64x2_mul(self.twiddle1re, x4p15); |
| 2335 |
let t_a5_5 = f64x2_mul(self.twiddle6re, x5p14); |
| 2336 |
let t_a5_6 = f64x2_mul(self.twiddle8re, x6p13); |
| 2337 |
let t_a5_7 = f64x2_mul(self.twiddle3re, x7p12); |
| 2338 |
let t_a5_8 = f64x2_mul(self.twiddle2re, x8p11); |
| 2339 |
let t_a5_9 = f64x2_mul(self.twiddle7re, x9p10); |
| 2340 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p18); |
| 2341 |
let t_a6_2 = f64x2_mul(self.twiddle7re, x2p17); |
| 2342 |
let t_a6_3 = f64x2_mul(self.twiddle1re, x3p16); |
| 2343 |
let t_a6_4 = f64x2_mul(self.twiddle5re, x4p15); |
| 2344 |
let t_a6_5 = f64x2_mul(self.twiddle8re, x5p14); |
| 2345 |
let t_a6_6 = f64x2_mul(self.twiddle2re, x6p13); |
| 2346 |
let t_a6_7 = f64x2_mul(self.twiddle4re, x7p12); |
| 2347 |
let t_a6_8 = f64x2_mul(self.twiddle9re, x8p11); |
| 2348 |
let t_a6_9 = f64x2_mul(self.twiddle3re, x9p10); |
| 2349 |
let t_a7_1 = f64x2_mul(self.twiddle7re, x1p18); |
| 2350 |
let t_a7_2 = f64x2_mul(self.twiddle5re, x2p17); |
| 2351 |
let t_a7_3 = f64x2_mul(self.twiddle2re, x3p16); |
| 2352 |
let t_a7_4 = f64x2_mul(self.twiddle9re, x4p15); |
| 2353 |
let t_a7_5 = f64x2_mul(self.twiddle3re, x5p14); |
| 2354 |
let t_a7_6 = f64x2_mul(self.twiddle4re, x6p13); |
| 2355 |
let t_a7_7 = f64x2_mul(self.twiddle8re, x7p12); |
| 2356 |
let t_a7_8 = f64x2_mul(self.twiddle1re, x8p11); |
| 2357 |
let t_a7_9 = f64x2_mul(self.twiddle6re, x9p10); |
| 2358 |
let t_a8_1 = f64x2_mul(self.twiddle8re, x1p18); |
| 2359 |
let t_a8_2 = f64x2_mul(self.twiddle3re, x2p17); |
| 2360 |
let t_a8_3 = f64x2_mul(self.twiddle5re, x3p16); |
| 2361 |
let t_a8_4 = f64x2_mul(self.twiddle6re, x4p15); |
| 2362 |
let t_a8_5 = f64x2_mul(self.twiddle2re, x5p14); |
| 2363 |
let t_a8_6 = f64x2_mul(self.twiddle9re, x6p13); |
| 2364 |
let t_a8_7 = f64x2_mul(self.twiddle1re, x7p12); |
| 2365 |
let t_a8_8 = f64x2_mul(self.twiddle7re, x8p11); |
| 2366 |
let t_a8_9 = f64x2_mul(self.twiddle4re, x9p10); |
| 2367 |
let t_a9_1 = f64x2_mul(self.twiddle9re, x1p18); |
| 2368 |
let t_a9_2 = f64x2_mul(self.twiddle1re, x2p17); |
| 2369 |
let t_a9_3 = f64x2_mul(self.twiddle8re, x3p16); |
| 2370 |
let t_a9_4 = f64x2_mul(self.twiddle2re, x4p15); |
| 2371 |
let t_a9_5 = f64x2_mul(self.twiddle7re, x5p14); |
| 2372 |
let t_a9_6 = f64x2_mul(self.twiddle3re, x6p13); |
| 2373 |
let t_a9_7 = f64x2_mul(self.twiddle6re, x7p12); |
| 2374 |
let t_a9_8 = f64x2_mul(self.twiddle4re, x8p11); |
| 2375 |
let t_a9_9 = f64x2_mul(self.twiddle5re, x9p10); |
| 2376 |
|
| 2377 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m18); |
| 2378 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m17); |
| 2379 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m16); |
| 2380 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m15); |
| 2381 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m14); |
| 2382 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m13); |
| 2383 |
let t_b1_7 = f64x2_mul(self.twiddle7im, x7m12); |
| 2384 |
let t_b1_8 = f64x2_mul(self.twiddle8im, x8m11); |
| 2385 |
let t_b1_9 = f64x2_mul(self.twiddle9im, x9m10); |
| 2386 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m18); |
| 2387 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m17); |
| 2388 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m16); |
| 2389 |
let t_b2_4 = f64x2_mul(self.twiddle8im, x4m15); |
| 2390 |
let t_b2_5 = f64x2_mul(self.twiddle9im, x5m14); |
| 2391 |
let t_b2_6 = f64x2_mul(self.twiddle7im, x6m13); |
| 2392 |
let t_b2_7 = f64x2_mul(self.twiddle5im, x7m12); |
| 2393 |
let t_b2_8 = f64x2_mul(self.twiddle3im, x8m11); |
| 2394 |
let t_b2_9 = f64x2_mul(self.twiddle1im, x9m10); |
| 2395 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m18); |
| 2396 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m17); |
| 2397 |
let t_b3_3 = f64x2_mul(self.twiddle9im, x3m16); |
| 2398 |
let t_b3_4 = f64x2_mul(self.twiddle7im, x4m15); |
| 2399 |
let t_b3_5 = f64x2_mul(self.twiddle4im, x5m14); |
| 2400 |
let t_b3_6 = f64x2_mul(self.twiddle1im, x6m13); |
| 2401 |
let t_b3_7 = f64x2_mul(self.twiddle2im, x7m12); |
| 2402 |
let t_b3_8 = f64x2_mul(self.twiddle5im, x8m11); |
| 2403 |
let t_b3_9 = f64x2_mul(self.twiddle8im, x9m10); |
| 2404 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m18); |
| 2405 |
let t_b4_2 = f64x2_mul(self.twiddle8im, x2m17); |
| 2406 |
let t_b4_3 = f64x2_mul(self.twiddle7im, x3m16); |
| 2407 |
let t_b4_4 = f64x2_mul(self.twiddle3im, x4m15); |
| 2408 |
let t_b4_5 = f64x2_mul(self.twiddle1im, x5m14); |
| 2409 |
let t_b4_6 = f64x2_mul(self.twiddle5im, x6m13); |
| 2410 |
let t_b4_7 = f64x2_mul(self.twiddle9im, x7m12); |
| 2411 |
let t_b4_8 = f64x2_mul(self.twiddle6im, x8m11); |
| 2412 |
let t_b4_9 = f64x2_mul(self.twiddle2im, x9m10); |
| 2413 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m18); |
| 2414 |
let t_b5_2 = f64x2_mul(self.twiddle9im, x2m17); |
| 2415 |
let t_b5_3 = f64x2_mul(self.twiddle4im, x3m16); |
| 2416 |
let t_b5_4 = f64x2_mul(self.twiddle1im, x4m15); |
| 2417 |
let t_b5_5 = f64x2_mul(self.twiddle6im, x5m14); |
| 2418 |
let t_b5_6 = f64x2_mul(self.twiddle8im, x6m13); |
| 2419 |
let t_b5_7 = f64x2_mul(self.twiddle3im, x7m12); |
| 2420 |
let t_b5_8 = f64x2_mul(self.twiddle2im, x8m11); |
| 2421 |
let t_b5_9 = f64x2_mul(self.twiddle7im, x9m10); |
| 2422 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m18); |
| 2423 |
let t_b6_2 = f64x2_mul(self.twiddle7im, x2m17); |
| 2424 |
let t_b6_3 = f64x2_mul(self.twiddle1im, x3m16); |
| 2425 |
let t_b6_4 = f64x2_mul(self.twiddle5im, x4m15); |
| 2426 |
let t_b6_5 = f64x2_mul(self.twiddle8im, x5m14); |
| 2427 |
let t_b6_6 = f64x2_mul(self.twiddle2im, x6m13); |
| 2428 |
let t_b6_7 = f64x2_mul(self.twiddle4im, x7m12); |
| 2429 |
let t_b6_8 = f64x2_mul(self.twiddle9im, x8m11); |
| 2430 |
let t_b6_9 = f64x2_mul(self.twiddle3im, x9m10); |
| 2431 |
let t_b7_1 = f64x2_mul(self.twiddle7im, x1m18); |
| 2432 |
let t_b7_2 = f64x2_mul(self.twiddle5im, x2m17); |
| 2433 |
let t_b7_3 = f64x2_mul(self.twiddle2im, x3m16); |
| 2434 |
let t_b7_4 = f64x2_mul(self.twiddle9im, x4m15); |
| 2435 |
let t_b7_5 = f64x2_mul(self.twiddle3im, x5m14); |
| 2436 |
let t_b7_6 = f64x2_mul(self.twiddle4im, x6m13); |
| 2437 |
let t_b7_7 = f64x2_mul(self.twiddle8im, x7m12); |
| 2438 |
let t_b7_8 = f64x2_mul(self.twiddle1im, x8m11); |
| 2439 |
let t_b7_9 = f64x2_mul(self.twiddle6im, x9m10); |
| 2440 |
let t_b8_1 = f64x2_mul(self.twiddle8im, x1m18); |
| 2441 |
let t_b8_2 = f64x2_mul(self.twiddle3im, x2m17); |
| 2442 |
let t_b8_3 = f64x2_mul(self.twiddle5im, x3m16); |
| 2443 |
let t_b8_4 = f64x2_mul(self.twiddle6im, x4m15); |
| 2444 |
let t_b8_5 = f64x2_mul(self.twiddle2im, x5m14); |
| 2445 |
let t_b8_6 = f64x2_mul(self.twiddle9im, x6m13); |
| 2446 |
let t_b8_7 = f64x2_mul(self.twiddle1im, x7m12); |
| 2447 |
let t_b8_8 = f64x2_mul(self.twiddle7im, x8m11); |
| 2448 |
let t_b8_9 = f64x2_mul(self.twiddle4im, x9m10); |
| 2449 |
let t_b9_1 = f64x2_mul(self.twiddle9im, x1m18); |
| 2450 |
let t_b9_2 = f64x2_mul(self.twiddle1im, x2m17); |
| 2451 |
let t_b9_3 = f64x2_mul(self.twiddle8im, x3m16); |
| 2452 |
let t_b9_4 = f64x2_mul(self.twiddle2im, x4m15); |
| 2453 |
let t_b9_5 = f64x2_mul(self.twiddle7im, x5m14); |
| 2454 |
let t_b9_6 = f64x2_mul(self.twiddle3im, x6m13); |
| 2455 |
let t_b9_7 = f64x2_mul(self.twiddle6im, x7m12); |
| 2456 |
let t_b9_8 = f64x2_mul(self.twiddle4im, x8m11); |
| 2457 |
let t_b9_9 = f64x2_mul(self.twiddle5im, x9m10); |
| 2458 |
|
| 2459 |
let x0 = values[0]; |
| 2460 |
let t_a1 = calc_f64!( |
| 2461 |
x0 + t_a1_1 + t_a1_2 + t_a1_3 + t_a1_4 + t_a1_5 + t_a1_6 + t_a1_7 + t_a1_8 + t_a1_9 |
| 2462 |
); |
| 2463 |
let t_a2 = calc_f64!( |
| 2464 |
x0 + t_a2_1 + t_a2_2 + t_a2_3 + t_a2_4 + t_a2_5 + t_a2_6 + t_a2_7 + t_a2_8 + t_a2_9 |
| 2465 |
); |
| 2466 |
let t_a3 = calc_f64!( |
| 2467 |
x0 + t_a3_1 + t_a3_2 + t_a3_3 + t_a3_4 + t_a3_5 + t_a3_6 + t_a3_7 + t_a3_8 + t_a3_9 |
| 2468 |
); |
| 2469 |
let t_a4 = calc_f64!( |
| 2470 |
x0 + t_a4_1 + t_a4_2 + t_a4_3 + t_a4_4 + t_a4_5 + t_a4_6 + t_a4_7 + t_a4_8 + t_a4_9 |
| 2471 |
); |
| 2472 |
let t_a5 = calc_f64!( |
| 2473 |
x0 + t_a5_1 + t_a5_2 + t_a5_3 + t_a5_4 + t_a5_5 + t_a5_6 + t_a5_7 + t_a5_8 + t_a5_9 |
| 2474 |
); |
| 2475 |
let t_a6 = calc_f64!( |
| 2476 |
x0 + t_a6_1 + t_a6_2 + t_a6_3 + t_a6_4 + t_a6_5 + t_a6_6 + t_a6_7 + t_a6_8 + t_a6_9 |
| 2477 |
); |
| 2478 |
let t_a7 = calc_f64!( |
| 2479 |
x0 + t_a7_1 + t_a7_2 + t_a7_3 + t_a7_4 + t_a7_5 + t_a7_6 + t_a7_7 + t_a7_8 + t_a7_9 |
| 2480 |
); |
| 2481 |
let t_a8 = calc_f64!( |
| 2482 |
x0 + t_a8_1 + t_a8_2 + t_a8_3 + t_a8_4 + t_a8_5 + t_a8_6 + t_a8_7 + t_a8_8 + t_a8_9 |
| 2483 |
); |
| 2484 |
let t_a9 = calc_f64!( |
| 2485 |
x0 + t_a9_1 + t_a9_2 + t_a9_3 + t_a9_4 + t_a9_5 + t_a9_6 + t_a9_7 + t_a9_8 + t_a9_9 |
| 2486 |
); |
| 2487 |
|
| 2488 |
let t_b1 = calc_f64!( |
| 2489 |
t_b1_1 + t_b1_2 + t_b1_3 + t_b1_4 + t_b1_5 + t_b1_6 + t_b1_7 + t_b1_8 + t_b1_9 |
| 2490 |
); |
| 2491 |
let t_b2 = calc_f64!( |
| 2492 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 - t_b2_5 - t_b2_6 - t_b2_7 - t_b2_8 - t_b2_9 |
| 2493 |
); |
| 2494 |
let t_b3 = calc_f64!( |
| 2495 |
t_b3_1 + t_b3_2 + t_b3_3 - t_b3_4 - t_b3_5 - t_b3_6 + t_b3_7 + t_b3_8 + t_b3_9 |
| 2496 |
); |
| 2497 |
let t_b4 = calc_f64!( |
| 2498 |
t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 + t_b4_5 + t_b4_6 + t_b4_7 - t_b4_8 - t_b4_9 |
| 2499 |
); |
| 2500 |
let t_b5 = calc_f64!( |
| 2501 |
t_b5_1 - t_b5_2 - t_b5_3 + t_b5_4 + t_b5_5 - t_b5_6 - t_b5_7 + t_b5_8 + t_b5_9 |
| 2502 |
); |
| 2503 |
let t_b6 = calc_f64!( |
| 2504 |
t_b6_1 - t_b6_2 - t_b6_3 + t_b6_4 - t_b6_5 - t_b6_6 + t_b6_7 - t_b6_8 - t_b6_9 |
| 2505 |
); |
| 2506 |
let t_b7 = calc_f64!( |
| 2507 |
t_b7_1 - t_b7_2 + t_b7_3 + t_b7_4 - t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 + t_b7_9 |
| 2508 |
); |
| 2509 |
let t_b8 = calc_f64!( |
| 2510 |
t_b8_1 - t_b8_2 + t_b8_3 - t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 - t_b8_9 |
| 2511 |
); |
| 2512 |
let t_b9 = calc_f64!( |
| 2513 |
t_b9_1 - t_b9_2 + t_b9_3 - t_b9_4 + t_b9_5 - t_b9_6 + t_b9_7 - t_b9_8 + t_b9_9 |
| 2514 |
); |
| 2515 |
|
| 2516 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 2517 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 2518 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 2519 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 2520 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 2521 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 2522 |
let t_b7_rot = self.rotate.rotate(t_b7); |
| 2523 |
let t_b8_rot = self.rotate.rotate(t_b8); |
| 2524 |
let t_b9_rot = self.rotate.rotate(t_b9); |
| 2525 |
|
| 2526 |
let y0 = |
| 2527 |
calc_f64!(x0 + x1p18 + x2p17 + x3p16 + x4p15 + x5p14 + x6p13 + x7p12 + x8p11 + x9p10); |
| 2528 |
let [y1, y18] = solo_fft2_f64(t_a1, t_b1_rot); |
| 2529 |
let [y2, y17] = solo_fft2_f64(t_a2, t_b2_rot); |
| 2530 |
let [y3, y16] = solo_fft2_f64(t_a3, t_b3_rot); |
| 2531 |
let [y4, y15] = solo_fft2_f64(t_a4, t_b4_rot); |
| 2532 |
let [y5, y14] = solo_fft2_f64(t_a5, t_b5_rot); |
| 2533 |
let [y6, y13] = solo_fft2_f64(t_a6, t_b6_rot); |
| 2534 |
let [y7, y12] = solo_fft2_f64(t_a7, t_b7_rot); |
| 2535 |
let [y8, y11] = solo_fft2_f64(t_a8, t_b8_rot); |
| 2536 |
let [y9, y10] = solo_fft2_f64(t_a9, t_b9_rot); |
| 2537 |
[ |
| 2538 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 2539 |
] |
| 2540 |
} |
| 2541 |
} |
| 2542 |
|
| 2543 |
// ____ _____ _________ _ _ _ |
| 2544 |
// |___ \|___ / |___ /___ \| |__ (_) |_ |
| 2545 |
// __) | |_ \ _____ |_ \ __) | '_ \| | __| |
| 2546 |
// / __/ ___) | |_____| ___) / __/| |_) | | |_ |
| 2547 |
// |_____|____/ |____/_____|_.__/|_|\__| |
| 2548 |
// |
| 2549 |
|
| 2550 |
pub struct WasmSimdF32Butterfly23<T> { |
| 2551 |
direction: FftDirection, |
| 2552 |
_phantom: std::marker::PhantomData<T>, |
| 2553 |
rotate: Rotate90F32, |
| 2554 |
twiddle1re: v128, |
| 2555 |
twiddle1im: v128, |
| 2556 |
twiddle2re: v128, |
| 2557 |
twiddle2im: v128, |
| 2558 |
twiddle3re: v128, |
| 2559 |
twiddle3im: v128, |
| 2560 |
twiddle4re: v128, |
| 2561 |
twiddle4im: v128, |
| 2562 |
twiddle5re: v128, |
| 2563 |
twiddle5im: v128, |
| 2564 |
twiddle6re: v128, |
| 2565 |
twiddle6im: v128, |
| 2566 |
twiddle7re: v128, |
| 2567 |
twiddle7im: v128, |
| 2568 |
twiddle8re: v128, |
| 2569 |
twiddle8im: v128, |
| 2570 |
twiddle9re: v128, |
| 2571 |
twiddle9im: v128, |
| 2572 |
twiddle10re: v128, |
| 2573 |
twiddle10im: v128, |
| 2574 |
twiddle11re: v128, |
| 2575 |
twiddle11im: v128, |
| 2576 |
} |
| 2577 |
|
| 2578 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 2579 |
WasmSimdF32Butterfly23, |
| 2580 |
23, |
| 2581 |
|this: &WasmSimdF32Butterfly23<_>| this.direction |
| 2582 |
); |
| 2583 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 2584 |
WasmSimdF32Butterfly23, |
| 2585 |
23, |
| 2586 |
|this: &WasmSimdF32Butterfly23<_>| this.direction |
| 2587 |
); |
| 2588 |
impl<T: FftNum> WasmSimdF32Butterfly23<T> { |
| 2589 |
#[inline(always)] |
| 2590 |
pub fn new(direction: FftDirection) -> Self { |
| 2591 |
assert_f32::<T>(); |
| 2592 |
let rotate = Rotate90F32::new(true); |
| 2593 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 23, direction); |
| 2594 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 23, direction); |
| 2595 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 23, direction); |
| 2596 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 23, direction); |
| 2597 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 23, direction); |
| 2598 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 23, direction); |
| 2599 |
let tw7: Complex<f32> = twiddles::compute_twiddle(7, 23, direction); |
| 2600 |
let tw8: Complex<f32> = twiddles::compute_twiddle(8, 23, direction); |
| 2601 |
let tw9: Complex<f32> = twiddles::compute_twiddle(9, 23, direction); |
| 2602 |
let tw10: Complex<f32> = twiddles::compute_twiddle(10, 23, direction); |
| 2603 |
let tw11: Complex<f32> = twiddles::compute_twiddle(11, 23, direction); |
| 2604 |
let twiddle1re = f32x4_splat(tw1.re); |
| 2605 |
let twiddle1im = f32x4_splat(tw1.im); |
| 2606 |
let twiddle2re = f32x4_splat(tw2.re); |
| 2607 |
let twiddle2im = f32x4_splat(tw2.im); |
| 2608 |
let twiddle3re = f32x4_splat(tw3.re); |
| 2609 |
let twiddle3im = f32x4_splat(tw3.im); |
| 2610 |
let twiddle4re = f32x4_splat(tw4.re); |
| 2611 |
let twiddle4im = f32x4_splat(tw4.im); |
| 2612 |
let twiddle5re = f32x4_splat(tw5.re); |
| 2613 |
let twiddle5im = f32x4_splat(tw5.im); |
| 2614 |
let twiddle6re = f32x4_splat(tw6.re); |
| 2615 |
let twiddle6im = f32x4_splat(tw6.im); |
| 2616 |
let twiddle7re = f32x4_splat(tw7.re); |
| 2617 |
let twiddle7im = f32x4_splat(tw7.im); |
| 2618 |
let twiddle8re = f32x4_splat(tw8.re); |
| 2619 |
let twiddle8im = f32x4_splat(tw8.im); |
| 2620 |
let twiddle9re = f32x4_splat(tw9.re); |
| 2621 |
let twiddle9im = f32x4_splat(tw9.im); |
| 2622 |
let twiddle10re = f32x4_splat(tw10.re); |
| 2623 |
let twiddle10im = f32x4_splat(tw10.im); |
| 2624 |
let twiddle11re = f32x4_splat(tw11.re); |
| 2625 |
let twiddle11im = f32x4_splat(tw11.im); |
| 2626 |
|
| 2627 |
Self { |
| 2628 |
direction, |
| 2629 |
_phantom: std::marker::PhantomData, |
| 2630 |
rotate, |
| 2631 |
twiddle1re, |
| 2632 |
twiddle1im, |
| 2633 |
twiddle2re, |
| 2634 |
twiddle2im, |
| 2635 |
twiddle3re, |
| 2636 |
twiddle3im, |
| 2637 |
twiddle4re, |
| 2638 |
twiddle4im, |
| 2639 |
twiddle5re, |
| 2640 |
twiddle5im, |
| 2641 |
twiddle6re, |
| 2642 |
twiddle6im, |
| 2643 |
twiddle7re, |
| 2644 |
twiddle7im, |
| 2645 |
twiddle8re, |
| 2646 |
twiddle8im, |
| 2647 |
twiddle9re, |
| 2648 |
twiddle9im, |
| 2649 |
twiddle10re, |
| 2650 |
twiddle10im, |
| 2651 |
twiddle11re, |
| 2652 |
twiddle11im, |
| 2653 |
} |
| 2654 |
} |
| 2655 |
|
| 2656 |
#[inline(always)] |
| 2657 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 2658 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}); |
| 2659 |
|
| 2660 |
let out = self.perform_parallel_fft_direct(values); |
| 2661 |
|
| 2662 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}); |
| 2663 |
} |
| 2664 |
|
| 2665 |
#[inline(always)] |
| 2666 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 2667 |
&self, |
| 2668 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 2669 |
) { |
| 2670 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44}); |
| 2671 |
|
| 2672 |
let values = [ |
| 2673 |
extract_lo_hi_f32(input_packed[0], input_packed[11]), |
| 2674 |
extract_hi_lo_f32(input_packed[0], input_packed[12]), |
| 2675 |
extract_lo_hi_f32(input_packed[1], input_packed[12]), |
| 2676 |
extract_hi_lo_f32(input_packed[1], input_packed[13]), |
| 2677 |
extract_lo_hi_f32(input_packed[2], input_packed[13]), |
| 2678 |
extract_hi_lo_f32(input_packed[2], input_packed[14]), |
| 2679 |
extract_lo_hi_f32(input_packed[3], input_packed[14]), |
| 2680 |
extract_hi_lo_f32(input_packed[3], input_packed[15]), |
| 2681 |
extract_lo_hi_f32(input_packed[4], input_packed[15]), |
| 2682 |
extract_hi_lo_f32(input_packed[4], input_packed[16]), |
| 2683 |
extract_lo_hi_f32(input_packed[5], input_packed[16]), |
| 2684 |
extract_hi_lo_f32(input_packed[5], input_packed[17]), |
| 2685 |
extract_lo_hi_f32(input_packed[6], input_packed[17]), |
| 2686 |
extract_hi_lo_f32(input_packed[6], input_packed[18]), |
| 2687 |
extract_lo_hi_f32(input_packed[7], input_packed[18]), |
| 2688 |
extract_hi_lo_f32(input_packed[7], input_packed[19]), |
| 2689 |
extract_lo_hi_f32(input_packed[8], input_packed[19]), |
| 2690 |
extract_hi_lo_f32(input_packed[8], input_packed[20]), |
| 2691 |
extract_lo_hi_f32(input_packed[9], input_packed[20]), |
| 2692 |
extract_hi_lo_f32(input_packed[9], input_packed[21]), |
| 2693 |
extract_lo_hi_f32(input_packed[10], input_packed[21]), |
| 2694 |
extract_hi_lo_f32(input_packed[10], input_packed[22]), |
| 2695 |
extract_lo_hi_f32(input_packed[11], input_packed[22]), |
| 2696 |
]; |
| 2697 |
|
| 2698 |
let out = self.perform_parallel_fft_direct(values); |
| 2699 |
|
| 2700 |
let out_packed = [ |
| 2701 |
extract_lo_lo_f32(out[0], out[1]), |
| 2702 |
extract_lo_lo_f32(out[2], out[3]), |
| 2703 |
extract_lo_lo_f32(out[4], out[5]), |
| 2704 |
extract_lo_lo_f32(out[6], out[7]), |
| 2705 |
extract_lo_lo_f32(out[8], out[9]), |
| 2706 |
extract_lo_lo_f32(out[10], out[11]), |
| 2707 |
extract_lo_lo_f32(out[12], out[13]), |
| 2708 |
extract_lo_lo_f32(out[14], out[15]), |
| 2709 |
extract_lo_lo_f32(out[16], out[17]), |
| 2710 |
extract_lo_lo_f32(out[18], out[19]), |
| 2711 |
extract_lo_lo_f32(out[20], out[21]), |
| 2712 |
extract_lo_hi_f32(out[22], out[0]), |
| 2713 |
extract_hi_hi_f32(out[1], out[2]), |
| 2714 |
extract_hi_hi_f32(out[3], out[4]), |
| 2715 |
extract_hi_hi_f32(out[5], out[6]), |
| 2716 |
extract_hi_hi_f32(out[7], out[8]), |
| 2717 |
extract_hi_hi_f32(out[9], out[10]), |
| 2718 |
extract_hi_hi_f32(out[11], out[12]), |
| 2719 |
extract_hi_hi_f32(out[13], out[14]), |
| 2720 |
extract_hi_hi_f32(out[15], out[16]), |
| 2721 |
extract_hi_hi_f32(out[17], out[18]), |
| 2722 |
extract_hi_hi_f32(out[19], out[20]), |
| 2723 |
extract_hi_hi_f32(out[21], out[22]), |
| 2724 |
]; |
| 2725 |
|
| 2726 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}); |
| 2727 |
} |
| 2728 |
|
| 2729 |
#[inline(always)] |
| 2730 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 23]) -> [v128; 23] { |
| 2731 |
let [x1p22, x1m22] = parallel_fft2_interleaved_f32(values[1], values[22]); |
| 2732 |
let [x2p21, x2m21] = parallel_fft2_interleaved_f32(values[2], values[21]); |
| 2733 |
let [x3p20, x3m20] = parallel_fft2_interleaved_f32(values[3], values[20]); |
| 2734 |
let [x4p19, x4m19] = parallel_fft2_interleaved_f32(values[4], values[19]); |
| 2735 |
let [x5p18, x5m18] = parallel_fft2_interleaved_f32(values[5], values[18]); |
| 2736 |
let [x6p17, x6m17] = parallel_fft2_interleaved_f32(values[6], values[17]); |
| 2737 |
let [x7p16, x7m16] = parallel_fft2_interleaved_f32(values[7], values[16]); |
| 2738 |
let [x8p15, x8m15] = parallel_fft2_interleaved_f32(values[8], values[15]); |
| 2739 |
let [x9p14, x9m14] = parallel_fft2_interleaved_f32(values[9], values[14]); |
| 2740 |
let [x10p13, x10m13] = parallel_fft2_interleaved_f32(values[10], values[13]); |
| 2741 |
let [x11p12, x11m12] = parallel_fft2_interleaved_f32(values[11], values[12]); |
| 2742 |
|
| 2743 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p22); |
| 2744 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p21); |
| 2745 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p20); |
| 2746 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p19); |
| 2747 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p18); |
| 2748 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p17); |
| 2749 |
let t_a1_7 = f32x4_mul(self.twiddle7re, x7p16); |
| 2750 |
let t_a1_8 = f32x4_mul(self.twiddle8re, x8p15); |
| 2751 |
let t_a1_9 = f32x4_mul(self.twiddle9re, x9p14); |
| 2752 |
let t_a1_10 = f32x4_mul(self.twiddle10re, x10p13); |
| 2753 |
let t_a1_11 = f32x4_mul(self.twiddle11re, x11p12); |
| 2754 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p22); |
| 2755 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p21); |
| 2756 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p20); |
| 2757 |
let t_a2_4 = f32x4_mul(self.twiddle8re, x4p19); |
| 2758 |
let t_a2_5 = f32x4_mul(self.twiddle10re, x5p18); |
| 2759 |
let t_a2_6 = f32x4_mul(self.twiddle11re, x6p17); |
| 2760 |
let t_a2_7 = f32x4_mul(self.twiddle9re, x7p16); |
| 2761 |
let t_a2_8 = f32x4_mul(self.twiddle7re, x8p15); |
| 2762 |
let t_a2_9 = f32x4_mul(self.twiddle5re, x9p14); |
| 2763 |
let t_a2_10 = f32x4_mul(self.twiddle3re, x10p13); |
| 2764 |
let t_a2_11 = f32x4_mul(self.twiddle1re, x11p12); |
| 2765 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p22); |
| 2766 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p21); |
| 2767 |
let t_a3_3 = f32x4_mul(self.twiddle9re, x3p20); |
| 2768 |
let t_a3_4 = f32x4_mul(self.twiddle11re, x4p19); |
| 2769 |
let t_a3_5 = f32x4_mul(self.twiddle8re, x5p18); |
| 2770 |
let t_a3_6 = f32x4_mul(self.twiddle5re, x6p17); |
| 2771 |
let t_a3_7 = f32x4_mul(self.twiddle2re, x7p16); |
| 2772 |
let t_a3_8 = f32x4_mul(self.twiddle1re, x8p15); |
| 2773 |
let t_a3_9 = f32x4_mul(self.twiddle4re, x9p14); |
| 2774 |
let t_a3_10 = f32x4_mul(self.twiddle7re, x10p13); |
| 2775 |
let t_a3_11 = f32x4_mul(self.twiddle10re, x11p12); |
| 2776 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p22); |
| 2777 |
let t_a4_2 = f32x4_mul(self.twiddle8re, x2p21); |
| 2778 |
let t_a4_3 = f32x4_mul(self.twiddle11re, x3p20); |
| 2779 |
let t_a4_4 = f32x4_mul(self.twiddle7re, x4p19); |
| 2780 |
let t_a4_5 = f32x4_mul(self.twiddle3re, x5p18); |
| 2781 |
let t_a4_6 = f32x4_mul(self.twiddle1re, x6p17); |
| 2782 |
let t_a4_7 = f32x4_mul(self.twiddle5re, x7p16); |
| 2783 |
let t_a4_8 = f32x4_mul(self.twiddle9re, x8p15); |
| 2784 |
let t_a4_9 = f32x4_mul(self.twiddle10re, x9p14); |
| 2785 |
let t_a4_10 = f32x4_mul(self.twiddle6re, x10p13); |
| 2786 |
let t_a4_11 = f32x4_mul(self.twiddle2re, x11p12); |
| 2787 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p22); |
| 2788 |
let t_a5_2 = f32x4_mul(self.twiddle10re, x2p21); |
| 2789 |
let t_a5_3 = f32x4_mul(self.twiddle8re, x3p20); |
| 2790 |
let t_a5_4 = f32x4_mul(self.twiddle3re, x4p19); |
| 2791 |
let t_a5_5 = f32x4_mul(self.twiddle2re, x5p18); |
| 2792 |
let t_a5_6 = f32x4_mul(self.twiddle7re, x6p17); |
| 2793 |
let t_a5_7 = f32x4_mul(self.twiddle11re, x7p16); |
| 2794 |
let t_a5_8 = f32x4_mul(self.twiddle6re, x8p15); |
| 2795 |
let t_a5_9 = f32x4_mul(self.twiddle1re, x9p14); |
| 2796 |
let t_a5_10 = f32x4_mul(self.twiddle4re, x10p13); |
| 2797 |
let t_a5_11 = f32x4_mul(self.twiddle9re, x11p12); |
| 2798 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p22); |
| 2799 |
let t_a6_2 = f32x4_mul(self.twiddle11re, x2p21); |
| 2800 |
let t_a6_3 = f32x4_mul(self.twiddle5re, x3p20); |
| 2801 |
let t_a6_4 = f32x4_mul(self.twiddle1re, x4p19); |
| 2802 |
let t_a6_5 = f32x4_mul(self.twiddle7re, x5p18); |
| 2803 |
let t_a6_6 = f32x4_mul(self.twiddle10re, x6p17); |
| 2804 |
let t_a6_7 = f32x4_mul(self.twiddle4re, x7p16); |
| 2805 |
let t_a6_8 = f32x4_mul(self.twiddle2re, x8p15); |
| 2806 |
let t_a6_9 = f32x4_mul(self.twiddle8re, x9p14); |
| 2807 |
let t_a6_10 = f32x4_mul(self.twiddle9re, x10p13); |
| 2808 |
let t_a6_11 = f32x4_mul(self.twiddle3re, x11p12); |
| 2809 |
let t_a7_1 = f32x4_mul(self.twiddle7re, x1p22); |
| 2810 |
let t_a7_2 = f32x4_mul(self.twiddle9re, x2p21); |
| 2811 |
let t_a7_3 = f32x4_mul(self.twiddle2re, x3p20); |
| 2812 |
let t_a7_4 = f32x4_mul(self.twiddle5re, x4p19); |
| 2813 |
let t_a7_5 = f32x4_mul(self.twiddle11re, x5p18); |
| 2814 |
let t_a7_6 = f32x4_mul(self.twiddle4re, x6p17); |
| 2815 |
let t_a7_7 = f32x4_mul(self.twiddle3re, x7p16); |
| 2816 |
let t_a7_8 = f32x4_mul(self.twiddle10re, x8p15); |
| 2817 |
let t_a7_9 = f32x4_mul(self.twiddle6re, x9p14); |
| 2818 |
let t_a7_10 = f32x4_mul(self.twiddle1re, x10p13); |
| 2819 |
let t_a7_11 = f32x4_mul(self.twiddle8re, x11p12); |
| 2820 |
let t_a8_1 = f32x4_mul(self.twiddle8re, x1p22); |
| 2821 |
let t_a8_2 = f32x4_mul(self.twiddle7re, x2p21); |
| 2822 |
let t_a8_3 = f32x4_mul(self.twiddle1re, x3p20); |
| 2823 |
let t_a8_4 = f32x4_mul(self.twiddle9re, x4p19); |
| 2824 |
let t_a8_5 = f32x4_mul(self.twiddle6re, x5p18); |
| 2825 |
let t_a8_6 = f32x4_mul(self.twiddle2re, x6p17); |
| 2826 |
let t_a8_7 = f32x4_mul(self.twiddle10re, x7p16); |
| 2827 |
let t_a8_8 = f32x4_mul(self.twiddle5re, x8p15); |
| 2828 |
let t_a8_9 = f32x4_mul(self.twiddle3re, x9p14); |
| 2829 |
let t_a8_10 = f32x4_mul(self.twiddle11re, x10p13); |
| 2830 |
let t_a8_11 = f32x4_mul(self.twiddle4re, x11p12); |
| 2831 |
let t_a9_1 = f32x4_mul(self.twiddle9re, x1p22); |
| 2832 |
let t_a9_2 = f32x4_mul(self.twiddle5re, x2p21); |
| 2833 |
let t_a9_3 = f32x4_mul(self.twiddle4re, x3p20); |
| 2834 |
let t_a9_4 = f32x4_mul(self.twiddle10re, x4p19); |
| 2835 |
let t_a9_5 = f32x4_mul(self.twiddle1re, x5p18); |
| 2836 |
let t_a9_6 = f32x4_mul(self.twiddle8re, x6p17); |
| 2837 |
let t_a9_7 = f32x4_mul(self.twiddle6re, x7p16); |
| 2838 |
let t_a9_8 = f32x4_mul(self.twiddle3re, x8p15); |
| 2839 |
let t_a9_9 = f32x4_mul(self.twiddle11re, x9p14); |
| 2840 |
let t_a9_10 = f32x4_mul(self.twiddle2re, x10p13); |
| 2841 |
let t_a9_11 = f32x4_mul(self.twiddle7re, x11p12); |
| 2842 |
let t_a10_1 = f32x4_mul(self.twiddle10re, x1p22); |
| 2843 |
let t_a10_2 = f32x4_mul(self.twiddle3re, x2p21); |
| 2844 |
let t_a10_3 = f32x4_mul(self.twiddle7re, x3p20); |
| 2845 |
let t_a10_4 = f32x4_mul(self.twiddle6re, x4p19); |
| 2846 |
let t_a10_5 = f32x4_mul(self.twiddle4re, x5p18); |
| 2847 |
let t_a10_6 = f32x4_mul(self.twiddle9re, x6p17); |
| 2848 |
let t_a10_7 = f32x4_mul(self.twiddle1re, x7p16); |
| 2849 |
let t_a10_8 = f32x4_mul(self.twiddle11re, x8p15); |
| 2850 |
let t_a10_9 = f32x4_mul(self.twiddle2re, x9p14); |
| 2851 |
let t_a10_10 = f32x4_mul(self.twiddle8re, x10p13); |
| 2852 |
let t_a10_11 = f32x4_mul(self.twiddle5re, x11p12); |
| 2853 |
let t_a11_1 = f32x4_mul(self.twiddle11re, x1p22); |
| 2854 |
let t_a11_2 = f32x4_mul(self.twiddle1re, x2p21); |
| 2855 |
let t_a11_3 = f32x4_mul(self.twiddle10re, x3p20); |
| 2856 |
let t_a11_4 = f32x4_mul(self.twiddle2re, x4p19); |
| 2857 |
let t_a11_5 = f32x4_mul(self.twiddle9re, x5p18); |
| 2858 |
let t_a11_6 = f32x4_mul(self.twiddle3re, x6p17); |
| 2859 |
let t_a11_7 = f32x4_mul(self.twiddle8re, x7p16); |
| 2860 |
let t_a11_8 = f32x4_mul(self.twiddle4re, x8p15); |
| 2861 |
let t_a11_9 = f32x4_mul(self.twiddle7re, x9p14); |
| 2862 |
let t_a11_10 = f32x4_mul(self.twiddle5re, x10p13); |
| 2863 |
let t_a11_11 = f32x4_mul(self.twiddle6re, x11p12); |
| 2864 |
|
| 2865 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m22); |
| 2866 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m21); |
| 2867 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m20); |
| 2868 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m19); |
| 2869 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m18); |
| 2870 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m17); |
| 2871 |
let t_b1_7 = f32x4_mul(self.twiddle7im, x7m16); |
| 2872 |
let t_b1_8 = f32x4_mul(self.twiddle8im, x8m15); |
| 2873 |
let t_b1_9 = f32x4_mul(self.twiddle9im, x9m14); |
| 2874 |
let t_b1_10 = f32x4_mul(self.twiddle10im, x10m13); |
| 2875 |
let t_b1_11 = f32x4_mul(self.twiddle11im, x11m12); |
| 2876 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m22); |
| 2877 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m21); |
| 2878 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m20); |
| 2879 |
let t_b2_4 = f32x4_mul(self.twiddle8im, x4m19); |
| 2880 |
let t_b2_5 = f32x4_mul(self.twiddle10im, x5m18); |
| 2881 |
let t_b2_6 = f32x4_mul(self.twiddle11im, x6m17); |
| 2882 |
let t_b2_7 = f32x4_mul(self.twiddle9im, x7m16); |
| 2883 |
let t_b2_8 = f32x4_mul(self.twiddle7im, x8m15); |
| 2884 |
let t_b2_9 = f32x4_mul(self.twiddle5im, x9m14); |
| 2885 |
let t_b2_10 = f32x4_mul(self.twiddle3im, x10m13); |
| 2886 |
let t_b2_11 = f32x4_mul(self.twiddle1im, x11m12); |
| 2887 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m22); |
| 2888 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m21); |
| 2889 |
let t_b3_3 = f32x4_mul(self.twiddle9im, x3m20); |
| 2890 |
let t_b3_4 = f32x4_mul(self.twiddle11im, x4m19); |
| 2891 |
let t_b3_5 = f32x4_mul(self.twiddle8im, x5m18); |
| 2892 |
let t_b3_6 = f32x4_mul(self.twiddle5im, x6m17); |
| 2893 |
let t_b3_7 = f32x4_mul(self.twiddle2im, x7m16); |
| 2894 |
let t_b3_8 = f32x4_mul(self.twiddle1im, x8m15); |
| 2895 |
let t_b3_9 = f32x4_mul(self.twiddle4im, x9m14); |
| 2896 |
let t_b3_10 = f32x4_mul(self.twiddle7im, x10m13); |
| 2897 |
let t_b3_11 = f32x4_mul(self.twiddle10im, x11m12); |
| 2898 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m22); |
| 2899 |
let t_b4_2 = f32x4_mul(self.twiddle8im, x2m21); |
| 2900 |
let t_b4_3 = f32x4_mul(self.twiddle11im, x3m20); |
| 2901 |
let t_b4_4 = f32x4_mul(self.twiddle7im, x4m19); |
| 2902 |
let t_b4_5 = f32x4_mul(self.twiddle3im, x5m18); |
| 2903 |
let t_b4_6 = f32x4_mul(self.twiddle1im, x6m17); |
| 2904 |
let t_b4_7 = f32x4_mul(self.twiddle5im, x7m16); |
| 2905 |
let t_b4_8 = f32x4_mul(self.twiddle9im, x8m15); |
| 2906 |
let t_b4_9 = f32x4_mul(self.twiddle10im, x9m14); |
| 2907 |
let t_b4_10 = f32x4_mul(self.twiddle6im, x10m13); |
| 2908 |
let t_b4_11 = f32x4_mul(self.twiddle2im, x11m12); |
| 2909 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m22); |
| 2910 |
let t_b5_2 = f32x4_mul(self.twiddle10im, x2m21); |
| 2911 |
let t_b5_3 = f32x4_mul(self.twiddle8im, x3m20); |
| 2912 |
let t_b5_4 = f32x4_mul(self.twiddle3im, x4m19); |
| 2913 |
let t_b5_5 = f32x4_mul(self.twiddle2im, x5m18); |
| 2914 |
let t_b5_6 = f32x4_mul(self.twiddle7im, x6m17); |
| 2915 |
let t_b5_7 = f32x4_mul(self.twiddle11im, x7m16); |
| 2916 |
let t_b5_8 = f32x4_mul(self.twiddle6im, x8m15); |
| 2917 |
let t_b5_9 = f32x4_mul(self.twiddle1im, x9m14); |
| 2918 |
let t_b5_10 = f32x4_mul(self.twiddle4im, x10m13); |
| 2919 |
let t_b5_11 = f32x4_mul(self.twiddle9im, x11m12); |
| 2920 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m22); |
| 2921 |
let t_b6_2 = f32x4_mul(self.twiddle11im, x2m21); |
| 2922 |
let t_b6_3 = f32x4_mul(self.twiddle5im, x3m20); |
| 2923 |
let t_b6_4 = f32x4_mul(self.twiddle1im, x4m19); |
| 2924 |
let t_b6_5 = f32x4_mul(self.twiddle7im, x5m18); |
| 2925 |
let t_b6_6 = f32x4_mul(self.twiddle10im, x6m17); |
| 2926 |
let t_b6_7 = f32x4_mul(self.twiddle4im, x7m16); |
| 2927 |
let t_b6_8 = f32x4_mul(self.twiddle2im, x8m15); |
| 2928 |
let t_b6_9 = f32x4_mul(self.twiddle8im, x9m14); |
| 2929 |
let t_b6_10 = f32x4_mul(self.twiddle9im, x10m13); |
| 2930 |
let t_b6_11 = f32x4_mul(self.twiddle3im, x11m12); |
| 2931 |
let t_b7_1 = f32x4_mul(self.twiddle7im, x1m22); |
| 2932 |
let t_b7_2 = f32x4_mul(self.twiddle9im, x2m21); |
| 2933 |
let t_b7_3 = f32x4_mul(self.twiddle2im, x3m20); |
| 2934 |
let t_b7_4 = f32x4_mul(self.twiddle5im, x4m19); |
| 2935 |
let t_b7_5 = f32x4_mul(self.twiddle11im, x5m18); |
| 2936 |
let t_b7_6 = f32x4_mul(self.twiddle4im, x6m17); |
| 2937 |
let t_b7_7 = f32x4_mul(self.twiddle3im, x7m16); |
| 2938 |
let t_b7_8 = f32x4_mul(self.twiddle10im, x8m15); |
| 2939 |
let t_b7_9 = f32x4_mul(self.twiddle6im, x9m14); |
| 2940 |
let t_b7_10 = f32x4_mul(self.twiddle1im, x10m13); |
| 2941 |
let t_b7_11 = f32x4_mul(self.twiddle8im, x11m12); |
| 2942 |
let t_b8_1 = f32x4_mul(self.twiddle8im, x1m22); |
| 2943 |
let t_b8_2 = f32x4_mul(self.twiddle7im, x2m21); |
| 2944 |
let t_b8_3 = f32x4_mul(self.twiddle1im, x3m20); |
| 2945 |
let t_b8_4 = f32x4_mul(self.twiddle9im, x4m19); |
| 2946 |
let t_b8_5 = f32x4_mul(self.twiddle6im, x5m18); |
| 2947 |
let t_b8_6 = f32x4_mul(self.twiddle2im, x6m17); |
| 2948 |
let t_b8_7 = f32x4_mul(self.twiddle10im, x7m16); |
| 2949 |
let t_b8_8 = f32x4_mul(self.twiddle5im, x8m15); |
| 2950 |
let t_b8_9 = f32x4_mul(self.twiddle3im, x9m14); |
| 2951 |
let t_b8_10 = f32x4_mul(self.twiddle11im, x10m13); |
| 2952 |
let t_b8_11 = f32x4_mul(self.twiddle4im, x11m12); |
| 2953 |
let t_b9_1 = f32x4_mul(self.twiddle9im, x1m22); |
| 2954 |
let t_b9_2 = f32x4_mul(self.twiddle5im, x2m21); |
| 2955 |
let t_b9_3 = f32x4_mul(self.twiddle4im, x3m20); |
| 2956 |
let t_b9_4 = f32x4_mul(self.twiddle10im, x4m19); |
| 2957 |
let t_b9_5 = f32x4_mul(self.twiddle1im, x5m18); |
| 2958 |
let t_b9_6 = f32x4_mul(self.twiddle8im, x6m17); |
| 2959 |
let t_b9_7 = f32x4_mul(self.twiddle6im, x7m16); |
| 2960 |
let t_b9_8 = f32x4_mul(self.twiddle3im, x8m15); |
| 2961 |
let t_b9_9 = f32x4_mul(self.twiddle11im, x9m14); |
| 2962 |
let t_b9_10 = f32x4_mul(self.twiddle2im, x10m13); |
| 2963 |
let t_b9_11 = f32x4_mul(self.twiddle7im, x11m12); |
| 2964 |
let t_b10_1 = f32x4_mul(self.twiddle10im, x1m22); |
| 2965 |
let t_b10_2 = f32x4_mul(self.twiddle3im, x2m21); |
| 2966 |
let t_b10_3 = f32x4_mul(self.twiddle7im, x3m20); |
| 2967 |
let t_b10_4 = f32x4_mul(self.twiddle6im, x4m19); |
| 2968 |
let t_b10_5 = f32x4_mul(self.twiddle4im, x5m18); |
| 2969 |
let t_b10_6 = f32x4_mul(self.twiddle9im, x6m17); |
| 2970 |
let t_b10_7 = f32x4_mul(self.twiddle1im, x7m16); |
| 2971 |
let t_b10_8 = f32x4_mul(self.twiddle11im, x8m15); |
| 2972 |
let t_b10_9 = f32x4_mul(self.twiddle2im, x9m14); |
| 2973 |
let t_b10_10 = f32x4_mul(self.twiddle8im, x10m13); |
| 2974 |
let t_b10_11 = f32x4_mul(self.twiddle5im, x11m12); |
| 2975 |
let t_b11_1 = f32x4_mul(self.twiddle11im, x1m22); |
| 2976 |
let t_b11_2 = f32x4_mul(self.twiddle1im, x2m21); |
| 2977 |
let t_b11_3 = f32x4_mul(self.twiddle10im, x3m20); |
| 2978 |
let t_b11_4 = f32x4_mul(self.twiddle2im, x4m19); |
| 2979 |
let t_b11_5 = f32x4_mul(self.twiddle9im, x5m18); |
| 2980 |
let t_b11_6 = f32x4_mul(self.twiddle3im, x6m17); |
| 2981 |
let t_b11_7 = f32x4_mul(self.twiddle8im, x7m16); |
| 2982 |
let t_b11_8 = f32x4_mul(self.twiddle4im, x8m15); |
| 2983 |
let t_b11_9 = f32x4_mul(self.twiddle7im, x9m14); |
| 2984 |
let t_b11_10 = f32x4_mul(self.twiddle5im, x10m13); |
| 2985 |
let t_b11_11 = f32x4_mul(self.twiddle6im, x11m12); |
| 2986 |
|
| 2987 |
let x0 = values[0]; |
| 2988 |
let t_a1 = calc_f32!( |
| 2989 |
x0 + t_a1_1 |
| 2990 |
+ t_a1_2 |
| 2991 |
+ t_a1_3 |
| 2992 |
+ t_a1_4 |
| 2993 |
+ t_a1_5 |
| 2994 |
+ t_a1_6 |
| 2995 |
+ t_a1_7 |
| 2996 |
+ t_a1_8 |
| 2997 |
+ t_a1_9 |
| 2998 |
+ t_a1_10 |
| 2999 |
+ t_a1_11 |
| 3000 |
); |
| 3001 |
let t_a2 = calc_f32!( |
| 3002 |
x0 + t_a2_1 |
| 3003 |
+ t_a2_2 |
| 3004 |
+ t_a2_3 |
| 3005 |
+ t_a2_4 |
| 3006 |
+ t_a2_5 |
| 3007 |
+ t_a2_6 |
| 3008 |
+ t_a2_7 |
| 3009 |
+ t_a2_8 |
| 3010 |
+ t_a2_9 |
| 3011 |
+ t_a2_10 |
| 3012 |
+ t_a2_11 |
| 3013 |
); |
| 3014 |
let t_a3 = calc_f32!( |
| 3015 |
x0 + t_a3_1 |
| 3016 |
+ t_a3_2 |
| 3017 |
+ t_a3_3 |
| 3018 |
+ t_a3_4 |
| 3019 |
+ t_a3_5 |
| 3020 |
+ t_a3_6 |
| 3021 |
+ t_a3_7 |
| 3022 |
+ t_a3_8 |
| 3023 |
+ t_a3_9 |
| 3024 |
+ t_a3_10 |
| 3025 |
+ t_a3_11 |
| 3026 |
); |
| 3027 |
let t_a4 = calc_f32!( |
| 3028 |
x0 + t_a4_1 |
| 3029 |
+ t_a4_2 |
| 3030 |
+ t_a4_3 |
| 3031 |
+ t_a4_4 |
| 3032 |
+ t_a4_5 |
| 3033 |
+ t_a4_6 |
| 3034 |
+ t_a4_7 |
| 3035 |
+ t_a4_8 |
| 3036 |
+ t_a4_9 |
| 3037 |
+ t_a4_10 |
| 3038 |
+ t_a4_11 |
| 3039 |
); |
| 3040 |
let t_a5 = calc_f32!( |
| 3041 |
x0 + t_a5_1 |
| 3042 |
+ t_a5_2 |
| 3043 |
+ t_a5_3 |
| 3044 |
+ t_a5_4 |
| 3045 |
+ t_a5_5 |
| 3046 |
+ t_a5_6 |
| 3047 |
+ t_a5_7 |
| 3048 |
+ t_a5_8 |
| 3049 |
+ t_a5_9 |
| 3050 |
+ t_a5_10 |
| 3051 |
+ t_a5_11 |
| 3052 |
); |
| 3053 |
let t_a6 = calc_f32!( |
| 3054 |
x0 + t_a6_1 |
| 3055 |
+ t_a6_2 |
| 3056 |
+ t_a6_3 |
| 3057 |
+ t_a6_4 |
| 3058 |
+ t_a6_5 |
| 3059 |
+ t_a6_6 |
| 3060 |
+ t_a6_7 |
| 3061 |
+ t_a6_8 |
| 3062 |
+ t_a6_9 |
| 3063 |
+ t_a6_10 |
| 3064 |
+ t_a6_11 |
| 3065 |
); |
| 3066 |
let t_a7 = calc_f32!( |
| 3067 |
x0 + t_a7_1 |
| 3068 |
+ t_a7_2 |
| 3069 |
+ t_a7_3 |
| 3070 |
+ t_a7_4 |
| 3071 |
+ t_a7_5 |
| 3072 |
+ t_a7_6 |
| 3073 |
+ t_a7_7 |
| 3074 |
+ t_a7_8 |
| 3075 |
+ t_a7_9 |
| 3076 |
+ t_a7_10 |
| 3077 |
+ t_a7_11 |
| 3078 |
); |
| 3079 |
let t_a8 = calc_f32!( |
| 3080 |
x0 + t_a8_1 |
| 3081 |
+ t_a8_2 |
| 3082 |
+ t_a8_3 |
| 3083 |
+ t_a8_4 |
| 3084 |
+ t_a8_5 |
| 3085 |
+ t_a8_6 |
| 3086 |
+ t_a8_7 |
| 3087 |
+ t_a8_8 |
| 3088 |
+ t_a8_9 |
| 3089 |
+ t_a8_10 |
| 3090 |
+ t_a8_11 |
| 3091 |
); |
| 3092 |
let t_a9 = calc_f32!( |
| 3093 |
x0 + t_a9_1 |
| 3094 |
+ t_a9_2 |
| 3095 |
+ t_a9_3 |
| 3096 |
+ t_a9_4 |
| 3097 |
+ t_a9_5 |
| 3098 |
+ t_a9_6 |
| 3099 |
+ t_a9_7 |
| 3100 |
+ t_a9_8 |
| 3101 |
+ t_a9_9 |
| 3102 |
+ t_a9_10 |
| 3103 |
+ t_a9_11 |
| 3104 |
); |
| 3105 |
let t_a10 = calc_f32!( |
| 3106 |
x0 + t_a10_1 |
| 3107 |
+ t_a10_2 |
| 3108 |
+ t_a10_3 |
| 3109 |
+ t_a10_4 |
| 3110 |
+ t_a10_5 |
| 3111 |
+ t_a10_6 |
| 3112 |
+ t_a10_7 |
| 3113 |
+ t_a10_8 |
| 3114 |
+ t_a10_9 |
| 3115 |
+ t_a10_10 |
| 3116 |
+ t_a10_11 |
| 3117 |
); |
| 3118 |
let t_a11 = calc_f32!( |
| 3119 |
x0 + t_a11_1 |
| 3120 |
+ t_a11_2 |
| 3121 |
+ t_a11_3 |
| 3122 |
+ t_a11_4 |
| 3123 |
+ t_a11_5 |
| 3124 |
+ t_a11_6 |
| 3125 |
+ t_a11_7 |
| 3126 |
+ t_a11_8 |
| 3127 |
+ t_a11_9 |
| 3128 |
+ t_a11_10 |
| 3129 |
+ t_a11_11 |
| 3130 |
); |
| 3131 |
|
| 3132 |
let t_b1 = calc_f32!( |
| 3133 |
t_b1_1 |
| 3134 |
+ t_b1_2 |
| 3135 |
+ t_b1_3 |
| 3136 |
+ t_b1_4 |
| 3137 |
+ t_b1_5 |
| 3138 |
+ t_b1_6 |
| 3139 |
+ t_b1_7 |
| 3140 |
+ t_b1_8 |
| 3141 |
+ t_b1_9 |
| 3142 |
+ t_b1_10 |
| 3143 |
+ t_b1_11 |
| 3144 |
); |
| 3145 |
let t_b2 = calc_f32!( |
| 3146 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 |
| 3147 |
- t_b2_6 |
| 3148 |
- t_b2_7 |
| 3149 |
- t_b2_8 |
| 3150 |
- t_b2_9 |
| 3151 |
- t_b2_10 |
| 3152 |
- t_b2_11 |
| 3153 |
); |
| 3154 |
let t_b3 = calc_f32!( |
| 3155 |
t_b3_1 + t_b3_2 + t_b3_3 - t_b3_4 - t_b3_5 - t_b3_6 - t_b3_7 |
| 3156 |
+ t_b3_8 |
| 3157 |
+ t_b3_9 |
| 3158 |
+ t_b3_10 |
| 3159 |
+ t_b3_11 |
| 3160 |
); |
| 3161 |
let t_b4 = calc_f32!( |
| 3162 |
t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 - t_b4_5 + t_b4_6 + t_b4_7 + t_b4_8 |
| 3163 |
- t_b4_9 |
| 3164 |
- t_b4_10 |
| 3165 |
- t_b4_11 |
| 3166 |
); |
| 3167 |
let t_b5 = calc_f32!( |
| 3168 |
t_b5_1 + t_b5_2 - t_b5_3 - t_b5_4 + t_b5_5 + t_b5_6 - t_b5_7 - t_b5_8 - t_b5_9 |
| 3169 |
+ t_b5_10 |
| 3170 |
+ t_b5_11 |
| 3171 |
); |
| 3172 |
let t_b6 = calc_f32!( |
| 3173 |
t_b6_1 - t_b6_2 - t_b6_3 + t_b6_4 + t_b6_5 - t_b6_6 - t_b6_7 + t_b6_8 + t_b6_9 |
| 3174 |
- t_b6_10 |
| 3175 |
- t_b6_11 |
| 3176 |
); |
| 3177 |
let t_b7 = calc_f32!( |
| 3178 |
t_b7_1 - t_b7_2 - t_b7_3 + t_b7_4 - t_b7_5 - t_b7_6 + t_b7_7 + t_b7_8 - t_b7_9 |
| 3179 |
+ t_b7_10 |
| 3180 |
+ t_b7_11 |
| 3181 |
); |
| 3182 |
let t_b8 = calc_f32!( |
| 3183 |
t_b8_1 - t_b8_2 + t_b8_3 + t_b8_4 - t_b8_5 + t_b8_6 + t_b8_7 - t_b8_8 |
| 3184 |
+ t_b8_9 |
| 3185 |
+ t_b8_10 |
| 3186 |
- t_b8_11 |
| 3187 |
); |
| 3188 |
let t_b9 = calc_f32!( |
| 3189 |
t_b9_1 - t_b9_2 + t_b9_3 - t_b9_4 - t_b9_5 + t_b9_6 - t_b9_7 + t_b9_8 |
| 3190 |
- t_b9_9 |
| 3191 |
- t_b9_10 |
| 3192 |
+ t_b9_11 |
| 3193 |
); |
| 3194 |
let t_b10 = calc_f32!( |
| 3195 |
t_b10_1 - t_b10_2 + t_b10_3 - t_b10_4 + t_b10_5 - t_b10_6 + t_b10_7 + t_b10_8 - t_b10_9 |
| 3196 |
+ t_b10_10 |
| 3197 |
- t_b10_11 |
| 3198 |
); |
| 3199 |
let t_b11 = calc_f32!( |
| 3200 |
t_b11_1 - t_b11_2 + t_b11_3 - t_b11_4 + t_b11_5 - t_b11_6 + t_b11_7 - t_b11_8 + t_b11_9 |
| 3201 |
- t_b11_10 |
| 3202 |
+ t_b11_11 |
| 3203 |
); |
| 3204 |
|
| 3205 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 3206 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 3207 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 3208 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 3209 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 3210 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 3211 |
let t_b7_rot = self.rotate.rotate_both(t_b7); |
| 3212 |
let t_b8_rot = self.rotate.rotate_both(t_b8); |
| 3213 |
let t_b9_rot = self.rotate.rotate_both(t_b9); |
| 3214 |
let t_b10_rot = self.rotate.rotate_both(t_b10); |
| 3215 |
let t_b11_rot = self.rotate.rotate_both(t_b11); |
| 3216 |
|
| 3217 |
let y0 = calc_f32!( |
| 3218 |
x0 + x1p22 |
| 3219 |
+ x2p21 |
| 3220 |
+ x3p20 |
| 3221 |
+ x4p19 |
| 3222 |
+ x5p18 |
| 3223 |
+ x6p17 |
| 3224 |
+ x7p16 |
| 3225 |
+ x8p15 |
| 3226 |
+ x9p14 |
| 3227 |
+ x10p13 |
| 3228 |
+ x11p12 |
| 3229 |
); |
| 3230 |
let [y1, y22] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 3231 |
let [y2, y21] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 3232 |
let [y3, y20] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 3233 |
let [y4, y19] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 3234 |
let [y5, y18] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 3235 |
let [y6, y17] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 3236 |
let [y7, y16] = parallel_fft2_interleaved_f32(t_a7, t_b7_rot); |
| 3237 |
let [y8, y15] = parallel_fft2_interleaved_f32(t_a8, t_b8_rot); |
| 3238 |
let [y9, y14] = parallel_fft2_interleaved_f32(t_a9, t_b9_rot); |
| 3239 |
let [y10, y13] = parallel_fft2_interleaved_f32(t_a10, t_b10_rot); |
| 3240 |
let [y11, y12] = parallel_fft2_interleaved_f32(t_a11, t_b11_rot); |
| 3241 |
[ |
| 3242 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 3243 |
y19, y20, y21, y22, |
| 3244 |
] |
| 3245 |
} |
| 3246 |
} |
| 3247 |
|
| 3248 |
// ____ _____ __ _ _ _ _ _ |
| 3249 |
// |___ \|___ / / /_ | || | | |__ (_) |_ |
| 3250 |
// __) | |_ \ _____ | '_ \| || |_| '_ \| | __| |
| 3251 |
// / __/ ___) | |_____| | (_) |__ _| |_) | | |_ |
| 3252 |
// |_____|____/ \___/ |_| |_.__/|_|\__| |
| 3253 |
// |
| 3254 |
|
| 3255 |
pub struct WasmSimdF64Butterfly23<T> { |
| 3256 |
direction: FftDirection, |
| 3257 |
_phantom: std::marker::PhantomData<T>, |
| 3258 |
rotate: Rotate90F64, |
| 3259 |
twiddle1re: v128, |
| 3260 |
twiddle1im: v128, |
| 3261 |
twiddle2re: v128, |
| 3262 |
twiddle2im: v128, |
| 3263 |
twiddle3re: v128, |
| 3264 |
twiddle3im: v128, |
| 3265 |
twiddle4re: v128, |
| 3266 |
twiddle4im: v128, |
| 3267 |
twiddle5re: v128, |
| 3268 |
twiddle5im: v128, |
| 3269 |
twiddle6re: v128, |
| 3270 |
twiddle6im: v128, |
| 3271 |
twiddle7re: v128, |
| 3272 |
twiddle7im: v128, |
| 3273 |
twiddle8re: v128, |
| 3274 |
twiddle8im: v128, |
| 3275 |
twiddle9re: v128, |
| 3276 |
twiddle9im: v128, |
| 3277 |
twiddle10re: v128, |
| 3278 |
twiddle10im: v128, |
| 3279 |
twiddle11re: v128, |
| 3280 |
twiddle11im: v128, |
| 3281 |
} |
| 3282 |
|
| 3283 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 3284 |
WasmSimdF64Butterfly23, |
| 3285 |
23, |
| 3286 |
|this: &WasmSimdF64Butterfly23<_>| this.direction |
| 3287 |
); |
| 3288 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 3289 |
WasmSimdF64Butterfly23, |
| 3290 |
23, |
| 3291 |
|this: &WasmSimdF64Butterfly23<_>| this.direction |
| 3292 |
); |
| 3293 |
impl<T: FftNum> WasmSimdF64Butterfly23<T> { |
| 3294 |
#[inline(always)] |
| 3295 |
pub fn new(direction: FftDirection) -> Self { |
| 3296 |
assert_f64::<T>(); |
| 3297 |
let rotate = Rotate90F64::new(true); |
| 3298 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 23, direction); |
| 3299 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 23, direction); |
| 3300 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 23, direction); |
| 3301 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 23, direction); |
| 3302 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 23, direction); |
| 3303 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 23, direction); |
| 3304 |
let tw7: Complex<f64> = twiddles::compute_twiddle(7, 23, direction); |
| 3305 |
let tw8: Complex<f64> = twiddles::compute_twiddle(8, 23, direction); |
| 3306 |
let tw9: Complex<f64> = twiddles::compute_twiddle(9, 23, direction); |
| 3307 |
let tw10: Complex<f64> = twiddles::compute_twiddle(10, 23, direction); |
| 3308 |
let tw11: Complex<f64> = twiddles::compute_twiddle(11, 23, direction); |
| 3309 |
let twiddle1re = f64x2_splat(tw1.re); |
| 3310 |
let twiddle1im = f64x2_splat(tw1.im); |
| 3311 |
let twiddle2re = f64x2_splat(tw2.re); |
| 3312 |
let twiddle2im = f64x2_splat(tw2.im); |
| 3313 |
let twiddle3re = f64x2_splat(tw3.re); |
| 3314 |
let twiddle3im = f64x2_splat(tw3.im); |
| 3315 |
let twiddle4re = f64x2_splat(tw4.re); |
| 3316 |
let twiddle4im = f64x2_splat(tw4.im); |
| 3317 |
let twiddle5re = f64x2_splat(tw5.re); |
| 3318 |
let twiddle5im = f64x2_splat(tw5.im); |
| 3319 |
let twiddle6re = f64x2_splat(tw6.re); |
| 3320 |
let twiddle6im = f64x2_splat(tw6.im); |
| 3321 |
let twiddle7re = f64x2_splat(tw7.re); |
| 3322 |
let twiddle7im = f64x2_splat(tw7.im); |
| 3323 |
let twiddle8re = f64x2_splat(tw8.re); |
| 3324 |
let twiddle8im = f64x2_splat(tw8.im); |
| 3325 |
let twiddle9re = f64x2_splat(tw9.re); |
| 3326 |
let twiddle9im = f64x2_splat(tw9.im); |
| 3327 |
let twiddle10re = f64x2_splat(tw10.re); |
| 3328 |
let twiddle10im = f64x2_splat(tw10.im); |
| 3329 |
let twiddle11re = f64x2_splat(tw11.re); |
| 3330 |
let twiddle11im = f64x2_splat(tw11.im); |
| 3331 |
|
| 3332 |
Self { |
| 3333 |
direction, |
| 3334 |
_phantom: std::marker::PhantomData, |
| 3335 |
rotate, |
| 3336 |
twiddle1re, |
| 3337 |
twiddle1im, |
| 3338 |
twiddle2re, |
| 3339 |
twiddle2im, |
| 3340 |
twiddle3re, |
| 3341 |
twiddle3im, |
| 3342 |
twiddle4re, |
| 3343 |
twiddle4im, |
| 3344 |
twiddle5re, |
| 3345 |
twiddle5im, |
| 3346 |
twiddle6re, |
| 3347 |
twiddle6im, |
| 3348 |
twiddle7re, |
| 3349 |
twiddle7im, |
| 3350 |
twiddle8re, |
| 3351 |
twiddle8im, |
| 3352 |
twiddle9re, |
| 3353 |
twiddle9im, |
| 3354 |
twiddle10re, |
| 3355 |
twiddle10im, |
| 3356 |
twiddle11re, |
| 3357 |
twiddle11im, |
| 3358 |
} |
| 3359 |
} |
| 3360 |
|
| 3361 |
#[inline(always)] |
| 3362 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 3363 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}); |
| 3364 |
|
| 3365 |
let out = self.perform_fft_direct(values); |
| 3366 |
|
| 3367 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22}); |
| 3368 |
} |
| 3369 |
|
| 3370 |
#[inline(always)] |
| 3371 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 23]) -> [v128; 23] { |
| 3372 |
let [x1p22, x1m22] = solo_fft2_f64(values[1], values[22]); |
| 3373 |
let [x2p21, x2m21] = solo_fft2_f64(values[2], values[21]); |
| 3374 |
let [x3p20, x3m20] = solo_fft2_f64(values[3], values[20]); |
| 3375 |
let [x4p19, x4m19] = solo_fft2_f64(values[4], values[19]); |
| 3376 |
let [x5p18, x5m18] = solo_fft2_f64(values[5], values[18]); |
| 3377 |
let [x6p17, x6m17] = solo_fft2_f64(values[6], values[17]); |
| 3378 |
let [x7p16, x7m16] = solo_fft2_f64(values[7], values[16]); |
| 3379 |
let [x8p15, x8m15] = solo_fft2_f64(values[8], values[15]); |
| 3380 |
let [x9p14, x9m14] = solo_fft2_f64(values[9], values[14]); |
| 3381 |
let [x10p13, x10m13] = solo_fft2_f64(values[10], values[13]); |
| 3382 |
let [x11p12, x11m12] = solo_fft2_f64(values[11], values[12]); |
| 3383 |
|
| 3384 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p22); |
| 3385 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p21); |
| 3386 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p20); |
| 3387 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p19); |
| 3388 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p18); |
| 3389 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p17); |
| 3390 |
let t_a1_7 = f64x2_mul(self.twiddle7re, x7p16); |
| 3391 |
let t_a1_8 = f64x2_mul(self.twiddle8re, x8p15); |
| 3392 |
let t_a1_9 = f64x2_mul(self.twiddle9re, x9p14); |
| 3393 |
let t_a1_10 = f64x2_mul(self.twiddle10re, x10p13); |
| 3394 |
let t_a1_11 = f64x2_mul(self.twiddle11re, x11p12); |
| 3395 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p22); |
| 3396 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p21); |
| 3397 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p20); |
| 3398 |
let t_a2_4 = f64x2_mul(self.twiddle8re, x4p19); |
| 3399 |
let t_a2_5 = f64x2_mul(self.twiddle10re, x5p18); |
| 3400 |
let t_a2_6 = f64x2_mul(self.twiddle11re, x6p17); |
| 3401 |
let t_a2_7 = f64x2_mul(self.twiddle9re, x7p16); |
| 3402 |
let t_a2_8 = f64x2_mul(self.twiddle7re, x8p15); |
| 3403 |
let t_a2_9 = f64x2_mul(self.twiddle5re, x9p14); |
| 3404 |
let t_a2_10 = f64x2_mul(self.twiddle3re, x10p13); |
| 3405 |
let t_a2_11 = f64x2_mul(self.twiddle1re, x11p12); |
| 3406 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p22); |
| 3407 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p21); |
| 3408 |
let t_a3_3 = f64x2_mul(self.twiddle9re, x3p20); |
| 3409 |
let t_a3_4 = f64x2_mul(self.twiddle11re, x4p19); |
| 3410 |
let t_a3_5 = f64x2_mul(self.twiddle8re, x5p18); |
| 3411 |
let t_a3_6 = f64x2_mul(self.twiddle5re, x6p17); |
| 3412 |
let t_a3_7 = f64x2_mul(self.twiddle2re, x7p16); |
| 3413 |
let t_a3_8 = f64x2_mul(self.twiddle1re, x8p15); |
| 3414 |
let t_a3_9 = f64x2_mul(self.twiddle4re, x9p14); |
| 3415 |
let t_a3_10 = f64x2_mul(self.twiddle7re, x10p13); |
| 3416 |
let t_a3_11 = f64x2_mul(self.twiddle10re, x11p12); |
| 3417 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p22); |
| 3418 |
let t_a4_2 = f64x2_mul(self.twiddle8re, x2p21); |
| 3419 |
let t_a4_3 = f64x2_mul(self.twiddle11re, x3p20); |
| 3420 |
let t_a4_4 = f64x2_mul(self.twiddle7re, x4p19); |
| 3421 |
let t_a4_5 = f64x2_mul(self.twiddle3re, x5p18); |
| 3422 |
let t_a4_6 = f64x2_mul(self.twiddle1re, x6p17); |
| 3423 |
let t_a4_7 = f64x2_mul(self.twiddle5re, x7p16); |
| 3424 |
let t_a4_8 = f64x2_mul(self.twiddle9re, x8p15); |
| 3425 |
let t_a4_9 = f64x2_mul(self.twiddle10re, x9p14); |
| 3426 |
let t_a4_10 = f64x2_mul(self.twiddle6re, x10p13); |
| 3427 |
let t_a4_11 = f64x2_mul(self.twiddle2re, x11p12); |
| 3428 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p22); |
| 3429 |
let t_a5_2 = f64x2_mul(self.twiddle10re, x2p21); |
| 3430 |
let t_a5_3 = f64x2_mul(self.twiddle8re, x3p20); |
| 3431 |
let t_a5_4 = f64x2_mul(self.twiddle3re, x4p19); |
| 3432 |
let t_a5_5 = f64x2_mul(self.twiddle2re, x5p18); |
| 3433 |
let t_a5_6 = f64x2_mul(self.twiddle7re, x6p17); |
| 3434 |
let t_a5_7 = f64x2_mul(self.twiddle11re, x7p16); |
| 3435 |
let t_a5_8 = f64x2_mul(self.twiddle6re, x8p15); |
| 3436 |
let t_a5_9 = f64x2_mul(self.twiddle1re, x9p14); |
| 3437 |
let t_a5_10 = f64x2_mul(self.twiddle4re, x10p13); |
| 3438 |
let t_a5_11 = f64x2_mul(self.twiddle9re, x11p12); |
| 3439 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p22); |
| 3440 |
let t_a6_2 = f64x2_mul(self.twiddle11re, x2p21); |
| 3441 |
let t_a6_3 = f64x2_mul(self.twiddle5re, x3p20); |
| 3442 |
let t_a6_4 = f64x2_mul(self.twiddle1re, x4p19); |
| 3443 |
let t_a6_5 = f64x2_mul(self.twiddle7re, x5p18); |
| 3444 |
let t_a6_6 = f64x2_mul(self.twiddle10re, x6p17); |
| 3445 |
let t_a6_7 = f64x2_mul(self.twiddle4re, x7p16); |
| 3446 |
let t_a6_8 = f64x2_mul(self.twiddle2re, x8p15); |
| 3447 |
let t_a6_9 = f64x2_mul(self.twiddle8re, x9p14); |
| 3448 |
let t_a6_10 = f64x2_mul(self.twiddle9re, x10p13); |
| 3449 |
let t_a6_11 = f64x2_mul(self.twiddle3re, x11p12); |
| 3450 |
let t_a7_1 = f64x2_mul(self.twiddle7re, x1p22); |
| 3451 |
let t_a7_2 = f64x2_mul(self.twiddle9re, x2p21); |
| 3452 |
let t_a7_3 = f64x2_mul(self.twiddle2re, x3p20); |
| 3453 |
let t_a7_4 = f64x2_mul(self.twiddle5re, x4p19); |
| 3454 |
let t_a7_5 = f64x2_mul(self.twiddle11re, x5p18); |
| 3455 |
let t_a7_6 = f64x2_mul(self.twiddle4re, x6p17); |
| 3456 |
let t_a7_7 = f64x2_mul(self.twiddle3re, x7p16); |
| 3457 |
let t_a7_8 = f64x2_mul(self.twiddle10re, x8p15); |
| 3458 |
let t_a7_9 = f64x2_mul(self.twiddle6re, x9p14); |
| 3459 |
let t_a7_10 = f64x2_mul(self.twiddle1re, x10p13); |
| 3460 |
let t_a7_11 = f64x2_mul(self.twiddle8re, x11p12); |
| 3461 |
let t_a8_1 = f64x2_mul(self.twiddle8re, x1p22); |
| 3462 |
let t_a8_2 = f64x2_mul(self.twiddle7re, x2p21); |
| 3463 |
let t_a8_3 = f64x2_mul(self.twiddle1re, x3p20); |
| 3464 |
let t_a8_4 = f64x2_mul(self.twiddle9re, x4p19); |
| 3465 |
let t_a8_5 = f64x2_mul(self.twiddle6re, x5p18); |
| 3466 |
let t_a8_6 = f64x2_mul(self.twiddle2re, x6p17); |
| 3467 |
let t_a8_7 = f64x2_mul(self.twiddle10re, x7p16); |
| 3468 |
let t_a8_8 = f64x2_mul(self.twiddle5re, x8p15); |
| 3469 |
let t_a8_9 = f64x2_mul(self.twiddle3re, x9p14); |
| 3470 |
let t_a8_10 = f64x2_mul(self.twiddle11re, x10p13); |
| 3471 |
let t_a8_11 = f64x2_mul(self.twiddle4re, x11p12); |
| 3472 |
let t_a9_1 = f64x2_mul(self.twiddle9re, x1p22); |
| 3473 |
let t_a9_2 = f64x2_mul(self.twiddle5re, x2p21); |
| 3474 |
let t_a9_3 = f64x2_mul(self.twiddle4re, x3p20); |
| 3475 |
let t_a9_4 = f64x2_mul(self.twiddle10re, x4p19); |
| 3476 |
let t_a9_5 = f64x2_mul(self.twiddle1re, x5p18); |
| 3477 |
let t_a9_6 = f64x2_mul(self.twiddle8re, x6p17); |
| 3478 |
let t_a9_7 = f64x2_mul(self.twiddle6re, x7p16); |
| 3479 |
let t_a9_8 = f64x2_mul(self.twiddle3re, x8p15); |
| 3480 |
let t_a9_9 = f64x2_mul(self.twiddle11re, x9p14); |
| 3481 |
let t_a9_10 = f64x2_mul(self.twiddle2re, x10p13); |
| 3482 |
let t_a9_11 = f64x2_mul(self.twiddle7re, x11p12); |
| 3483 |
let t_a10_1 = f64x2_mul(self.twiddle10re, x1p22); |
| 3484 |
let t_a10_2 = f64x2_mul(self.twiddle3re, x2p21); |
| 3485 |
let t_a10_3 = f64x2_mul(self.twiddle7re, x3p20); |
| 3486 |
let t_a10_4 = f64x2_mul(self.twiddle6re, x4p19); |
| 3487 |
let t_a10_5 = f64x2_mul(self.twiddle4re, x5p18); |
| 3488 |
let t_a10_6 = f64x2_mul(self.twiddle9re, x6p17); |
| 3489 |
let t_a10_7 = f64x2_mul(self.twiddle1re, x7p16); |
| 3490 |
let t_a10_8 = f64x2_mul(self.twiddle11re, x8p15); |
| 3491 |
let t_a10_9 = f64x2_mul(self.twiddle2re, x9p14); |
| 3492 |
let t_a10_10 = f64x2_mul(self.twiddle8re, x10p13); |
| 3493 |
let t_a10_11 = f64x2_mul(self.twiddle5re, x11p12); |
| 3494 |
let t_a11_1 = f64x2_mul(self.twiddle11re, x1p22); |
| 3495 |
let t_a11_2 = f64x2_mul(self.twiddle1re, x2p21); |
| 3496 |
let t_a11_3 = f64x2_mul(self.twiddle10re, x3p20); |
| 3497 |
let t_a11_4 = f64x2_mul(self.twiddle2re, x4p19); |
| 3498 |
let t_a11_5 = f64x2_mul(self.twiddle9re, x5p18); |
| 3499 |
let t_a11_6 = f64x2_mul(self.twiddle3re, x6p17); |
| 3500 |
let t_a11_7 = f64x2_mul(self.twiddle8re, x7p16); |
| 3501 |
let t_a11_8 = f64x2_mul(self.twiddle4re, x8p15); |
| 3502 |
let t_a11_9 = f64x2_mul(self.twiddle7re, x9p14); |
| 3503 |
let t_a11_10 = f64x2_mul(self.twiddle5re, x10p13); |
| 3504 |
let t_a11_11 = f64x2_mul(self.twiddle6re, x11p12); |
| 3505 |
|
| 3506 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m22); |
| 3507 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m21); |
| 3508 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m20); |
| 3509 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m19); |
| 3510 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m18); |
| 3511 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m17); |
| 3512 |
let t_b1_7 = f64x2_mul(self.twiddle7im, x7m16); |
| 3513 |
let t_b1_8 = f64x2_mul(self.twiddle8im, x8m15); |
| 3514 |
let t_b1_9 = f64x2_mul(self.twiddle9im, x9m14); |
| 3515 |
let t_b1_10 = f64x2_mul(self.twiddle10im, x10m13); |
| 3516 |
let t_b1_11 = f64x2_mul(self.twiddle11im, x11m12); |
| 3517 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m22); |
| 3518 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m21); |
| 3519 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m20); |
| 3520 |
let t_b2_4 = f64x2_mul(self.twiddle8im, x4m19); |
| 3521 |
let t_b2_5 = f64x2_mul(self.twiddle10im, x5m18); |
| 3522 |
let t_b2_6 = f64x2_mul(self.twiddle11im, x6m17); |
| 3523 |
let t_b2_7 = f64x2_mul(self.twiddle9im, x7m16); |
| 3524 |
let t_b2_8 = f64x2_mul(self.twiddle7im, x8m15); |
| 3525 |
let t_b2_9 = f64x2_mul(self.twiddle5im, x9m14); |
| 3526 |
let t_b2_10 = f64x2_mul(self.twiddle3im, x10m13); |
| 3527 |
let t_b2_11 = f64x2_mul(self.twiddle1im, x11m12); |
| 3528 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m22); |
| 3529 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m21); |
| 3530 |
let t_b3_3 = f64x2_mul(self.twiddle9im, x3m20); |
| 3531 |
let t_b3_4 = f64x2_mul(self.twiddle11im, x4m19); |
| 3532 |
let t_b3_5 = f64x2_mul(self.twiddle8im, x5m18); |
| 3533 |
let t_b3_6 = f64x2_mul(self.twiddle5im, x6m17); |
| 3534 |
let t_b3_7 = f64x2_mul(self.twiddle2im, x7m16); |
| 3535 |
let t_b3_8 = f64x2_mul(self.twiddle1im, x8m15); |
| 3536 |
let t_b3_9 = f64x2_mul(self.twiddle4im, x9m14); |
| 3537 |
let t_b3_10 = f64x2_mul(self.twiddle7im, x10m13); |
| 3538 |
let t_b3_11 = f64x2_mul(self.twiddle10im, x11m12); |
| 3539 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m22); |
| 3540 |
let t_b4_2 = f64x2_mul(self.twiddle8im, x2m21); |
| 3541 |
let t_b4_3 = f64x2_mul(self.twiddle11im, x3m20); |
| 3542 |
let t_b4_4 = f64x2_mul(self.twiddle7im, x4m19); |
| 3543 |
let t_b4_5 = f64x2_mul(self.twiddle3im, x5m18); |
| 3544 |
let t_b4_6 = f64x2_mul(self.twiddle1im, x6m17); |
| 3545 |
let t_b4_7 = f64x2_mul(self.twiddle5im, x7m16); |
| 3546 |
let t_b4_8 = f64x2_mul(self.twiddle9im, x8m15); |
| 3547 |
let t_b4_9 = f64x2_mul(self.twiddle10im, x9m14); |
| 3548 |
let t_b4_10 = f64x2_mul(self.twiddle6im, x10m13); |
| 3549 |
let t_b4_11 = f64x2_mul(self.twiddle2im, x11m12); |
| 3550 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m22); |
| 3551 |
let t_b5_2 = f64x2_mul(self.twiddle10im, x2m21); |
| 3552 |
let t_b5_3 = f64x2_mul(self.twiddle8im, x3m20); |
| 3553 |
let t_b5_4 = f64x2_mul(self.twiddle3im, x4m19); |
| 3554 |
let t_b5_5 = f64x2_mul(self.twiddle2im, x5m18); |
| 3555 |
let t_b5_6 = f64x2_mul(self.twiddle7im, x6m17); |
| 3556 |
let t_b5_7 = f64x2_mul(self.twiddle11im, x7m16); |
| 3557 |
let t_b5_8 = f64x2_mul(self.twiddle6im, x8m15); |
| 3558 |
let t_b5_9 = f64x2_mul(self.twiddle1im, x9m14); |
| 3559 |
let t_b5_10 = f64x2_mul(self.twiddle4im, x10m13); |
| 3560 |
let t_b5_11 = f64x2_mul(self.twiddle9im, x11m12); |
| 3561 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m22); |
| 3562 |
let t_b6_2 = f64x2_mul(self.twiddle11im, x2m21); |
| 3563 |
let t_b6_3 = f64x2_mul(self.twiddle5im, x3m20); |
| 3564 |
let t_b6_4 = f64x2_mul(self.twiddle1im, x4m19); |
| 3565 |
let t_b6_5 = f64x2_mul(self.twiddle7im, x5m18); |
| 3566 |
let t_b6_6 = f64x2_mul(self.twiddle10im, x6m17); |
| 3567 |
let t_b6_7 = f64x2_mul(self.twiddle4im, x7m16); |
| 3568 |
let t_b6_8 = f64x2_mul(self.twiddle2im, x8m15); |
| 3569 |
let t_b6_9 = f64x2_mul(self.twiddle8im, x9m14); |
| 3570 |
let t_b6_10 = f64x2_mul(self.twiddle9im, x10m13); |
| 3571 |
let t_b6_11 = f64x2_mul(self.twiddle3im, x11m12); |
| 3572 |
let t_b7_1 = f64x2_mul(self.twiddle7im, x1m22); |
| 3573 |
let t_b7_2 = f64x2_mul(self.twiddle9im, x2m21); |
| 3574 |
let t_b7_3 = f64x2_mul(self.twiddle2im, x3m20); |
| 3575 |
let t_b7_4 = f64x2_mul(self.twiddle5im, x4m19); |
| 3576 |
let t_b7_5 = f64x2_mul(self.twiddle11im, x5m18); |
| 3577 |
let t_b7_6 = f64x2_mul(self.twiddle4im, x6m17); |
| 3578 |
let t_b7_7 = f64x2_mul(self.twiddle3im, x7m16); |
| 3579 |
let t_b7_8 = f64x2_mul(self.twiddle10im, x8m15); |
| 3580 |
let t_b7_9 = f64x2_mul(self.twiddle6im, x9m14); |
| 3581 |
let t_b7_10 = f64x2_mul(self.twiddle1im, x10m13); |
| 3582 |
let t_b7_11 = f64x2_mul(self.twiddle8im, x11m12); |
| 3583 |
let t_b8_1 = f64x2_mul(self.twiddle8im, x1m22); |
| 3584 |
let t_b8_2 = f64x2_mul(self.twiddle7im, x2m21); |
| 3585 |
let t_b8_3 = f64x2_mul(self.twiddle1im, x3m20); |
| 3586 |
let t_b8_4 = f64x2_mul(self.twiddle9im, x4m19); |
| 3587 |
let t_b8_5 = f64x2_mul(self.twiddle6im, x5m18); |
| 3588 |
let t_b8_6 = f64x2_mul(self.twiddle2im, x6m17); |
| 3589 |
let t_b8_7 = f64x2_mul(self.twiddle10im, x7m16); |
| 3590 |
let t_b8_8 = f64x2_mul(self.twiddle5im, x8m15); |
| 3591 |
let t_b8_9 = f64x2_mul(self.twiddle3im, x9m14); |
| 3592 |
let t_b8_10 = f64x2_mul(self.twiddle11im, x10m13); |
| 3593 |
let t_b8_11 = f64x2_mul(self.twiddle4im, x11m12); |
| 3594 |
let t_b9_1 = f64x2_mul(self.twiddle9im, x1m22); |
| 3595 |
let t_b9_2 = f64x2_mul(self.twiddle5im, x2m21); |
| 3596 |
let t_b9_3 = f64x2_mul(self.twiddle4im, x3m20); |
| 3597 |
let t_b9_4 = f64x2_mul(self.twiddle10im, x4m19); |
| 3598 |
let t_b9_5 = f64x2_mul(self.twiddle1im, x5m18); |
| 3599 |
let t_b9_6 = f64x2_mul(self.twiddle8im, x6m17); |
| 3600 |
let t_b9_7 = f64x2_mul(self.twiddle6im, x7m16); |
| 3601 |
let t_b9_8 = f64x2_mul(self.twiddle3im, x8m15); |
| 3602 |
let t_b9_9 = f64x2_mul(self.twiddle11im, x9m14); |
| 3603 |
let t_b9_10 = f64x2_mul(self.twiddle2im, x10m13); |
| 3604 |
let t_b9_11 = f64x2_mul(self.twiddle7im, x11m12); |
| 3605 |
let t_b10_1 = f64x2_mul(self.twiddle10im, x1m22); |
| 3606 |
let t_b10_2 = f64x2_mul(self.twiddle3im, x2m21); |
| 3607 |
let t_b10_3 = f64x2_mul(self.twiddle7im, x3m20); |
| 3608 |
let t_b10_4 = f64x2_mul(self.twiddle6im, x4m19); |
| 3609 |
let t_b10_5 = f64x2_mul(self.twiddle4im, x5m18); |
| 3610 |
let t_b10_6 = f64x2_mul(self.twiddle9im, x6m17); |
| 3611 |
let t_b10_7 = f64x2_mul(self.twiddle1im, x7m16); |
| 3612 |
let t_b10_8 = f64x2_mul(self.twiddle11im, x8m15); |
| 3613 |
let t_b10_9 = f64x2_mul(self.twiddle2im, x9m14); |
| 3614 |
let t_b10_10 = f64x2_mul(self.twiddle8im, x10m13); |
| 3615 |
let t_b10_11 = f64x2_mul(self.twiddle5im, x11m12); |
| 3616 |
let t_b11_1 = f64x2_mul(self.twiddle11im, x1m22); |
| 3617 |
let t_b11_2 = f64x2_mul(self.twiddle1im, x2m21); |
| 3618 |
let t_b11_3 = f64x2_mul(self.twiddle10im, x3m20); |
| 3619 |
let t_b11_4 = f64x2_mul(self.twiddle2im, x4m19); |
| 3620 |
let t_b11_5 = f64x2_mul(self.twiddle9im, x5m18); |
| 3621 |
let t_b11_6 = f64x2_mul(self.twiddle3im, x6m17); |
| 3622 |
let t_b11_7 = f64x2_mul(self.twiddle8im, x7m16); |
| 3623 |
let t_b11_8 = f64x2_mul(self.twiddle4im, x8m15); |
| 3624 |
let t_b11_9 = f64x2_mul(self.twiddle7im, x9m14); |
| 3625 |
let t_b11_10 = f64x2_mul(self.twiddle5im, x10m13); |
| 3626 |
let t_b11_11 = f64x2_mul(self.twiddle6im, x11m12); |
| 3627 |
|
| 3628 |
let x0 = values[0]; |
| 3629 |
let t_a1 = calc_f64!( |
| 3630 |
x0 + t_a1_1 |
| 3631 |
+ t_a1_2 |
| 3632 |
+ t_a1_3 |
| 3633 |
+ t_a1_4 |
| 3634 |
+ t_a1_5 |
| 3635 |
+ t_a1_6 |
| 3636 |
+ t_a1_7 |
| 3637 |
+ t_a1_8 |
| 3638 |
+ t_a1_9 |
| 3639 |
+ t_a1_10 |
| 3640 |
+ t_a1_11 |
| 3641 |
); |
| 3642 |
let t_a2 = calc_f64!( |
| 3643 |
x0 + t_a2_1 |
| 3644 |
+ t_a2_2 |
| 3645 |
+ t_a2_3 |
| 3646 |
+ t_a2_4 |
| 3647 |
+ t_a2_5 |
| 3648 |
+ t_a2_6 |
| 3649 |
+ t_a2_7 |
| 3650 |
+ t_a2_8 |
| 3651 |
+ t_a2_9 |
| 3652 |
+ t_a2_10 |
| 3653 |
+ t_a2_11 |
| 3654 |
); |
| 3655 |
let t_a3 = calc_f64!( |
| 3656 |
x0 + t_a3_1 |
| 3657 |
+ t_a3_2 |
| 3658 |
+ t_a3_3 |
| 3659 |
+ t_a3_4 |
| 3660 |
+ t_a3_5 |
| 3661 |
+ t_a3_6 |
| 3662 |
+ t_a3_7 |
| 3663 |
+ t_a3_8 |
| 3664 |
+ t_a3_9 |
| 3665 |
+ t_a3_10 |
| 3666 |
+ t_a3_11 |
| 3667 |
); |
| 3668 |
let t_a4 = calc_f64!( |
| 3669 |
x0 + t_a4_1 |
| 3670 |
+ t_a4_2 |
| 3671 |
+ t_a4_3 |
| 3672 |
+ t_a4_4 |
| 3673 |
+ t_a4_5 |
| 3674 |
+ t_a4_6 |
| 3675 |
+ t_a4_7 |
| 3676 |
+ t_a4_8 |
| 3677 |
+ t_a4_9 |
| 3678 |
+ t_a4_10 |
| 3679 |
+ t_a4_11 |
| 3680 |
); |
| 3681 |
let t_a5 = calc_f64!( |
| 3682 |
x0 + t_a5_1 |
| 3683 |
+ t_a5_2 |
| 3684 |
+ t_a5_3 |
| 3685 |
+ t_a5_4 |
| 3686 |
+ t_a5_5 |
| 3687 |
+ t_a5_6 |
| 3688 |
+ t_a5_7 |
| 3689 |
+ t_a5_8 |
| 3690 |
+ t_a5_9 |
| 3691 |
+ t_a5_10 |
| 3692 |
+ t_a5_11 |
| 3693 |
); |
| 3694 |
let t_a6 = calc_f64!( |
| 3695 |
x0 + t_a6_1 |
| 3696 |
+ t_a6_2 |
| 3697 |
+ t_a6_3 |
| 3698 |
+ t_a6_4 |
| 3699 |
+ t_a6_5 |
| 3700 |
+ t_a6_6 |
| 3701 |
+ t_a6_7 |
| 3702 |
+ t_a6_8 |
| 3703 |
+ t_a6_9 |
| 3704 |
+ t_a6_10 |
| 3705 |
+ t_a6_11 |
| 3706 |
); |
| 3707 |
let t_a7 = calc_f64!( |
| 3708 |
x0 + t_a7_1 |
| 3709 |
+ t_a7_2 |
| 3710 |
+ t_a7_3 |
| 3711 |
+ t_a7_4 |
| 3712 |
+ t_a7_5 |
| 3713 |
+ t_a7_6 |
| 3714 |
+ t_a7_7 |
| 3715 |
+ t_a7_8 |
| 3716 |
+ t_a7_9 |
| 3717 |
+ t_a7_10 |
| 3718 |
+ t_a7_11 |
| 3719 |
); |
| 3720 |
let t_a8 = calc_f64!( |
| 3721 |
x0 + t_a8_1 |
| 3722 |
+ t_a8_2 |
| 3723 |
+ t_a8_3 |
| 3724 |
+ t_a8_4 |
| 3725 |
+ t_a8_5 |
| 3726 |
+ t_a8_6 |
| 3727 |
+ t_a8_7 |
| 3728 |
+ t_a8_8 |
| 3729 |
+ t_a8_9 |
| 3730 |
+ t_a8_10 |
| 3731 |
+ t_a8_11 |
| 3732 |
); |
| 3733 |
let t_a9 = calc_f64!( |
| 3734 |
x0 + t_a9_1 |
| 3735 |
+ t_a9_2 |
| 3736 |
+ t_a9_3 |
| 3737 |
+ t_a9_4 |
| 3738 |
+ t_a9_5 |
| 3739 |
+ t_a9_6 |
| 3740 |
+ t_a9_7 |
| 3741 |
+ t_a9_8 |
| 3742 |
+ t_a9_9 |
| 3743 |
+ t_a9_10 |
| 3744 |
+ t_a9_11 |
| 3745 |
); |
| 3746 |
let t_a10 = calc_f64!( |
| 3747 |
x0 + t_a10_1 |
| 3748 |
+ t_a10_2 |
| 3749 |
+ t_a10_3 |
| 3750 |
+ t_a10_4 |
| 3751 |
+ t_a10_5 |
| 3752 |
+ t_a10_6 |
| 3753 |
+ t_a10_7 |
| 3754 |
+ t_a10_8 |
| 3755 |
+ t_a10_9 |
| 3756 |
+ t_a10_10 |
| 3757 |
+ t_a10_11 |
| 3758 |
); |
| 3759 |
let t_a11 = calc_f64!( |
| 3760 |
x0 + t_a11_1 |
| 3761 |
+ t_a11_2 |
| 3762 |
+ t_a11_3 |
| 3763 |
+ t_a11_4 |
| 3764 |
+ t_a11_5 |
| 3765 |
+ t_a11_6 |
| 3766 |
+ t_a11_7 |
| 3767 |
+ t_a11_8 |
| 3768 |
+ t_a11_9 |
| 3769 |
+ t_a11_10 |
| 3770 |
+ t_a11_11 |
| 3771 |
); |
| 3772 |
|
| 3773 |
let t_b1 = calc_f64!( |
| 3774 |
t_b1_1 |
| 3775 |
+ t_b1_2 |
| 3776 |
+ t_b1_3 |
| 3777 |
+ t_b1_4 |
| 3778 |
+ t_b1_5 |
| 3779 |
+ t_b1_6 |
| 3780 |
+ t_b1_7 |
| 3781 |
+ t_b1_8 |
| 3782 |
+ t_b1_9 |
| 3783 |
+ t_b1_10 |
| 3784 |
+ t_b1_11 |
| 3785 |
); |
| 3786 |
let t_b2 = calc_f64!( |
| 3787 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 |
| 3788 |
- t_b2_6 |
| 3789 |
- t_b2_7 |
| 3790 |
- t_b2_8 |
| 3791 |
- t_b2_9 |
| 3792 |
- t_b2_10 |
| 3793 |
- t_b2_11 |
| 3794 |
); |
| 3795 |
let t_b3 = calc_f64!( |
| 3796 |
t_b3_1 + t_b3_2 + t_b3_3 - t_b3_4 - t_b3_5 - t_b3_6 - t_b3_7 |
| 3797 |
+ t_b3_8 |
| 3798 |
+ t_b3_9 |
| 3799 |
+ t_b3_10 |
| 3800 |
+ t_b3_11 |
| 3801 |
); |
| 3802 |
let t_b4 = calc_f64!( |
| 3803 |
t_b4_1 + t_b4_2 - t_b4_3 - t_b4_4 - t_b4_5 + t_b4_6 + t_b4_7 + t_b4_8 |
| 3804 |
- t_b4_9 |
| 3805 |
- t_b4_10 |
| 3806 |
- t_b4_11 |
| 3807 |
); |
| 3808 |
let t_b5 = calc_f64!( |
| 3809 |
t_b5_1 + t_b5_2 - t_b5_3 - t_b5_4 + t_b5_5 + t_b5_6 - t_b5_7 - t_b5_8 - t_b5_9 |
| 3810 |
+ t_b5_10 |
| 3811 |
+ t_b5_11 |
| 3812 |
); |
| 3813 |
let t_b6 = calc_f64!( |
| 3814 |
t_b6_1 - t_b6_2 - t_b6_3 + t_b6_4 + t_b6_5 - t_b6_6 - t_b6_7 + t_b6_8 + t_b6_9 |
| 3815 |
- t_b6_10 |
| 3816 |
- t_b6_11 |
| 3817 |
); |
| 3818 |
let t_b7 = calc_f64!( |
| 3819 |
t_b7_1 - t_b7_2 - t_b7_3 + t_b7_4 - t_b7_5 - t_b7_6 + t_b7_7 + t_b7_8 - t_b7_9 |
| 3820 |
+ t_b7_10 |
| 3821 |
+ t_b7_11 |
| 3822 |
); |
| 3823 |
let t_b8 = calc_f64!( |
| 3824 |
t_b8_1 - t_b8_2 + t_b8_3 + t_b8_4 - t_b8_5 + t_b8_6 + t_b8_7 - t_b8_8 |
| 3825 |
+ t_b8_9 |
| 3826 |
+ t_b8_10 |
| 3827 |
- t_b8_11 |
| 3828 |
); |
| 3829 |
let t_b9 = calc_f64!( |
| 3830 |
t_b9_1 - t_b9_2 + t_b9_3 - t_b9_4 - t_b9_5 + t_b9_6 - t_b9_7 + t_b9_8 |
| 3831 |
- t_b9_9 |
| 3832 |
- t_b9_10 |
| 3833 |
+ t_b9_11 |
| 3834 |
); |
| 3835 |
let t_b10 = calc_f64!( |
| 3836 |
t_b10_1 - t_b10_2 + t_b10_3 - t_b10_4 + t_b10_5 - t_b10_6 + t_b10_7 + t_b10_8 - t_b10_9 |
| 3837 |
+ t_b10_10 |
| 3838 |
- t_b10_11 |
| 3839 |
); |
| 3840 |
let t_b11 = calc_f64!( |
| 3841 |
t_b11_1 - t_b11_2 + t_b11_3 - t_b11_4 + t_b11_5 - t_b11_6 + t_b11_7 - t_b11_8 + t_b11_9 |
| 3842 |
- t_b11_10 |
| 3843 |
+ t_b11_11 |
| 3844 |
); |
| 3845 |
|
| 3846 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 3847 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 3848 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 3849 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 3850 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 3851 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 3852 |
let t_b7_rot = self.rotate.rotate(t_b7); |
| 3853 |
let t_b8_rot = self.rotate.rotate(t_b8); |
| 3854 |
let t_b9_rot = self.rotate.rotate(t_b9); |
| 3855 |
let t_b10_rot = self.rotate.rotate(t_b10); |
| 3856 |
let t_b11_rot = self.rotate.rotate(t_b11); |
| 3857 |
|
| 3858 |
let y0 = calc_f64!( |
| 3859 |
x0 + x1p22 |
| 3860 |
+ x2p21 |
| 3861 |
+ x3p20 |
| 3862 |
+ x4p19 |
| 3863 |
+ x5p18 |
| 3864 |
+ x6p17 |
| 3865 |
+ x7p16 |
| 3866 |
+ x8p15 |
| 3867 |
+ x9p14 |
| 3868 |
+ x10p13 |
| 3869 |
+ x11p12 |
| 3870 |
); |
| 3871 |
let [y1, y22] = solo_fft2_f64(t_a1, t_b1_rot); |
| 3872 |
let [y2, y21] = solo_fft2_f64(t_a2, t_b2_rot); |
| 3873 |
let [y3, y20] = solo_fft2_f64(t_a3, t_b3_rot); |
| 3874 |
let [y4, y19] = solo_fft2_f64(t_a4, t_b4_rot); |
| 3875 |
let [y5, y18] = solo_fft2_f64(t_a5, t_b5_rot); |
| 3876 |
let [y6, y17] = solo_fft2_f64(t_a6, t_b6_rot); |
| 3877 |
let [y7, y16] = solo_fft2_f64(t_a7, t_b7_rot); |
| 3878 |
let [y8, y15] = solo_fft2_f64(t_a8, t_b8_rot); |
| 3879 |
let [y9, y14] = solo_fft2_f64(t_a9, t_b9_rot); |
| 3880 |
let [y10, y13] = solo_fft2_f64(t_a10, t_b10_rot); |
| 3881 |
let [y11, y12] = solo_fft2_f64(t_a11, t_b11_rot); |
| 3882 |
[ |
| 3883 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 3884 |
y19, y20, y21, y22, |
| 3885 |
] |
| 3886 |
} |
| 3887 |
} |
| 3888 |
|
| 3889 |
// ____ ___ _________ _ _ _ |
| 3890 |
// |___ \ / _ \ |___ /___ \| |__ (_) |_ |
| 3891 |
// __) | (_) | _____ |_ \ __) | '_ \| | __| |
| 3892 |
// / __/ \__, | |_____| ___) / __/| |_) | | |_ |
| 3893 |
// |_____| /_/ |____/_____|_.__/|_|\__| |
| 3894 |
// |
| 3895 |
|
| 3896 |
pub struct WasmSimdF32Butterfly29<T> { |
| 3897 |
direction: FftDirection, |
| 3898 |
_phantom: std::marker::PhantomData<T>, |
| 3899 |
rotate: Rotate90F32, |
| 3900 |
twiddle1re: v128, |
| 3901 |
twiddle1im: v128, |
| 3902 |
twiddle2re: v128, |
| 3903 |
twiddle2im: v128, |
| 3904 |
twiddle3re: v128, |
| 3905 |
twiddle3im: v128, |
| 3906 |
twiddle4re: v128, |
| 3907 |
twiddle4im: v128, |
| 3908 |
twiddle5re: v128, |
| 3909 |
twiddle5im: v128, |
| 3910 |
twiddle6re: v128, |
| 3911 |
twiddle6im: v128, |
| 3912 |
twiddle7re: v128, |
| 3913 |
twiddle7im: v128, |
| 3914 |
twiddle8re: v128, |
| 3915 |
twiddle8im: v128, |
| 3916 |
twiddle9re: v128, |
| 3917 |
twiddle9im: v128, |
| 3918 |
twiddle10re: v128, |
| 3919 |
twiddle10im: v128, |
| 3920 |
twiddle11re: v128, |
| 3921 |
twiddle11im: v128, |
| 3922 |
twiddle12re: v128, |
| 3923 |
twiddle12im: v128, |
| 3924 |
twiddle13re: v128, |
| 3925 |
twiddle13im: v128, |
| 3926 |
twiddle14re: v128, |
| 3927 |
twiddle14im: v128, |
| 3928 |
} |
| 3929 |
|
| 3930 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 3931 |
WasmSimdF32Butterfly29, |
| 3932 |
29, |
| 3933 |
|this: &WasmSimdF32Butterfly29<_>| this.direction |
| 3934 |
); |
| 3935 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 3936 |
WasmSimdF32Butterfly29, |
| 3937 |
29, |
| 3938 |
|this: &WasmSimdF32Butterfly29<_>| this.direction |
| 3939 |
); |
| 3940 |
impl<T: FftNum> WasmSimdF32Butterfly29<T> { |
| 3941 |
#[inline(always)] |
| 3942 |
pub fn new(direction: FftDirection) -> Self { |
| 3943 |
assert_f32::<T>(); |
| 3944 |
let rotate = Rotate90F32::new(true); |
| 3945 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 29, direction); |
| 3946 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 29, direction); |
| 3947 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 29, direction); |
| 3948 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 29, direction); |
| 3949 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 29, direction); |
| 3950 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 29, direction); |
| 3951 |
let tw7: Complex<f32> = twiddles::compute_twiddle(7, 29, direction); |
| 3952 |
let tw8: Complex<f32> = twiddles::compute_twiddle(8, 29, direction); |
| 3953 |
let tw9: Complex<f32> = twiddles::compute_twiddle(9, 29, direction); |
| 3954 |
let tw10: Complex<f32> = twiddles::compute_twiddle(10, 29, direction); |
| 3955 |
let tw11: Complex<f32> = twiddles::compute_twiddle(11, 29, direction); |
| 3956 |
let tw12: Complex<f32> = twiddles::compute_twiddle(12, 29, direction); |
| 3957 |
let tw13: Complex<f32> = twiddles::compute_twiddle(13, 29, direction); |
| 3958 |
let tw14: Complex<f32> = twiddles::compute_twiddle(14, 29, direction); |
| 3959 |
let twiddle1re = f32x4_splat(tw1.re); |
| 3960 |
let twiddle1im = f32x4_splat(tw1.im); |
| 3961 |
let twiddle2re = f32x4_splat(tw2.re); |
| 3962 |
let twiddle2im = f32x4_splat(tw2.im); |
| 3963 |
let twiddle3re = f32x4_splat(tw3.re); |
| 3964 |
let twiddle3im = f32x4_splat(tw3.im); |
| 3965 |
let twiddle4re = f32x4_splat(tw4.re); |
| 3966 |
let twiddle4im = f32x4_splat(tw4.im); |
| 3967 |
let twiddle5re = f32x4_splat(tw5.re); |
| 3968 |
let twiddle5im = f32x4_splat(tw5.im); |
| 3969 |
let twiddle6re = f32x4_splat(tw6.re); |
| 3970 |
let twiddle6im = f32x4_splat(tw6.im); |
| 3971 |
let twiddle7re = f32x4_splat(tw7.re); |
| 3972 |
let twiddle7im = f32x4_splat(tw7.im); |
| 3973 |
let twiddle8re = f32x4_splat(tw8.re); |
| 3974 |
let twiddle8im = f32x4_splat(tw8.im); |
| 3975 |
let twiddle9re = f32x4_splat(tw9.re); |
| 3976 |
let twiddle9im = f32x4_splat(tw9.im); |
| 3977 |
let twiddle10re = f32x4_splat(tw10.re); |
| 3978 |
let twiddle10im = f32x4_splat(tw10.im); |
| 3979 |
let twiddle11re = f32x4_splat(tw11.re); |
| 3980 |
let twiddle11im = f32x4_splat(tw11.im); |
| 3981 |
let twiddle12re = f32x4_splat(tw12.re); |
| 3982 |
let twiddle12im = f32x4_splat(tw12.im); |
| 3983 |
let twiddle13re = f32x4_splat(tw13.re); |
| 3984 |
let twiddle13im = f32x4_splat(tw13.im); |
| 3985 |
let twiddle14re = f32x4_splat(tw14.re); |
| 3986 |
let twiddle14im = f32x4_splat(tw14.im); |
| 3987 |
|
| 3988 |
Self { |
| 3989 |
direction, |
| 3990 |
_phantom: std::marker::PhantomData, |
| 3991 |
rotate, |
| 3992 |
twiddle1re, |
| 3993 |
twiddle1im, |
| 3994 |
twiddle2re, |
| 3995 |
twiddle2im, |
| 3996 |
twiddle3re, |
| 3997 |
twiddle3im, |
| 3998 |
twiddle4re, |
| 3999 |
twiddle4im, |
| 4000 |
twiddle5re, |
| 4001 |
twiddle5im, |
| 4002 |
twiddle6re, |
| 4003 |
twiddle6im, |
| 4004 |
twiddle7re, |
| 4005 |
twiddle7im, |
| 4006 |
twiddle8re, |
| 4007 |
twiddle8im, |
| 4008 |
twiddle9re, |
| 4009 |
twiddle9im, |
| 4010 |
twiddle10re, |
| 4011 |
twiddle10im, |
| 4012 |
twiddle11re, |
| 4013 |
twiddle11im, |
| 4014 |
twiddle12re, |
| 4015 |
twiddle12im, |
| 4016 |
twiddle13re, |
| 4017 |
twiddle13im, |
| 4018 |
twiddle14re, |
| 4019 |
twiddle14im, |
| 4020 |
} |
| 4021 |
} |
| 4022 |
|
| 4023 |
#[inline(always)] |
| 4024 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 4025 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28}); |
| 4026 |
|
| 4027 |
let out = self.perform_parallel_fft_direct(values); |
| 4028 |
|
| 4029 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28}); |
| 4030 |
} |
| 4031 |
|
| 4032 |
#[inline(always)] |
| 4033 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 4034 |
&self, |
| 4035 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 4036 |
) { |
| 4037 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56}); |
| 4038 |
|
| 4039 |
let values = [ |
| 4040 |
extract_lo_hi_f32(input_packed[0], input_packed[14]), |
| 4041 |
extract_hi_lo_f32(input_packed[0], input_packed[15]), |
| 4042 |
extract_lo_hi_f32(input_packed[1], input_packed[15]), |
| 4043 |
extract_hi_lo_f32(input_packed[1], input_packed[16]), |
| 4044 |
extract_lo_hi_f32(input_packed[2], input_packed[16]), |
| 4045 |
extract_hi_lo_f32(input_packed[2], input_packed[17]), |
| 4046 |
extract_lo_hi_f32(input_packed[3], input_packed[17]), |
| 4047 |
extract_hi_lo_f32(input_packed[3], input_packed[18]), |
| 4048 |
extract_lo_hi_f32(input_packed[4], input_packed[18]), |
| 4049 |
extract_hi_lo_f32(input_packed[4], input_packed[19]), |
| 4050 |
extract_lo_hi_f32(input_packed[5], input_packed[19]), |
| 4051 |
extract_hi_lo_f32(input_packed[5], input_packed[20]), |
| 4052 |
extract_lo_hi_f32(input_packed[6], input_packed[20]), |
| 4053 |
extract_hi_lo_f32(input_packed[6], input_packed[21]), |
| 4054 |
extract_lo_hi_f32(input_packed[7], input_packed[21]), |
| 4055 |
extract_hi_lo_f32(input_packed[7], input_packed[22]), |
| 4056 |
extract_lo_hi_f32(input_packed[8], input_packed[22]), |
| 4057 |
extract_hi_lo_f32(input_packed[8], input_packed[23]), |
| 4058 |
extract_lo_hi_f32(input_packed[9], input_packed[23]), |
| 4059 |
extract_hi_lo_f32(input_packed[9], input_packed[24]), |
| 4060 |
extract_lo_hi_f32(input_packed[10], input_packed[24]), |
| 4061 |
extract_hi_lo_f32(input_packed[10], input_packed[25]), |
| 4062 |
extract_lo_hi_f32(input_packed[11], input_packed[25]), |
| 4063 |
extract_hi_lo_f32(input_packed[11], input_packed[26]), |
| 4064 |
extract_lo_hi_f32(input_packed[12], input_packed[26]), |
| 4065 |
extract_hi_lo_f32(input_packed[12], input_packed[27]), |
| 4066 |
extract_lo_hi_f32(input_packed[13], input_packed[27]), |
| 4067 |
extract_hi_lo_f32(input_packed[13], input_packed[28]), |
| 4068 |
extract_lo_hi_f32(input_packed[14], input_packed[28]), |
| 4069 |
]; |
| 4070 |
|
| 4071 |
let out = self.perform_parallel_fft_direct(values); |
| 4072 |
|
| 4073 |
let out_packed = [ |
| 4074 |
extract_lo_lo_f32(out[0], out[1]), |
| 4075 |
extract_lo_lo_f32(out[2], out[3]), |
| 4076 |
extract_lo_lo_f32(out[4], out[5]), |
| 4077 |
extract_lo_lo_f32(out[6], out[7]), |
| 4078 |
extract_lo_lo_f32(out[8], out[9]), |
| 4079 |
extract_lo_lo_f32(out[10], out[11]), |
| 4080 |
extract_lo_lo_f32(out[12], out[13]), |
| 4081 |
extract_lo_lo_f32(out[14], out[15]), |
| 4082 |
extract_lo_lo_f32(out[16], out[17]), |
| 4083 |
extract_lo_lo_f32(out[18], out[19]), |
| 4084 |
extract_lo_lo_f32(out[20], out[21]), |
| 4085 |
extract_lo_lo_f32(out[22], out[23]), |
| 4086 |
extract_lo_lo_f32(out[24], out[25]), |
| 4087 |
extract_lo_lo_f32(out[26], out[27]), |
| 4088 |
extract_lo_hi_f32(out[28], out[0]), |
| 4089 |
extract_hi_hi_f32(out[1], out[2]), |
| 4090 |
extract_hi_hi_f32(out[3], out[4]), |
| 4091 |
extract_hi_hi_f32(out[5], out[6]), |
| 4092 |
extract_hi_hi_f32(out[7], out[8]), |
| 4093 |
extract_hi_hi_f32(out[9], out[10]), |
| 4094 |
extract_hi_hi_f32(out[11], out[12]), |
| 4095 |
extract_hi_hi_f32(out[13], out[14]), |
| 4096 |
extract_hi_hi_f32(out[15], out[16]), |
| 4097 |
extract_hi_hi_f32(out[17], out[18]), |
| 4098 |
extract_hi_hi_f32(out[19], out[20]), |
| 4099 |
extract_hi_hi_f32(out[21], out[22]), |
| 4100 |
extract_hi_hi_f32(out[23], out[24]), |
| 4101 |
extract_hi_hi_f32(out[25], out[26]), |
| 4102 |
extract_hi_hi_f32(out[27], out[28]), |
| 4103 |
]; |
| 4104 |
|
| 4105 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28}); |
| 4106 |
} |
| 4107 |
|
| 4108 |
#[inline(always)] |
| 4109 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 29]) -> [v128; 29] { |
| 4110 |
let [x1p28, x1m28] = parallel_fft2_interleaved_f32(values[1], values[28]); |
| 4111 |
let [x2p27, x2m27] = parallel_fft2_interleaved_f32(values[2], values[27]); |
| 4112 |
let [x3p26, x3m26] = parallel_fft2_interleaved_f32(values[3], values[26]); |
| 4113 |
let [x4p25, x4m25] = parallel_fft2_interleaved_f32(values[4], values[25]); |
| 4114 |
let [x5p24, x5m24] = parallel_fft2_interleaved_f32(values[5], values[24]); |
| 4115 |
let [x6p23, x6m23] = parallel_fft2_interleaved_f32(values[6], values[23]); |
| 4116 |
let [x7p22, x7m22] = parallel_fft2_interleaved_f32(values[7], values[22]); |
| 4117 |
let [x8p21, x8m21] = parallel_fft2_interleaved_f32(values[8], values[21]); |
| 4118 |
let [x9p20, x9m20] = parallel_fft2_interleaved_f32(values[9], values[20]); |
| 4119 |
let [x10p19, x10m19] = parallel_fft2_interleaved_f32(values[10], values[19]); |
| 4120 |
let [x11p18, x11m18] = parallel_fft2_interleaved_f32(values[11], values[18]); |
| 4121 |
let [x12p17, x12m17] = parallel_fft2_interleaved_f32(values[12], values[17]); |
| 4122 |
let [x13p16, x13m16] = parallel_fft2_interleaved_f32(values[13], values[16]); |
| 4123 |
let [x14p15, x14m15] = parallel_fft2_interleaved_f32(values[14], values[15]); |
| 4124 |
|
| 4125 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p28); |
| 4126 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p27); |
| 4127 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p26); |
| 4128 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p25); |
| 4129 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p24); |
| 4130 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p23); |
| 4131 |
let t_a1_7 = f32x4_mul(self.twiddle7re, x7p22); |
| 4132 |
let t_a1_8 = f32x4_mul(self.twiddle8re, x8p21); |
| 4133 |
let t_a1_9 = f32x4_mul(self.twiddle9re, x9p20); |
| 4134 |
let t_a1_10 = f32x4_mul(self.twiddle10re, x10p19); |
| 4135 |
let t_a1_11 = f32x4_mul(self.twiddle11re, x11p18); |
| 4136 |
let t_a1_12 = f32x4_mul(self.twiddle12re, x12p17); |
| 4137 |
let t_a1_13 = f32x4_mul(self.twiddle13re, x13p16); |
| 4138 |
let t_a1_14 = f32x4_mul(self.twiddle14re, x14p15); |
| 4139 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p28); |
| 4140 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p27); |
| 4141 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p26); |
| 4142 |
let t_a2_4 = f32x4_mul(self.twiddle8re, x4p25); |
| 4143 |
let t_a2_5 = f32x4_mul(self.twiddle10re, x5p24); |
| 4144 |
let t_a2_6 = f32x4_mul(self.twiddle12re, x6p23); |
| 4145 |
let t_a2_7 = f32x4_mul(self.twiddle14re, x7p22); |
| 4146 |
let t_a2_8 = f32x4_mul(self.twiddle13re, x8p21); |
| 4147 |
let t_a2_9 = f32x4_mul(self.twiddle11re, x9p20); |
| 4148 |
let t_a2_10 = f32x4_mul(self.twiddle9re, x10p19); |
| 4149 |
let t_a2_11 = f32x4_mul(self.twiddle7re, x11p18); |
| 4150 |
let t_a2_12 = f32x4_mul(self.twiddle5re, x12p17); |
| 4151 |
let t_a2_13 = f32x4_mul(self.twiddle3re, x13p16); |
| 4152 |
let t_a2_14 = f32x4_mul(self.twiddle1re, x14p15); |
| 4153 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p28); |
| 4154 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p27); |
| 4155 |
let t_a3_3 = f32x4_mul(self.twiddle9re, x3p26); |
| 4156 |
let t_a3_4 = f32x4_mul(self.twiddle12re, x4p25); |
| 4157 |
let t_a3_5 = f32x4_mul(self.twiddle14re, x5p24); |
| 4158 |
let t_a3_6 = f32x4_mul(self.twiddle11re, x6p23); |
| 4159 |
let t_a3_7 = f32x4_mul(self.twiddle8re, x7p22); |
| 4160 |
let t_a3_8 = f32x4_mul(self.twiddle5re, x8p21); |
| 4161 |
let t_a3_9 = f32x4_mul(self.twiddle2re, x9p20); |
| 4162 |
let t_a3_10 = f32x4_mul(self.twiddle1re, x10p19); |
| 4163 |
let t_a3_11 = f32x4_mul(self.twiddle4re, x11p18); |
| 4164 |
let t_a3_12 = f32x4_mul(self.twiddle7re, x12p17); |
| 4165 |
let t_a3_13 = f32x4_mul(self.twiddle10re, x13p16); |
| 4166 |
let t_a3_14 = f32x4_mul(self.twiddle13re, x14p15); |
| 4167 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p28); |
| 4168 |
let t_a4_2 = f32x4_mul(self.twiddle8re, x2p27); |
| 4169 |
let t_a4_3 = f32x4_mul(self.twiddle12re, x3p26); |
| 4170 |
let t_a4_4 = f32x4_mul(self.twiddle13re, x4p25); |
| 4171 |
let t_a4_5 = f32x4_mul(self.twiddle9re, x5p24); |
| 4172 |
let t_a4_6 = f32x4_mul(self.twiddle5re, x6p23); |
| 4173 |
let t_a4_7 = f32x4_mul(self.twiddle1re, x7p22); |
| 4174 |
let t_a4_8 = f32x4_mul(self.twiddle3re, x8p21); |
| 4175 |
let t_a4_9 = f32x4_mul(self.twiddle7re, x9p20); |
| 4176 |
let t_a4_10 = f32x4_mul(self.twiddle11re, x10p19); |
| 4177 |
let t_a4_11 = f32x4_mul(self.twiddle14re, x11p18); |
| 4178 |
let t_a4_12 = f32x4_mul(self.twiddle10re, x12p17); |
| 4179 |
let t_a4_13 = f32x4_mul(self.twiddle6re, x13p16); |
| 4180 |
let t_a4_14 = f32x4_mul(self.twiddle2re, x14p15); |
| 4181 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p28); |
| 4182 |
let t_a5_2 = f32x4_mul(self.twiddle10re, x2p27); |
| 4183 |
let t_a5_3 = f32x4_mul(self.twiddle14re, x3p26); |
| 4184 |
let t_a5_4 = f32x4_mul(self.twiddle9re, x4p25); |
| 4185 |
let t_a5_5 = f32x4_mul(self.twiddle4re, x5p24); |
| 4186 |
let t_a5_6 = f32x4_mul(self.twiddle1re, x6p23); |
| 4187 |
let t_a5_7 = f32x4_mul(self.twiddle6re, x7p22); |
| 4188 |
let t_a5_8 = f32x4_mul(self.twiddle11re, x8p21); |
| 4189 |
let t_a5_9 = f32x4_mul(self.twiddle13re, x9p20); |
| 4190 |
let t_a5_10 = f32x4_mul(self.twiddle8re, x10p19); |
| 4191 |
let t_a5_11 = f32x4_mul(self.twiddle3re, x11p18); |
| 4192 |
let t_a5_12 = f32x4_mul(self.twiddle2re, x12p17); |
| 4193 |
let t_a5_13 = f32x4_mul(self.twiddle7re, x13p16); |
| 4194 |
let t_a5_14 = f32x4_mul(self.twiddle12re, x14p15); |
| 4195 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p28); |
| 4196 |
let t_a6_2 = f32x4_mul(self.twiddle12re, x2p27); |
| 4197 |
let t_a6_3 = f32x4_mul(self.twiddle11re, x3p26); |
| 4198 |
let t_a6_4 = f32x4_mul(self.twiddle5re, x4p25); |
| 4199 |
let t_a6_5 = f32x4_mul(self.twiddle1re, x5p24); |
| 4200 |
let t_a6_6 = f32x4_mul(self.twiddle7re, x6p23); |
| 4201 |
let t_a6_7 = f32x4_mul(self.twiddle13re, x7p22); |
| 4202 |
let t_a6_8 = f32x4_mul(self.twiddle10re, x8p21); |
| 4203 |
let t_a6_9 = f32x4_mul(self.twiddle4re, x9p20); |
| 4204 |
let t_a6_10 = f32x4_mul(self.twiddle2re, x10p19); |
| 4205 |
let t_a6_11 = f32x4_mul(self.twiddle8re, x11p18); |
| 4206 |
let t_a6_12 = f32x4_mul(self.twiddle14re, x12p17); |
| 4207 |
let t_a6_13 = f32x4_mul(self.twiddle9re, x13p16); |
| 4208 |
let t_a6_14 = f32x4_mul(self.twiddle3re, x14p15); |
| 4209 |
let t_a7_1 = f32x4_mul(self.twiddle7re, x1p28); |
| 4210 |
let t_a7_2 = f32x4_mul(self.twiddle14re, x2p27); |
| 4211 |
let t_a7_3 = f32x4_mul(self.twiddle8re, x3p26); |
| 4212 |
let t_a7_4 = f32x4_mul(self.twiddle1re, x4p25); |
| 4213 |
let t_a7_5 = f32x4_mul(self.twiddle6re, x5p24); |
| 4214 |
let t_a7_6 = f32x4_mul(self.twiddle13re, x6p23); |
| 4215 |
let t_a7_7 = f32x4_mul(self.twiddle9re, x7p22); |
| 4216 |
let t_a7_8 = f32x4_mul(self.twiddle2re, x8p21); |
| 4217 |
let t_a7_9 = f32x4_mul(self.twiddle5re, x9p20); |
| 4218 |
let t_a7_10 = f32x4_mul(self.twiddle12re, x10p19); |
| 4219 |
let t_a7_11 = f32x4_mul(self.twiddle10re, x11p18); |
| 4220 |
let t_a7_12 = f32x4_mul(self.twiddle3re, x12p17); |
| 4221 |
let t_a7_13 = f32x4_mul(self.twiddle4re, x13p16); |
| 4222 |
let t_a7_14 = f32x4_mul(self.twiddle11re, x14p15); |
| 4223 |
let t_a8_1 = f32x4_mul(self.twiddle8re, x1p28); |
| 4224 |
let t_a8_2 = f32x4_mul(self.twiddle13re, x2p27); |
| 4225 |
let t_a8_3 = f32x4_mul(self.twiddle5re, x3p26); |
| 4226 |
let t_a8_4 = f32x4_mul(self.twiddle3re, x4p25); |
| 4227 |
let t_a8_5 = f32x4_mul(self.twiddle11re, x5p24); |
| 4228 |
let t_a8_6 = f32x4_mul(self.twiddle10re, x6p23); |
| 4229 |
let t_a8_7 = f32x4_mul(self.twiddle2re, x7p22); |
| 4230 |
let t_a8_8 = f32x4_mul(self.twiddle6re, x8p21); |
| 4231 |
let t_a8_9 = f32x4_mul(self.twiddle14re, x9p20); |
| 4232 |
let t_a8_10 = f32x4_mul(self.twiddle7re, x10p19); |
| 4233 |
let t_a8_11 = f32x4_mul(self.twiddle1re, x11p18); |
| 4234 |
let t_a8_12 = f32x4_mul(self.twiddle9re, x12p17); |
| 4235 |
let t_a8_13 = f32x4_mul(self.twiddle12re, x13p16); |
| 4236 |
let t_a8_14 = f32x4_mul(self.twiddle4re, x14p15); |
| 4237 |
let t_a9_1 = f32x4_mul(self.twiddle9re, x1p28); |
| 4238 |
let t_a9_2 = f32x4_mul(self.twiddle11re, x2p27); |
| 4239 |
let t_a9_3 = f32x4_mul(self.twiddle2re, x3p26); |
| 4240 |
let t_a9_4 = f32x4_mul(self.twiddle7re, x4p25); |
| 4241 |
let t_a9_5 = f32x4_mul(self.twiddle13re, x5p24); |
| 4242 |
let t_a9_6 = f32x4_mul(self.twiddle4re, x6p23); |
| 4243 |
let t_a9_7 = f32x4_mul(self.twiddle5re, x7p22); |
| 4244 |
let t_a9_8 = f32x4_mul(self.twiddle14re, x8p21); |
| 4245 |
let t_a9_9 = f32x4_mul(self.twiddle6re, x9p20); |
| 4246 |
let t_a9_10 = f32x4_mul(self.twiddle3re, x10p19); |
| 4247 |
let t_a9_11 = f32x4_mul(self.twiddle12re, x11p18); |
| 4248 |
let t_a9_12 = f32x4_mul(self.twiddle8re, x12p17); |
| 4249 |
let t_a9_13 = f32x4_mul(self.twiddle1re, x13p16); |
| 4250 |
let t_a9_14 = f32x4_mul(self.twiddle10re, x14p15); |
| 4251 |
let t_a10_1 = f32x4_mul(self.twiddle10re, x1p28); |
| 4252 |
let t_a10_2 = f32x4_mul(self.twiddle9re, x2p27); |
| 4253 |
let t_a10_3 = f32x4_mul(self.twiddle1re, x3p26); |
| 4254 |
let t_a10_4 = f32x4_mul(self.twiddle11re, x4p25); |
| 4255 |
let t_a10_5 = f32x4_mul(self.twiddle8re, x5p24); |
| 4256 |
let t_a10_6 = f32x4_mul(self.twiddle2re, x6p23); |
| 4257 |
let t_a10_7 = f32x4_mul(self.twiddle12re, x7p22); |
| 4258 |
let t_a10_8 = f32x4_mul(self.twiddle7re, x8p21); |
| 4259 |
let t_a10_9 = f32x4_mul(self.twiddle3re, x9p20); |
| 4260 |
let t_a10_10 = f32x4_mul(self.twiddle13re, x10p19); |
| 4261 |
let t_a10_11 = f32x4_mul(self.twiddle6re, x11p18); |
| 4262 |
let t_a10_12 = f32x4_mul(self.twiddle4re, x12p17); |
| 4263 |
let t_a10_13 = f32x4_mul(self.twiddle14re, x13p16); |
| 4264 |
let t_a10_14 = f32x4_mul(self.twiddle5re, x14p15); |
| 4265 |
let t_a11_1 = f32x4_mul(self.twiddle11re, x1p28); |
| 4266 |
let t_a11_2 = f32x4_mul(self.twiddle7re, x2p27); |
| 4267 |
let t_a11_3 = f32x4_mul(self.twiddle4re, x3p26); |
| 4268 |
let t_a11_4 = f32x4_mul(self.twiddle14re, x4p25); |
| 4269 |
let t_a11_5 = f32x4_mul(self.twiddle3re, x5p24); |
| 4270 |
let t_a11_6 = f32x4_mul(self.twiddle8re, x6p23); |
| 4271 |
let t_a11_7 = f32x4_mul(self.twiddle10re, x7p22); |
| 4272 |
let t_a11_8 = f32x4_mul(self.twiddle1re, x8p21); |
| 4273 |
let t_a11_9 = f32x4_mul(self.twiddle12re, x9p20); |
| 4274 |
let t_a11_10 = f32x4_mul(self.twiddle6re, x10p19); |
| 4275 |
let t_a11_11 = f32x4_mul(self.twiddle5re, x11p18); |
| 4276 |
let t_a11_12 = f32x4_mul(self.twiddle13re, x12p17); |
| 4277 |
let t_a11_13 = f32x4_mul(self.twiddle2re, x13p16); |
| 4278 |
let t_a11_14 = f32x4_mul(self.twiddle9re, x14p15); |
| 4279 |
let t_a12_1 = f32x4_mul(self.twiddle12re, x1p28); |
| 4280 |
let t_a12_2 = f32x4_mul(self.twiddle5re, x2p27); |
| 4281 |
let t_a12_3 = f32x4_mul(self.twiddle7re, x3p26); |
| 4282 |
let t_a12_4 = f32x4_mul(self.twiddle10re, x4p25); |
| 4283 |
let t_a12_5 = f32x4_mul(self.twiddle2re, x5p24); |
| 4284 |
let t_a12_6 = f32x4_mul(self.twiddle14re, x6p23); |
| 4285 |
let t_a12_7 = f32x4_mul(self.twiddle3re, x7p22); |
| 4286 |
let t_a12_8 = f32x4_mul(self.twiddle9re, x8p21); |
| 4287 |
let t_a12_9 = f32x4_mul(self.twiddle8re, x9p20); |
| 4288 |
let t_a12_10 = f32x4_mul(self.twiddle4re, x10p19); |
| 4289 |
let t_a12_11 = f32x4_mul(self.twiddle13re, x11p18); |
| 4290 |
let t_a12_12 = f32x4_mul(self.twiddle1re, x12p17); |
| 4291 |
let t_a12_13 = f32x4_mul(self.twiddle11re, x13p16); |
| 4292 |
let t_a12_14 = f32x4_mul(self.twiddle6re, x14p15); |
| 4293 |
let t_a13_1 = f32x4_mul(self.twiddle13re, x1p28); |
| 4294 |
let t_a13_2 = f32x4_mul(self.twiddle3re, x2p27); |
| 4295 |
let t_a13_3 = f32x4_mul(self.twiddle10re, x3p26); |
| 4296 |
let t_a13_4 = f32x4_mul(self.twiddle6re, x4p25); |
| 4297 |
let t_a13_5 = f32x4_mul(self.twiddle7re, x5p24); |
| 4298 |
let t_a13_6 = f32x4_mul(self.twiddle9re, x6p23); |
| 4299 |
let t_a13_7 = f32x4_mul(self.twiddle4re, x7p22); |
| 4300 |
let t_a13_8 = f32x4_mul(self.twiddle12re, x8p21); |
| 4301 |
let t_a13_9 = f32x4_mul(self.twiddle1re, x9p20); |
| 4302 |
let t_a13_10 = f32x4_mul(self.twiddle14re, x10p19); |
| 4303 |
let t_a13_11 = f32x4_mul(self.twiddle2re, x11p18); |
| 4304 |
let t_a13_12 = f32x4_mul(self.twiddle11re, x12p17); |
| 4305 |
let t_a13_13 = f32x4_mul(self.twiddle5re, x13p16); |
| 4306 |
let t_a13_14 = f32x4_mul(self.twiddle8re, x14p15); |
| 4307 |
let t_a14_1 = f32x4_mul(self.twiddle14re, x1p28); |
| 4308 |
let t_a14_2 = f32x4_mul(self.twiddle1re, x2p27); |
| 4309 |
let t_a14_3 = f32x4_mul(self.twiddle13re, x3p26); |
| 4310 |
let t_a14_4 = f32x4_mul(self.twiddle2re, x4p25); |
| 4311 |
let t_a14_5 = f32x4_mul(self.twiddle12re, x5p24); |
| 4312 |
let t_a14_6 = f32x4_mul(self.twiddle3re, x6p23); |
| 4313 |
let t_a14_7 = f32x4_mul(self.twiddle11re, x7p22); |
| 4314 |
let t_a14_8 = f32x4_mul(self.twiddle4re, x8p21); |
| 4315 |
let t_a14_9 = f32x4_mul(self.twiddle10re, x9p20); |
| 4316 |
let t_a14_10 = f32x4_mul(self.twiddle5re, x10p19); |
| 4317 |
let t_a14_11 = f32x4_mul(self.twiddle9re, x11p18); |
| 4318 |
let t_a14_12 = f32x4_mul(self.twiddle6re, x12p17); |
| 4319 |
let t_a14_13 = f32x4_mul(self.twiddle8re, x13p16); |
| 4320 |
let t_a14_14 = f32x4_mul(self.twiddle7re, x14p15); |
| 4321 |
|
| 4322 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m28); |
| 4323 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m27); |
| 4324 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m26); |
| 4325 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m25); |
| 4326 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m24); |
| 4327 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m23); |
| 4328 |
let t_b1_7 = f32x4_mul(self.twiddle7im, x7m22); |
| 4329 |
let t_b1_8 = f32x4_mul(self.twiddle8im, x8m21); |
| 4330 |
let t_b1_9 = f32x4_mul(self.twiddle9im, x9m20); |
| 4331 |
let t_b1_10 = f32x4_mul(self.twiddle10im, x10m19); |
| 4332 |
let t_b1_11 = f32x4_mul(self.twiddle11im, x11m18); |
| 4333 |
let t_b1_12 = f32x4_mul(self.twiddle12im, x12m17); |
| 4334 |
let t_b1_13 = f32x4_mul(self.twiddle13im, x13m16); |
| 4335 |
let t_b1_14 = f32x4_mul(self.twiddle14im, x14m15); |
| 4336 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m28); |
| 4337 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m27); |
| 4338 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m26); |
| 4339 |
let t_b2_4 = f32x4_mul(self.twiddle8im, x4m25); |
| 4340 |
let t_b2_5 = f32x4_mul(self.twiddle10im, x5m24); |
| 4341 |
let t_b2_6 = f32x4_mul(self.twiddle12im, x6m23); |
| 4342 |
let t_b2_7 = f32x4_mul(self.twiddle14im, x7m22); |
| 4343 |
let t_b2_8 = f32x4_mul(self.twiddle13im, x8m21); |
| 4344 |
let t_b2_9 = f32x4_mul(self.twiddle11im, x9m20); |
| 4345 |
let t_b2_10 = f32x4_mul(self.twiddle9im, x10m19); |
| 4346 |
let t_b2_11 = f32x4_mul(self.twiddle7im, x11m18); |
| 4347 |
let t_b2_12 = f32x4_mul(self.twiddle5im, x12m17); |
| 4348 |
let t_b2_13 = f32x4_mul(self.twiddle3im, x13m16); |
| 4349 |
let t_b2_14 = f32x4_mul(self.twiddle1im, x14m15); |
| 4350 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m28); |
| 4351 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m27); |
| 4352 |
let t_b3_3 = f32x4_mul(self.twiddle9im, x3m26); |
| 4353 |
let t_b3_4 = f32x4_mul(self.twiddle12im, x4m25); |
| 4354 |
let t_b3_5 = f32x4_mul(self.twiddle14im, x5m24); |
| 4355 |
let t_b3_6 = f32x4_mul(self.twiddle11im, x6m23); |
| 4356 |
let t_b3_7 = f32x4_mul(self.twiddle8im, x7m22); |
| 4357 |
let t_b3_8 = f32x4_mul(self.twiddle5im, x8m21); |
| 4358 |
let t_b3_9 = f32x4_mul(self.twiddle2im, x9m20); |
| 4359 |
let t_b3_10 = f32x4_mul(self.twiddle1im, x10m19); |
| 4360 |
let t_b3_11 = f32x4_mul(self.twiddle4im, x11m18); |
| 4361 |
let t_b3_12 = f32x4_mul(self.twiddle7im, x12m17); |
| 4362 |
let t_b3_13 = f32x4_mul(self.twiddle10im, x13m16); |
| 4363 |
let t_b3_14 = f32x4_mul(self.twiddle13im, x14m15); |
| 4364 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m28); |
| 4365 |
let t_b4_2 = f32x4_mul(self.twiddle8im, x2m27); |
| 4366 |
let t_b4_3 = f32x4_mul(self.twiddle12im, x3m26); |
| 4367 |
let t_b4_4 = f32x4_mul(self.twiddle13im, x4m25); |
| 4368 |
let t_b4_5 = f32x4_mul(self.twiddle9im, x5m24); |
| 4369 |
let t_b4_6 = f32x4_mul(self.twiddle5im, x6m23); |
| 4370 |
let t_b4_7 = f32x4_mul(self.twiddle1im, x7m22); |
| 4371 |
let t_b4_8 = f32x4_mul(self.twiddle3im, x8m21); |
| 4372 |
let t_b4_9 = f32x4_mul(self.twiddle7im, x9m20); |
| 4373 |
let t_b4_10 = f32x4_mul(self.twiddle11im, x10m19); |
| 4374 |
let t_b4_11 = f32x4_mul(self.twiddle14im, x11m18); |
| 4375 |
let t_b4_12 = f32x4_mul(self.twiddle10im, x12m17); |
| 4376 |
let t_b4_13 = f32x4_mul(self.twiddle6im, x13m16); |
| 4377 |
let t_b4_14 = f32x4_mul(self.twiddle2im, x14m15); |
| 4378 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m28); |
| 4379 |
let t_b5_2 = f32x4_mul(self.twiddle10im, x2m27); |
| 4380 |
let t_b5_3 = f32x4_mul(self.twiddle14im, x3m26); |
| 4381 |
let t_b5_4 = f32x4_mul(self.twiddle9im, x4m25); |
| 4382 |
let t_b5_5 = f32x4_mul(self.twiddle4im, x5m24); |
| 4383 |
let t_b5_6 = f32x4_mul(self.twiddle1im, x6m23); |
| 4384 |
let t_b5_7 = f32x4_mul(self.twiddle6im, x7m22); |
| 4385 |
let t_b5_8 = f32x4_mul(self.twiddle11im, x8m21); |
| 4386 |
let t_b5_9 = f32x4_mul(self.twiddle13im, x9m20); |
| 4387 |
let t_b5_10 = f32x4_mul(self.twiddle8im, x10m19); |
| 4388 |
let t_b5_11 = f32x4_mul(self.twiddle3im, x11m18); |
| 4389 |
let t_b5_12 = f32x4_mul(self.twiddle2im, x12m17); |
| 4390 |
let t_b5_13 = f32x4_mul(self.twiddle7im, x13m16); |
| 4391 |
let t_b5_14 = f32x4_mul(self.twiddle12im, x14m15); |
| 4392 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m28); |
| 4393 |
let t_b6_2 = f32x4_mul(self.twiddle12im, x2m27); |
| 4394 |
let t_b6_3 = f32x4_mul(self.twiddle11im, x3m26); |
| 4395 |
let t_b6_4 = f32x4_mul(self.twiddle5im, x4m25); |
| 4396 |
let t_b6_5 = f32x4_mul(self.twiddle1im, x5m24); |
| 4397 |
let t_b6_6 = f32x4_mul(self.twiddle7im, x6m23); |
| 4398 |
let t_b6_7 = f32x4_mul(self.twiddle13im, x7m22); |
| 4399 |
let t_b6_8 = f32x4_mul(self.twiddle10im, x8m21); |
| 4400 |
let t_b6_9 = f32x4_mul(self.twiddle4im, x9m20); |
| 4401 |
let t_b6_10 = f32x4_mul(self.twiddle2im, x10m19); |
| 4402 |
let t_b6_11 = f32x4_mul(self.twiddle8im, x11m18); |
| 4403 |
let t_b6_12 = f32x4_mul(self.twiddle14im, x12m17); |
| 4404 |
let t_b6_13 = f32x4_mul(self.twiddle9im, x13m16); |
| 4405 |
let t_b6_14 = f32x4_mul(self.twiddle3im, x14m15); |
| 4406 |
let t_b7_1 = f32x4_mul(self.twiddle7im, x1m28); |
| 4407 |
let t_b7_2 = f32x4_mul(self.twiddle14im, x2m27); |
| 4408 |
let t_b7_3 = f32x4_mul(self.twiddle8im, x3m26); |
| 4409 |
let t_b7_4 = f32x4_mul(self.twiddle1im, x4m25); |
| 4410 |
let t_b7_5 = f32x4_mul(self.twiddle6im, x5m24); |
| 4411 |
let t_b7_6 = f32x4_mul(self.twiddle13im, x6m23); |
| 4412 |
let t_b7_7 = f32x4_mul(self.twiddle9im, x7m22); |
| 4413 |
let t_b7_8 = f32x4_mul(self.twiddle2im, x8m21); |
| 4414 |
let t_b7_9 = f32x4_mul(self.twiddle5im, x9m20); |
| 4415 |
let t_b7_10 = f32x4_mul(self.twiddle12im, x10m19); |
| 4416 |
let t_b7_11 = f32x4_mul(self.twiddle10im, x11m18); |
| 4417 |
let t_b7_12 = f32x4_mul(self.twiddle3im, x12m17); |
| 4418 |
let t_b7_13 = f32x4_mul(self.twiddle4im, x13m16); |
| 4419 |
let t_b7_14 = f32x4_mul(self.twiddle11im, x14m15); |
| 4420 |
let t_b8_1 = f32x4_mul(self.twiddle8im, x1m28); |
| 4421 |
let t_b8_2 = f32x4_mul(self.twiddle13im, x2m27); |
| 4422 |
let t_b8_3 = f32x4_mul(self.twiddle5im, x3m26); |
| 4423 |
let t_b8_4 = f32x4_mul(self.twiddle3im, x4m25); |
| 4424 |
let t_b8_5 = f32x4_mul(self.twiddle11im, x5m24); |
| 4425 |
let t_b8_6 = f32x4_mul(self.twiddle10im, x6m23); |
| 4426 |
let t_b8_7 = f32x4_mul(self.twiddle2im, x7m22); |
| 4427 |
let t_b8_8 = f32x4_mul(self.twiddle6im, x8m21); |
| 4428 |
let t_b8_9 = f32x4_mul(self.twiddle14im, x9m20); |
| 4429 |
let t_b8_10 = f32x4_mul(self.twiddle7im, x10m19); |
| 4430 |
let t_b8_11 = f32x4_mul(self.twiddle1im, x11m18); |
| 4431 |
let t_b8_12 = f32x4_mul(self.twiddle9im, x12m17); |
| 4432 |
let t_b8_13 = f32x4_mul(self.twiddle12im, x13m16); |
| 4433 |
let t_b8_14 = f32x4_mul(self.twiddle4im, x14m15); |
| 4434 |
let t_b9_1 = f32x4_mul(self.twiddle9im, x1m28); |
| 4435 |
let t_b9_2 = f32x4_mul(self.twiddle11im, x2m27); |
| 4436 |
let t_b9_3 = f32x4_mul(self.twiddle2im, x3m26); |
| 4437 |
let t_b9_4 = f32x4_mul(self.twiddle7im, x4m25); |
| 4438 |
let t_b9_5 = f32x4_mul(self.twiddle13im, x5m24); |
| 4439 |
let t_b9_6 = f32x4_mul(self.twiddle4im, x6m23); |
| 4440 |
let t_b9_7 = f32x4_mul(self.twiddle5im, x7m22); |
| 4441 |
let t_b9_8 = f32x4_mul(self.twiddle14im, x8m21); |
| 4442 |
let t_b9_9 = f32x4_mul(self.twiddle6im, x9m20); |
| 4443 |
let t_b9_10 = f32x4_mul(self.twiddle3im, x10m19); |
| 4444 |
let t_b9_11 = f32x4_mul(self.twiddle12im, x11m18); |
| 4445 |
let t_b9_12 = f32x4_mul(self.twiddle8im, x12m17); |
| 4446 |
let t_b9_13 = f32x4_mul(self.twiddle1im, x13m16); |
| 4447 |
let t_b9_14 = f32x4_mul(self.twiddle10im, x14m15); |
| 4448 |
let t_b10_1 = f32x4_mul(self.twiddle10im, x1m28); |
| 4449 |
let t_b10_2 = f32x4_mul(self.twiddle9im, x2m27); |
| 4450 |
let t_b10_3 = f32x4_mul(self.twiddle1im, x3m26); |
| 4451 |
let t_b10_4 = f32x4_mul(self.twiddle11im, x4m25); |
| 4452 |
let t_b10_5 = f32x4_mul(self.twiddle8im, x5m24); |
| 4453 |
let t_b10_6 = f32x4_mul(self.twiddle2im, x6m23); |
| 4454 |
let t_b10_7 = f32x4_mul(self.twiddle12im, x7m22); |
| 4455 |
let t_b10_8 = f32x4_mul(self.twiddle7im, x8m21); |
| 4456 |
let t_b10_9 = f32x4_mul(self.twiddle3im, x9m20); |
| 4457 |
let t_b10_10 = f32x4_mul(self.twiddle13im, x10m19); |
| 4458 |
let t_b10_11 = f32x4_mul(self.twiddle6im, x11m18); |
| 4459 |
let t_b10_12 = f32x4_mul(self.twiddle4im, x12m17); |
| 4460 |
let t_b10_13 = f32x4_mul(self.twiddle14im, x13m16); |
| 4461 |
let t_b10_14 = f32x4_mul(self.twiddle5im, x14m15); |
| 4462 |
let t_b11_1 = f32x4_mul(self.twiddle11im, x1m28); |
| 4463 |
let t_b11_2 = f32x4_mul(self.twiddle7im, x2m27); |
| 4464 |
let t_b11_3 = f32x4_mul(self.twiddle4im, x3m26); |
| 4465 |
let t_b11_4 = f32x4_mul(self.twiddle14im, x4m25); |
| 4466 |
let t_b11_5 = f32x4_mul(self.twiddle3im, x5m24); |
| 4467 |
let t_b11_6 = f32x4_mul(self.twiddle8im, x6m23); |
| 4468 |
let t_b11_7 = f32x4_mul(self.twiddle10im, x7m22); |
| 4469 |
let t_b11_8 = f32x4_mul(self.twiddle1im, x8m21); |
| 4470 |
let t_b11_9 = f32x4_mul(self.twiddle12im, x9m20); |
| 4471 |
let t_b11_10 = f32x4_mul(self.twiddle6im, x10m19); |
| 4472 |
let t_b11_11 = f32x4_mul(self.twiddle5im, x11m18); |
| 4473 |
let t_b11_12 = f32x4_mul(self.twiddle13im, x12m17); |
| 4474 |
let t_b11_13 = f32x4_mul(self.twiddle2im, x13m16); |
| 4475 |
let t_b11_14 = f32x4_mul(self.twiddle9im, x14m15); |
| 4476 |
let t_b12_1 = f32x4_mul(self.twiddle12im, x1m28); |
| 4477 |
let t_b12_2 = f32x4_mul(self.twiddle5im, x2m27); |
| 4478 |
let t_b12_3 = f32x4_mul(self.twiddle7im, x3m26); |
| 4479 |
let t_b12_4 = f32x4_mul(self.twiddle10im, x4m25); |
| 4480 |
let t_b12_5 = f32x4_mul(self.twiddle2im, x5m24); |
| 4481 |
let t_b12_6 = f32x4_mul(self.twiddle14im, x6m23); |
| 4482 |
let t_b12_7 = f32x4_mul(self.twiddle3im, x7m22); |
| 4483 |
let t_b12_8 = f32x4_mul(self.twiddle9im, x8m21); |
| 4484 |
let t_b12_9 = f32x4_mul(self.twiddle8im, x9m20); |
| 4485 |
let t_b12_10 = f32x4_mul(self.twiddle4im, x10m19); |
| 4486 |
let t_b12_11 = f32x4_mul(self.twiddle13im, x11m18); |
| 4487 |
let t_b12_12 = f32x4_mul(self.twiddle1im, x12m17); |
| 4488 |
let t_b12_13 = f32x4_mul(self.twiddle11im, x13m16); |
| 4489 |
let t_b12_14 = f32x4_mul(self.twiddle6im, x14m15); |
| 4490 |
let t_b13_1 = f32x4_mul(self.twiddle13im, x1m28); |
| 4491 |
let t_b13_2 = f32x4_mul(self.twiddle3im, x2m27); |
| 4492 |
let t_b13_3 = f32x4_mul(self.twiddle10im, x3m26); |
| 4493 |
let t_b13_4 = f32x4_mul(self.twiddle6im, x4m25); |
| 4494 |
let t_b13_5 = f32x4_mul(self.twiddle7im, x5m24); |
| 4495 |
let t_b13_6 = f32x4_mul(self.twiddle9im, x6m23); |
| 4496 |
let t_b13_7 = f32x4_mul(self.twiddle4im, x7m22); |
| 4497 |
let t_b13_8 = f32x4_mul(self.twiddle12im, x8m21); |
| 4498 |
let t_b13_9 = f32x4_mul(self.twiddle1im, x9m20); |
| 4499 |
let t_b13_10 = f32x4_mul(self.twiddle14im, x10m19); |
| 4500 |
let t_b13_11 = f32x4_mul(self.twiddle2im, x11m18); |
| 4501 |
let t_b13_12 = f32x4_mul(self.twiddle11im, x12m17); |
| 4502 |
let t_b13_13 = f32x4_mul(self.twiddle5im, x13m16); |
| 4503 |
let t_b13_14 = f32x4_mul(self.twiddle8im, x14m15); |
| 4504 |
let t_b14_1 = f32x4_mul(self.twiddle14im, x1m28); |
| 4505 |
let t_b14_2 = f32x4_mul(self.twiddle1im, x2m27); |
| 4506 |
let t_b14_3 = f32x4_mul(self.twiddle13im, x3m26); |
| 4507 |
let t_b14_4 = f32x4_mul(self.twiddle2im, x4m25); |
| 4508 |
let t_b14_5 = f32x4_mul(self.twiddle12im, x5m24); |
| 4509 |
let t_b14_6 = f32x4_mul(self.twiddle3im, x6m23); |
| 4510 |
let t_b14_7 = f32x4_mul(self.twiddle11im, x7m22); |
| 4511 |
let t_b14_8 = f32x4_mul(self.twiddle4im, x8m21); |
| 4512 |
let t_b14_9 = f32x4_mul(self.twiddle10im, x9m20); |
| 4513 |
let t_b14_10 = f32x4_mul(self.twiddle5im, x10m19); |
| 4514 |
let t_b14_11 = f32x4_mul(self.twiddle9im, x11m18); |
| 4515 |
let t_b14_12 = f32x4_mul(self.twiddle6im, x12m17); |
| 4516 |
let t_b14_13 = f32x4_mul(self.twiddle8im, x13m16); |
| 4517 |
let t_b14_14 = f32x4_mul(self.twiddle7im, x14m15); |
| 4518 |
|
| 4519 |
let x0 = values[0]; |
| 4520 |
let t_a1 = calc_f32!( |
| 4521 |
x0 + t_a1_1 |
| 4522 |
+ t_a1_2 |
| 4523 |
+ t_a1_3 |
| 4524 |
+ t_a1_4 |
| 4525 |
+ t_a1_5 |
| 4526 |
+ t_a1_6 |
| 4527 |
+ t_a1_7 |
| 4528 |
+ t_a1_8 |
| 4529 |
+ t_a1_9 |
| 4530 |
+ t_a1_10 |
| 4531 |
+ t_a1_11 |
| 4532 |
+ t_a1_12 |
| 4533 |
+ t_a1_13 |
| 4534 |
+ t_a1_14 |
| 4535 |
); |
| 4536 |
let t_a2 = calc_f32!( |
| 4537 |
x0 + t_a2_1 |
| 4538 |
+ t_a2_2 |
| 4539 |
+ t_a2_3 |
| 4540 |
+ t_a2_4 |
| 4541 |
+ t_a2_5 |
| 4542 |
+ t_a2_6 |
| 4543 |
+ t_a2_7 |
| 4544 |
+ t_a2_8 |
| 4545 |
+ t_a2_9 |
| 4546 |
+ t_a2_10 |
| 4547 |
+ t_a2_11 |
| 4548 |
+ t_a2_12 |
| 4549 |
+ t_a2_13 |
| 4550 |
+ t_a2_14 |
| 4551 |
); |
| 4552 |
let t_a3 = calc_f32!( |
| 4553 |
x0 + t_a3_1 |
| 4554 |
+ t_a3_2 |
| 4555 |
+ t_a3_3 |
| 4556 |
+ t_a3_4 |
| 4557 |
+ t_a3_5 |
| 4558 |
+ t_a3_6 |
| 4559 |
+ t_a3_7 |
| 4560 |
+ t_a3_8 |
| 4561 |
+ t_a3_9 |
| 4562 |
+ t_a3_10 |
| 4563 |
+ t_a3_11 |
| 4564 |
+ t_a3_12 |
| 4565 |
+ t_a3_13 |
| 4566 |
+ t_a3_14 |
| 4567 |
); |
| 4568 |
let t_a4 = calc_f32!( |
| 4569 |
x0 + t_a4_1 |
| 4570 |
+ t_a4_2 |
| 4571 |
+ t_a4_3 |
| 4572 |
+ t_a4_4 |
| 4573 |
+ t_a4_5 |
| 4574 |
+ t_a4_6 |
| 4575 |
+ t_a4_7 |
| 4576 |
+ t_a4_8 |
| 4577 |
+ t_a4_9 |
| 4578 |
+ t_a4_10 |
| 4579 |
+ t_a4_11 |
| 4580 |
+ t_a4_12 |
| 4581 |
+ t_a4_13 |
| 4582 |
+ t_a4_14 |
| 4583 |
); |
| 4584 |
let t_a5 = calc_f32!( |
| 4585 |
x0 + t_a5_1 |
| 4586 |
+ t_a5_2 |
| 4587 |
+ t_a5_3 |
| 4588 |
+ t_a5_4 |
| 4589 |
+ t_a5_5 |
| 4590 |
+ t_a5_6 |
| 4591 |
+ t_a5_7 |
| 4592 |
+ t_a5_8 |
| 4593 |
+ t_a5_9 |
| 4594 |
+ t_a5_10 |
| 4595 |
+ t_a5_11 |
| 4596 |
+ t_a5_12 |
| 4597 |
+ t_a5_13 |
| 4598 |
+ t_a5_14 |
| 4599 |
); |
| 4600 |
let t_a6 = calc_f32!( |
| 4601 |
x0 + t_a6_1 |
| 4602 |
+ t_a6_2 |
| 4603 |
+ t_a6_3 |
| 4604 |
+ t_a6_4 |
| 4605 |
+ t_a6_5 |
| 4606 |
+ t_a6_6 |
| 4607 |
+ t_a6_7 |
| 4608 |
+ t_a6_8 |
| 4609 |
+ t_a6_9 |
| 4610 |
+ t_a6_10 |
| 4611 |
+ t_a6_11 |
| 4612 |
+ t_a6_12 |
| 4613 |
+ t_a6_13 |
| 4614 |
+ t_a6_14 |
| 4615 |
); |
| 4616 |
let t_a7 = calc_f32!( |
| 4617 |
x0 + t_a7_1 |
| 4618 |
+ t_a7_2 |
| 4619 |
+ t_a7_3 |
| 4620 |
+ t_a7_4 |
| 4621 |
+ t_a7_5 |
| 4622 |
+ t_a7_6 |
| 4623 |
+ t_a7_7 |
| 4624 |
+ t_a7_8 |
| 4625 |
+ t_a7_9 |
| 4626 |
+ t_a7_10 |
| 4627 |
+ t_a7_11 |
| 4628 |
+ t_a7_12 |
| 4629 |
+ t_a7_13 |
| 4630 |
+ t_a7_14 |
| 4631 |
); |
| 4632 |
let t_a8 = calc_f32!( |
| 4633 |
x0 + t_a8_1 |
| 4634 |
+ t_a8_2 |
| 4635 |
+ t_a8_3 |
| 4636 |
+ t_a8_4 |
| 4637 |
+ t_a8_5 |
| 4638 |
+ t_a8_6 |
| 4639 |
+ t_a8_7 |
| 4640 |
+ t_a8_8 |
| 4641 |
+ t_a8_9 |
| 4642 |
+ t_a8_10 |
| 4643 |
+ t_a8_11 |
| 4644 |
+ t_a8_12 |
| 4645 |
+ t_a8_13 |
| 4646 |
+ t_a8_14 |
| 4647 |
); |
| 4648 |
let t_a9 = calc_f32!( |
| 4649 |
x0 + t_a9_1 |
| 4650 |
+ t_a9_2 |
| 4651 |
+ t_a9_3 |
| 4652 |
+ t_a9_4 |
| 4653 |
+ t_a9_5 |
| 4654 |
+ t_a9_6 |
| 4655 |
+ t_a9_7 |
| 4656 |
+ t_a9_8 |
| 4657 |
+ t_a9_9 |
| 4658 |
+ t_a9_10 |
| 4659 |
+ t_a9_11 |
| 4660 |
+ t_a9_12 |
| 4661 |
+ t_a9_13 |
| 4662 |
+ t_a9_14 |
| 4663 |
); |
| 4664 |
let t_a10 = calc_f32!( |
| 4665 |
x0 + t_a10_1 |
| 4666 |
+ t_a10_2 |
| 4667 |
+ t_a10_3 |
| 4668 |
+ t_a10_4 |
| 4669 |
+ t_a10_5 |
| 4670 |
+ t_a10_6 |
| 4671 |
+ t_a10_7 |
| 4672 |
+ t_a10_8 |
| 4673 |
+ t_a10_9 |
| 4674 |
+ t_a10_10 |
| 4675 |
+ t_a10_11 |
| 4676 |
+ t_a10_12 |
| 4677 |
+ t_a10_13 |
| 4678 |
+ t_a10_14 |
| 4679 |
); |
| 4680 |
let t_a11 = calc_f32!( |
| 4681 |
x0 + t_a11_1 |
| 4682 |
+ t_a11_2 |
| 4683 |
+ t_a11_3 |
| 4684 |
+ t_a11_4 |
| 4685 |
+ t_a11_5 |
| 4686 |
+ t_a11_6 |
| 4687 |
+ t_a11_7 |
| 4688 |
+ t_a11_8 |
| 4689 |
+ t_a11_9 |
| 4690 |
+ t_a11_10 |
| 4691 |
+ t_a11_11 |
| 4692 |
+ t_a11_12 |
| 4693 |
+ t_a11_13 |
| 4694 |
+ t_a11_14 |
| 4695 |
); |
| 4696 |
let t_a12 = calc_f32!( |
| 4697 |
x0 + t_a12_1 |
| 4698 |
+ t_a12_2 |
| 4699 |
+ t_a12_3 |
| 4700 |
+ t_a12_4 |
| 4701 |
+ t_a12_5 |
| 4702 |
+ t_a12_6 |
| 4703 |
+ t_a12_7 |
| 4704 |
+ t_a12_8 |
| 4705 |
+ t_a12_9 |
| 4706 |
+ t_a12_10 |
| 4707 |
+ t_a12_11 |
| 4708 |
+ t_a12_12 |
| 4709 |
+ t_a12_13 |
| 4710 |
+ t_a12_14 |
| 4711 |
); |
| 4712 |
let t_a13 = calc_f32!( |
| 4713 |
x0 + t_a13_1 |
| 4714 |
+ t_a13_2 |
| 4715 |
+ t_a13_3 |
| 4716 |
+ t_a13_4 |
| 4717 |
+ t_a13_5 |
| 4718 |
+ t_a13_6 |
| 4719 |
+ t_a13_7 |
| 4720 |
+ t_a13_8 |
| 4721 |
+ t_a13_9 |
| 4722 |
+ t_a13_10 |
| 4723 |
+ t_a13_11 |
| 4724 |
+ t_a13_12 |
| 4725 |
+ t_a13_13 |
| 4726 |
+ t_a13_14 |
| 4727 |
); |
| 4728 |
let t_a14 = calc_f32!( |
| 4729 |
x0 + t_a14_1 |
| 4730 |
+ t_a14_2 |
| 4731 |
+ t_a14_3 |
| 4732 |
+ t_a14_4 |
| 4733 |
+ t_a14_5 |
| 4734 |
+ t_a14_6 |
| 4735 |
+ t_a14_7 |
| 4736 |
+ t_a14_8 |
| 4737 |
+ t_a14_9 |
| 4738 |
+ t_a14_10 |
| 4739 |
+ t_a14_11 |
| 4740 |
+ t_a14_12 |
| 4741 |
+ t_a14_13 |
| 4742 |
+ t_a14_14 |
| 4743 |
); |
| 4744 |
|
| 4745 |
let t_b1 = calc_f32!( |
| 4746 |
t_b1_1 |
| 4747 |
+ t_b1_2 |
| 4748 |
+ t_b1_3 |
| 4749 |
+ t_b1_4 |
| 4750 |
+ t_b1_5 |
| 4751 |
+ t_b1_6 |
| 4752 |
+ t_b1_7 |
| 4753 |
+ t_b1_8 |
| 4754 |
+ t_b1_9 |
| 4755 |
+ t_b1_10 |
| 4756 |
+ t_b1_11 |
| 4757 |
+ t_b1_12 |
| 4758 |
+ t_b1_13 |
| 4759 |
+ t_b1_14 |
| 4760 |
); |
| 4761 |
let t_b2 = calc_f32!( |
| 4762 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 + t_b2_6 + t_b2_7 |
| 4763 |
- t_b2_8 |
| 4764 |
- t_b2_9 |
| 4765 |
- t_b2_10 |
| 4766 |
- t_b2_11 |
| 4767 |
- t_b2_12 |
| 4768 |
- t_b2_13 |
| 4769 |
- t_b2_14 |
| 4770 |
); |
| 4771 |
let t_b3 = calc_f32!( |
| 4772 |
t_b3_1 + t_b3_2 + t_b3_3 + t_b3_4 - t_b3_5 - t_b3_6 - t_b3_7 - t_b3_8 - t_b3_9 |
| 4773 |
+ t_b3_10 |
| 4774 |
+ t_b3_11 |
| 4775 |
+ t_b3_12 |
| 4776 |
+ t_b3_13 |
| 4777 |
+ t_b3_14 |
| 4778 |
); |
| 4779 |
let t_b4 = calc_f32!( |
| 4780 |
t_b4_1 + t_b4_2 + t_b4_3 - t_b4_4 - t_b4_5 - t_b4_6 - t_b4_7 |
| 4781 |
+ t_b4_8 |
| 4782 |
+ t_b4_9 |
| 4783 |
+ t_b4_10 |
| 4784 |
- t_b4_11 |
| 4785 |
- t_b4_12 |
| 4786 |
- t_b4_13 |
| 4787 |
- t_b4_14 |
| 4788 |
); |
| 4789 |
let t_b5 = calc_f32!( |
| 4790 |
t_b5_1 + t_b5_2 - t_b5_3 - t_b5_4 - t_b5_5 + t_b5_6 + t_b5_7 + t_b5_8 |
| 4791 |
- t_b5_9 |
| 4792 |
- t_b5_10 |
| 4793 |
- t_b5_11 |
| 4794 |
+ t_b5_12 |
| 4795 |
+ t_b5_13 |
| 4796 |
+ t_b5_14 |
| 4797 |
); |
| 4798 |
let t_b6 = calc_f32!( |
| 4799 |
t_b6_1 + t_b6_2 - t_b6_3 - t_b6_4 + t_b6_5 + t_b6_6 + t_b6_7 - t_b6_8 - t_b6_9 |
| 4800 |
+ t_b6_10 |
| 4801 |
+ t_b6_11 |
| 4802 |
+ t_b6_12 |
| 4803 |
- t_b6_13 |
| 4804 |
- t_b6_14 |
| 4805 |
); |
| 4806 |
let t_b7 = calc_f32!( |
| 4807 |
t_b7_1 + t_b7_2 - t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 |
| 4808 |
+ t_b7_9 |
| 4809 |
+ t_b7_10 |
| 4810 |
- t_b7_11 |
| 4811 |
- t_b7_12 |
| 4812 |
+ t_b7_13 |
| 4813 |
+ t_b7_14 |
| 4814 |
); |
| 4815 |
let t_b8 = calc_f32!( |
| 4816 |
t_b8_1 - t_b8_2 - t_b8_3 + t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 + t_b8_9 |
| 4817 |
- t_b8_10 |
| 4818 |
+ t_b8_11 |
| 4819 |
+ t_b8_12 |
| 4820 |
- t_b8_13 |
| 4821 |
- t_b8_14 |
| 4822 |
); |
| 4823 |
let t_b9 = calc_f32!( |
| 4824 |
t_b9_1 - t_b9_2 - t_b9_3 + t_b9_4 - t_b9_5 - t_b9_6 + t_b9_7 + t_b9_8 - t_b9_9 |
| 4825 |
+ t_b9_10 |
| 4826 |
+ t_b9_11 |
| 4827 |
- t_b9_12 |
| 4828 |
+ t_b9_13 |
| 4829 |
+ t_b9_14 |
| 4830 |
); |
| 4831 |
let t_b10 = calc_f32!( |
| 4832 |
t_b10_1 - t_b10_2 + t_b10_3 + t_b10_4 - t_b10_5 + t_b10_6 + t_b10_7 - t_b10_8 |
| 4833 |
+ t_b10_9 |
| 4834 |
+ t_b10_10 |
| 4835 |
- t_b10_11 |
| 4836 |
+ t_b10_12 |
| 4837 |
+ t_b10_13 |
| 4838 |
- t_b10_14 |
| 4839 |
); |
| 4840 |
let t_b11 = calc_f32!( |
| 4841 |
t_b11_1 - t_b11_2 + t_b11_3 - t_b11_4 - t_b11_5 + t_b11_6 - t_b11_7 + t_b11_8 + t_b11_9 |
| 4842 |
- t_b11_10 |
| 4843 |
+ t_b11_11 |
| 4844 |
- t_b11_12 |
| 4845 |
- t_b11_13 |
| 4846 |
+ t_b11_14 |
| 4847 |
); |
| 4848 |
let t_b12 = calc_f32!( |
| 4849 |
t_b12_1 - t_b12_2 + t_b12_3 - t_b12_4 + t_b12_5 + t_b12_6 - t_b12_7 + t_b12_8 - t_b12_9 |
| 4850 |
+ t_b12_10 |
| 4851 |
- t_b12_11 |
| 4852 |
- t_b12_12 |
| 4853 |
+ t_b12_13 |
| 4854 |
- t_b12_14 |
| 4855 |
); |
| 4856 |
let t_b13 = calc_f32!( |
| 4857 |
t_b13_1 - t_b13_2 + t_b13_3 - t_b13_4 + t_b13_5 - t_b13_6 + t_b13_7 - t_b13_8 |
| 4858 |
+ t_b13_9 |
| 4859 |
+ t_b13_10 |
| 4860 |
- t_b13_11 |
| 4861 |
+ t_b13_12 |
| 4862 |
- t_b13_13 |
| 4863 |
+ t_b13_14 |
| 4864 |
); |
| 4865 |
let t_b14 = calc_f32!( |
| 4866 |
t_b14_1 - t_b14_2 + t_b14_3 - t_b14_4 + t_b14_5 - t_b14_6 + t_b14_7 - t_b14_8 + t_b14_9 |
| 4867 |
- t_b14_10 |
| 4868 |
+ t_b14_11 |
| 4869 |
- t_b14_12 |
| 4870 |
+ t_b14_13 |
| 4871 |
- t_b14_14 |
| 4872 |
); |
| 4873 |
|
| 4874 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 4875 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 4876 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 4877 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 4878 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 4879 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 4880 |
let t_b7_rot = self.rotate.rotate_both(t_b7); |
| 4881 |
let t_b8_rot = self.rotate.rotate_both(t_b8); |
| 4882 |
let t_b9_rot = self.rotate.rotate_both(t_b9); |
| 4883 |
let t_b10_rot = self.rotate.rotate_both(t_b10); |
| 4884 |
let t_b11_rot = self.rotate.rotate_both(t_b11); |
| 4885 |
let t_b12_rot = self.rotate.rotate_both(t_b12); |
| 4886 |
let t_b13_rot = self.rotate.rotate_both(t_b13); |
| 4887 |
let t_b14_rot = self.rotate.rotate_both(t_b14); |
| 4888 |
|
| 4889 |
let y0 = calc_f32!( |
| 4890 |
x0 + x1p28 |
| 4891 |
+ x2p27 |
| 4892 |
+ x3p26 |
| 4893 |
+ x4p25 |
| 4894 |
+ x5p24 |
| 4895 |
+ x6p23 |
| 4896 |
+ x7p22 |
| 4897 |
+ x8p21 |
| 4898 |
+ x9p20 |
| 4899 |
+ x10p19 |
| 4900 |
+ x11p18 |
| 4901 |
+ x12p17 |
| 4902 |
+ x13p16 |
| 4903 |
+ x14p15 |
| 4904 |
); |
| 4905 |
let [y1, y28] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 4906 |
let [y2, y27] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 4907 |
let [y3, y26] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 4908 |
let [y4, y25] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 4909 |
let [y5, y24] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 4910 |
let [y6, y23] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 4911 |
let [y7, y22] = parallel_fft2_interleaved_f32(t_a7, t_b7_rot); |
| 4912 |
let [y8, y21] = parallel_fft2_interleaved_f32(t_a8, t_b8_rot); |
| 4913 |
let [y9, y20] = parallel_fft2_interleaved_f32(t_a9, t_b9_rot); |
| 4914 |
let [y10, y19] = parallel_fft2_interleaved_f32(t_a10, t_b10_rot); |
| 4915 |
let [y11, y18] = parallel_fft2_interleaved_f32(t_a11, t_b11_rot); |
| 4916 |
let [y12, y17] = parallel_fft2_interleaved_f32(t_a12, t_b12_rot); |
| 4917 |
let [y13, y16] = parallel_fft2_interleaved_f32(t_a13, t_b13_rot); |
| 4918 |
let [y14, y15] = parallel_fft2_interleaved_f32(t_a14, t_b14_rot); |
| 4919 |
[ |
| 4920 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 4921 |
y19, y20, y21, y22, y23, y24, y25, y26, y27, y28, |
| 4922 |
] |
| 4923 |
} |
| 4924 |
} |
| 4925 |
|
| 4926 |
// ____ ___ __ _ _ _ _ _ |
| 4927 |
// |___ \ / _ \ / /_ | || | | |__ (_) |_ |
| 4928 |
// __) | (_) | _____ | '_ \| || |_| '_ \| | __| |
| 4929 |
// / __/ \__, | |_____| | (_) |__ _| |_) | | |_ |
| 4930 |
// |_____| /_/ \___/ |_| |_.__/|_|\__| |
| 4931 |
// |
| 4932 |
|
| 4933 |
pub struct WasmSimdF64Butterfly29<T> { |
| 4934 |
direction: FftDirection, |
| 4935 |
_phantom: std::marker::PhantomData<T>, |
| 4936 |
rotate: Rotate90F64, |
| 4937 |
twiddle1re: v128, |
| 4938 |
twiddle1im: v128, |
| 4939 |
twiddle2re: v128, |
| 4940 |
twiddle2im: v128, |
| 4941 |
twiddle3re: v128, |
| 4942 |
twiddle3im: v128, |
| 4943 |
twiddle4re: v128, |
| 4944 |
twiddle4im: v128, |
| 4945 |
twiddle5re: v128, |
| 4946 |
twiddle5im: v128, |
| 4947 |
twiddle6re: v128, |
| 4948 |
twiddle6im: v128, |
| 4949 |
twiddle7re: v128, |
| 4950 |
twiddle7im: v128, |
| 4951 |
twiddle8re: v128, |
| 4952 |
twiddle8im: v128, |
| 4953 |
twiddle9re: v128, |
| 4954 |
twiddle9im: v128, |
| 4955 |
twiddle10re: v128, |
| 4956 |
twiddle10im: v128, |
| 4957 |
twiddle11re: v128, |
| 4958 |
twiddle11im: v128, |
| 4959 |
twiddle12re: v128, |
| 4960 |
twiddle12im: v128, |
| 4961 |
twiddle13re: v128, |
| 4962 |
twiddle13im: v128, |
| 4963 |
twiddle14re: v128, |
| 4964 |
twiddle14im: v128, |
| 4965 |
} |
| 4966 |
|
| 4967 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 4968 |
WasmSimdF64Butterfly29, |
| 4969 |
29, |
| 4970 |
|this: &WasmSimdF64Butterfly29<_>| this.direction |
| 4971 |
); |
| 4972 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 4973 |
WasmSimdF64Butterfly29, |
| 4974 |
29, |
| 4975 |
|this: &WasmSimdF64Butterfly29<_>| this.direction |
| 4976 |
); |
| 4977 |
impl<T: FftNum> WasmSimdF64Butterfly29<T> { |
| 4978 |
#[inline(always)] |
| 4979 |
pub fn new(direction: FftDirection) -> Self { |
| 4980 |
assert_f64::<T>(); |
| 4981 |
let rotate = Rotate90F64::new(true); |
| 4982 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 29, direction); |
| 4983 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 29, direction); |
| 4984 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 29, direction); |
| 4985 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 29, direction); |
| 4986 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 29, direction); |
| 4987 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 29, direction); |
| 4988 |
let tw7: Complex<f64> = twiddles::compute_twiddle(7, 29, direction); |
| 4989 |
let tw8: Complex<f64> = twiddles::compute_twiddle(8, 29, direction); |
| 4990 |
let tw9: Complex<f64> = twiddles::compute_twiddle(9, 29, direction); |
| 4991 |
let tw10: Complex<f64> = twiddles::compute_twiddle(10, 29, direction); |
| 4992 |
let tw11: Complex<f64> = twiddles::compute_twiddle(11, 29, direction); |
| 4993 |
let tw12: Complex<f64> = twiddles::compute_twiddle(12, 29, direction); |
| 4994 |
let tw13: Complex<f64> = twiddles::compute_twiddle(13, 29, direction); |
| 4995 |
let tw14: Complex<f64> = twiddles::compute_twiddle(14, 29, direction); |
| 4996 |
let twiddle1re = f64x2_splat(tw1.re); |
| 4997 |
let twiddle1im = f64x2_splat(tw1.im); |
| 4998 |
let twiddle2re = f64x2_splat(tw2.re); |
| 4999 |
let twiddle2im = f64x2_splat(tw2.im); |
| 5000 |
let twiddle3re = f64x2_splat(tw3.re); |
| 5001 |
let twiddle3im = f64x2_splat(tw3.im); |
| 5002 |
let twiddle4re = f64x2_splat(tw4.re); |
| 5003 |
let twiddle4im = f64x2_splat(tw4.im); |
| 5004 |
let twiddle5re = f64x2_splat(tw5.re); |
| 5005 |
let twiddle5im = f64x2_splat(tw5.im); |
| 5006 |
let twiddle6re = f64x2_splat(tw6.re); |
| 5007 |
let twiddle6im = f64x2_splat(tw6.im); |
| 5008 |
let twiddle7re = f64x2_splat(tw7.re); |
| 5009 |
let twiddle7im = f64x2_splat(tw7.im); |
| 5010 |
let twiddle8re = f64x2_splat(tw8.re); |
| 5011 |
let twiddle8im = f64x2_splat(tw8.im); |
| 5012 |
let twiddle9re = f64x2_splat(tw9.re); |
| 5013 |
let twiddle9im = f64x2_splat(tw9.im); |
| 5014 |
let twiddle10re = f64x2_splat(tw10.re); |
| 5015 |
let twiddle10im = f64x2_splat(tw10.im); |
| 5016 |
let twiddle11re = f64x2_splat(tw11.re); |
| 5017 |
let twiddle11im = f64x2_splat(tw11.im); |
| 5018 |
let twiddle12re = f64x2_splat(tw12.re); |
| 5019 |
let twiddle12im = f64x2_splat(tw12.im); |
| 5020 |
let twiddle13re = f64x2_splat(tw13.re); |
| 5021 |
let twiddle13im = f64x2_splat(tw13.im); |
| 5022 |
let twiddle14re = f64x2_splat(tw14.re); |
| 5023 |
let twiddle14im = f64x2_splat(tw14.im); |
| 5024 |
|
| 5025 |
Self { |
| 5026 |
direction, |
| 5027 |
_phantom: std::marker::PhantomData, |
| 5028 |
rotate, |
| 5029 |
twiddle1re, |
| 5030 |
twiddle1im, |
| 5031 |
twiddle2re, |
| 5032 |
twiddle2im, |
| 5033 |
twiddle3re, |
| 5034 |
twiddle3im, |
| 5035 |
twiddle4re, |
| 5036 |
twiddle4im, |
| 5037 |
twiddle5re, |
| 5038 |
twiddle5im, |
| 5039 |
twiddle6re, |
| 5040 |
twiddle6im, |
| 5041 |
twiddle7re, |
| 5042 |
twiddle7im, |
| 5043 |
twiddle8re, |
| 5044 |
twiddle8im, |
| 5045 |
twiddle9re, |
| 5046 |
twiddle9im, |
| 5047 |
twiddle10re, |
| 5048 |
twiddle10im, |
| 5049 |
twiddle11re, |
| 5050 |
twiddle11im, |
| 5051 |
twiddle12re, |
| 5052 |
twiddle12im, |
| 5053 |
twiddle13re, |
| 5054 |
twiddle13im, |
| 5055 |
twiddle14re, |
| 5056 |
twiddle14im, |
| 5057 |
} |
| 5058 |
} |
| 5059 |
|
| 5060 |
#[inline(always)] |
| 5061 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 5062 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28}); |
| 5063 |
|
| 5064 |
let out = self.perform_fft_direct(values); |
| 5065 |
|
| 5066 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28}); |
| 5067 |
} |
| 5068 |
|
| 5069 |
#[inline(always)] |
| 5070 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 29]) -> [v128; 29] { |
| 5071 |
let [x1p28, x1m28] = solo_fft2_f64(values[1], values[28]); |
| 5072 |
let [x2p27, x2m27] = solo_fft2_f64(values[2], values[27]); |
| 5073 |
let [x3p26, x3m26] = solo_fft2_f64(values[3], values[26]); |
| 5074 |
let [x4p25, x4m25] = solo_fft2_f64(values[4], values[25]); |
| 5075 |
let [x5p24, x5m24] = solo_fft2_f64(values[5], values[24]); |
| 5076 |
let [x6p23, x6m23] = solo_fft2_f64(values[6], values[23]); |
| 5077 |
let [x7p22, x7m22] = solo_fft2_f64(values[7], values[22]); |
| 5078 |
let [x8p21, x8m21] = solo_fft2_f64(values[8], values[21]); |
| 5079 |
let [x9p20, x9m20] = solo_fft2_f64(values[9], values[20]); |
| 5080 |
let [x10p19, x10m19] = solo_fft2_f64(values[10], values[19]); |
| 5081 |
let [x11p18, x11m18] = solo_fft2_f64(values[11], values[18]); |
| 5082 |
let [x12p17, x12m17] = solo_fft2_f64(values[12], values[17]); |
| 5083 |
let [x13p16, x13m16] = solo_fft2_f64(values[13], values[16]); |
| 5084 |
let [x14p15, x14m15] = solo_fft2_f64(values[14], values[15]); |
| 5085 |
|
| 5086 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p28); |
| 5087 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p27); |
| 5088 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p26); |
| 5089 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p25); |
| 5090 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p24); |
| 5091 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p23); |
| 5092 |
let t_a1_7 = f64x2_mul(self.twiddle7re, x7p22); |
| 5093 |
let t_a1_8 = f64x2_mul(self.twiddle8re, x8p21); |
| 5094 |
let t_a1_9 = f64x2_mul(self.twiddle9re, x9p20); |
| 5095 |
let t_a1_10 = f64x2_mul(self.twiddle10re, x10p19); |
| 5096 |
let t_a1_11 = f64x2_mul(self.twiddle11re, x11p18); |
| 5097 |
let t_a1_12 = f64x2_mul(self.twiddle12re, x12p17); |
| 5098 |
let t_a1_13 = f64x2_mul(self.twiddle13re, x13p16); |
| 5099 |
let t_a1_14 = f64x2_mul(self.twiddle14re, x14p15); |
| 5100 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p28); |
| 5101 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p27); |
| 5102 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p26); |
| 5103 |
let t_a2_4 = f64x2_mul(self.twiddle8re, x4p25); |
| 5104 |
let t_a2_5 = f64x2_mul(self.twiddle10re, x5p24); |
| 5105 |
let t_a2_6 = f64x2_mul(self.twiddle12re, x6p23); |
| 5106 |
let t_a2_7 = f64x2_mul(self.twiddle14re, x7p22); |
| 5107 |
let t_a2_8 = f64x2_mul(self.twiddle13re, x8p21); |
| 5108 |
let t_a2_9 = f64x2_mul(self.twiddle11re, x9p20); |
| 5109 |
let t_a2_10 = f64x2_mul(self.twiddle9re, x10p19); |
| 5110 |
let t_a2_11 = f64x2_mul(self.twiddle7re, x11p18); |
| 5111 |
let t_a2_12 = f64x2_mul(self.twiddle5re, x12p17); |
| 5112 |
let t_a2_13 = f64x2_mul(self.twiddle3re, x13p16); |
| 5113 |
let t_a2_14 = f64x2_mul(self.twiddle1re, x14p15); |
| 5114 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p28); |
| 5115 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p27); |
| 5116 |
let t_a3_3 = f64x2_mul(self.twiddle9re, x3p26); |
| 5117 |
let t_a3_4 = f64x2_mul(self.twiddle12re, x4p25); |
| 5118 |
let t_a3_5 = f64x2_mul(self.twiddle14re, x5p24); |
| 5119 |
let t_a3_6 = f64x2_mul(self.twiddle11re, x6p23); |
| 5120 |
let t_a3_7 = f64x2_mul(self.twiddle8re, x7p22); |
| 5121 |
let t_a3_8 = f64x2_mul(self.twiddle5re, x8p21); |
| 5122 |
let t_a3_9 = f64x2_mul(self.twiddle2re, x9p20); |
| 5123 |
let t_a3_10 = f64x2_mul(self.twiddle1re, x10p19); |
| 5124 |
let t_a3_11 = f64x2_mul(self.twiddle4re, x11p18); |
| 5125 |
let t_a3_12 = f64x2_mul(self.twiddle7re, x12p17); |
| 5126 |
let t_a3_13 = f64x2_mul(self.twiddle10re, x13p16); |
| 5127 |
let t_a3_14 = f64x2_mul(self.twiddle13re, x14p15); |
| 5128 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p28); |
| 5129 |
let t_a4_2 = f64x2_mul(self.twiddle8re, x2p27); |
| 5130 |
let t_a4_3 = f64x2_mul(self.twiddle12re, x3p26); |
| 5131 |
let t_a4_4 = f64x2_mul(self.twiddle13re, x4p25); |
| 5132 |
let t_a4_5 = f64x2_mul(self.twiddle9re, x5p24); |
| 5133 |
let t_a4_6 = f64x2_mul(self.twiddle5re, x6p23); |
| 5134 |
let t_a4_7 = f64x2_mul(self.twiddle1re, x7p22); |
| 5135 |
let t_a4_8 = f64x2_mul(self.twiddle3re, x8p21); |
| 5136 |
let t_a4_9 = f64x2_mul(self.twiddle7re, x9p20); |
| 5137 |
let t_a4_10 = f64x2_mul(self.twiddle11re, x10p19); |
| 5138 |
let t_a4_11 = f64x2_mul(self.twiddle14re, x11p18); |
| 5139 |
let t_a4_12 = f64x2_mul(self.twiddle10re, x12p17); |
| 5140 |
let t_a4_13 = f64x2_mul(self.twiddle6re, x13p16); |
| 5141 |
let t_a4_14 = f64x2_mul(self.twiddle2re, x14p15); |
| 5142 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p28); |
| 5143 |
let t_a5_2 = f64x2_mul(self.twiddle10re, x2p27); |
| 5144 |
let t_a5_3 = f64x2_mul(self.twiddle14re, x3p26); |
| 5145 |
let t_a5_4 = f64x2_mul(self.twiddle9re, x4p25); |
| 5146 |
let t_a5_5 = f64x2_mul(self.twiddle4re, x5p24); |
| 5147 |
let t_a5_6 = f64x2_mul(self.twiddle1re, x6p23); |
| 5148 |
let t_a5_7 = f64x2_mul(self.twiddle6re, x7p22); |
| 5149 |
let t_a5_8 = f64x2_mul(self.twiddle11re, x8p21); |
| 5150 |
let t_a5_9 = f64x2_mul(self.twiddle13re, x9p20); |
| 5151 |
let t_a5_10 = f64x2_mul(self.twiddle8re, x10p19); |
| 5152 |
let t_a5_11 = f64x2_mul(self.twiddle3re, x11p18); |
| 5153 |
let t_a5_12 = f64x2_mul(self.twiddle2re, x12p17); |
| 5154 |
let t_a5_13 = f64x2_mul(self.twiddle7re, x13p16); |
| 5155 |
let t_a5_14 = f64x2_mul(self.twiddle12re, x14p15); |
| 5156 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p28); |
| 5157 |
let t_a6_2 = f64x2_mul(self.twiddle12re, x2p27); |
| 5158 |
let t_a6_3 = f64x2_mul(self.twiddle11re, x3p26); |
| 5159 |
let t_a6_4 = f64x2_mul(self.twiddle5re, x4p25); |
| 5160 |
let t_a6_5 = f64x2_mul(self.twiddle1re, x5p24); |
| 5161 |
let t_a6_6 = f64x2_mul(self.twiddle7re, x6p23); |
| 5162 |
let t_a6_7 = f64x2_mul(self.twiddle13re, x7p22); |
| 5163 |
let t_a6_8 = f64x2_mul(self.twiddle10re, x8p21); |
| 5164 |
let t_a6_9 = f64x2_mul(self.twiddle4re, x9p20); |
| 5165 |
let t_a6_10 = f64x2_mul(self.twiddle2re, x10p19); |
| 5166 |
let t_a6_11 = f64x2_mul(self.twiddle8re, x11p18); |
| 5167 |
let t_a6_12 = f64x2_mul(self.twiddle14re, x12p17); |
| 5168 |
let t_a6_13 = f64x2_mul(self.twiddle9re, x13p16); |
| 5169 |
let t_a6_14 = f64x2_mul(self.twiddle3re, x14p15); |
| 5170 |
let t_a7_1 = f64x2_mul(self.twiddle7re, x1p28); |
| 5171 |
let t_a7_2 = f64x2_mul(self.twiddle14re, x2p27); |
| 5172 |
let t_a7_3 = f64x2_mul(self.twiddle8re, x3p26); |
| 5173 |
let t_a7_4 = f64x2_mul(self.twiddle1re, x4p25); |
| 5174 |
let t_a7_5 = f64x2_mul(self.twiddle6re, x5p24); |
| 5175 |
let t_a7_6 = f64x2_mul(self.twiddle13re, x6p23); |
| 5176 |
let t_a7_7 = f64x2_mul(self.twiddle9re, x7p22); |
| 5177 |
let t_a7_8 = f64x2_mul(self.twiddle2re, x8p21); |
| 5178 |
let t_a7_9 = f64x2_mul(self.twiddle5re, x9p20); |
| 5179 |
let t_a7_10 = f64x2_mul(self.twiddle12re, x10p19); |
| 5180 |
let t_a7_11 = f64x2_mul(self.twiddle10re, x11p18); |
| 5181 |
let t_a7_12 = f64x2_mul(self.twiddle3re, x12p17); |
| 5182 |
let t_a7_13 = f64x2_mul(self.twiddle4re, x13p16); |
| 5183 |
let t_a7_14 = f64x2_mul(self.twiddle11re, x14p15); |
| 5184 |
let t_a8_1 = f64x2_mul(self.twiddle8re, x1p28); |
| 5185 |
let t_a8_2 = f64x2_mul(self.twiddle13re, x2p27); |
| 5186 |
let t_a8_3 = f64x2_mul(self.twiddle5re, x3p26); |
| 5187 |
let t_a8_4 = f64x2_mul(self.twiddle3re, x4p25); |
| 5188 |
let t_a8_5 = f64x2_mul(self.twiddle11re, x5p24); |
| 5189 |
let t_a8_6 = f64x2_mul(self.twiddle10re, x6p23); |
| 5190 |
let t_a8_7 = f64x2_mul(self.twiddle2re, x7p22); |
| 5191 |
let t_a8_8 = f64x2_mul(self.twiddle6re, x8p21); |
| 5192 |
let t_a8_9 = f64x2_mul(self.twiddle14re, x9p20); |
| 5193 |
let t_a8_10 = f64x2_mul(self.twiddle7re, x10p19); |
| 5194 |
let t_a8_11 = f64x2_mul(self.twiddle1re, x11p18); |
| 5195 |
let t_a8_12 = f64x2_mul(self.twiddle9re, x12p17); |
| 5196 |
let t_a8_13 = f64x2_mul(self.twiddle12re, x13p16); |
| 5197 |
let t_a8_14 = f64x2_mul(self.twiddle4re, x14p15); |
| 5198 |
let t_a9_1 = f64x2_mul(self.twiddle9re, x1p28); |
| 5199 |
let t_a9_2 = f64x2_mul(self.twiddle11re, x2p27); |
| 5200 |
let t_a9_3 = f64x2_mul(self.twiddle2re, x3p26); |
| 5201 |
let t_a9_4 = f64x2_mul(self.twiddle7re, x4p25); |
| 5202 |
let t_a9_5 = f64x2_mul(self.twiddle13re, x5p24); |
| 5203 |
let t_a9_6 = f64x2_mul(self.twiddle4re, x6p23); |
| 5204 |
let t_a9_7 = f64x2_mul(self.twiddle5re, x7p22); |
| 5205 |
let t_a9_8 = f64x2_mul(self.twiddle14re, x8p21); |
| 5206 |
let t_a9_9 = f64x2_mul(self.twiddle6re, x9p20); |
| 5207 |
let t_a9_10 = f64x2_mul(self.twiddle3re, x10p19); |
| 5208 |
let t_a9_11 = f64x2_mul(self.twiddle12re, x11p18); |
| 5209 |
let t_a9_12 = f64x2_mul(self.twiddle8re, x12p17); |
| 5210 |
let t_a9_13 = f64x2_mul(self.twiddle1re, x13p16); |
| 5211 |
let t_a9_14 = f64x2_mul(self.twiddle10re, x14p15); |
| 5212 |
let t_a10_1 = f64x2_mul(self.twiddle10re, x1p28); |
| 5213 |
let t_a10_2 = f64x2_mul(self.twiddle9re, x2p27); |
| 5214 |
let t_a10_3 = f64x2_mul(self.twiddle1re, x3p26); |
| 5215 |
let t_a10_4 = f64x2_mul(self.twiddle11re, x4p25); |
| 5216 |
let t_a10_5 = f64x2_mul(self.twiddle8re, x5p24); |
| 5217 |
let t_a10_6 = f64x2_mul(self.twiddle2re, x6p23); |
| 5218 |
let t_a10_7 = f64x2_mul(self.twiddle12re, x7p22); |
| 5219 |
let t_a10_8 = f64x2_mul(self.twiddle7re, x8p21); |
| 5220 |
let t_a10_9 = f64x2_mul(self.twiddle3re, x9p20); |
| 5221 |
let t_a10_10 = f64x2_mul(self.twiddle13re, x10p19); |
| 5222 |
let t_a10_11 = f64x2_mul(self.twiddle6re, x11p18); |
| 5223 |
let t_a10_12 = f64x2_mul(self.twiddle4re, x12p17); |
| 5224 |
let t_a10_13 = f64x2_mul(self.twiddle14re, x13p16); |
| 5225 |
let t_a10_14 = f64x2_mul(self.twiddle5re, x14p15); |
| 5226 |
let t_a11_1 = f64x2_mul(self.twiddle11re, x1p28); |
| 5227 |
let t_a11_2 = f64x2_mul(self.twiddle7re, x2p27); |
| 5228 |
let t_a11_3 = f64x2_mul(self.twiddle4re, x3p26); |
| 5229 |
let t_a11_4 = f64x2_mul(self.twiddle14re, x4p25); |
| 5230 |
let t_a11_5 = f64x2_mul(self.twiddle3re, x5p24); |
| 5231 |
let t_a11_6 = f64x2_mul(self.twiddle8re, x6p23); |
| 5232 |
let t_a11_7 = f64x2_mul(self.twiddle10re, x7p22); |
| 5233 |
let t_a11_8 = f64x2_mul(self.twiddle1re, x8p21); |
| 5234 |
let t_a11_9 = f64x2_mul(self.twiddle12re, x9p20); |
| 5235 |
let t_a11_10 = f64x2_mul(self.twiddle6re, x10p19); |
| 5236 |
let t_a11_11 = f64x2_mul(self.twiddle5re, x11p18); |
| 5237 |
let t_a11_12 = f64x2_mul(self.twiddle13re, x12p17); |
| 5238 |
let t_a11_13 = f64x2_mul(self.twiddle2re, x13p16); |
| 5239 |
let t_a11_14 = f64x2_mul(self.twiddle9re, x14p15); |
| 5240 |
let t_a12_1 = f64x2_mul(self.twiddle12re, x1p28); |
| 5241 |
let t_a12_2 = f64x2_mul(self.twiddle5re, x2p27); |
| 5242 |
let t_a12_3 = f64x2_mul(self.twiddle7re, x3p26); |
| 5243 |
let t_a12_4 = f64x2_mul(self.twiddle10re, x4p25); |
| 5244 |
let t_a12_5 = f64x2_mul(self.twiddle2re, x5p24); |
| 5245 |
let t_a12_6 = f64x2_mul(self.twiddle14re, x6p23); |
| 5246 |
let t_a12_7 = f64x2_mul(self.twiddle3re, x7p22); |
| 5247 |
let t_a12_8 = f64x2_mul(self.twiddle9re, x8p21); |
| 5248 |
let t_a12_9 = f64x2_mul(self.twiddle8re, x9p20); |
| 5249 |
let t_a12_10 = f64x2_mul(self.twiddle4re, x10p19); |
| 5250 |
let t_a12_11 = f64x2_mul(self.twiddle13re, x11p18); |
| 5251 |
let t_a12_12 = f64x2_mul(self.twiddle1re, x12p17); |
| 5252 |
let t_a12_13 = f64x2_mul(self.twiddle11re, x13p16); |
| 5253 |
let t_a12_14 = f64x2_mul(self.twiddle6re, x14p15); |
| 5254 |
let t_a13_1 = f64x2_mul(self.twiddle13re, x1p28); |
| 5255 |
let t_a13_2 = f64x2_mul(self.twiddle3re, x2p27); |
| 5256 |
let t_a13_3 = f64x2_mul(self.twiddle10re, x3p26); |
| 5257 |
let t_a13_4 = f64x2_mul(self.twiddle6re, x4p25); |
| 5258 |
let t_a13_5 = f64x2_mul(self.twiddle7re, x5p24); |
| 5259 |
let t_a13_6 = f64x2_mul(self.twiddle9re, x6p23); |
| 5260 |
let t_a13_7 = f64x2_mul(self.twiddle4re, x7p22); |
| 5261 |
let t_a13_8 = f64x2_mul(self.twiddle12re, x8p21); |
| 5262 |
let t_a13_9 = f64x2_mul(self.twiddle1re, x9p20); |
| 5263 |
let t_a13_10 = f64x2_mul(self.twiddle14re, x10p19); |
| 5264 |
let t_a13_11 = f64x2_mul(self.twiddle2re, x11p18); |
| 5265 |
let t_a13_12 = f64x2_mul(self.twiddle11re, x12p17); |
| 5266 |
let t_a13_13 = f64x2_mul(self.twiddle5re, x13p16); |
| 5267 |
let t_a13_14 = f64x2_mul(self.twiddle8re, x14p15); |
| 5268 |
let t_a14_1 = f64x2_mul(self.twiddle14re, x1p28); |
| 5269 |
let t_a14_2 = f64x2_mul(self.twiddle1re, x2p27); |
| 5270 |
let t_a14_3 = f64x2_mul(self.twiddle13re, x3p26); |
| 5271 |
let t_a14_4 = f64x2_mul(self.twiddle2re, x4p25); |
| 5272 |
let t_a14_5 = f64x2_mul(self.twiddle12re, x5p24); |
| 5273 |
let t_a14_6 = f64x2_mul(self.twiddle3re, x6p23); |
| 5274 |
let t_a14_7 = f64x2_mul(self.twiddle11re, x7p22); |
| 5275 |
let t_a14_8 = f64x2_mul(self.twiddle4re, x8p21); |
| 5276 |
let t_a14_9 = f64x2_mul(self.twiddle10re, x9p20); |
| 5277 |
let t_a14_10 = f64x2_mul(self.twiddle5re, x10p19); |
| 5278 |
let t_a14_11 = f64x2_mul(self.twiddle9re, x11p18); |
| 5279 |
let t_a14_12 = f64x2_mul(self.twiddle6re, x12p17); |
| 5280 |
let t_a14_13 = f64x2_mul(self.twiddle8re, x13p16); |
| 5281 |
let t_a14_14 = f64x2_mul(self.twiddle7re, x14p15); |
| 5282 |
|
| 5283 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m28); |
| 5284 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m27); |
| 5285 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m26); |
| 5286 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m25); |
| 5287 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m24); |
| 5288 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m23); |
| 5289 |
let t_b1_7 = f64x2_mul(self.twiddle7im, x7m22); |
| 5290 |
let t_b1_8 = f64x2_mul(self.twiddle8im, x8m21); |
| 5291 |
let t_b1_9 = f64x2_mul(self.twiddle9im, x9m20); |
| 5292 |
let t_b1_10 = f64x2_mul(self.twiddle10im, x10m19); |
| 5293 |
let t_b1_11 = f64x2_mul(self.twiddle11im, x11m18); |
| 5294 |
let t_b1_12 = f64x2_mul(self.twiddle12im, x12m17); |
| 5295 |
let t_b1_13 = f64x2_mul(self.twiddle13im, x13m16); |
| 5296 |
let t_b1_14 = f64x2_mul(self.twiddle14im, x14m15); |
| 5297 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m28); |
| 5298 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m27); |
| 5299 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m26); |
| 5300 |
let t_b2_4 = f64x2_mul(self.twiddle8im, x4m25); |
| 5301 |
let t_b2_5 = f64x2_mul(self.twiddle10im, x5m24); |
| 5302 |
let t_b2_6 = f64x2_mul(self.twiddle12im, x6m23); |
| 5303 |
let t_b2_7 = f64x2_mul(self.twiddle14im, x7m22); |
| 5304 |
let t_b2_8 = f64x2_mul(self.twiddle13im, x8m21); |
| 5305 |
let t_b2_9 = f64x2_mul(self.twiddle11im, x9m20); |
| 5306 |
let t_b2_10 = f64x2_mul(self.twiddle9im, x10m19); |
| 5307 |
let t_b2_11 = f64x2_mul(self.twiddle7im, x11m18); |
| 5308 |
let t_b2_12 = f64x2_mul(self.twiddle5im, x12m17); |
| 5309 |
let t_b2_13 = f64x2_mul(self.twiddle3im, x13m16); |
| 5310 |
let t_b2_14 = f64x2_mul(self.twiddle1im, x14m15); |
| 5311 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m28); |
| 5312 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m27); |
| 5313 |
let t_b3_3 = f64x2_mul(self.twiddle9im, x3m26); |
| 5314 |
let t_b3_4 = f64x2_mul(self.twiddle12im, x4m25); |
| 5315 |
let t_b3_5 = f64x2_mul(self.twiddle14im, x5m24); |
| 5316 |
let t_b3_6 = f64x2_mul(self.twiddle11im, x6m23); |
| 5317 |
let t_b3_7 = f64x2_mul(self.twiddle8im, x7m22); |
| 5318 |
let t_b3_8 = f64x2_mul(self.twiddle5im, x8m21); |
| 5319 |
let t_b3_9 = f64x2_mul(self.twiddle2im, x9m20); |
| 5320 |
let t_b3_10 = f64x2_mul(self.twiddle1im, x10m19); |
| 5321 |
let t_b3_11 = f64x2_mul(self.twiddle4im, x11m18); |
| 5322 |
let t_b3_12 = f64x2_mul(self.twiddle7im, x12m17); |
| 5323 |
let t_b3_13 = f64x2_mul(self.twiddle10im, x13m16); |
| 5324 |
let t_b3_14 = f64x2_mul(self.twiddle13im, x14m15); |
| 5325 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m28); |
| 5326 |
let t_b4_2 = f64x2_mul(self.twiddle8im, x2m27); |
| 5327 |
let t_b4_3 = f64x2_mul(self.twiddle12im, x3m26); |
| 5328 |
let t_b4_4 = f64x2_mul(self.twiddle13im, x4m25); |
| 5329 |
let t_b4_5 = f64x2_mul(self.twiddle9im, x5m24); |
| 5330 |
let t_b4_6 = f64x2_mul(self.twiddle5im, x6m23); |
| 5331 |
let t_b4_7 = f64x2_mul(self.twiddle1im, x7m22); |
| 5332 |
let t_b4_8 = f64x2_mul(self.twiddle3im, x8m21); |
| 5333 |
let t_b4_9 = f64x2_mul(self.twiddle7im, x9m20); |
| 5334 |
let t_b4_10 = f64x2_mul(self.twiddle11im, x10m19); |
| 5335 |
let t_b4_11 = f64x2_mul(self.twiddle14im, x11m18); |
| 5336 |
let t_b4_12 = f64x2_mul(self.twiddle10im, x12m17); |
| 5337 |
let t_b4_13 = f64x2_mul(self.twiddle6im, x13m16); |
| 5338 |
let t_b4_14 = f64x2_mul(self.twiddle2im, x14m15); |
| 5339 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m28); |
| 5340 |
let t_b5_2 = f64x2_mul(self.twiddle10im, x2m27); |
| 5341 |
let t_b5_3 = f64x2_mul(self.twiddle14im, x3m26); |
| 5342 |
let t_b5_4 = f64x2_mul(self.twiddle9im, x4m25); |
| 5343 |
let t_b5_5 = f64x2_mul(self.twiddle4im, x5m24); |
| 5344 |
let t_b5_6 = f64x2_mul(self.twiddle1im, x6m23); |
| 5345 |
let t_b5_7 = f64x2_mul(self.twiddle6im, x7m22); |
| 5346 |
let t_b5_8 = f64x2_mul(self.twiddle11im, x8m21); |
| 5347 |
let t_b5_9 = f64x2_mul(self.twiddle13im, x9m20); |
| 5348 |
let t_b5_10 = f64x2_mul(self.twiddle8im, x10m19); |
| 5349 |
let t_b5_11 = f64x2_mul(self.twiddle3im, x11m18); |
| 5350 |
let t_b5_12 = f64x2_mul(self.twiddle2im, x12m17); |
| 5351 |
let t_b5_13 = f64x2_mul(self.twiddle7im, x13m16); |
| 5352 |
let t_b5_14 = f64x2_mul(self.twiddle12im, x14m15); |
| 5353 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m28); |
| 5354 |
let t_b6_2 = f64x2_mul(self.twiddle12im, x2m27); |
| 5355 |
let t_b6_3 = f64x2_mul(self.twiddle11im, x3m26); |
| 5356 |
let t_b6_4 = f64x2_mul(self.twiddle5im, x4m25); |
| 5357 |
let t_b6_5 = f64x2_mul(self.twiddle1im, x5m24); |
| 5358 |
let t_b6_6 = f64x2_mul(self.twiddle7im, x6m23); |
| 5359 |
let t_b6_7 = f64x2_mul(self.twiddle13im, x7m22); |
| 5360 |
let t_b6_8 = f64x2_mul(self.twiddle10im, x8m21); |
| 5361 |
let t_b6_9 = f64x2_mul(self.twiddle4im, x9m20); |
| 5362 |
let t_b6_10 = f64x2_mul(self.twiddle2im, x10m19); |
| 5363 |
let t_b6_11 = f64x2_mul(self.twiddle8im, x11m18); |
| 5364 |
let t_b6_12 = f64x2_mul(self.twiddle14im, x12m17); |
| 5365 |
let t_b6_13 = f64x2_mul(self.twiddle9im, x13m16); |
| 5366 |
let t_b6_14 = f64x2_mul(self.twiddle3im, x14m15); |
| 5367 |
let t_b7_1 = f64x2_mul(self.twiddle7im, x1m28); |
| 5368 |
let t_b7_2 = f64x2_mul(self.twiddle14im, x2m27); |
| 5369 |
let t_b7_3 = f64x2_mul(self.twiddle8im, x3m26); |
| 5370 |
let t_b7_4 = f64x2_mul(self.twiddle1im, x4m25); |
| 5371 |
let t_b7_5 = f64x2_mul(self.twiddle6im, x5m24); |
| 5372 |
let t_b7_6 = f64x2_mul(self.twiddle13im, x6m23); |
| 5373 |
let t_b7_7 = f64x2_mul(self.twiddle9im, x7m22); |
| 5374 |
let t_b7_8 = f64x2_mul(self.twiddle2im, x8m21); |
| 5375 |
let t_b7_9 = f64x2_mul(self.twiddle5im, x9m20); |
| 5376 |
let t_b7_10 = f64x2_mul(self.twiddle12im, x10m19); |
| 5377 |
let t_b7_11 = f64x2_mul(self.twiddle10im, x11m18); |
| 5378 |
let t_b7_12 = f64x2_mul(self.twiddle3im, x12m17); |
| 5379 |
let t_b7_13 = f64x2_mul(self.twiddle4im, x13m16); |
| 5380 |
let t_b7_14 = f64x2_mul(self.twiddle11im, x14m15); |
| 5381 |
let t_b8_1 = f64x2_mul(self.twiddle8im, x1m28); |
| 5382 |
let t_b8_2 = f64x2_mul(self.twiddle13im, x2m27); |
| 5383 |
let t_b8_3 = f64x2_mul(self.twiddle5im, x3m26); |
| 5384 |
let t_b8_4 = f64x2_mul(self.twiddle3im, x4m25); |
| 5385 |
let t_b8_5 = f64x2_mul(self.twiddle11im, x5m24); |
| 5386 |
let t_b8_6 = f64x2_mul(self.twiddle10im, x6m23); |
| 5387 |
let t_b8_7 = f64x2_mul(self.twiddle2im, x7m22); |
| 5388 |
let t_b8_8 = f64x2_mul(self.twiddle6im, x8m21); |
| 5389 |
let t_b8_9 = f64x2_mul(self.twiddle14im, x9m20); |
| 5390 |
let t_b8_10 = f64x2_mul(self.twiddle7im, x10m19); |
| 5391 |
let t_b8_11 = f64x2_mul(self.twiddle1im, x11m18); |
| 5392 |
let t_b8_12 = f64x2_mul(self.twiddle9im, x12m17); |
| 5393 |
let t_b8_13 = f64x2_mul(self.twiddle12im, x13m16); |
| 5394 |
let t_b8_14 = f64x2_mul(self.twiddle4im, x14m15); |
| 5395 |
let t_b9_1 = f64x2_mul(self.twiddle9im, x1m28); |
| 5396 |
let t_b9_2 = f64x2_mul(self.twiddle11im, x2m27); |
| 5397 |
let t_b9_3 = f64x2_mul(self.twiddle2im, x3m26); |
| 5398 |
let t_b9_4 = f64x2_mul(self.twiddle7im, x4m25); |
| 5399 |
let t_b9_5 = f64x2_mul(self.twiddle13im, x5m24); |
| 5400 |
let t_b9_6 = f64x2_mul(self.twiddle4im, x6m23); |
| 5401 |
let t_b9_7 = f64x2_mul(self.twiddle5im, x7m22); |
| 5402 |
let t_b9_8 = f64x2_mul(self.twiddle14im, x8m21); |
| 5403 |
let t_b9_9 = f64x2_mul(self.twiddle6im, x9m20); |
| 5404 |
let t_b9_10 = f64x2_mul(self.twiddle3im, x10m19); |
| 5405 |
let t_b9_11 = f64x2_mul(self.twiddle12im, x11m18); |
| 5406 |
let t_b9_12 = f64x2_mul(self.twiddle8im, x12m17); |
| 5407 |
let t_b9_13 = f64x2_mul(self.twiddle1im, x13m16); |
| 5408 |
let t_b9_14 = f64x2_mul(self.twiddle10im, x14m15); |
| 5409 |
let t_b10_1 = f64x2_mul(self.twiddle10im, x1m28); |
| 5410 |
let t_b10_2 = f64x2_mul(self.twiddle9im, x2m27); |
| 5411 |
let t_b10_3 = f64x2_mul(self.twiddle1im, x3m26); |
| 5412 |
let t_b10_4 = f64x2_mul(self.twiddle11im, x4m25); |
| 5413 |
let t_b10_5 = f64x2_mul(self.twiddle8im, x5m24); |
| 5414 |
let t_b10_6 = f64x2_mul(self.twiddle2im, x6m23); |
| 5415 |
let t_b10_7 = f64x2_mul(self.twiddle12im, x7m22); |
| 5416 |
let t_b10_8 = f64x2_mul(self.twiddle7im, x8m21); |
| 5417 |
let t_b10_9 = f64x2_mul(self.twiddle3im, x9m20); |
| 5418 |
let t_b10_10 = f64x2_mul(self.twiddle13im, x10m19); |
| 5419 |
let t_b10_11 = f64x2_mul(self.twiddle6im, x11m18); |
| 5420 |
let t_b10_12 = f64x2_mul(self.twiddle4im, x12m17); |
| 5421 |
let t_b10_13 = f64x2_mul(self.twiddle14im, x13m16); |
| 5422 |
let t_b10_14 = f64x2_mul(self.twiddle5im, x14m15); |
| 5423 |
let t_b11_1 = f64x2_mul(self.twiddle11im, x1m28); |
| 5424 |
let t_b11_2 = f64x2_mul(self.twiddle7im, x2m27); |
| 5425 |
let t_b11_3 = f64x2_mul(self.twiddle4im, x3m26); |
| 5426 |
let t_b11_4 = f64x2_mul(self.twiddle14im, x4m25); |
| 5427 |
let t_b11_5 = f64x2_mul(self.twiddle3im, x5m24); |
| 5428 |
let t_b11_6 = f64x2_mul(self.twiddle8im, x6m23); |
| 5429 |
let t_b11_7 = f64x2_mul(self.twiddle10im, x7m22); |
| 5430 |
let t_b11_8 = f64x2_mul(self.twiddle1im, x8m21); |
| 5431 |
let t_b11_9 = f64x2_mul(self.twiddle12im, x9m20); |
| 5432 |
let t_b11_10 = f64x2_mul(self.twiddle6im, x10m19); |
| 5433 |
let t_b11_11 = f64x2_mul(self.twiddle5im, x11m18); |
| 5434 |
let t_b11_12 = f64x2_mul(self.twiddle13im, x12m17); |
| 5435 |
let t_b11_13 = f64x2_mul(self.twiddle2im, x13m16); |
| 5436 |
let t_b11_14 = f64x2_mul(self.twiddle9im, x14m15); |
| 5437 |
let t_b12_1 = f64x2_mul(self.twiddle12im, x1m28); |
| 5438 |
let t_b12_2 = f64x2_mul(self.twiddle5im, x2m27); |
| 5439 |
let t_b12_3 = f64x2_mul(self.twiddle7im, x3m26); |
| 5440 |
let t_b12_4 = f64x2_mul(self.twiddle10im, x4m25); |
| 5441 |
let t_b12_5 = f64x2_mul(self.twiddle2im, x5m24); |
| 5442 |
let t_b12_6 = f64x2_mul(self.twiddle14im, x6m23); |
| 5443 |
let t_b12_7 = f64x2_mul(self.twiddle3im, x7m22); |
| 5444 |
let t_b12_8 = f64x2_mul(self.twiddle9im, x8m21); |
| 5445 |
let t_b12_9 = f64x2_mul(self.twiddle8im, x9m20); |
| 5446 |
let t_b12_10 = f64x2_mul(self.twiddle4im, x10m19); |
| 5447 |
let t_b12_11 = f64x2_mul(self.twiddle13im, x11m18); |
| 5448 |
let t_b12_12 = f64x2_mul(self.twiddle1im, x12m17); |
| 5449 |
let t_b12_13 = f64x2_mul(self.twiddle11im, x13m16); |
| 5450 |
let t_b12_14 = f64x2_mul(self.twiddle6im, x14m15); |
| 5451 |
let t_b13_1 = f64x2_mul(self.twiddle13im, x1m28); |
| 5452 |
let t_b13_2 = f64x2_mul(self.twiddle3im, x2m27); |
| 5453 |
let t_b13_3 = f64x2_mul(self.twiddle10im, x3m26); |
| 5454 |
let t_b13_4 = f64x2_mul(self.twiddle6im, x4m25); |
| 5455 |
let t_b13_5 = f64x2_mul(self.twiddle7im, x5m24); |
| 5456 |
let t_b13_6 = f64x2_mul(self.twiddle9im, x6m23); |
| 5457 |
let t_b13_7 = f64x2_mul(self.twiddle4im, x7m22); |
| 5458 |
let t_b13_8 = f64x2_mul(self.twiddle12im, x8m21); |
| 5459 |
let t_b13_9 = f64x2_mul(self.twiddle1im, x9m20); |
| 5460 |
let t_b13_10 = f64x2_mul(self.twiddle14im, x10m19); |
| 5461 |
let t_b13_11 = f64x2_mul(self.twiddle2im, x11m18); |
| 5462 |
let t_b13_12 = f64x2_mul(self.twiddle11im, x12m17); |
| 5463 |
let t_b13_13 = f64x2_mul(self.twiddle5im, x13m16); |
| 5464 |
let t_b13_14 = f64x2_mul(self.twiddle8im, x14m15); |
| 5465 |
let t_b14_1 = f64x2_mul(self.twiddle14im, x1m28); |
| 5466 |
let t_b14_2 = f64x2_mul(self.twiddle1im, x2m27); |
| 5467 |
let t_b14_3 = f64x2_mul(self.twiddle13im, x3m26); |
| 5468 |
let t_b14_4 = f64x2_mul(self.twiddle2im, x4m25); |
| 5469 |
let t_b14_5 = f64x2_mul(self.twiddle12im, x5m24); |
| 5470 |
let t_b14_6 = f64x2_mul(self.twiddle3im, x6m23); |
| 5471 |
let t_b14_7 = f64x2_mul(self.twiddle11im, x7m22); |
| 5472 |
let t_b14_8 = f64x2_mul(self.twiddle4im, x8m21); |
| 5473 |
let t_b14_9 = f64x2_mul(self.twiddle10im, x9m20); |
| 5474 |
let t_b14_10 = f64x2_mul(self.twiddle5im, x10m19); |
| 5475 |
let t_b14_11 = f64x2_mul(self.twiddle9im, x11m18); |
| 5476 |
let t_b14_12 = f64x2_mul(self.twiddle6im, x12m17); |
| 5477 |
let t_b14_13 = f64x2_mul(self.twiddle8im, x13m16); |
| 5478 |
let t_b14_14 = f64x2_mul(self.twiddle7im, x14m15); |
| 5479 |
|
| 5480 |
let x0 = values[0]; |
| 5481 |
let t_a1 = calc_f64!( |
| 5482 |
x0 + t_a1_1 |
| 5483 |
+ t_a1_2 |
| 5484 |
+ t_a1_3 |
| 5485 |
+ t_a1_4 |
| 5486 |
+ t_a1_5 |
| 5487 |
+ t_a1_6 |
| 5488 |
+ t_a1_7 |
| 5489 |
+ t_a1_8 |
| 5490 |
+ t_a1_9 |
| 5491 |
+ t_a1_10 |
| 5492 |
+ t_a1_11 |
| 5493 |
+ t_a1_12 |
| 5494 |
+ t_a1_13 |
| 5495 |
+ t_a1_14 |
| 5496 |
); |
| 5497 |
let t_a2 = calc_f64!( |
| 5498 |
x0 + t_a2_1 |
| 5499 |
+ t_a2_2 |
| 5500 |
+ t_a2_3 |
| 5501 |
+ t_a2_4 |
| 5502 |
+ t_a2_5 |
| 5503 |
+ t_a2_6 |
| 5504 |
+ t_a2_7 |
| 5505 |
+ t_a2_8 |
| 5506 |
+ t_a2_9 |
| 5507 |
+ t_a2_10 |
| 5508 |
+ t_a2_11 |
| 5509 |
+ t_a2_12 |
| 5510 |
+ t_a2_13 |
| 5511 |
+ t_a2_14 |
| 5512 |
); |
| 5513 |
let t_a3 = calc_f64!( |
| 5514 |
x0 + t_a3_1 |
| 5515 |
+ t_a3_2 |
| 5516 |
+ t_a3_3 |
| 5517 |
+ t_a3_4 |
| 5518 |
+ t_a3_5 |
| 5519 |
+ t_a3_6 |
| 5520 |
+ t_a3_7 |
| 5521 |
+ t_a3_8 |
| 5522 |
+ t_a3_9 |
| 5523 |
+ t_a3_10 |
| 5524 |
+ t_a3_11 |
| 5525 |
+ t_a3_12 |
| 5526 |
+ t_a3_13 |
| 5527 |
+ t_a3_14 |
| 5528 |
); |
| 5529 |
let t_a4 = calc_f64!( |
| 5530 |
x0 + t_a4_1 |
| 5531 |
+ t_a4_2 |
| 5532 |
+ t_a4_3 |
| 5533 |
+ t_a4_4 |
| 5534 |
+ t_a4_5 |
| 5535 |
+ t_a4_6 |
| 5536 |
+ t_a4_7 |
| 5537 |
+ t_a4_8 |
| 5538 |
+ t_a4_9 |
| 5539 |
+ t_a4_10 |
| 5540 |
+ t_a4_11 |
| 5541 |
+ t_a4_12 |
| 5542 |
+ t_a4_13 |
| 5543 |
+ t_a4_14 |
| 5544 |
); |
| 5545 |
let t_a5 = calc_f64!( |
| 5546 |
x0 + t_a5_1 |
| 5547 |
+ t_a5_2 |
| 5548 |
+ t_a5_3 |
| 5549 |
+ t_a5_4 |
| 5550 |
+ t_a5_5 |
| 5551 |
+ t_a5_6 |
| 5552 |
+ t_a5_7 |
| 5553 |
+ t_a5_8 |
| 5554 |
+ t_a5_9 |
| 5555 |
+ t_a5_10 |
| 5556 |
+ t_a5_11 |
| 5557 |
+ t_a5_12 |
| 5558 |
+ t_a5_13 |
| 5559 |
+ t_a5_14 |
| 5560 |
); |
| 5561 |
let t_a6 = calc_f64!( |
| 5562 |
x0 + t_a6_1 |
| 5563 |
+ t_a6_2 |
| 5564 |
+ t_a6_3 |
| 5565 |
+ t_a6_4 |
| 5566 |
+ t_a6_5 |
| 5567 |
+ t_a6_6 |
| 5568 |
+ t_a6_7 |
| 5569 |
+ t_a6_8 |
| 5570 |
+ t_a6_9 |
| 5571 |
+ t_a6_10 |
| 5572 |
+ t_a6_11 |
| 5573 |
+ t_a6_12 |
| 5574 |
+ t_a6_13 |
| 5575 |
+ t_a6_14 |
| 5576 |
); |
| 5577 |
let t_a7 = calc_f64!( |
| 5578 |
x0 + t_a7_1 |
| 5579 |
+ t_a7_2 |
| 5580 |
+ t_a7_3 |
| 5581 |
+ t_a7_4 |
| 5582 |
+ t_a7_5 |
| 5583 |
+ t_a7_6 |
| 5584 |
+ t_a7_7 |
| 5585 |
+ t_a7_8 |
| 5586 |
+ t_a7_9 |
| 5587 |
+ t_a7_10 |
| 5588 |
+ t_a7_11 |
| 5589 |
+ t_a7_12 |
| 5590 |
+ t_a7_13 |
| 5591 |
+ t_a7_14 |
| 5592 |
); |
| 5593 |
let t_a8 = calc_f64!( |
| 5594 |
x0 + t_a8_1 |
| 5595 |
+ t_a8_2 |
| 5596 |
+ t_a8_3 |
| 5597 |
+ t_a8_4 |
| 5598 |
+ t_a8_5 |
| 5599 |
+ t_a8_6 |
| 5600 |
+ t_a8_7 |
| 5601 |
+ t_a8_8 |
| 5602 |
+ t_a8_9 |
| 5603 |
+ t_a8_10 |
| 5604 |
+ t_a8_11 |
| 5605 |
+ t_a8_12 |
| 5606 |
+ t_a8_13 |
| 5607 |
+ t_a8_14 |
| 5608 |
); |
| 5609 |
let t_a9 = calc_f64!( |
| 5610 |
x0 + t_a9_1 |
| 5611 |
+ t_a9_2 |
| 5612 |
+ t_a9_3 |
| 5613 |
+ t_a9_4 |
| 5614 |
+ t_a9_5 |
| 5615 |
+ t_a9_6 |
| 5616 |
+ t_a9_7 |
| 5617 |
+ t_a9_8 |
| 5618 |
+ t_a9_9 |
| 5619 |
+ t_a9_10 |
| 5620 |
+ t_a9_11 |
| 5621 |
+ t_a9_12 |
| 5622 |
+ t_a9_13 |
| 5623 |
+ t_a9_14 |
| 5624 |
); |
| 5625 |
let t_a10 = calc_f64!( |
| 5626 |
x0 + t_a10_1 |
| 5627 |
+ t_a10_2 |
| 5628 |
+ t_a10_3 |
| 5629 |
+ t_a10_4 |
| 5630 |
+ t_a10_5 |
| 5631 |
+ t_a10_6 |
| 5632 |
+ t_a10_7 |
| 5633 |
+ t_a10_8 |
| 5634 |
+ t_a10_9 |
| 5635 |
+ t_a10_10 |
| 5636 |
+ t_a10_11 |
| 5637 |
+ t_a10_12 |
| 5638 |
+ t_a10_13 |
| 5639 |
+ t_a10_14 |
| 5640 |
); |
| 5641 |
let t_a11 = calc_f64!( |
| 5642 |
x0 + t_a11_1 |
| 5643 |
+ t_a11_2 |
| 5644 |
+ t_a11_3 |
| 5645 |
+ t_a11_4 |
| 5646 |
+ t_a11_5 |
| 5647 |
+ t_a11_6 |
| 5648 |
+ t_a11_7 |
| 5649 |
+ t_a11_8 |
| 5650 |
+ t_a11_9 |
| 5651 |
+ t_a11_10 |
| 5652 |
+ t_a11_11 |
| 5653 |
+ t_a11_12 |
| 5654 |
+ t_a11_13 |
| 5655 |
+ t_a11_14 |
| 5656 |
); |
| 5657 |
let t_a12 = calc_f64!( |
| 5658 |
x0 + t_a12_1 |
| 5659 |
+ t_a12_2 |
| 5660 |
+ t_a12_3 |
| 5661 |
+ t_a12_4 |
| 5662 |
+ t_a12_5 |
| 5663 |
+ t_a12_6 |
| 5664 |
+ t_a12_7 |
| 5665 |
+ t_a12_8 |
| 5666 |
+ t_a12_9 |
| 5667 |
+ t_a12_10 |
| 5668 |
+ t_a12_11 |
| 5669 |
+ t_a12_12 |
| 5670 |
+ t_a12_13 |
| 5671 |
+ t_a12_14 |
| 5672 |
); |
| 5673 |
let t_a13 = calc_f64!( |
| 5674 |
x0 + t_a13_1 |
| 5675 |
+ t_a13_2 |
| 5676 |
+ t_a13_3 |
| 5677 |
+ t_a13_4 |
| 5678 |
+ t_a13_5 |
| 5679 |
+ t_a13_6 |
| 5680 |
+ t_a13_7 |
| 5681 |
+ t_a13_8 |
| 5682 |
+ t_a13_9 |
| 5683 |
+ t_a13_10 |
| 5684 |
+ t_a13_11 |
| 5685 |
+ t_a13_12 |
| 5686 |
+ t_a13_13 |
| 5687 |
+ t_a13_14 |
| 5688 |
); |
| 5689 |
let t_a14 = calc_f64!( |
| 5690 |
x0 + t_a14_1 |
| 5691 |
+ t_a14_2 |
| 5692 |
+ t_a14_3 |
| 5693 |
+ t_a14_4 |
| 5694 |
+ t_a14_5 |
| 5695 |
+ t_a14_6 |
| 5696 |
+ t_a14_7 |
| 5697 |
+ t_a14_8 |
| 5698 |
+ t_a14_9 |
| 5699 |
+ t_a14_10 |
| 5700 |
+ t_a14_11 |
| 5701 |
+ t_a14_12 |
| 5702 |
+ t_a14_13 |
| 5703 |
+ t_a14_14 |
| 5704 |
); |
| 5705 |
|
| 5706 |
let t_b1 = calc_f64!( |
| 5707 |
t_b1_1 |
| 5708 |
+ t_b1_2 |
| 5709 |
+ t_b1_3 |
| 5710 |
+ t_b1_4 |
| 5711 |
+ t_b1_5 |
| 5712 |
+ t_b1_6 |
| 5713 |
+ t_b1_7 |
| 5714 |
+ t_b1_8 |
| 5715 |
+ t_b1_9 |
| 5716 |
+ t_b1_10 |
| 5717 |
+ t_b1_11 |
| 5718 |
+ t_b1_12 |
| 5719 |
+ t_b1_13 |
| 5720 |
+ t_b1_14 |
| 5721 |
); |
| 5722 |
let t_b2 = calc_f64!( |
| 5723 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 + t_b2_6 + t_b2_7 |
| 5724 |
- t_b2_8 |
| 5725 |
- t_b2_9 |
| 5726 |
- t_b2_10 |
| 5727 |
- t_b2_11 |
| 5728 |
- t_b2_12 |
| 5729 |
- t_b2_13 |
| 5730 |
- t_b2_14 |
| 5731 |
); |
| 5732 |
let t_b3 = calc_f64!( |
| 5733 |
t_b3_1 + t_b3_2 + t_b3_3 + t_b3_4 - t_b3_5 - t_b3_6 - t_b3_7 - t_b3_8 - t_b3_9 |
| 5734 |
+ t_b3_10 |
| 5735 |
+ t_b3_11 |
| 5736 |
+ t_b3_12 |
| 5737 |
+ t_b3_13 |
| 5738 |
+ t_b3_14 |
| 5739 |
); |
| 5740 |
let t_b4 = calc_f64!( |
| 5741 |
t_b4_1 + t_b4_2 + t_b4_3 - t_b4_4 - t_b4_5 - t_b4_6 - t_b4_7 |
| 5742 |
+ t_b4_8 |
| 5743 |
+ t_b4_9 |
| 5744 |
+ t_b4_10 |
| 5745 |
- t_b4_11 |
| 5746 |
- t_b4_12 |
| 5747 |
- t_b4_13 |
| 5748 |
- t_b4_14 |
| 5749 |
); |
| 5750 |
let t_b5 = calc_f64!( |
| 5751 |
t_b5_1 + t_b5_2 - t_b5_3 - t_b5_4 - t_b5_5 + t_b5_6 + t_b5_7 + t_b5_8 |
| 5752 |
- t_b5_9 |
| 5753 |
- t_b5_10 |
| 5754 |
- t_b5_11 |
| 5755 |
+ t_b5_12 |
| 5756 |
+ t_b5_13 |
| 5757 |
+ t_b5_14 |
| 5758 |
); |
| 5759 |
let t_b6 = calc_f64!( |
| 5760 |
t_b6_1 + t_b6_2 - t_b6_3 - t_b6_4 + t_b6_5 + t_b6_6 + t_b6_7 - t_b6_8 - t_b6_9 |
| 5761 |
+ t_b6_10 |
| 5762 |
+ t_b6_11 |
| 5763 |
+ t_b6_12 |
| 5764 |
- t_b6_13 |
| 5765 |
- t_b6_14 |
| 5766 |
); |
| 5767 |
let t_b7 = calc_f64!( |
| 5768 |
t_b7_1 + t_b7_2 - t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 |
| 5769 |
+ t_b7_9 |
| 5770 |
+ t_b7_10 |
| 5771 |
- t_b7_11 |
| 5772 |
- t_b7_12 |
| 5773 |
+ t_b7_13 |
| 5774 |
+ t_b7_14 |
| 5775 |
); |
| 5776 |
let t_b8 = calc_f64!( |
| 5777 |
t_b8_1 - t_b8_2 - t_b8_3 + t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 + t_b8_9 |
| 5778 |
- t_b8_10 |
| 5779 |
+ t_b8_11 |
| 5780 |
+ t_b8_12 |
| 5781 |
- t_b8_13 |
| 5782 |
- t_b8_14 |
| 5783 |
); |
| 5784 |
let t_b9 = calc_f64!( |
| 5785 |
t_b9_1 - t_b9_2 - t_b9_3 + t_b9_4 - t_b9_5 - t_b9_6 + t_b9_7 + t_b9_8 - t_b9_9 |
| 5786 |
+ t_b9_10 |
| 5787 |
+ t_b9_11 |
| 5788 |
- t_b9_12 |
| 5789 |
+ t_b9_13 |
| 5790 |
+ t_b9_14 |
| 5791 |
); |
| 5792 |
let t_b10 = calc_f64!( |
| 5793 |
t_b10_1 - t_b10_2 + t_b10_3 + t_b10_4 - t_b10_5 + t_b10_6 + t_b10_7 - t_b10_8 |
| 5794 |
+ t_b10_9 |
| 5795 |
+ t_b10_10 |
| 5796 |
- t_b10_11 |
| 5797 |
+ t_b10_12 |
| 5798 |
+ t_b10_13 |
| 5799 |
- t_b10_14 |
| 5800 |
); |
| 5801 |
let t_b11 = calc_f64!( |
| 5802 |
t_b11_1 - t_b11_2 + t_b11_3 - t_b11_4 - t_b11_5 + t_b11_6 - t_b11_7 + t_b11_8 + t_b11_9 |
| 5803 |
- t_b11_10 |
| 5804 |
+ t_b11_11 |
| 5805 |
- t_b11_12 |
| 5806 |
- t_b11_13 |
| 5807 |
+ t_b11_14 |
| 5808 |
); |
| 5809 |
let t_b12 = calc_f64!( |
| 5810 |
t_b12_1 - t_b12_2 + t_b12_3 - t_b12_4 + t_b12_5 + t_b12_6 - t_b12_7 + t_b12_8 - t_b12_9 |
| 5811 |
+ t_b12_10 |
| 5812 |
- t_b12_11 |
| 5813 |
- t_b12_12 |
| 5814 |
+ t_b12_13 |
| 5815 |
- t_b12_14 |
| 5816 |
); |
| 5817 |
let t_b13 = calc_f64!( |
| 5818 |
t_b13_1 - t_b13_2 + t_b13_3 - t_b13_4 + t_b13_5 - t_b13_6 + t_b13_7 - t_b13_8 |
| 5819 |
+ t_b13_9 |
| 5820 |
+ t_b13_10 |
| 5821 |
- t_b13_11 |
| 5822 |
+ t_b13_12 |
| 5823 |
- t_b13_13 |
| 5824 |
+ t_b13_14 |
| 5825 |
); |
| 5826 |
let t_b14 = calc_f64!( |
| 5827 |
t_b14_1 - t_b14_2 + t_b14_3 - t_b14_4 + t_b14_5 - t_b14_6 + t_b14_7 - t_b14_8 + t_b14_9 |
| 5828 |
- t_b14_10 |
| 5829 |
+ t_b14_11 |
| 5830 |
- t_b14_12 |
| 5831 |
+ t_b14_13 |
| 5832 |
- t_b14_14 |
| 5833 |
); |
| 5834 |
|
| 5835 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 5836 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 5837 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 5838 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 5839 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 5840 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 5841 |
let t_b7_rot = self.rotate.rotate(t_b7); |
| 5842 |
let t_b8_rot = self.rotate.rotate(t_b8); |
| 5843 |
let t_b9_rot = self.rotate.rotate(t_b9); |
| 5844 |
let t_b10_rot = self.rotate.rotate(t_b10); |
| 5845 |
let t_b11_rot = self.rotate.rotate(t_b11); |
| 5846 |
let t_b12_rot = self.rotate.rotate(t_b12); |
| 5847 |
let t_b13_rot = self.rotate.rotate(t_b13); |
| 5848 |
let t_b14_rot = self.rotate.rotate(t_b14); |
| 5849 |
|
| 5850 |
let y0 = calc_f64!( |
| 5851 |
x0 + x1p28 |
| 5852 |
+ x2p27 |
| 5853 |
+ x3p26 |
| 5854 |
+ x4p25 |
| 5855 |
+ x5p24 |
| 5856 |
+ x6p23 |
| 5857 |
+ x7p22 |
| 5858 |
+ x8p21 |
| 5859 |
+ x9p20 |
| 5860 |
+ x10p19 |
| 5861 |
+ x11p18 |
| 5862 |
+ x12p17 |
| 5863 |
+ x13p16 |
| 5864 |
+ x14p15 |
| 5865 |
); |
| 5866 |
let [y1, y28] = solo_fft2_f64(t_a1, t_b1_rot); |
| 5867 |
let [y2, y27] = solo_fft2_f64(t_a2, t_b2_rot); |
| 5868 |
let [y3, y26] = solo_fft2_f64(t_a3, t_b3_rot); |
| 5869 |
let [y4, y25] = solo_fft2_f64(t_a4, t_b4_rot); |
| 5870 |
let [y5, y24] = solo_fft2_f64(t_a5, t_b5_rot); |
| 5871 |
let [y6, y23] = solo_fft2_f64(t_a6, t_b6_rot); |
| 5872 |
let [y7, y22] = solo_fft2_f64(t_a7, t_b7_rot); |
| 5873 |
let [y8, y21] = solo_fft2_f64(t_a8, t_b8_rot); |
| 5874 |
let [y9, y20] = solo_fft2_f64(t_a9, t_b9_rot); |
| 5875 |
let [y10, y19] = solo_fft2_f64(t_a10, t_b10_rot); |
| 5876 |
let [y11, y18] = solo_fft2_f64(t_a11, t_b11_rot); |
| 5877 |
let [y12, y17] = solo_fft2_f64(t_a12, t_b12_rot); |
| 5878 |
let [y13, y16] = solo_fft2_f64(t_a13, t_b13_rot); |
| 5879 |
let [y14, y15] = solo_fft2_f64(t_a14, t_b14_rot); |
| 5880 |
[ |
| 5881 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 5882 |
y19, y20, y21, y22, y23, y24, y25, y26, y27, y28, |
| 5883 |
] |
| 5884 |
} |
| 5885 |
} |
| 5886 |
|
| 5887 |
// _____ _ _________ _ _ _ |
| 5888 |
// |___ // | |___ /___ \| |__ (_) |_ |
| 5889 |
// |_ \| | _____ |_ \ __) | '_ \| | __| |
| 5890 |
// ___) | | |_____| ___) / __/| |_) | | |_ |
| 5891 |
// |____/|_| |____/_____|_.__/|_|\__| |
| 5892 |
// |
| 5893 |
pub struct WasmSimdF32Butterfly31<T> { |
| 5894 |
direction: FftDirection, |
| 5895 |
_phantom: std::marker::PhantomData<T>, |
| 5896 |
rotate: Rotate90F32, |
| 5897 |
twiddle1re: v128, |
| 5898 |
twiddle1im: v128, |
| 5899 |
twiddle2re: v128, |
| 5900 |
twiddle2im: v128, |
| 5901 |
twiddle3re: v128, |
| 5902 |
twiddle3im: v128, |
| 5903 |
twiddle4re: v128, |
| 5904 |
twiddle4im: v128, |
| 5905 |
twiddle5re: v128, |
| 5906 |
twiddle5im: v128, |
| 5907 |
twiddle6re: v128, |
| 5908 |
twiddle6im: v128, |
| 5909 |
twiddle7re: v128, |
| 5910 |
twiddle7im: v128, |
| 5911 |
twiddle8re: v128, |
| 5912 |
twiddle8im: v128, |
| 5913 |
twiddle9re: v128, |
| 5914 |
twiddle9im: v128, |
| 5915 |
twiddle10re: v128, |
| 5916 |
twiddle10im: v128, |
| 5917 |
twiddle11re: v128, |
| 5918 |
twiddle11im: v128, |
| 5919 |
twiddle12re: v128, |
| 5920 |
twiddle12im: v128, |
| 5921 |
twiddle13re: v128, |
| 5922 |
twiddle13im: v128, |
| 5923 |
twiddle14re: v128, |
| 5924 |
twiddle14im: v128, |
| 5925 |
twiddle15re: v128, |
| 5926 |
twiddle15im: v128, |
| 5927 |
} |
| 5928 |
|
| 5929 |
boilerplate_fft_wasm_simd_f32_butterfly!( |
| 5930 |
WasmSimdF32Butterfly31, |
| 5931 |
31, |
| 5932 |
|this: &WasmSimdF32Butterfly31<_>| this.direction |
| 5933 |
); |
| 5934 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 5935 |
WasmSimdF32Butterfly31, |
| 5936 |
31, |
| 5937 |
|this: &WasmSimdF32Butterfly31<_>| this.direction |
| 5938 |
); |
| 5939 |
impl<T: FftNum> WasmSimdF32Butterfly31<T> { |
| 5940 |
#[inline(always)] |
| 5941 |
pub fn new(direction: FftDirection) -> Self { |
| 5942 |
assert_f32::<T>(); |
| 5943 |
let rotate = Rotate90F32::new(true); |
| 5944 |
let tw1: Complex<f32> = twiddles::compute_twiddle(1, 31, direction); |
| 5945 |
let tw2: Complex<f32> = twiddles::compute_twiddle(2, 31, direction); |
| 5946 |
let tw3: Complex<f32> = twiddles::compute_twiddle(3, 31, direction); |
| 5947 |
let tw4: Complex<f32> = twiddles::compute_twiddle(4, 31, direction); |
| 5948 |
let tw5: Complex<f32> = twiddles::compute_twiddle(5, 31, direction); |
| 5949 |
let tw6: Complex<f32> = twiddles::compute_twiddle(6, 31, direction); |
| 5950 |
let tw7: Complex<f32> = twiddles::compute_twiddle(7, 31, direction); |
| 5951 |
let tw8: Complex<f32> = twiddles::compute_twiddle(8, 31, direction); |
| 5952 |
let tw9: Complex<f32> = twiddles::compute_twiddle(9, 31, direction); |
| 5953 |
let tw10: Complex<f32> = twiddles::compute_twiddle(10, 31, direction); |
| 5954 |
let tw11: Complex<f32> = twiddles::compute_twiddle(11, 31, direction); |
| 5955 |
let tw12: Complex<f32> = twiddles::compute_twiddle(12, 31, direction); |
| 5956 |
let tw13: Complex<f32> = twiddles::compute_twiddle(13, 31, direction); |
| 5957 |
let tw14: Complex<f32> = twiddles::compute_twiddle(14, 31, direction); |
| 5958 |
let tw15: Complex<f32> = twiddles::compute_twiddle(15, 31, direction); |
| 5959 |
let twiddle1re = f32x4_splat(tw1.re); |
| 5960 |
let twiddle1im = f32x4_splat(tw1.im); |
| 5961 |
let twiddle2re = f32x4_splat(tw2.re); |
| 5962 |
let twiddle2im = f32x4_splat(tw2.im); |
| 5963 |
let twiddle3re = f32x4_splat(tw3.re); |
| 5964 |
let twiddle3im = f32x4_splat(tw3.im); |
| 5965 |
let twiddle4re = f32x4_splat(tw4.re); |
| 5966 |
let twiddle4im = f32x4_splat(tw4.im); |
| 5967 |
let twiddle5re = f32x4_splat(tw5.re); |
| 5968 |
let twiddle5im = f32x4_splat(tw5.im); |
| 5969 |
let twiddle6re = f32x4_splat(tw6.re); |
| 5970 |
let twiddle6im = f32x4_splat(tw6.im); |
| 5971 |
let twiddle7re = f32x4_splat(tw7.re); |
| 5972 |
let twiddle7im = f32x4_splat(tw7.im); |
| 5973 |
let twiddle8re = f32x4_splat(tw8.re); |
| 5974 |
let twiddle8im = f32x4_splat(tw8.im); |
| 5975 |
let twiddle9re = f32x4_splat(tw9.re); |
| 5976 |
let twiddle9im = f32x4_splat(tw9.im); |
| 5977 |
let twiddle10re = f32x4_splat(tw10.re); |
| 5978 |
let twiddle10im = f32x4_splat(tw10.im); |
| 5979 |
let twiddle11re = f32x4_splat(tw11.re); |
| 5980 |
let twiddle11im = f32x4_splat(tw11.im); |
| 5981 |
let twiddle12re = f32x4_splat(tw12.re); |
| 5982 |
let twiddle12im = f32x4_splat(tw12.im); |
| 5983 |
let twiddle13re = f32x4_splat(tw13.re); |
| 5984 |
let twiddle13im = f32x4_splat(tw13.im); |
| 5985 |
let twiddle14re = f32x4_splat(tw14.re); |
| 5986 |
let twiddle14im = f32x4_splat(tw14.im); |
| 5987 |
let twiddle15re = f32x4_splat(tw15.re); |
| 5988 |
let twiddle15im = f32x4_splat(tw15.im); |
| 5989 |
|
| 5990 |
Self { |
| 5991 |
direction, |
| 5992 |
_phantom: std::marker::PhantomData, |
| 5993 |
rotate, |
| 5994 |
twiddle1re, |
| 5995 |
twiddle1im, |
| 5996 |
twiddle2re, |
| 5997 |
twiddle2im, |
| 5998 |
twiddle3re, |
| 5999 |
twiddle3im, |
| 6000 |
twiddle4re, |
| 6001 |
twiddle4im, |
| 6002 |
twiddle5re, |
| 6003 |
twiddle5im, |
| 6004 |
twiddle6re, |
| 6005 |
twiddle6im, |
| 6006 |
twiddle7re, |
| 6007 |
twiddle7im, |
| 6008 |
twiddle8re, |
| 6009 |
twiddle8im, |
| 6010 |
twiddle9re, |
| 6011 |
twiddle9im, |
| 6012 |
twiddle10re, |
| 6013 |
twiddle10im, |
| 6014 |
twiddle11re, |
| 6015 |
twiddle11im, |
| 6016 |
twiddle12re, |
| 6017 |
twiddle12im, |
| 6018 |
twiddle13re, |
| 6019 |
twiddle13im, |
| 6020 |
twiddle14re, |
| 6021 |
twiddle14im, |
| 6022 |
twiddle15re, |
| 6023 |
twiddle15im, |
| 6024 |
} |
| 6025 |
} |
| 6026 |
|
| 6027 |
#[inline(always)] |
| 6028 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f32>) { |
| 6029 |
let values = read_partial1_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30}); |
| 6030 |
|
| 6031 |
let out = self.perform_parallel_fft_direct(values); |
| 6032 |
|
| 6033 |
write_partial_lo_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30}); |
| 6034 |
} |
| 6035 |
|
| 6036 |
#[inline(always)] |
| 6037 |
pub(crate) unsafe fn perform_parallel_fft_contiguous( |
| 6038 |
&self, |
| 6039 |
mut buffer: impl WasmSimdArrayMut<f32>, |
| 6040 |
) { |
| 6041 |
let input_packed = read_complex_to_array!(buffer, {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, 50, 52, 54, 56, 58, 60}); |
| 6042 |
|
| 6043 |
let values = [ |
| 6044 |
extract_lo_hi_f32(input_packed[0], input_packed[15]), |
| 6045 |
extract_hi_lo_f32(input_packed[0], input_packed[16]), |
| 6046 |
extract_lo_hi_f32(input_packed[1], input_packed[16]), |
| 6047 |
extract_hi_lo_f32(input_packed[1], input_packed[17]), |
| 6048 |
extract_lo_hi_f32(input_packed[2], input_packed[17]), |
| 6049 |
extract_hi_lo_f32(input_packed[2], input_packed[18]), |
| 6050 |
extract_lo_hi_f32(input_packed[3], input_packed[18]), |
| 6051 |
extract_hi_lo_f32(input_packed[3], input_packed[19]), |
| 6052 |
extract_lo_hi_f32(input_packed[4], input_packed[19]), |
| 6053 |
extract_hi_lo_f32(input_packed[4], input_packed[20]), |
| 6054 |
extract_lo_hi_f32(input_packed[5], input_packed[20]), |
| 6055 |
extract_hi_lo_f32(input_packed[5], input_packed[21]), |
| 6056 |
extract_lo_hi_f32(input_packed[6], input_packed[21]), |
| 6057 |
extract_hi_lo_f32(input_packed[6], input_packed[22]), |
| 6058 |
extract_lo_hi_f32(input_packed[7], input_packed[22]), |
| 6059 |
extract_hi_lo_f32(input_packed[7], input_packed[23]), |
| 6060 |
extract_lo_hi_f32(input_packed[8], input_packed[23]), |
| 6061 |
extract_hi_lo_f32(input_packed[8], input_packed[24]), |
| 6062 |
extract_lo_hi_f32(input_packed[9], input_packed[24]), |
| 6063 |
extract_hi_lo_f32(input_packed[9], input_packed[25]), |
| 6064 |
extract_lo_hi_f32(input_packed[10], input_packed[25]), |
| 6065 |
extract_hi_lo_f32(input_packed[10], input_packed[26]), |
| 6066 |
extract_lo_hi_f32(input_packed[11], input_packed[26]), |
| 6067 |
extract_hi_lo_f32(input_packed[11], input_packed[27]), |
| 6068 |
extract_lo_hi_f32(input_packed[12], input_packed[27]), |
| 6069 |
extract_hi_lo_f32(input_packed[12], input_packed[28]), |
| 6070 |
extract_lo_hi_f32(input_packed[13], input_packed[28]), |
| 6071 |
extract_hi_lo_f32(input_packed[13], input_packed[29]), |
| 6072 |
extract_lo_hi_f32(input_packed[14], input_packed[29]), |
| 6073 |
extract_hi_lo_f32(input_packed[14], input_packed[30]), |
| 6074 |
extract_lo_hi_f32(input_packed[15], input_packed[30]), |
| 6075 |
]; |
| 6076 |
|
| 6077 |
let out = self.perform_parallel_fft_direct(values); |
| 6078 |
|
| 6079 |
let out_packed = [ |
| 6080 |
extract_lo_lo_f32(out[0], out[1]), |
| 6081 |
extract_lo_lo_f32(out[2], out[3]), |
| 6082 |
extract_lo_lo_f32(out[4], out[5]), |
| 6083 |
extract_lo_lo_f32(out[6], out[7]), |
| 6084 |
extract_lo_lo_f32(out[8], out[9]), |
| 6085 |
extract_lo_lo_f32(out[10], out[11]), |
| 6086 |
extract_lo_lo_f32(out[12], out[13]), |
| 6087 |
extract_lo_lo_f32(out[14], out[15]), |
| 6088 |
extract_lo_lo_f32(out[16], out[17]), |
| 6089 |
extract_lo_lo_f32(out[18], out[19]), |
| 6090 |
extract_lo_lo_f32(out[20], out[21]), |
| 6091 |
extract_lo_lo_f32(out[22], out[23]), |
| 6092 |
extract_lo_lo_f32(out[24], out[25]), |
| 6093 |
extract_lo_lo_f32(out[26], out[27]), |
| 6094 |
extract_lo_lo_f32(out[28], out[29]), |
| 6095 |
extract_lo_hi_f32(out[30], out[0]), |
| 6096 |
extract_hi_hi_f32(out[1], out[2]), |
| 6097 |
extract_hi_hi_f32(out[3], out[4]), |
| 6098 |
extract_hi_hi_f32(out[5], out[6]), |
| 6099 |
extract_hi_hi_f32(out[7], out[8]), |
| 6100 |
extract_hi_hi_f32(out[9], out[10]), |
| 6101 |
extract_hi_hi_f32(out[11], out[12]), |
| 6102 |
extract_hi_hi_f32(out[13], out[14]), |
| 6103 |
extract_hi_hi_f32(out[15], out[16]), |
| 6104 |
extract_hi_hi_f32(out[17], out[18]), |
| 6105 |
extract_hi_hi_f32(out[19], out[20]), |
| 6106 |
extract_hi_hi_f32(out[21], out[22]), |
| 6107 |
extract_hi_hi_f32(out[23], out[24]), |
| 6108 |
extract_hi_hi_f32(out[25], out[26]), |
| 6109 |
extract_hi_hi_f32(out[27], out[28]), |
| 6110 |
extract_hi_hi_f32(out[29], out[30]), |
| 6111 |
]; |
| 6112 |
|
| 6113 |
write_complex_to_array_strided!(out_packed, buffer, 2, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30}); |
| 6114 |
} |
| 6115 |
|
| 6116 |
#[inline(always)] |
| 6117 |
pub(crate) unsafe fn perform_parallel_fft_direct(&self, values: [v128; 31]) -> [v128; 31] { |
| 6118 |
let [x1p30, x1m30] = parallel_fft2_interleaved_f32(values[1], values[30]); |
| 6119 |
let [x2p29, x2m29] = parallel_fft2_interleaved_f32(values[2], values[29]); |
| 6120 |
let [x3p28, x3m28] = parallel_fft2_interleaved_f32(values[3], values[28]); |
| 6121 |
let [x4p27, x4m27] = parallel_fft2_interleaved_f32(values[4], values[27]); |
| 6122 |
let [x5p26, x5m26] = parallel_fft2_interleaved_f32(values[5], values[26]); |
| 6123 |
let [x6p25, x6m25] = parallel_fft2_interleaved_f32(values[6], values[25]); |
| 6124 |
let [x7p24, x7m24] = parallel_fft2_interleaved_f32(values[7], values[24]); |
| 6125 |
let [x8p23, x8m23] = parallel_fft2_interleaved_f32(values[8], values[23]); |
| 6126 |
let [x9p22, x9m22] = parallel_fft2_interleaved_f32(values[9], values[22]); |
| 6127 |
let [x10p21, x10m21] = parallel_fft2_interleaved_f32(values[10], values[21]); |
| 6128 |
let [x11p20, x11m20] = parallel_fft2_interleaved_f32(values[11], values[20]); |
| 6129 |
let [x12p19, x12m19] = parallel_fft2_interleaved_f32(values[12], values[19]); |
| 6130 |
let [x13p18, x13m18] = parallel_fft2_interleaved_f32(values[13], values[18]); |
| 6131 |
let [x14p17, x14m17] = parallel_fft2_interleaved_f32(values[14], values[17]); |
| 6132 |
let [x15p16, x15m16] = parallel_fft2_interleaved_f32(values[15], values[16]); |
| 6133 |
|
| 6134 |
let t_a1_1 = f32x4_mul(self.twiddle1re, x1p30); |
| 6135 |
let t_a1_2 = f32x4_mul(self.twiddle2re, x2p29); |
| 6136 |
let t_a1_3 = f32x4_mul(self.twiddle3re, x3p28); |
| 6137 |
let t_a1_4 = f32x4_mul(self.twiddle4re, x4p27); |
| 6138 |
let t_a1_5 = f32x4_mul(self.twiddle5re, x5p26); |
| 6139 |
let t_a1_6 = f32x4_mul(self.twiddle6re, x6p25); |
| 6140 |
let t_a1_7 = f32x4_mul(self.twiddle7re, x7p24); |
| 6141 |
let t_a1_8 = f32x4_mul(self.twiddle8re, x8p23); |
| 6142 |
let t_a1_9 = f32x4_mul(self.twiddle9re, x9p22); |
| 6143 |
let t_a1_10 = f32x4_mul(self.twiddle10re, x10p21); |
| 6144 |
let t_a1_11 = f32x4_mul(self.twiddle11re, x11p20); |
| 6145 |
let t_a1_12 = f32x4_mul(self.twiddle12re, x12p19); |
| 6146 |
let t_a1_13 = f32x4_mul(self.twiddle13re, x13p18); |
| 6147 |
let t_a1_14 = f32x4_mul(self.twiddle14re, x14p17); |
| 6148 |
let t_a1_15 = f32x4_mul(self.twiddle15re, x15p16); |
| 6149 |
let t_a2_1 = f32x4_mul(self.twiddle2re, x1p30); |
| 6150 |
let t_a2_2 = f32x4_mul(self.twiddle4re, x2p29); |
| 6151 |
let t_a2_3 = f32x4_mul(self.twiddle6re, x3p28); |
| 6152 |
let t_a2_4 = f32x4_mul(self.twiddle8re, x4p27); |
| 6153 |
let t_a2_5 = f32x4_mul(self.twiddle10re, x5p26); |
| 6154 |
let t_a2_6 = f32x4_mul(self.twiddle12re, x6p25); |
| 6155 |
let t_a2_7 = f32x4_mul(self.twiddle14re, x7p24); |
| 6156 |
let t_a2_8 = f32x4_mul(self.twiddle15re, x8p23); |
| 6157 |
let t_a2_9 = f32x4_mul(self.twiddle13re, x9p22); |
| 6158 |
let t_a2_10 = f32x4_mul(self.twiddle11re, x10p21); |
| 6159 |
let t_a2_11 = f32x4_mul(self.twiddle9re, x11p20); |
| 6160 |
let t_a2_12 = f32x4_mul(self.twiddle7re, x12p19); |
| 6161 |
let t_a2_13 = f32x4_mul(self.twiddle5re, x13p18); |
| 6162 |
let t_a2_14 = f32x4_mul(self.twiddle3re, x14p17); |
| 6163 |
let t_a2_15 = f32x4_mul(self.twiddle1re, x15p16); |
| 6164 |
let t_a3_1 = f32x4_mul(self.twiddle3re, x1p30); |
| 6165 |
let t_a3_2 = f32x4_mul(self.twiddle6re, x2p29); |
| 6166 |
let t_a3_3 = f32x4_mul(self.twiddle9re, x3p28); |
| 6167 |
let t_a3_4 = f32x4_mul(self.twiddle12re, x4p27); |
| 6168 |
let t_a3_5 = f32x4_mul(self.twiddle15re, x5p26); |
| 6169 |
let t_a3_6 = f32x4_mul(self.twiddle13re, x6p25); |
| 6170 |
let t_a3_7 = f32x4_mul(self.twiddle10re, x7p24); |
| 6171 |
let t_a3_8 = f32x4_mul(self.twiddle7re, x8p23); |
| 6172 |
let t_a3_9 = f32x4_mul(self.twiddle4re, x9p22); |
| 6173 |
let t_a3_10 = f32x4_mul(self.twiddle1re, x10p21); |
| 6174 |
let t_a3_11 = f32x4_mul(self.twiddle2re, x11p20); |
| 6175 |
let t_a3_12 = f32x4_mul(self.twiddle5re, x12p19); |
| 6176 |
let t_a3_13 = f32x4_mul(self.twiddle8re, x13p18); |
| 6177 |
let t_a3_14 = f32x4_mul(self.twiddle11re, x14p17); |
| 6178 |
let t_a3_15 = f32x4_mul(self.twiddle14re, x15p16); |
| 6179 |
let t_a4_1 = f32x4_mul(self.twiddle4re, x1p30); |
| 6180 |
let t_a4_2 = f32x4_mul(self.twiddle8re, x2p29); |
| 6181 |
let t_a4_3 = f32x4_mul(self.twiddle12re, x3p28); |
| 6182 |
let t_a4_4 = f32x4_mul(self.twiddle15re, x4p27); |
| 6183 |
let t_a4_5 = f32x4_mul(self.twiddle11re, x5p26); |
| 6184 |
let t_a4_6 = f32x4_mul(self.twiddle7re, x6p25); |
| 6185 |
let t_a4_7 = f32x4_mul(self.twiddle3re, x7p24); |
| 6186 |
let t_a4_8 = f32x4_mul(self.twiddle1re, x8p23); |
| 6187 |
let t_a4_9 = f32x4_mul(self.twiddle5re, x9p22); |
| 6188 |
let t_a4_10 = f32x4_mul(self.twiddle9re, x10p21); |
| 6189 |
let t_a4_11 = f32x4_mul(self.twiddle13re, x11p20); |
| 6190 |
let t_a4_12 = f32x4_mul(self.twiddle14re, x12p19); |
| 6191 |
let t_a4_13 = f32x4_mul(self.twiddle10re, x13p18); |
| 6192 |
let t_a4_14 = f32x4_mul(self.twiddle6re, x14p17); |
| 6193 |
let t_a4_15 = f32x4_mul(self.twiddle2re, x15p16); |
| 6194 |
let t_a5_1 = f32x4_mul(self.twiddle5re, x1p30); |
| 6195 |
let t_a5_2 = f32x4_mul(self.twiddle10re, x2p29); |
| 6196 |
let t_a5_3 = f32x4_mul(self.twiddle15re, x3p28); |
| 6197 |
let t_a5_4 = f32x4_mul(self.twiddle11re, x4p27); |
| 6198 |
let t_a5_5 = f32x4_mul(self.twiddle6re, x5p26); |
| 6199 |
let t_a5_6 = f32x4_mul(self.twiddle1re, x6p25); |
| 6200 |
let t_a5_7 = f32x4_mul(self.twiddle4re, x7p24); |
| 6201 |
let t_a5_8 = f32x4_mul(self.twiddle9re, x8p23); |
| 6202 |
let t_a5_9 = f32x4_mul(self.twiddle14re, x9p22); |
| 6203 |
let t_a5_10 = f32x4_mul(self.twiddle12re, x10p21); |
| 6204 |
let t_a5_11 = f32x4_mul(self.twiddle7re, x11p20); |
| 6205 |
let t_a5_12 = f32x4_mul(self.twiddle2re, x12p19); |
| 6206 |
let t_a5_13 = f32x4_mul(self.twiddle3re, x13p18); |
| 6207 |
let t_a5_14 = f32x4_mul(self.twiddle8re, x14p17); |
| 6208 |
let t_a5_15 = f32x4_mul(self.twiddle13re, x15p16); |
| 6209 |
let t_a6_1 = f32x4_mul(self.twiddle6re, x1p30); |
| 6210 |
let t_a6_2 = f32x4_mul(self.twiddle12re, x2p29); |
| 6211 |
let t_a6_3 = f32x4_mul(self.twiddle13re, x3p28); |
| 6212 |
let t_a6_4 = f32x4_mul(self.twiddle7re, x4p27); |
| 6213 |
let t_a6_5 = f32x4_mul(self.twiddle1re, x5p26); |
| 6214 |
let t_a6_6 = f32x4_mul(self.twiddle5re, x6p25); |
| 6215 |
let t_a6_7 = f32x4_mul(self.twiddle11re, x7p24); |
| 6216 |
let t_a6_8 = f32x4_mul(self.twiddle14re, x8p23); |
| 6217 |
let t_a6_9 = f32x4_mul(self.twiddle8re, x9p22); |
| 6218 |
let t_a6_10 = f32x4_mul(self.twiddle2re, x10p21); |
| 6219 |
let t_a6_11 = f32x4_mul(self.twiddle4re, x11p20); |
| 6220 |
let t_a6_12 = f32x4_mul(self.twiddle10re, x12p19); |
| 6221 |
let t_a6_13 = f32x4_mul(self.twiddle15re, x13p18); |
| 6222 |
let t_a6_14 = f32x4_mul(self.twiddle9re, x14p17); |
| 6223 |
let t_a6_15 = f32x4_mul(self.twiddle3re, x15p16); |
| 6224 |
let t_a7_1 = f32x4_mul(self.twiddle7re, x1p30); |
| 6225 |
let t_a7_2 = f32x4_mul(self.twiddle14re, x2p29); |
| 6226 |
let t_a7_3 = f32x4_mul(self.twiddle10re, x3p28); |
| 6227 |
let t_a7_4 = f32x4_mul(self.twiddle3re, x4p27); |
| 6228 |
let t_a7_5 = f32x4_mul(self.twiddle4re, x5p26); |
| 6229 |
let t_a7_6 = f32x4_mul(self.twiddle11re, x6p25); |
| 6230 |
let t_a7_7 = f32x4_mul(self.twiddle13re, x7p24); |
| 6231 |
let t_a7_8 = f32x4_mul(self.twiddle6re, x8p23); |
| 6232 |
let t_a7_9 = f32x4_mul(self.twiddle1re, x9p22); |
| 6233 |
let t_a7_10 = f32x4_mul(self.twiddle8re, x10p21); |
| 6234 |
let t_a7_11 = f32x4_mul(self.twiddle15re, x11p20); |
| 6235 |
let t_a7_12 = f32x4_mul(self.twiddle9re, x12p19); |
| 6236 |
let t_a7_13 = f32x4_mul(self.twiddle2re, x13p18); |
| 6237 |
let t_a7_14 = f32x4_mul(self.twiddle5re, x14p17); |
| 6238 |
let t_a7_15 = f32x4_mul(self.twiddle12re, x15p16); |
| 6239 |
let t_a8_1 = f32x4_mul(self.twiddle8re, x1p30); |
| 6240 |
let t_a8_2 = f32x4_mul(self.twiddle15re, x2p29); |
| 6241 |
let t_a8_3 = f32x4_mul(self.twiddle7re, x3p28); |
| 6242 |
let t_a8_4 = f32x4_mul(self.twiddle1re, x4p27); |
| 6243 |
let t_a8_5 = f32x4_mul(self.twiddle9re, x5p26); |
| 6244 |
let t_a8_6 = f32x4_mul(self.twiddle14re, x6p25); |
| 6245 |
let t_a8_7 = f32x4_mul(self.twiddle6re, x7p24); |
| 6246 |
let t_a8_8 = f32x4_mul(self.twiddle2re, x8p23); |
| 6247 |
let t_a8_9 = f32x4_mul(self.twiddle10re, x9p22); |
| 6248 |
let t_a8_10 = f32x4_mul(self.twiddle13re, x10p21); |
| 6249 |
let t_a8_11 = f32x4_mul(self.twiddle5re, x11p20); |
| 6250 |
let t_a8_12 = f32x4_mul(self.twiddle3re, x12p19); |
| 6251 |
let t_a8_13 = f32x4_mul(self.twiddle11re, x13p18); |
| 6252 |
let t_a8_14 = f32x4_mul(self.twiddle12re, x14p17); |
| 6253 |
let t_a8_15 = f32x4_mul(self.twiddle4re, x15p16); |
| 6254 |
let t_a9_1 = f32x4_mul(self.twiddle9re, x1p30); |
| 6255 |
let t_a9_2 = f32x4_mul(self.twiddle13re, x2p29); |
| 6256 |
let t_a9_3 = f32x4_mul(self.twiddle4re, x3p28); |
| 6257 |
let t_a9_4 = f32x4_mul(self.twiddle5re, x4p27); |
| 6258 |
let t_a9_5 = f32x4_mul(self.twiddle14re, x5p26); |
| 6259 |
let t_a9_6 = f32x4_mul(self.twiddle8re, x6p25); |
| 6260 |
let t_a9_7 = f32x4_mul(self.twiddle1re, x7p24); |
| 6261 |
let t_a9_8 = f32x4_mul(self.twiddle10re, x8p23); |
| 6262 |
let t_a9_9 = f32x4_mul(self.twiddle12re, x9p22); |
| 6263 |
let t_a9_10 = f32x4_mul(self.twiddle3re, x10p21); |
| 6264 |
let t_a9_11 = f32x4_mul(self.twiddle6re, x11p20); |
| 6265 |
let t_a9_12 = f32x4_mul(self.twiddle15re, x12p19); |
| 6266 |
let t_a9_13 = f32x4_mul(self.twiddle7re, x13p18); |
| 6267 |
let t_a9_14 = f32x4_mul(self.twiddle2re, x14p17); |
| 6268 |
let t_a9_15 = f32x4_mul(self.twiddle11re, x15p16); |
| 6269 |
let t_a10_1 = f32x4_mul(self.twiddle10re, x1p30); |
| 6270 |
let t_a10_2 = f32x4_mul(self.twiddle11re, x2p29); |
| 6271 |
let t_a10_3 = f32x4_mul(self.twiddle1re, x3p28); |
| 6272 |
let t_a10_4 = f32x4_mul(self.twiddle9re, x4p27); |
| 6273 |
let t_a10_5 = f32x4_mul(self.twiddle12re, x5p26); |
| 6274 |
let t_a10_6 = f32x4_mul(self.twiddle2re, x6p25); |
| 6275 |
let t_a10_7 = f32x4_mul(self.twiddle8re, x7p24); |
| 6276 |
let t_a10_8 = f32x4_mul(self.twiddle13re, x8p23); |
| 6277 |
let t_a10_9 = f32x4_mul(self.twiddle3re, x9p22); |
| 6278 |
let t_a10_10 = f32x4_mul(self.twiddle7re, x10p21); |
| 6279 |
let t_a10_11 = f32x4_mul(self.twiddle14re, x11p20); |
| 6280 |
let t_a10_12 = f32x4_mul(self.twiddle4re, x12p19); |
| 6281 |
let t_a10_13 = f32x4_mul(self.twiddle6re, x13p18); |
| 6282 |
let t_a10_14 = f32x4_mul(self.twiddle15re, x14p17); |
| 6283 |
let t_a10_15 = f32x4_mul(self.twiddle5re, x15p16); |
| 6284 |
let t_a11_1 = f32x4_mul(self.twiddle11re, x1p30); |
| 6285 |
let t_a11_2 = f32x4_mul(self.twiddle9re, x2p29); |
| 6286 |
let t_a11_3 = f32x4_mul(self.twiddle2re, x3p28); |
| 6287 |
let t_a11_4 = f32x4_mul(self.twiddle13re, x4p27); |
| 6288 |
let t_a11_5 = f32x4_mul(self.twiddle7re, x5p26); |
| 6289 |
let t_a11_6 = f32x4_mul(self.twiddle4re, x6p25); |
| 6290 |
let t_a11_7 = f32x4_mul(self.twiddle15re, x7p24); |
| 6291 |
let t_a11_8 = f32x4_mul(self.twiddle5re, x8p23); |
| 6292 |
let t_a11_9 = f32x4_mul(self.twiddle6re, x9p22); |
| 6293 |
let t_a11_10 = f32x4_mul(self.twiddle14re, x10p21); |
| 6294 |
let t_a11_11 = f32x4_mul(self.twiddle3re, x11p20); |
| 6295 |
let t_a11_12 = f32x4_mul(self.twiddle8re, x12p19); |
| 6296 |
let t_a11_13 = f32x4_mul(self.twiddle12re, x13p18); |
| 6297 |
let t_a11_14 = f32x4_mul(self.twiddle1re, x14p17); |
| 6298 |
let t_a11_15 = f32x4_mul(self.twiddle10re, x15p16); |
| 6299 |
let t_a12_1 = f32x4_mul(self.twiddle12re, x1p30); |
| 6300 |
let t_a12_2 = f32x4_mul(self.twiddle7re, x2p29); |
| 6301 |
let t_a12_3 = f32x4_mul(self.twiddle5re, x3p28); |
| 6302 |
let t_a12_4 = f32x4_mul(self.twiddle14re, x4p27); |
| 6303 |
let t_a12_5 = f32x4_mul(self.twiddle2re, x5p26); |
| 6304 |
let t_a12_6 = f32x4_mul(self.twiddle10re, x6p25); |
| 6305 |
let t_a12_7 = f32x4_mul(self.twiddle9re, x7p24); |
| 6306 |
let t_a12_8 = f32x4_mul(self.twiddle3re, x8p23); |
| 6307 |
let t_a12_9 = f32x4_mul(self.twiddle15re, x9p22); |
| 6308 |
let t_a12_10 = f32x4_mul(self.twiddle4re, x10p21); |
| 6309 |
let t_a12_11 = f32x4_mul(self.twiddle8re, x11p20); |
| 6310 |
let t_a12_12 = f32x4_mul(self.twiddle11re, x12p19); |
| 6311 |
let t_a12_13 = f32x4_mul(self.twiddle1re, x13p18); |
| 6312 |
let t_a12_14 = f32x4_mul(self.twiddle13re, x14p17); |
| 6313 |
let t_a12_15 = f32x4_mul(self.twiddle6re, x15p16); |
| 6314 |
let t_a13_1 = f32x4_mul(self.twiddle13re, x1p30); |
| 6315 |
let t_a13_2 = f32x4_mul(self.twiddle5re, x2p29); |
| 6316 |
let t_a13_3 = f32x4_mul(self.twiddle8re, x3p28); |
| 6317 |
let t_a13_4 = f32x4_mul(self.twiddle10re, x4p27); |
| 6318 |
let t_a13_5 = f32x4_mul(self.twiddle3re, x5p26); |
| 6319 |
let t_a13_6 = f32x4_mul(self.twiddle15re, x6p25); |
| 6320 |
let t_a13_7 = f32x4_mul(self.twiddle2re, x7p24); |
| 6321 |
let t_a13_8 = f32x4_mul(self.twiddle11re, x8p23); |
| 6322 |
let t_a13_9 = f32x4_mul(self.twiddle7re, x9p22); |
| 6323 |
let t_a13_10 = f32x4_mul(self.twiddle6re, x10p21); |
| 6324 |
let t_a13_11 = f32x4_mul(self.twiddle12re, x11p20); |
| 6325 |
let t_a13_12 = f32x4_mul(self.twiddle1re, x12p19); |
| 6326 |
let t_a13_13 = f32x4_mul(self.twiddle14re, x13p18); |
| 6327 |
let t_a13_14 = f32x4_mul(self.twiddle4re, x14p17); |
| 6328 |
let t_a13_15 = f32x4_mul(self.twiddle9re, x15p16); |
| 6329 |
let t_a14_1 = f32x4_mul(self.twiddle14re, x1p30); |
| 6330 |
let t_a14_2 = f32x4_mul(self.twiddle3re, x2p29); |
| 6331 |
let t_a14_3 = f32x4_mul(self.twiddle11re, x3p28); |
| 6332 |
let t_a14_4 = f32x4_mul(self.twiddle6re, x4p27); |
| 6333 |
let t_a14_5 = f32x4_mul(self.twiddle8re, x5p26); |
| 6334 |
let t_a14_6 = f32x4_mul(self.twiddle9re, x6p25); |
| 6335 |
let t_a14_7 = f32x4_mul(self.twiddle5re, x7p24); |
| 6336 |
let t_a14_8 = f32x4_mul(self.twiddle12re, x8p23); |
| 6337 |
let t_a14_9 = f32x4_mul(self.twiddle2re, x9p22); |
| 6338 |
let t_a14_10 = f32x4_mul(self.twiddle15re, x10p21); |
| 6339 |
let t_a14_11 = f32x4_mul(self.twiddle1re, x11p20); |
| 6340 |
let t_a14_12 = f32x4_mul(self.twiddle13re, x12p19); |
| 6341 |
let t_a14_13 = f32x4_mul(self.twiddle4re, x13p18); |
| 6342 |
let t_a14_14 = f32x4_mul(self.twiddle10re, x14p17); |
| 6343 |
let t_a14_15 = f32x4_mul(self.twiddle7re, x15p16); |
| 6344 |
let t_a15_1 = f32x4_mul(self.twiddle15re, x1p30); |
| 6345 |
let t_a15_2 = f32x4_mul(self.twiddle1re, x2p29); |
| 6346 |
let t_a15_3 = f32x4_mul(self.twiddle14re, x3p28); |
| 6347 |
let t_a15_4 = f32x4_mul(self.twiddle2re, x4p27); |
| 6348 |
let t_a15_5 = f32x4_mul(self.twiddle13re, x5p26); |
| 6349 |
let t_a15_6 = f32x4_mul(self.twiddle3re, x6p25); |
| 6350 |
let t_a15_7 = f32x4_mul(self.twiddle12re, x7p24); |
| 6351 |
let t_a15_8 = f32x4_mul(self.twiddle4re, x8p23); |
| 6352 |
let t_a15_9 = f32x4_mul(self.twiddle11re, x9p22); |
| 6353 |
let t_a15_10 = f32x4_mul(self.twiddle5re, x10p21); |
| 6354 |
let t_a15_11 = f32x4_mul(self.twiddle10re, x11p20); |
| 6355 |
let t_a15_12 = f32x4_mul(self.twiddle6re, x12p19); |
| 6356 |
let t_a15_13 = f32x4_mul(self.twiddle9re, x13p18); |
| 6357 |
let t_a15_14 = f32x4_mul(self.twiddle7re, x14p17); |
| 6358 |
let t_a15_15 = f32x4_mul(self.twiddle8re, x15p16); |
| 6359 |
|
| 6360 |
let t_b1_1 = f32x4_mul(self.twiddle1im, x1m30); |
| 6361 |
let t_b1_2 = f32x4_mul(self.twiddle2im, x2m29); |
| 6362 |
let t_b1_3 = f32x4_mul(self.twiddle3im, x3m28); |
| 6363 |
let t_b1_4 = f32x4_mul(self.twiddle4im, x4m27); |
| 6364 |
let t_b1_5 = f32x4_mul(self.twiddle5im, x5m26); |
| 6365 |
let t_b1_6 = f32x4_mul(self.twiddle6im, x6m25); |
| 6366 |
let t_b1_7 = f32x4_mul(self.twiddle7im, x7m24); |
| 6367 |
let t_b1_8 = f32x4_mul(self.twiddle8im, x8m23); |
| 6368 |
let t_b1_9 = f32x4_mul(self.twiddle9im, x9m22); |
| 6369 |
let t_b1_10 = f32x4_mul(self.twiddle10im, x10m21); |
| 6370 |
let t_b1_11 = f32x4_mul(self.twiddle11im, x11m20); |
| 6371 |
let t_b1_12 = f32x4_mul(self.twiddle12im, x12m19); |
| 6372 |
let t_b1_13 = f32x4_mul(self.twiddle13im, x13m18); |
| 6373 |
let t_b1_14 = f32x4_mul(self.twiddle14im, x14m17); |
| 6374 |
let t_b1_15 = f32x4_mul(self.twiddle15im, x15m16); |
| 6375 |
let t_b2_1 = f32x4_mul(self.twiddle2im, x1m30); |
| 6376 |
let t_b2_2 = f32x4_mul(self.twiddle4im, x2m29); |
| 6377 |
let t_b2_3 = f32x4_mul(self.twiddle6im, x3m28); |
| 6378 |
let t_b2_4 = f32x4_mul(self.twiddle8im, x4m27); |
| 6379 |
let t_b2_5 = f32x4_mul(self.twiddle10im, x5m26); |
| 6380 |
let t_b2_6 = f32x4_mul(self.twiddle12im, x6m25); |
| 6381 |
let t_b2_7 = f32x4_mul(self.twiddle14im, x7m24); |
| 6382 |
let t_b2_8 = f32x4_mul(self.twiddle15im, x8m23); |
| 6383 |
let t_b2_9 = f32x4_mul(self.twiddle13im, x9m22); |
| 6384 |
let t_b2_10 = f32x4_mul(self.twiddle11im, x10m21); |
| 6385 |
let t_b2_11 = f32x4_mul(self.twiddle9im, x11m20); |
| 6386 |
let t_b2_12 = f32x4_mul(self.twiddle7im, x12m19); |
| 6387 |
let t_b2_13 = f32x4_mul(self.twiddle5im, x13m18); |
| 6388 |
let t_b2_14 = f32x4_mul(self.twiddle3im, x14m17); |
| 6389 |
let t_b2_15 = f32x4_mul(self.twiddle1im, x15m16); |
| 6390 |
let t_b3_1 = f32x4_mul(self.twiddle3im, x1m30); |
| 6391 |
let t_b3_2 = f32x4_mul(self.twiddle6im, x2m29); |
| 6392 |
let t_b3_3 = f32x4_mul(self.twiddle9im, x3m28); |
| 6393 |
let t_b3_4 = f32x4_mul(self.twiddle12im, x4m27); |
| 6394 |
let t_b3_5 = f32x4_mul(self.twiddle15im, x5m26); |
| 6395 |
let t_b3_6 = f32x4_mul(self.twiddle13im, x6m25); |
| 6396 |
let t_b3_7 = f32x4_mul(self.twiddle10im, x7m24); |
| 6397 |
let t_b3_8 = f32x4_mul(self.twiddle7im, x8m23); |
| 6398 |
let t_b3_9 = f32x4_mul(self.twiddle4im, x9m22); |
| 6399 |
let t_b3_10 = f32x4_mul(self.twiddle1im, x10m21); |
| 6400 |
let t_b3_11 = f32x4_mul(self.twiddle2im, x11m20); |
| 6401 |
let t_b3_12 = f32x4_mul(self.twiddle5im, x12m19); |
| 6402 |
let t_b3_13 = f32x4_mul(self.twiddle8im, x13m18); |
| 6403 |
let t_b3_14 = f32x4_mul(self.twiddle11im, x14m17); |
| 6404 |
let t_b3_15 = f32x4_mul(self.twiddle14im, x15m16); |
| 6405 |
let t_b4_1 = f32x4_mul(self.twiddle4im, x1m30); |
| 6406 |
let t_b4_2 = f32x4_mul(self.twiddle8im, x2m29); |
| 6407 |
let t_b4_3 = f32x4_mul(self.twiddle12im, x3m28); |
| 6408 |
let t_b4_4 = f32x4_mul(self.twiddle15im, x4m27); |
| 6409 |
let t_b4_5 = f32x4_mul(self.twiddle11im, x5m26); |
| 6410 |
let t_b4_6 = f32x4_mul(self.twiddle7im, x6m25); |
| 6411 |
let t_b4_7 = f32x4_mul(self.twiddle3im, x7m24); |
| 6412 |
let t_b4_8 = f32x4_mul(self.twiddle1im, x8m23); |
| 6413 |
let t_b4_9 = f32x4_mul(self.twiddle5im, x9m22); |
| 6414 |
let t_b4_10 = f32x4_mul(self.twiddle9im, x10m21); |
| 6415 |
let t_b4_11 = f32x4_mul(self.twiddle13im, x11m20); |
| 6416 |
let t_b4_12 = f32x4_mul(self.twiddle14im, x12m19); |
| 6417 |
let t_b4_13 = f32x4_mul(self.twiddle10im, x13m18); |
| 6418 |
let t_b4_14 = f32x4_mul(self.twiddle6im, x14m17); |
| 6419 |
let t_b4_15 = f32x4_mul(self.twiddle2im, x15m16); |
| 6420 |
let t_b5_1 = f32x4_mul(self.twiddle5im, x1m30); |
| 6421 |
let t_b5_2 = f32x4_mul(self.twiddle10im, x2m29); |
| 6422 |
let t_b5_3 = f32x4_mul(self.twiddle15im, x3m28); |
| 6423 |
let t_b5_4 = f32x4_mul(self.twiddle11im, x4m27); |
| 6424 |
let t_b5_5 = f32x4_mul(self.twiddle6im, x5m26); |
| 6425 |
let t_b5_6 = f32x4_mul(self.twiddle1im, x6m25); |
| 6426 |
let t_b5_7 = f32x4_mul(self.twiddle4im, x7m24); |
| 6427 |
let t_b5_8 = f32x4_mul(self.twiddle9im, x8m23); |
| 6428 |
let t_b5_9 = f32x4_mul(self.twiddle14im, x9m22); |
| 6429 |
let t_b5_10 = f32x4_mul(self.twiddle12im, x10m21); |
| 6430 |
let t_b5_11 = f32x4_mul(self.twiddle7im, x11m20); |
| 6431 |
let t_b5_12 = f32x4_mul(self.twiddle2im, x12m19); |
| 6432 |
let t_b5_13 = f32x4_mul(self.twiddle3im, x13m18); |
| 6433 |
let t_b5_14 = f32x4_mul(self.twiddle8im, x14m17); |
| 6434 |
let t_b5_15 = f32x4_mul(self.twiddle13im, x15m16); |
| 6435 |
let t_b6_1 = f32x4_mul(self.twiddle6im, x1m30); |
| 6436 |
let t_b6_2 = f32x4_mul(self.twiddle12im, x2m29); |
| 6437 |
let t_b6_3 = f32x4_mul(self.twiddle13im, x3m28); |
| 6438 |
let t_b6_4 = f32x4_mul(self.twiddle7im, x4m27); |
| 6439 |
let t_b6_5 = f32x4_mul(self.twiddle1im, x5m26); |
| 6440 |
let t_b6_6 = f32x4_mul(self.twiddle5im, x6m25); |
| 6441 |
let t_b6_7 = f32x4_mul(self.twiddle11im, x7m24); |
| 6442 |
let t_b6_8 = f32x4_mul(self.twiddle14im, x8m23); |
| 6443 |
let t_b6_9 = f32x4_mul(self.twiddle8im, x9m22); |
| 6444 |
let t_b6_10 = f32x4_mul(self.twiddle2im, x10m21); |
| 6445 |
let t_b6_11 = f32x4_mul(self.twiddle4im, x11m20); |
| 6446 |
let t_b6_12 = f32x4_mul(self.twiddle10im, x12m19); |
| 6447 |
let t_b6_13 = f32x4_mul(self.twiddle15im, x13m18); |
| 6448 |
let t_b6_14 = f32x4_mul(self.twiddle9im, x14m17); |
| 6449 |
let t_b6_15 = f32x4_mul(self.twiddle3im, x15m16); |
| 6450 |
let t_b7_1 = f32x4_mul(self.twiddle7im, x1m30); |
| 6451 |
let t_b7_2 = f32x4_mul(self.twiddle14im, x2m29); |
| 6452 |
let t_b7_3 = f32x4_mul(self.twiddle10im, x3m28); |
| 6453 |
let t_b7_4 = f32x4_mul(self.twiddle3im, x4m27); |
| 6454 |
let t_b7_5 = f32x4_mul(self.twiddle4im, x5m26); |
| 6455 |
let t_b7_6 = f32x4_mul(self.twiddle11im, x6m25); |
| 6456 |
let t_b7_7 = f32x4_mul(self.twiddle13im, x7m24); |
| 6457 |
let t_b7_8 = f32x4_mul(self.twiddle6im, x8m23); |
| 6458 |
let t_b7_9 = f32x4_mul(self.twiddle1im, x9m22); |
| 6459 |
let t_b7_10 = f32x4_mul(self.twiddle8im, x10m21); |
| 6460 |
let t_b7_11 = f32x4_mul(self.twiddle15im, x11m20); |
| 6461 |
let t_b7_12 = f32x4_mul(self.twiddle9im, x12m19); |
| 6462 |
let t_b7_13 = f32x4_mul(self.twiddle2im, x13m18); |
| 6463 |
let t_b7_14 = f32x4_mul(self.twiddle5im, x14m17); |
| 6464 |
let t_b7_15 = f32x4_mul(self.twiddle12im, x15m16); |
| 6465 |
let t_b8_1 = f32x4_mul(self.twiddle8im, x1m30); |
| 6466 |
let t_b8_2 = f32x4_mul(self.twiddle15im, x2m29); |
| 6467 |
let t_b8_3 = f32x4_mul(self.twiddle7im, x3m28); |
| 6468 |
let t_b8_4 = f32x4_mul(self.twiddle1im, x4m27); |
| 6469 |
let t_b8_5 = f32x4_mul(self.twiddle9im, x5m26); |
| 6470 |
let t_b8_6 = f32x4_mul(self.twiddle14im, x6m25); |
| 6471 |
let t_b8_7 = f32x4_mul(self.twiddle6im, x7m24); |
| 6472 |
let t_b8_8 = f32x4_mul(self.twiddle2im, x8m23); |
| 6473 |
let t_b8_9 = f32x4_mul(self.twiddle10im, x9m22); |
| 6474 |
let t_b8_10 = f32x4_mul(self.twiddle13im, x10m21); |
| 6475 |
let t_b8_11 = f32x4_mul(self.twiddle5im, x11m20); |
| 6476 |
let t_b8_12 = f32x4_mul(self.twiddle3im, x12m19); |
| 6477 |
let t_b8_13 = f32x4_mul(self.twiddle11im, x13m18); |
| 6478 |
let t_b8_14 = f32x4_mul(self.twiddle12im, x14m17); |
| 6479 |
let t_b8_15 = f32x4_mul(self.twiddle4im, x15m16); |
| 6480 |
let t_b9_1 = f32x4_mul(self.twiddle9im, x1m30); |
| 6481 |
let t_b9_2 = f32x4_mul(self.twiddle13im, x2m29); |
| 6482 |
let t_b9_3 = f32x4_mul(self.twiddle4im, x3m28); |
| 6483 |
let t_b9_4 = f32x4_mul(self.twiddle5im, x4m27); |
| 6484 |
let t_b9_5 = f32x4_mul(self.twiddle14im, x5m26); |
| 6485 |
let t_b9_6 = f32x4_mul(self.twiddle8im, x6m25); |
| 6486 |
let t_b9_7 = f32x4_mul(self.twiddle1im, x7m24); |
| 6487 |
let t_b9_8 = f32x4_mul(self.twiddle10im, x8m23); |
| 6488 |
let t_b9_9 = f32x4_mul(self.twiddle12im, x9m22); |
| 6489 |
let t_b9_10 = f32x4_mul(self.twiddle3im, x10m21); |
| 6490 |
let t_b9_11 = f32x4_mul(self.twiddle6im, x11m20); |
| 6491 |
let t_b9_12 = f32x4_mul(self.twiddle15im, x12m19); |
| 6492 |
let t_b9_13 = f32x4_mul(self.twiddle7im, x13m18); |
| 6493 |
let t_b9_14 = f32x4_mul(self.twiddle2im, x14m17); |
| 6494 |
let t_b9_15 = f32x4_mul(self.twiddle11im, x15m16); |
| 6495 |
let t_b10_1 = f32x4_mul(self.twiddle10im, x1m30); |
| 6496 |
let t_b10_2 = f32x4_mul(self.twiddle11im, x2m29); |
| 6497 |
let t_b10_3 = f32x4_mul(self.twiddle1im, x3m28); |
| 6498 |
let t_b10_4 = f32x4_mul(self.twiddle9im, x4m27); |
| 6499 |
let t_b10_5 = f32x4_mul(self.twiddle12im, x5m26); |
| 6500 |
let t_b10_6 = f32x4_mul(self.twiddle2im, x6m25); |
| 6501 |
let t_b10_7 = f32x4_mul(self.twiddle8im, x7m24); |
| 6502 |
let t_b10_8 = f32x4_mul(self.twiddle13im, x8m23); |
| 6503 |
let t_b10_9 = f32x4_mul(self.twiddle3im, x9m22); |
| 6504 |
let t_b10_10 = f32x4_mul(self.twiddle7im, x10m21); |
| 6505 |
let t_b10_11 = f32x4_mul(self.twiddle14im, x11m20); |
| 6506 |
let t_b10_12 = f32x4_mul(self.twiddle4im, x12m19); |
| 6507 |
let t_b10_13 = f32x4_mul(self.twiddle6im, x13m18); |
| 6508 |
let t_b10_14 = f32x4_mul(self.twiddle15im, x14m17); |
| 6509 |
let t_b10_15 = f32x4_mul(self.twiddle5im, x15m16); |
| 6510 |
let t_b11_1 = f32x4_mul(self.twiddle11im, x1m30); |
| 6511 |
let t_b11_2 = f32x4_mul(self.twiddle9im, x2m29); |
| 6512 |
let t_b11_3 = f32x4_mul(self.twiddle2im, x3m28); |
| 6513 |
let t_b11_4 = f32x4_mul(self.twiddle13im, x4m27); |
| 6514 |
let t_b11_5 = f32x4_mul(self.twiddle7im, x5m26); |
| 6515 |
let t_b11_6 = f32x4_mul(self.twiddle4im, x6m25); |
| 6516 |
let t_b11_7 = f32x4_mul(self.twiddle15im, x7m24); |
| 6517 |
let t_b11_8 = f32x4_mul(self.twiddle5im, x8m23); |
| 6518 |
let t_b11_9 = f32x4_mul(self.twiddle6im, x9m22); |
| 6519 |
let t_b11_10 = f32x4_mul(self.twiddle14im, x10m21); |
| 6520 |
let t_b11_11 = f32x4_mul(self.twiddle3im, x11m20); |
| 6521 |
let t_b11_12 = f32x4_mul(self.twiddle8im, x12m19); |
| 6522 |
let t_b11_13 = f32x4_mul(self.twiddle12im, x13m18); |
| 6523 |
let t_b11_14 = f32x4_mul(self.twiddle1im, x14m17); |
| 6524 |
let t_b11_15 = f32x4_mul(self.twiddle10im, x15m16); |
| 6525 |
let t_b12_1 = f32x4_mul(self.twiddle12im, x1m30); |
| 6526 |
let t_b12_2 = f32x4_mul(self.twiddle7im, x2m29); |
| 6527 |
let t_b12_3 = f32x4_mul(self.twiddle5im, x3m28); |
| 6528 |
let t_b12_4 = f32x4_mul(self.twiddle14im, x4m27); |
| 6529 |
let t_b12_5 = f32x4_mul(self.twiddle2im, x5m26); |
| 6530 |
let t_b12_6 = f32x4_mul(self.twiddle10im, x6m25); |
| 6531 |
let t_b12_7 = f32x4_mul(self.twiddle9im, x7m24); |
| 6532 |
let t_b12_8 = f32x4_mul(self.twiddle3im, x8m23); |
| 6533 |
let t_b12_9 = f32x4_mul(self.twiddle15im, x9m22); |
| 6534 |
let t_b12_10 = f32x4_mul(self.twiddle4im, x10m21); |
| 6535 |
let t_b12_11 = f32x4_mul(self.twiddle8im, x11m20); |
| 6536 |
let t_b12_12 = f32x4_mul(self.twiddle11im, x12m19); |
| 6537 |
let t_b12_13 = f32x4_mul(self.twiddle1im, x13m18); |
| 6538 |
let t_b12_14 = f32x4_mul(self.twiddle13im, x14m17); |
| 6539 |
let t_b12_15 = f32x4_mul(self.twiddle6im, x15m16); |
| 6540 |
let t_b13_1 = f32x4_mul(self.twiddle13im, x1m30); |
| 6541 |
let t_b13_2 = f32x4_mul(self.twiddle5im, x2m29); |
| 6542 |
let t_b13_3 = f32x4_mul(self.twiddle8im, x3m28); |
| 6543 |
let t_b13_4 = f32x4_mul(self.twiddle10im, x4m27); |
| 6544 |
let t_b13_5 = f32x4_mul(self.twiddle3im, x5m26); |
| 6545 |
let t_b13_6 = f32x4_mul(self.twiddle15im, x6m25); |
| 6546 |
let t_b13_7 = f32x4_mul(self.twiddle2im, x7m24); |
| 6547 |
let t_b13_8 = f32x4_mul(self.twiddle11im, x8m23); |
| 6548 |
let t_b13_9 = f32x4_mul(self.twiddle7im, x9m22); |
| 6549 |
let t_b13_10 = f32x4_mul(self.twiddle6im, x10m21); |
| 6550 |
let t_b13_11 = f32x4_mul(self.twiddle12im, x11m20); |
| 6551 |
let t_b13_12 = f32x4_mul(self.twiddle1im, x12m19); |
| 6552 |
let t_b13_13 = f32x4_mul(self.twiddle14im, x13m18); |
| 6553 |
let t_b13_14 = f32x4_mul(self.twiddle4im, x14m17); |
| 6554 |
let t_b13_15 = f32x4_mul(self.twiddle9im, x15m16); |
| 6555 |
let t_b14_1 = f32x4_mul(self.twiddle14im, x1m30); |
| 6556 |
let t_b14_2 = f32x4_mul(self.twiddle3im, x2m29); |
| 6557 |
let t_b14_3 = f32x4_mul(self.twiddle11im, x3m28); |
| 6558 |
let t_b14_4 = f32x4_mul(self.twiddle6im, x4m27); |
| 6559 |
let t_b14_5 = f32x4_mul(self.twiddle8im, x5m26); |
| 6560 |
let t_b14_6 = f32x4_mul(self.twiddle9im, x6m25); |
| 6561 |
let t_b14_7 = f32x4_mul(self.twiddle5im, x7m24); |
| 6562 |
let t_b14_8 = f32x4_mul(self.twiddle12im, x8m23); |
| 6563 |
let t_b14_9 = f32x4_mul(self.twiddle2im, x9m22); |
| 6564 |
let t_b14_10 = f32x4_mul(self.twiddle15im, x10m21); |
| 6565 |
let t_b14_11 = f32x4_mul(self.twiddle1im, x11m20); |
| 6566 |
let t_b14_12 = f32x4_mul(self.twiddle13im, x12m19); |
| 6567 |
let t_b14_13 = f32x4_mul(self.twiddle4im, x13m18); |
| 6568 |
let t_b14_14 = f32x4_mul(self.twiddle10im, x14m17); |
| 6569 |
let t_b14_15 = f32x4_mul(self.twiddle7im, x15m16); |
| 6570 |
let t_b15_1 = f32x4_mul(self.twiddle15im, x1m30); |
| 6571 |
let t_b15_2 = f32x4_mul(self.twiddle1im, x2m29); |
| 6572 |
let t_b15_3 = f32x4_mul(self.twiddle14im, x3m28); |
| 6573 |
let t_b15_4 = f32x4_mul(self.twiddle2im, x4m27); |
| 6574 |
let t_b15_5 = f32x4_mul(self.twiddle13im, x5m26); |
| 6575 |
let t_b15_6 = f32x4_mul(self.twiddle3im, x6m25); |
| 6576 |
let t_b15_7 = f32x4_mul(self.twiddle12im, x7m24); |
| 6577 |
let t_b15_8 = f32x4_mul(self.twiddle4im, x8m23); |
| 6578 |
let t_b15_9 = f32x4_mul(self.twiddle11im, x9m22); |
| 6579 |
let t_b15_10 = f32x4_mul(self.twiddle5im, x10m21); |
| 6580 |
let t_b15_11 = f32x4_mul(self.twiddle10im, x11m20); |
| 6581 |
let t_b15_12 = f32x4_mul(self.twiddle6im, x12m19); |
| 6582 |
let t_b15_13 = f32x4_mul(self.twiddle9im, x13m18); |
| 6583 |
let t_b15_14 = f32x4_mul(self.twiddle7im, x14m17); |
| 6584 |
let t_b15_15 = f32x4_mul(self.twiddle8im, x15m16); |
| 6585 |
|
| 6586 |
let x0 = values[0]; |
| 6587 |
let t_a1 = calc_f32!( |
| 6588 |
x0 + t_a1_1 |
| 6589 |
+ t_a1_2 |
| 6590 |
+ t_a1_3 |
| 6591 |
+ t_a1_4 |
| 6592 |
+ t_a1_5 |
| 6593 |
+ t_a1_6 |
| 6594 |
+ t_a1_7 |
| 6595 |
+ t_a1_8 |
| 6596 |
+ t_a1_9 |
| 6597 |
+ t_a1_10 |
| 6598 |
+ t_a1_11 |
| 6599 |
+ t_a1_12 |
| 6600 |
+ t_a1_13 |
| 6601 |
+ t_a1_14 |
| 6602 |
+ t_a1_15 |
| 6603 |
); |
| 6604 |
let t_a2 = calc_f32!( |
| 6605 |
x0 + t_a2_1 |
| 6606 |
+ t_a2_2 |
| 6607 |
+ t_a2_3 |
| 6608 |
+ t_a2_4 |
| 6609 |
+ t_a2_5 |
| 6610 |
+ t_a2_6 |
| 6611 |
+ t_a2_7 |
| 6612 |
+ t_a2_8 |
| 6613 |
+ t_a2_9 |
| 6614 |
+ t_a2_10 |
| 6615 |
+ t_a2_11 |
| 6616 |
+ t_a2_12 |
| 6617 |
+ t_a2_13 |
| 6618 |
+ t_a2_14 |
| 6619 |
+ t_a2_15 |
| 6620 |
); |
| 6621 |
let t_a3 = calc_f32!( |
| 6622 |
x0 + t_a3_1 |
| 6623 |
+ t_a3_2 |
| 6624 |
+ t_a3_3 |
| 6625 |
+ t_a3_4 |
| 6626 |
+ t_a3_5 |
| 6627 |
+ t_a3_6 |
| 6628 |
+ t_a3_7 |
| 6629 |
+ t_a3_8 |
| 6630 |
+ t_a3_9 |
| 6631 |
+ t_a3_10 |
| 6632 |
+ t_a3_11 |
| 6633 |
+ t_a3_12 |
| 6634 |
+ t_a3_13 |
| 6635 |
+ t_a3_14 |
| 6636 |
+ t_a3_15 |
| 6637 |
); |
| 6638 |
let t_a4 = calc_f32!( |
| 6639 |
x0 + t_a4_1 |
| 6640 |
+ t_a4_2 |
| 6641 |
+ t_a4_3 |
| 6642 |
+ t_a4_4 |
| 6643 |
+ t_a4_5 |
| 6644 |
+ t_a4_6 |
| 6645 |
+ t_a4_7 |
| 6646 |
+ t_a4_8 |
| 6647 |
+ t_a4_9 |
| 6648 |
+ t_a4_10 |
| 6649 |
+ t_a4_11 |
| 6650 |
+ t_a4_12 |
| 6651 |
+ t_a4_13 |
| 6652 |
+ t_a4_14 |
| 6653 |
+ t_a4_15 |
| 6654 |
); |
| 6655 |
let t_a5 = calc_f32!( |
| 6656 |
x0 + t_a5_1 |
| 6657 |
+ t_a5_2 |
| 6658 |
+ t_a5_3 |
| 6659 |
+ t_a5_4 |
| 6660 |
+ t_a5_5 |
| 6661 |
+ t_a5_6 |
| 6662 |
+ t_a5_7 |
| 6663 |
+ t_a5_8 |
| 6664 |
+ t_a5_9 |
| 6665 |
+ t_a5_10 |
| 6666 |
+ t_a5_11 |
| 6667 |
+ t_a5_12 |
| 6668 |
+ t_a5_13 |
| 6669 |
+ t_a5_14 |
| 6670 |
+ t_a5_15 |
| 6671 |
); |
| 6672 |
let t_a6 = calc_f32!( |
| 6673 |
x0 + t_a6_1 |
| 6674 |
+ t_a6_2 |
| 6675 |
+ t_a6_3 |
| 6676 |
+ t_a6_4 |
| 6677 |
+ t_a6_5 |
| 6678 |
+ t_a6_6 |
| 6679 |
+ t_a6_7 |
| 6680 |
+ t_a6_8 |
| 6681 |
+ t_a6_9 |
| 6682 |
+ t_a6_10 |
| 6683 |
+ t_a6_11 |
| 6684 |
+ t_a6_12 |
| 6685 |
+ t_a6_13 |
| 6686 |
+ t_a6_14 |
| 6687 |
+ t_a6_15 |
| 6688 |
); |
| 6689 |
let t_a7 = calc_f32!( |
| 6690 |
x0 + t_a7_1 |
| 6691 |
+ t_a7_2 |
| 6692 |
+ t_a7_3 |
| 6693 |
+ t_a7_4 |
| 6694 |
+ t_a7_5 |
| 6695 |
+ t_a7_6 |
| 6696 |
+ t_a7_7 |
| 6697 |
+ t_a7_8 |
| 6698 |
+ t_a7_9 |
| 6699 |
+ t_a7_10 |
| 6700 |
+ t_a7_11 |
| 6701 |
+ t_a7_12 |
| 6702 |
+ t_a7_13 |
| 6703 |
+ t_a7_14 |
| 6704 |
+ t_a7_15 |
| 6705 |
); |
| 6706 |
let t_a8 = calc_f32!( |
| 6707 |
x0 + t_a8_1 |
| 6708 |
+ t_a8_2 |
| 6709 |
+ t_a8_3 |
| 6710 |
+ t_a8_4 |
| 6711 |
+ t_a8_5 |
| 6712 |
+ t_a8_6 |
| 6713 |
+ t_a8_7 |
| 6714 |
+ t_a8_8 |
| 6715 |
+ t_a8_9 |
| 6716 |
+ t_a8_10 |
| 6717 |
+ t_a8_11 |
| 6718 |
+ t_a8_12 |
| 6719 |
+ t_a8_13 |
| 6720 |
+ t_a8_14 |
| 6721 |
+ t_a8_15 |
| 6722 |
); |
| 6723 |
let t_a9 = calc_f32!( |
| 6724 |
x0 + t_a9_1 |
| 6725 |
+ t_a9_2 |
| 6726 |
+ t_a9_3 |
| 6727 |
+ t_a9_4 |
| 6728 |
+ t_a9_5 |
| 6729 |
+ t_a9_6 |
| 6730 |
+ t_a9_7 |
| 6731 |
+ t_a9_8 |
| 6732 |
+ t_a9_9 |
| 6733 |
+ t_a9_10 |
| 6734 |
+ t_a9_11 |
| 6735 |
+ t_a9_12 |
| 6736 |
+ t_a9_13 |
| 6737 |
+ t_a9_14 |
| 6738 |
+ t_a9_15 |
| 6739 |
); |
| 6740 |
let t_a10 = calc_f32!( |
| 6741 |
x0 + t_a10_1 |
| 6742 |
+ t_a10_2 |
| 6743 |
+ t_a10_3 |
| 6744 |
+ t_a10_4 |
| 6745 |
+ t_a10_5 |
| 6746 |
+ t_a10_6 |
| 6747 |
+ t_a10_7 |
| 6748 |
+ t_a10_8 |
| 6749 |
+ t_a10_9 |
| 6750 |
+ t_a10_10 |
| 6751 |
+ t_a10_11 |
| 6752 |
+ t_a10_12 |
| 6753 |
+ t_a10_13 |
| 6754 |
+ t_a10_14 |
| 6755 |
+ t_a10_15 |
| 6756 |
); |
| 6757 |
let t_a11 = calc_f32!( |
| 6758 |
x0 + t_a11_1 |
| 6759 |
+ t_a11_2 |
| 6760 |
+ t_a11_3 |
| 6761 |
+ t_a11_4 |
| 6762 |
+ t_a11_5 |
| 6763 |
+ t_a11_6 |
| 6764 |
+ t_a11_7 |
| 6765 |
+ t_a11_8 |
| 6766 |
+ t_a11_9 |
| 6767 |
+ t_a11_10 |
| 6768 |
+ t_a11_11 |
| 6769 |
+ t_a11_12 |
| 6770 |
+ t_a11_13 |
| 6771 |
+ t_a11_14 |
| 6772 |
+ t_a11_15 |
| 6773 |
); |
| 6774 |
let t_a12 = calc_f32!( |
| 6775 |
x0 + t_a12_1 |
| 6776 |
+ t_a12_2 |
| 6777 |
+ t_a12_3 |
| 6778 |
+ t_a12_4 |
| 6779 |
+ t_a12_5 |
| 6780 |
+ t_a12_6 |
| 6781 |
+ t_a12_7 |
| 6782 |
+ t_a12_8 |
| 6783 |
+ t_a12_9 |
| 6784 |
+ t_a12_10 |
| 6785 |
+ t_a12_11 |
| 6786 |
+ t_a12_12 |
| 6787 |
+ t_a12_13 |
| 6788 |
+ t_a12_14 |
| 6789 |
+ t_a12_15 |
| 6790 |
); |
| 6791 |
let t_a13 = calc_f32!( |
| 6792 |
x0 + t_a13_1 |
| 6793 |
+ t_a13_2 |
| 6794 |
+ t_a13_3 |
| 6795 |
+ t_a13_4 |
| 6796 |
+ t_a13_5 |
| 6797 |
+ t_a13_6 |
| 6798 |
+ t_a13_7 |
| 6799 |
+ t_a13_8 |
| 6800 |
+ t_a13_9 |
| 6801 |
+ t_a13_10 |
| 6802 |
+ t_a13_11 |
| 6803 |
+ t_a13_12 |
| 6804 |
+ t_a13_13 |
| 6805 |
+ t_a13_14 |
| 6806 |
+ t_a13_15 |
| 6807 |
); |
| 6808 |
let t_a14 = calc_f32!( |
| 6809 |
x0 + t_a14_1 |
| 6810 |
+ t_a14_2 |
| 6811 |
+ t_a14_3 |
| 6812 |
+ t_a14_4 |
| 6813 |
+ t_a14_5 |
| 6814 |
+ t_a14_6 |
| 6815 |
+ t_a14_7 |
| 6816 |
+ t_a14_8 |
| 6817 |
+ t_a14_9 |
| 6818 |
+ t_a14_10 |
| 6819 |
+ t_a14_11 |
| 6820 |
+ t_a14_12 |
| 6821 |
+ t_a14_13 |
| 6822 |
+ t_a14_14 |
| 6823 |
+ t_a14_15 |
| 6824 |
); |
| 6825 |
let t_a15 = calc_f32!( |
| 6826 |
x0 + t_a15_1 |
| 6827 |
+ t_a15_2 |
| 6828 |
+ t_a15_3 |
| 6829 |
+ t_a15_4 |
| 6830 |
+ t_a15_5 |
| 6831 |
+ t_a15_6 |
| 6832 |
+ t_a15_7 |
| 6833 |
+ t_a15_8 |
| 6834 |
+ t_a15_9 |
| 6835 |
+ t_a15_10 |
| 6836 |
+ t_a15_11 |
| 6837 |
+ t_a15_12 |
| 6838 |
+ t_a15_13 |
| 6839 |
+ t_a15_14 |
| 6840 |
+ t_a15_15 |
| 6841 |
); |
| 6842 |
|
| 6843 |
let t_b1 = calc_f32!( |
| 6844 |
t_b1_1 |
| 6845 |
+ t_b1_2 |
| 6846 |
+ t_b1_3 |
| 6847 |
+ t_b1_4 |
| 6848 |
+ t_b1_5 |
| 6849 |
+ t_b1_6 |
| 6850 |
+ t_b1_7 |
| 6851 |
+ t_b1_8 |
| 6852 |
+ t_b1_9 |
| 6853 |
+ t_b1_10 |
| 6854 |
+ t_b1_11 |
| 6855 |
+ t_b1_12 |
| 6856 |
+ t_b1_13 |
| 6857 |
+ t_b1_14 |
| 6858 |
+ t_b1_15 |
| 6859 |
); |
| 6860 |
let t_b2 = calc_f32!( |
| 6861 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 + t_b2_6 + t_b2_7 |
| 6862 |
- t_b2_8 |
| 6863 |
- t_b2_9 |
| 6864 |
- t_b2_10 |
| 6865 |
- t_b2_11 |
| 6866 |
- t_b2_12 |
| 6867 |
- t_b2_13 |
| 6868 |
- t_b2_14 |
| 6869 |
- t_b2_15 |
| 6870 |
); |
| 6871 |
let t_b3 = calc_f32!( |
| 6872 |
t_b3_1 + t_b3_2 + t_b3_3 + t_b3_4 + t_b3_5 |
| 6873 |
- t_b3_6 |
| 6874 |
- t_b3_7 |
| 6875 |
- t_b3_8 |
| 6876 |
- t_b3_9 |
| 6877 |
- t_b3_10 |
| 6878 |
+ t_b3_11 |
| 6879 |
+ t_b3_12 |
| 6880 |
+ t_b3_13 |
| 6881 |
+ t_b3_14 |
| 6882 |
+ t_b3_15 |
| 6883 |
); |
| 6884 |
let t_b4 = calc_f32!( |
| 6885 |
t_b4_1 + t_b4_2 + t_b4_3 - t_b4_4 - t_b4_5 - t_b4_6 - t_b4_7 |
| 6886 |
+ t_b4_8 |
| 6887 |
+ t_b4_9 |
| 6888 |
+ t_b4_10 |
| 6889 |
+ t_b4_11 |
| 6890 |
- t_b4_12 |
| 6891 |
- t_b4_13 |
| 6892 |
- t_b4_14 |
| 6893 |
- t_b4_15 |
| 6894 |
); |
| 6895 |
let t_b5 = calc_f32!( |
| 6896 |
t_b5_1 + t_b5_2 + t_b5_3 - t_b5_4 - t_b5_5 - t_b5_6 + t_b5_7 + t_b5_8 + t_b5_9 |
| 6897 |
- t_b5_10 |
| 6898 |
- t_b5_11 |
| 6899 |
- t_b5_12 |
| 6900 |
+ t_b5_13 |
| 6901 |
+ t_b5_14 |
| 6902 |
+ t_b5_15 |
| 6903 |
); |
| 6904 |
let t_b6 = calc_f32!( |
| 6905 |
t_b6_1 + t_b6_2 - t_b6_3 - t_b6_4 - t_b6_5 + t_b6_6 + t_b6_7 |
| 6906 |
- t_b6_8 |
| 6907 |
- t_b6_9 |
| 6908 |
- t_b6_10 |
| 6909 |
+ t_b6_11 |
| 6910 |
+ t_b6_12 |
| 6911 |
- t_b6_13 |
| 6912 |
- t_b6_14 |
| 6913 |
- t_b6_15 |
| 6914 |
); |
| 6915 |
let t_b7 = calc_f32!( |
| 6916 |
t_b7_1 + t_b7_2 - t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 |
| 6917 |
+ t_b7_9 |
| 6918 |
+ t_b7_10 |
| 6919 |
+ t_b7_11 |
| 6920 |
- t_b7_12 |
| 6921 |
- t_b7_13 |
| 6922 |
+ t_b7_14 |
| 6923 |
+ t_b7_15 |
| 6924 |
); |
| 6925 |
let t_b8 = calc_f32!( |
| 6926 |
t_b8_1 - t_b8_2 - t_b8_3 + t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 + t_b8_9 |
| 6927 |
- t_b8_10 |
| 6928 |
- t_b8_11 |
| 6929 |
+ t_b8_12 |
| 6930 |
+ t_b8_13 |
| 6931 |
- t_b8_14 |
| 6932 |
- t_b8_15 |
| 6933 |
); |
| 6934 |
let t_b9 = calc_f32!( |
| 6935 |
t_b9_1 - t_b9_2 - t_b9_3 + t_b9_4 + t_b9_5 - t_b9_6 + t_b9_7 + t_b9_8 |
| 6936 |
- t_b9_9 |
| 6937 |
- t_b9_10 |
| 6938 |
+ t_b9_11 |
| 6939 |
+ t_b9_12 |
| 6940 |
- t_b9_13 |
| 6941 |
+ t_b9_14 |
| 6942 |
+ t_b9_15 |
| 6943 |
); |
| 6944 |
let t_b10 = calc_f32!( |
| 6945 |
t_b10_1 - t_b10_2 - t_b10_3 + t_b10_4 - t_b10_5 - t_b10_6 + t_b10_7 - t_b10_8 - t_b10_9 |
| 6946 |
+ t_b10_10 |
| 6947 |
- t_b10_11 |
| 6948 |
- t_b10_12 |
| 6949 |
+ t_b10_13 |
| 6950 |
- t_b10_14 |
| 6951 |
- t_b10_15 |
| 6952 |
); |
| 6953 |
let t_b11 = calc_f32!( |
| 6954 |
t_b11_1 - t_b11_2 + t_b11_3 + t_b11_4 - t_b11_5 + t_b11_6 + t_b11_7 - t_b11_8 + t_b11_9 |
| 6955 |
- t_b11_10 |
| 6956 |
- t_b11_11 |
| 6957 |
+ t_b11_12 |
| 6958 |
- t_b11_13 |
| 6959 |
- t_b11_14 |
| 6960 |
+ t_b11_15 |
| 6961 |
); |
| 6962 |
let t_b12 = calc_f32!( |
| 6963 |
t_b12_1 - t_b12_2 + t_b12_3 - t_b12_4 - t_b12_5 + t_b12_6 - t_b12_7 + t_b12_8 + t_b12_9 |
| 6964 |
- t_b12_10 |
| 6965 |
+ t_b12_11 |
| 6966 |
- t_b12_12 |
| 6967 |
+ t_b12_13 |
| 6968 |
+ t_b12_14 |
| 6969 |
- t_b12_15 |
| 6970 |
); |
| 6971 |
let t_b13 = calc_f32!( |
| 6972 |
t_b13_1 - t_b13_2 + t_b13_3 - t_b13_4 + t_b13_5 - t_b13_6 - t_b13_7 + t_b13_8 - t_b13_9 |
| 6973 |
+ t_b13_10 |
| 6974 |
- t_b13_11 |
| 6975 |
+ t_b13_12 |
| 6976 |
+ t_b13_13 |
| 6977 |
- t_b13_14 |
| 6978 |
+ t_b13_15 |
| 6979 |
); |
| 6980 |
let t_b14 = calc_f32!( |
| 6981 |
t_b14_1 - t_b14_2 + t_b14_3 - t_b14_4 + t_b14_5 - t_b14_6 + t_b14_7 - t_b14_8 + t_b14_9 |
| 6982 |
- t_b14_10 |
| 6983 |
- t_b14_11 |
| 6984 |
+ t_b14_12 |
| 6985 |
- t_b14_13 |
| 6986 |
+ t_b14_14 |
| 6987 |
- t_b14_15 |
| 6988 |
); |
| 6989 |
let t_b15 = calc_f32!( |
| 6990 |
t_b15_1 - t_b15_2 + t_b15_3 - t_b15_4 + t_b15_5 - t_b15_6 + t_b15_7 - t_b15_8 + t_b15_9 |
| 6991 |
- t_b15_10 |
| 6992 |
+ t_b15_11 |
| 6993 |
- t_b15_12 |
| 6994 |
+ t_b15_13 |
| 6995 |
- t_b15_14 |
| 6996 |
+ t_b15_15 |
| 6997 |
); |
| 6998 |
|
| 6999 |
let t_b1_rot = self.rotate.rotate_both(t_b1); |
| 7000 |
let t_b2_rot = self.rotate.rotate_both(t_b2); |
| 7001 |
let t_b3_rot = self.rotate.rotate_both(t_b3); |
| 7002 |
let t_b4_rot = self.rotate.rotate_both(t_b4); |
| 7003 |
let t_b5_rot = self.rotate.rotate_both(t_b5); |
| 7004 |
let t_b6_rot = self.rotate.rotate_both(t_b6); |
| 7005 |
let t_b7_rot = self.rotate.rotate_both(t_b7); |
| 7006 |
let t_b8_rot = self.rotate.rotate_both(t_b8); |
| 7007 |
let t_b9_rot = self.rotate.rotate_both(t_b9); |
| 7008 |
let t_b10_rot = self.rotate.rotate_both(t_b10); |
| 7009 |
let t_b11_rot = self.rotate.rotate_both(t_b11); |
| 7010 |
let t_b12_rot = self.rotate.rotate_both(t_b12); |
| 7011 |
let t_b13_rot = self.rotate.rotate_both(t_b13); |
| 7012 |
let t_b14_rot = self.rotate.rotate_both(t_b14); |
| 7013 |
let t_b15_rot = self.rotate.rotate_both(t_b15); |
| 7014 |
|
| 7015 |
let y0 = calc_f32!( |
| 7016 |
x0 + x1p30 |
| 7017 |
+ x2p29 |
| 7018 |
+ x3p28 |
| 7019 |
+ x4p27 |
| 7020 |
+ x5p26 |
| 7021 |
+ x6p25 |
| 7022 |
+ x7p24 |
| 7023 |
+ x8p23 |
| 7024 |
+ x9p22 |
| 7025 |
+ x10p21 |
| 7026 |
+ x11p20 |
| 7027 |
+ x12p19 |
| 7028 |
+ x13p18 |
| 7029 |
+ x14p17 |
| 7030 |
+ x15p16 |
| 7031 |
); |
| 7032 |
let [y1, y30] = parallel_fft2_interleaved_f32(t_a1, t_b1_rot); |
| 7033 |
let [y2, y29] = parallel_fft2_interleaved_f32(t_a2, t_b2_rot); |
| 7034 |
let [y3, y28] = parallel_fft2_interleaved_f32(t_a3, t_b3_rot); |
| 7035 |
let [y4, y27] = parallel_fft2_interleaved_f32(t_a4, t_b4_rot); |
| 7036 |
let [y5, y26] = parallel_fft2_interleaved_f32(t_a5, t_b5_rot); |
| 7037 |
let [y6, y25] = parallel_fft2_interleaved_f32(t_a6, t_b6_rot); |
| 7038 |
let [y7, y24] = parallel_fft2_interleaved_f32(t_a7, t_b7_rot); |
| 7039 |
let [y8, y23] = parallel_fft2_interleaved_f32(t_a8, t_b8_rot); |
| 7040 |
let [y9, y22] = parallel_fft2_interleaved_f32(t_a9, t_b9_rot); |
| 7041 |
let [y10, y21] = parallel_fft2_interleaved_f32(t_a10, t_b10_rot); |
| 7042 |
let [y11, y20] = parallel_fft2_interleaved_f32(t_a11, t_b11_rot); |
| 7043 |
let [y12, y19] = parallel_fft2_interleaved_f32(t_a12, t_b12_rot); |
| 7044 |
let [y13, y18] = parallel_fft2_interleaved_f32(t_a13, t_b13_rot); |
| 7045 |
let [y14, y17] = parallel_fft2_interleaved_f32(t_a14, t_b14_rot); |
| 7046 |
let [y15, y16] = parallel_fft2_interleaved_f32(t_a15, t_b15_rot); |
| 7047 |
[ |
| 7048 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 7049 |
y19, y20, y21, y22, y23, y24, y25, y26, y27, y28, y29, y30, |
| 7050 |
] |
| 7051 |
} |
| 7052 |
} |
| 7053 |
|
| 7054 |
// _____ _ __ _ _ _ _ _ |
| 7055 |
// |___ // | / /_ | || | | |__ (_) |_ |
| 7056 |
// |_ \| | _____ | '_ \| || |_| '_ \| | __| |
| 7057 |
// ___) | | |_____| | (_) |__ _| |_) | | |_ |
| 7058 |
// |____/|_| \___/ |_| |_.__/|_|\__| |
| 7059 |
// |
| 7060 |
|
| 7061 |
pub struct WasmSimdF64Butterfly31<T> { |
| 7062 |
direction: FftDirection, |
| 7063 |
_phantom: std::marker::PhantomData<T>, |
| 7064 |
rotate: Rotate90F64, |
| 7065 |
twiddle1re: v128, |
| 7066 |
twiddle1im: v128, |
| 7067 |
twiddle2re: v128, |
| 7068 |
twiddle2im: v128, |
| 7069 |
twiddle3re: v128, |
| 7070 |
twiddle3im: v128, |
| 7071 |
twiddle4re: v128, |
| 7072 |
twiddle4im: v128, |
| 7073 |
twiddle5re: v128, |
| 7074 |
twiddle5im: v128, |
| 7075 |
twiddle6re: v128, |
| 7076 |
twiddle6im: v128, |
| 7077 |
twiddle7re: v128, |
| 7078 |
twiddle7im: v128, |
| 7079 |
twiddle8re: v128, |
| 7080 |
twiddle8im: v128, |
| 7081 |
twiddle9re: v128, |
| 7082 |
twiddle9im: v128, |
| 7083 |
twiddle10re: v128, |
| 7084 |
twiddle10im: v128, |
| 7085 |
twiddle11re: v128, |
| 7086 |
twiddle11im: v128, |
| 7087 |
twiddle12re: v128, |
| 7088 |
twiddle12im: v128, |
| 7089 |
twiddle13re: v128, |
| 7090 |
twiddle13im: v128, |
| 7091 |
twiddle14re: v128, |
| 7092 |
twiddle14im: v128, |
| 7093 |
twiddle15re: v128, |
| 7094 |
twiddle15im: v128, |
| 7095 |
} |
| 7096 |
|
| 7097 |
boilerplate_fft_wasm_simd_f64_butterfly!( |
| 7098 |
WasmSimdF64Butterfly31, |
| 7099 |
31, |
| 7100 |
|this: &WasmSimdF64Butterfly31<_>| this.direction |
| 7101 |
); |
| 7102 |
boilerplate_fft_wasm_simd_common_butterfly!( |
| 7103 |
WasmSimdF64Butterfly31, |
| 7104 |
31, |
| 7105 |
|this: &WasmSimdF64Butterfly31<_>| this.direction |
| 7106 |
); |
| 7107 |
impl<T: FftNum> WasmSimdF64Butterfly31<T> { |
| 7108 |
#[inline(always)] |
| 7109 |
pub fn new(direction: FftDirection) -> Self { |
| 7110 |
assert_f64::<T>(); |
| 7111 |
let rotate = Rotate90F64::new(true); |
| 7112 |
let tw1: Complex<f64> = twiddles::compute_twiddle(1, 31, direction); |
| 7113 |
let tw2: Complex<f64> = twiddles::compute_twiddle(2, 31, direction); |
| 7114 |
let tw3: Complex<f64> = twiddles::compute_twiddle(3, 31, direction); |
| 7115 |
let tw4: Complex<f64> = twiddles::compute_twiddle(4, 31, direction); |
| 7116 |
let tw5: Complex<f64> = twiddles::compute_twiddle(5, 31, direction); |
| 7117 |
let tw6: Complex<f64> = twiddles::compute_twiddle(6, 31, direction); |
| 7118 |
let tw7: Complex<f64> = twiddles::compute_twiddle(7, 31, direction); |
| 7119 |
let tw8: Complex<f64> = twiddles::compute_twiddle(8, 31, direction); |
| 7120 |
let tw9: Complex<f64> = twiddles::compute_twiddle(9, 31, direction); |
| 7121 |
let tw10: Complex<f64> = twiddles::compute_twiddle(10, 31, direction); |
| 7122 |
let tw11: Complex<f64> = twiddles::compute_twiddle(11, 31, direction); |
| 7123 |
let tw12: Complex<f64> = twiddles::compute_twiddle(12, 31, direction); |
| 7124 |
let tw13: Complex<f64> = twiddles::compute_twiddle(13, 31, direction); |
| 7125 |
let tw14: Complex<f64> = twiddles::compute_twiddle(14, 31, direction); |
| 7126 |
let tw15: Complex<f64> = twiddles::compute_twiddle(15, 31, direction); |
| 7127 |
let twiddle1re = f64x2_splat(tw1.re); |
| 7128 |
let twiddle1im = f64x2_splat(tw1.im); |
| 7129 |
let twiddle2re = f64x2_splat(tw2.re); |
| 7130 |
let twiddle2im = f64x2_splat(tw2.im); |
| 7131 |
let twiddle3re = f64x2_splat(tw3.re); |
| 7132 |
let twiddle3im = f64x2_splat(tw3.im); |
| 7133 |
let twiddle4re = f64x2_splat(tw4.re); |
| 7134 |
let twiddle4im = f64x2_splat(tw4.im); |
| 7135 |
let twiddle5re = f64x2_splat(tw5.re); |
| 7136 |
let twiddle5im = f64x2_splat(tw5.im); |
| 7137 |
let twiddle6re = f64x2_splat(tw6.re); |
| 7138 |
let twiddle6im = f64x2_splat(tw6.im); |
| 7139 |
let twiddle7re = f64x2_splat(tw7.re); |
| 7140 |
let twiddle7im = f64x2_splat(tw7.im); |
| 7141 |
let twiddle8re = f64x2_splat(tw8.re); |
| 7142 |
let twiddle8im = f64x2_splat(tw8.im); |
| 7143 |
let twiddle9re = f64x2_splat(tw9.re); |
| 7144 |
let twiddle9im = f64x2_splat(tw9.im); |
| 7145 |
let twiddle10re = f64x2_splat(tw10.re); |
| 7146 |
let twiddle10im = f64x2_splat(tw10.im); |
| 7147 |
let twiddle11re = f64x2_splat(tw11.re); |
| 7148 |
let twiddle11im = f64x2_splat(tw11.im); |
| 7149 |
let twiddle12re = f64x2_splat(tw12.re); |
| 7150 |
let twiddle12im = f64x2_splat(tw12.im); |
| 7151 |
let twiddle13re = f64x2_splat(tw13.re); |
| 7152 |
let twiddle13im = f64x2_splat(tw13.im); |
| 7153 |
let twiddle14re = f64x2_splat(tw14.re); |
| 7154 |
let twiddle14im = f64x2_splat(tw14.im); |
| 7155 |
let twiddle15re = f64x2_splat(tw15.re); |
| 7156 |
let twiddle15im = f64x2_splat(tw15.im); |
| 7157 |
|
| 7158 |
Self { |
| 7159 |
direction, |
| 7160 |
_phantom: std::marker::PhantomData, |
| 7161 |
rotate, |
| 7162 |
twiddle1re, |
| 7163 |
twiddle1im, |
| 7164 |
twiddle2re, |
| 7165 |
twiddle2im, |
| 7166 |
twiddle3re, |
| 7167 |
twiddle3im, |
| 7168 |
twiddle4re, |
| 7169 |
twiddle4im, |
| 7170 |
twiddle5re, |
| 7171 |
twiddle5im, |
| 7172 |
twiddle6re, |
| 7173 |
twiddle6im, |
| 7174 |
twiddle7re, |
| 7175 |
twiddle7im, |
| 7176 |
twiddle8re, |
| 7177 |
twiddle8im, |
| 7178 |
twiddle9re, |
| 7179 |
twiddle9im, |
| 7180 |
twiddle10re, |
| 7181 |
twiddle10im, |
| 7182 |
twiddle11re, |
| 7183 |
twiddle11im, |
| 7184 |
twiddle12re, |
| 7185 |
twiddle12im, |
| 7186 |
twiddle13re, |
| 7187 |
twiddle13im, |
| 7188 |
twiddle14re, |
| 7189 |
twiddle14im, |
| 7190 |
twiddle15re, |
| 7191 |
twiddle15im, |
| 7192 |
} |
| 7193 |
} |
| 7194 |
|
| 7195 |
#[inline(always)] |
| 7196 |
pub(crate) unsafe fn perform_fft_contiguous(&self, mut buffer: impl WasmSimdArrayMut<f64>) { |
| 7197 |
let values = read_complex_to_array!(buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30}); |
| 7198 |
|
| 7199 |
let out = self.perform_fft_direct(values); |
| 7200 |
|
| 7201 |
write_complex_to_array!(out, buffer, {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30}); |
| 7202 |
} |
| 7203 |
|
| 7204 |
#[inline(always)] |
| 7205 |
pub(crate) unsafe fn perform_fft_direct(&self, values: [v128; 31]) -> [v128; 31] { |
| 7206 |
let [x1p30, x1m30] = solo_fft2_f64(values[1], values[30]); |
| 7207 |
let [x2p29, x2m29] = solo_fft2_f64(values[2], values[29]); |
| 7208 |
let [x3p28, x3m28] = solo_fft2_f64(values[3], values[28]); |
| 7209 |
let [x4p27, x4m27] = solo_fft2_f64(values[4], values[27]); |
| 7210 |
let [x5p26, x5m26] = solo_fft2_f64(values[5], values[26]); |
| 7211 |
let [x6p25, x6m25] = solo_fft2_f64(values[6], values[25]); |
| 7212 |
let [x7p24, x7m24] = solo_fft2_f64(values[7], values[24]); |
| 7213 |
let [x8p23, x8m23] = solo_fft2_f64(values[8], values[23]); |
| 7214 |
let [x9p22, x9m22] = solo_fft2_f64(values[9], values[22]); |
| 7215 |
let [x10p21, x10m21] = solo_fft2_f64(values[10], values[21]); |
| 7216 |
let [x11p20, x11m20] = solo_fft2_f64(values[11], values[20]); |
| 7217 |
let [x12p19, x12m19] = solo_fft2_f64(values[12], values[19]); |
| 7218 |
let [x13p18, x13m18] = solo_fft2_f64(values[13], values[18]); |
| 7219 |
let [x14p17, x14m17] = solo_fft2_f64(values[14], values[17]); |
| 7220 |
let [x15p16, x15m16] = solo_fft2_f64(values[15], values[16]); |
| 7221 |
|
| 7222 |
let t_a1_1 = f64x2_mul(self.twiddle1re, x1p30); |
| 7223 |
let t_a1_2 = f64x2_mul(self.twiddle2re, x2p29); |
| 7224 |
let t_a1_3 = f64x2_mul(self.twiddle3re, x3p28); |
| 7225 |
let t_a1_4 = f64x2_mul(self.twiddle4re, x4p27); |
| 7226 |
let t_a1_5 = f64x2_mul(self.twiddle5re, x5p26); |
| 7227 |
let t_a1_6 = f64x2_mul(self.twiddle6re, x6p25); |
| 7228 |
let t_a1_7 = f64x2_mul(self.twiddle7re, x7p24); |
| 7229 |
let t_a1_8 = f64x2_mul(self.twiddle8re, x8p23); |
| 7230 |
let t_a1_9 = f64x2_mul(self.twiddle9re, x9p22); |
| 7231 |
let t_a1_10 = f64x2_mul(self.twiddle10re, x10p21); |
| 7232 |
let t_a1_11 = f64x2_mul(self.twiddle11re, x11p20); |
| 7233 |
let t_a1_12 = f64x2_mul(self.twiddle12re, x12p19); |
| 7234 |
let t_a1_13 = f64x2_mul(self.twiddle13re, x13p18); |
| 7235 |
let t_a1_14 = f64x2_mul(self.twiddle14re, x14p17); |
| 7236 |
let t_a1_15 = f64x2_mul(self.twiddle15re, x15p16); |
| 7237 |
let t_a2_1 = f64x2_mul(self.twiddle2re, x1p30); |
| 7238 |
let t_a2_2 = f64x2_mul(self.twiddle4re, x2p29); |
| 7239 |
let t_a2_3 = f64x2_mul(self.twiddle6re, x3p28); |
| 7240 |
let t_a2_4 = f64x2_mul(self.twiddle8re, x4p27); |
| 7241 |
let t_a2_5 = f64x2_mul(self.twiddle10re, x5p26); |
| 7242 |
let t_a2_6 = f64x2_mul(self.twiddle12re, x6p25); |
| 7243 |
let t_a2_7 = f64x2_mul(self.twiddle14re, x7p24); |
| 7244 |
let t_a2_8 = f64x2_mul(self.twiddle15re, x8p23); |
| 7245 |
let t_a2_9 = f64x2_mul(self.twiddle13re, x9p22); |
| 7246 |
let t_a2_10 = f64x2_mul(self.twiddle11re, x10p21); |
| 7247 |
let t_a2_11 = f64x2_mul(self.twiddle9re, x11p20); |
| 7248 |
let t_a2_12 = f64x2_mul(self.twiddle7re, x12p19); |
| 7249 |
let t_a2_13 = f64x2_mul(self.twiddle5re, x13p18); |
| 7250 |
let t_a2_14 = f64x2_mul(self.twiddle3re, x14p17); |
| 7251 |
let t_a2_15 = f64x2_mul(self.twiddle1re, x15p16); |
| 7252 |
let t_a3_1 = f64x2_mul(self.twiddle3re, x1p30); |
| 7253 |
let t_a3_2 = f64x2_mul(self.twiddle6re, x2p29); |
| 7254 |
let t_a3_3 = f64x2_mul(self.twiddle9re, x3p28); |
| 7255 |
let t_a3_4 = f64x2_mul(self.twiddle12re, x4p27); |
| 7256 |
let t_a3_5 = f64x2_mul(self.twiddle15re, x5p26); |
| 7257 |
let t_a3_6 = f64x2_mul(self.twiddle13re, x6p25); |
| 7258 |
let t_a3_7 = f64x2_mul(self.twiddle10re, x7p24); |
| 7259 |
let t_a3_8 = f64x2_mul(self.twiddle7re, x8p23); |
| 7260 |
let t_a3_9 = f64x2_mul(self.twiddle4re, x9p22); |
| 7261 |
let t_a3_10 = f64x2_mul(self.twiddle1re, x10p21); |
| 7262 |
let t_a3_11 = f64x2_mul(self.twiddle2re, x11p20); |
| 7263 |
let t_a3_12 = f64x2_mul(self.twiddle5re, x12p19); |
| 7264 |
let t_a3_13 = f64x2_mul(self.twiddle8re, x13p18); |
| 7265 |
let t_a3_14 = f64x2_mul(self.twiddle11re, x14p17); |
| 7266 |
let t_a3_15 = f64x2_mul(self.twiddle14re, x15p16); |
| 7267 |
let t_a4_1 = f64x2_mul(self.twiddle4re, x1p30); |
| 7268 |
let t_a4_2 = f64x2_mul(self.twiddle8re, x2p29); |
| 7269 |
let t_a4_3 = f64x2_mul(self.twiddle12re, x3p28); |
| 7270 |
let t_a4_4 = f64x2_mul(self.twiddle15re, x4p27); |
| 7271 |
let t_a4_5 = f64x2_mul(self.twiddle11re, x5p26); |
| 7272 |
let t_a4_6 = f64x2_mul(self.twiddle7re, x6p25); |
| 7273 |
let t_a4_7 = f64x2_mul(self.twiddle3re, x7p24); |
| 7274 |
let t_a4_8 = f64x2_mul(self.twiddle1re, x8p23); |
| 7275 |
let t_a4_9 = f64x2_mul(self.twiddle5re, x9p22); |
| 7276 |
let t_a4_10 = f64x2_mul(self.twiddle9re, x10p21); |
| 7277 |
let t_a4_11 = f64x2_mul(self.twiddle13re, x11p20); |
| 7278 |
let t_a4_12 = f64x2_mul(self.twiddle14re, x12p19); |
| 7279 |
let t_a4_13 = f64x2_mul(self.twiddle10re, x13p18); |
| 7280 |
let t_a4_14 = f64x2_mul(self.twiddle6re, x14p17); |
| 7281 |
let t_a4_15 = f64x2_mul(self.twiddle2re, x15p16); |
| 7282 |
let t_a5_1 = f64x2_mul(self.twiddle5re, x1p30); |
| 7283 |
let t_a5_2 = f64x2_mul(self.twiddle10re, x2p29); |
| 7284 |
let t_a5_3 = f64x2_mul(self.twiddle15re, x3p28); |
| 7285 |
let t_a5_4 = f64x2_mul(self.twiddle11re, x4p27); |
| 7286 |
let t_a5_5 = f64x2_mul(self.twiddle6re, x5p26); |
| 7287 |
let t_a5_6 = f64x2_mul(self.twiddle1re, x6p25); |
| 7288 |
let t_a5_7 = f64x2_mul(self.twiddle4re, x7p24); |
| 7289 |
let t_a5_8 = f64x2_mul(self.twiddle9re, x8p23); |
| 7290 |
let t_a5_9 = f64x2_mul(self.twiddle14re, x9p22); |
| 7291 |
let t_a5_10 = f64x2_mul(self.twiddle12re, x10p21); |
| 7292 |
let t_a5_11 = f64x2_mul(self.twiddle7re, x11p20); |
| 7293 |
let t_a5_12 = f64x2_mul(self.twiddle2re, x12p19); |
| 7294 |
let t_a5_13 = f64x2_mul(self.twiddle3re, x13p18); |
| 7295 |
let t_a5_14 = f64x2_mul(self.twiddle8re, x14p17); |
| 7296 |
let t_a5_15 = f64x2_mul(self.twiddle13re, x15p16); |
| 7297 |
let t_a6_1 = f64x2_mul(self.twiddle6re, x1p30); |
| 7298 |
let t_a6_2 = f64x2_mul(self.twiddle12re, x2p29); |
| 7299 |
let t_a6_3 = f64x2_mul(self.twiddle13re, x3p28); |
| 7300 |
let t_a6_4 = f64x2_mul(self.twiddle7re, x4p27); |
| 7301 |
let t_a6_5 = f64x2_mul(self.twiddle1re, x5p26); |
| 7302 |
let t_a6_6 = f64x2_mul(self.twiddle5re, x6p25); |
| 7303 |
let t_a6_7 = f64x2_mul(self.twiddle11re, x7p24); |
| 7304 |
let t_a6_8 = f64x2_mul(self.twiddle14re, x8p23); |
| 7305 |
let t_a6_9 = f64x2_mul(self.twiddle8re, x9p22); |
| 7306 |
let t_a6_10 = f64x2_mul(self.twiddle2re, x10p21); |
| 7307 |
let t_a6_11 = f64x2_mul(self.twiddle4re, x11p20); |
| 7308 |
let t_a6_12 = f64x2_mul(self.twiddle10re, x12p19); |
| 7309 |
let t_a6_13 = f64x2_mul(self.twiddle15re, x13p18); |
| 7310 |
let t_a6_14 = f64x2_mul(self.twiddle9re, x14p17); |
| 7311 |
let t_a6_15 = f64x2_mul(self.twiddle3re, x15p16); |
| 7312 |
let t_a7_1 = f64x2_mul(self.twiddle7re, x1p30); |
| 7313 |
let t_a7_2 = f64x2_mul(self.twiddle14re, x2p29); |
| 7314 |
let t_a7_3 = f64x2_mul(self.twiddle10re, x3p28); |
| 7315 |
let t_a7_4 = f64x2_mul(self.twiddle3re, x4p27); |
| 7316 |
let t_a7_5 = f64x2_mul(self.twiddle4re, x5p26); |
| 7317 |
let t_a7_6 = f64x2_mul(self.twiddle11re, x6p25); |
| 7318 |
let t_a7_7 = f64x2_mul(self.twiddle13re, x7p24); |
| 7319 |
let t_a7_8 = f64x2_mul(self.twiddle6re, x8p23); |
| 7320 |
let t_a7_9 = f64x2_mul(self.twiddle1re, x9p22); |
| 7321 |
let t_a7_10 = f64x2_mul(self.twiddle8re, x10p21); |
| 7322 |
let t_a7_11 = f64x2_mul(self.twiddle15re, x11p20); |
| 7323 |
let t_a7_12 = f64x2_mul(self.twiddle9re, x12p19); |
| 7324 |
let t_a7_13 = f64x2_mul(self.twiddle2re, x13p18); |
| 7325 |
let t_a7_14 = f64x2_mul(self.twiddle5re, x14p17); |
| 7326 |
let t_a7_15 = f64x2_mul(self.twiddle12re, x15p16); |
| 7327 |
let t_a8_1 = f64x2_mul(self.twiddle8re, x1p30); |
| 7328 |
let t_a8_2 = f64x2_mul(self.twiddle15re, x2p29); |
| 7329 |
let t_a8_3 = f64x2_mul(self.twiddle7re, x3p28); |
| 7330 |
let t_a8_4 = f64x2_mul(self.twiddle1re, x4p27); |
| 7331 |
let t_a8_5 = f64x2_mul(self.twiddle9re, x5p26); |
| 7332 |
let t_a8_6 = f64x2_mul(self.twiddle14re, x6p25); |
| 7333 |
let t_a8_7 = f64x2_mul(self.twiddle6re, x7p24); |
| 7334 |
let t_a8_8 = f64x2_mul(self.twiddle2re, x8p23); |
| 7335 |
let t_a8_9 = f64x2_mul(self.twiddle10re, x9p22); |
| 7336 |
let t_a8_10 = f64x2_mul(self.twiddle13re, x10p21); |
| 7337 |
let t_a8_11 = f64x2_mul(self.twiddle5re, x11p20); |
| 7338 |
let t_a8_12 = f64x2_mul(self.twiddle3re, x12p19); |
| 7339 |
let t_a8_13 = f64x2_mul(self.twiddle11re, x13p18); |
| 7340 |
let t_a8_14 = f64x2_mul(self.twiddle12re, x14p17); |
| 7341 |
let t_a8_15 = f64x2_mul(self.twiddle4re, x15p16); |
| 7342 |
let t_a9_1 = f64x2_mul(self.twiddle9re, x1p30); |
| 7343 |
let t_a9_2 = f64x2_mul(self.twiddle13re, x2p29); |
| 7344 |
let t_a9_3 = f64x2_mul(self.twiddle4re, x3p28); |
| 7345 |
let t_a9_4 = f64x2_mul(self.twiddle5re, x4p27); |
| 7346 |
let t_a9_5 = f64x2_mul(self.twiddle14re, x5p26); |
| 7347 |
let t_a9_6 = f64x2_mul(self.twiddle8re, x6p25); |
| 7348 |
let t_a9_7 = f64x2_mul(self.twiddle1re, x7p24); |
| 7349 |
let t_a9_8 = f64x2_mul(self.twiddle10re, x8p23); |
| 7350 |
let t_a9_9 = f64x2_mul(self.twiddle12re, x9p22); |
| 7351 |
let t_a9_10 = f64x2_mul(self.twiddle3re, x10p21); |
| 7352 |
let t_a9_11 = f64x2_mul(self.twiddle6re, x11p20); |
| 7353 |
let t_a9_12 = f64x2_mul(self.twiddle15re, x12p19); |
| 7354 |
let t_a9_13 = f64x2_mul(self.twiddle7re, x13p18); |
| 7355 |
let t_a9_14 = f64x2_mul(self.twiddle2re, x14p17); |
| 7356 |
let t_a9_15 = f64x2_mul(self.twiddle11re, x15p16); |
| 7357 |
let t_a10_1 = f64x2_mul(self.twiddle10re, x1p30); |
| 7358 |
let t_a10_2 = f64x2_mul(self.twiddle11re, x2p29); |
| 7359 |
let t_a10_3 = f64x2_mul(self.twiddle1re, x3p28); |
| 7360 |
let t_a10_4 = f64x2_mul(self.twiddle9re, x4p27); |
| 7361 |
let t_a10_5 = f64x2_mul(self.twiddle12re, x5p26); |
| 7362 |
let t_a10_6 = f64x2_mul(self.twiddle2re, x6p25); |
| 7363 |
let t_a10_7 = f64x2_mul(self.twiddle8re, x7p24); |
| 7364 |
let t_a10_8 = f64x2_mul(self.twiddle13re, x8p23); |
| 7365 |
let t_a10_9 = f64x2_mul(self.twiddle3re, x9p22); |
| 7366 |
let t_a10_10 = f64x2_mul(self.twiddle7re, x10p21); |
| 7367 |
let t_a10_11 = f64x2_mul(self.twiddle14re, x11p20); |
| 7368 |
let t_a10_12 = f64x2_mul(self.twiddle4re, x12p19); |
| 7369 |
let t_a10_13 = f64x2_mul(self.twiddle6re, x13p18); |
| 7370 |
let t_a10_14 = f64x2_mul(self.twiddle15re, x14p17); |
| 7371 |
let t_a10_15 = f64x2_mul(self.twiddle5re, x15p16); |
| 7372 |
let t_a11_1 = f64x2_mul(self.twiddle11re, x1p30); |
| 7373 |
let t_a11_2 = f64x2_mul(self.twiddle9re, x2p29); |
| 7374 |
let t_a11_3 = f64x2_mul(self.twiddle2re, x3p28); |
| 7375 |
let t_a11_4 = f64x2_mul(self.twiddle13re, x4p27); |
| 7376 |
let t_a11_5 = f64x2_mul(self.twiddle7re, x5p26); |
| 7377 |
let t_a11_6 = f64x2_mul(self.twiddle4re, x6p25); |
| 7378 |
let t_a11_7 = f64x2_mul(self.twiddle15re, x7p24); |
| 7379 |
let t_a11_8 = f64x2_mul(self.twiddle5re, x8p23); |
| 7380 |
let t_a11_9 = f64x2_mul(self.twiddle6re, x9p22); |
| 7381 |
let t_a11_10 = f64x2_mul(self.twiddle14re, x10p21); |
| 7382 |
let t_a11_11 = f64x2_mul(self.twiddle3re, x11p20); |
| 7383 |
let t_a11_12 = f64x2_mul(self.twiddle8re, x12p19); |
| 7384 |
let t_a11_13 = f64x2_mul(self.twiddle12re, x13p18); |
| 7385 |
let t_a11_14 = f64x2_mul(self.twiddle1re, x14p17); |
| 7386 |
let t_a11_15 = f64x2_mul(self.twiddle10re, x15p16); |
| 7387 |
let t_a12_1 = f64x2_mul(self.twiddle12re, x1p30); |
| 7388 |
let t_a12_2 = f64x2_mul(self.twiddle7re, x2p29); |
| 7389 |
let t_a12_3 = f64x2_mul(self.twiddle5re, x3p28); |
| 7390 |
let t_a12_4 = f64x2_mul(self.twiddle14re, x4p27); |
| 7391 |
let t_a12_5 = f64x2_mul(self.twiddle2re, x5p26); |
| 7392 |
let t_a12_6 = f64x2_mul(self.twiddle10re, x6p25); |
| 7393 |
let t_a12_7 = f64x2_mul(self.twiddle9re, x7p24); |
| 7394 |
let t_a12_8 = f64x2_mul(self.twiddle3re, x8p23); |
| 7395 |
let t_a12_9 = f64x2_mul(self.twiddle15re, x9p22); |
| 7396 |
let t_a12_10 = f64x2_mul(self.twiddle4re, x10p21); |
| 7397 |
let t_a12_11 = f64x2_mul(self.twiddle8re, x11p20); |
| 7398 |
let t_a12_12 = f64x2_mul(self.twiddle11re, x12p19); |
| 7399 |
let t_a12_13 = f64x2_mul(self.twiddle1re, x13p18); |
| 7400 |
let t_a12_14 = f64x2_mul(self.twiddle13re, x14p17); |
| 7401 |
let t_a12_15 = f64x2_mul(self.twiddle6re, x15p16); |
| 7402 |
let t_a13_1 = f64x2_mul(self.twiddle13re, x1p30); |
| 7403 |
let t_a13_2 = f64x2_mul(self.twiddle5re, x2p29); |
| 7404 |
let t_a13_3 = f64x2_mul(self.twiddle8re, x3p28); |
| 7405 |
let t_a13_4 = f64x2_mul(self.twiddle10re, x4p27); |
| 7406 |
let t_a13_5 = f64x2_mul(self.twiddle3re, x5p26); |
| 7407 |
let t_a13_6 = f64x2_mul(self.twiddle15re, x6p25); |
| 7408 |
let t_a13_7 = f64x2_mul(self.twiddle2re, x7p24); |
| 7409 |
let t_a13_8 = f64x2_mul(self.twiddle11re, x8p23); |
| 7410 |
let t_a13_9 = f64x2_mul(self.twiddle7re, x9p22); |
| 7411 |
let t_a13_10 = f64x2_mul(self.twiddle6re, x10p21); |
| 7412 |
let t_a13_11 = f64x2_mul(self.twiddle12re, x11p20); |
| 7413 |
let t_a13_12 = f64x2_mul(self.twiddle1re, x12p19); |
| 7414 |
let t_a13_13 = f64x2_mul(self.twiddle14re, x13p18); |
| 7415 |
let t_a13_14 = f64x2_mul(self.twiddle4re, x14p17); |
| 7416 |
let t_a13_15 = f64x2_mul(self.twiddle9re, x15p16); |
| 7417 |
let t_a14_1 = f64x2_mul(self.twiddle14re, x1p30); |
| 7418 |
let t_a14_2 = f64x2_mul(self.twiddle3re, x2p29); |
| 7419 |
let t_a14_3 = f64x2_mul(self.twiddle11re, x3p28); |
| 7420 |
let t_a14_4 = f64x2_mul(self.twiddle6re, x4p27); |
| 7421 |
let t_a14_5 = f64x2_mul(self.twiddle8re, x5p26); |
| 7422 |
let t_a14_6 = f64x2_mul(self.twiddle9re, x6p25); |
| 7423 |
let t_a14_7 = f64x2_mul(self.twiddle5re, x7p24); |
| 7424 |
let t_a14_8 = f64x2_mul(self.twiddle12re, x8p23); |
| 7425 |
let t_a14_9 = f64x2_mul(self.twiddle2re, x9p22); |
| 7426 |
let t_a14_10 = f64x2_mul(self.twiddle15re, x10p21); |
| 7427 |
let t_a14_11 = f64x2_mul(self.twiddle1re, x11p20); |
| 7428 |
let t_a14_12 = f64x2_mul(self.twiddle13re, x12p19); |
| 7429 |
let t_a14_13 = f64x2_mul(self.twiddle4re, x13p18); |
| 7430 |
let t_a14_14 = f64x2_mul(self.twiddle10re, x14p17); |
| 7431 |
let t_a14_15 = f64x2_mul(self.twiddle7re, x15p16); |
| 7432 |
let t_a15_1 = f64x2_mul(self.twiddle15re, x1p30); |
| 7433 |
let t_a15_2 = f64x2_mul(self.twiddle1re, x2p29); |
| 7434 |
let t_a15_3 = f64x2_mul(self.twiddle14re, x3p28); |
| 7435 |
let t_a15_4 = f64x2_mul(self.twiddle2re, x4p27); |
| 7436 |
let t_a15_5 = f64x2_mul(self.twiddle13re, x5p26); |
| 7437 |
let t_a15_6 = f64x2_mul(self.twiddle3re, x6p25); |
| 7438 |
let t_a15_7 = f64x2_mul(self.twiddle12re, x7p24); |
| 7439 |
let t_a15_8 = f64x2_mul(self.twiddle4re, x8p23); |
| 7440 |
let t_a15_9 = f64x2_mul(self.twiddle11re, x9p22); |
| 7441 |
let t_a15_10 = f64x2_mul(self.twiddle5re, x10p21); |
| 7442 |
let t_a15_11 = f64x2_mul(self.twiddle10re, x11p20); |
| 7443 |
let t_a15_12 = f64x2_mul(self.twiddle6re, x12p19); |
| 7444 |
let t_a15_13 = f64x2_mul(self.twiddle9re, x13p18); |
| 7445 |
let t_a15_14 = f64x2_mul(self.twiddle7re, x14p17); |
| 7446 |
let t_a15_15 = f64x2_mul(self.twiddle8re, x15p16); |
| 7447 |
|
| 7448 |
let t_b1_1 = f64x2_mul(self.twiddle1im, x1m30); |
| 7449 |
let t_b1_2 = f64x2_mul(self.twiddle2im, x2m29); |
| 7450 |
let t_b1_3 = f64x2_mul(self.twiddle3im, x3m28); |
| 7451 |
let t_b1_4 = f64x2_mul(self.twiddle4im, x4m27); |
| 7452 |
let t_b1_5 = f64x2_mul(self.twiddle5im, x5m26); |
| 7453 |
let t_b1_6 = f64x2_mul(self.twiddle6im, x6m25); |
| 7454 |
let t_b1_7 = f64x2_mul(self.twiddle7im, x7m24); |
| 7455 |
let t_b1_8 = f64x2_mul(self.twiddle8im, x8m23); |
| 7456 |
let t_b1_9 = f64x2_mul(self.twiddle9im, x9m22); |
| 7457 |
let t_b1_10 = f64x2_mul(self.twiddle10im, x10m21); |
| 7458 |
let t_b1_11 = f64x2_mul(self.twiddle11im, x11m20); |
| 7459 |
let t_b1_12 = f64x2_mul(self.twiddle12im, x12m19); |
| 7460 |
let t_b1_13 = f64x2_mul(self.twiddle13im, x13m18); |
| 7461 |
let t_b1_14 = f64x2_mul(self.twiddle14im, x14m17); |
| 7462 |
let t_b1_15 = f64x2_mul(self.twiddle15im, x15m16); |
| 7463 |
let t_b2_1 = f64x2_mul(self.twiddle2im, x1m30); |
| 7464 |
let t_b2_2 = f64x2_mul(self.twiddle4im, x2m29); |
| 7465 |
let t_b2_3 = f64x2_mul(self.twiddle6im, x3m28); |
| 7466 |
let t_b2_4 = f64x2_mul(self.twiddle8im, x4m27); |
| 7467 |
let t_b2_5 = f64x2_mul(self.twiddle10im, x5m26); |
| 7468 |
let t_b2_6 = f64x2_mul(self.twiddle12im, x6m25); |
| 7469 |
let t_b2_7 = f64x2_mul(self.twiddle14im, x7m24); |
| 7470 |
let t_b2_8 = f64x2_mul(self.twiddle15im, x8m23); |
| 7471 |
let t_b2_9 = f64x2_mul(self.twiddle13im, x9m22); |
| 7472 |
let t_b2_10 = f64x2_mul(self.twiddle11im, x10m21); |
| 7473 |
let t_b2_11 = f64x2_mul(self.twiddle9im, x11m20); |
| 7474 |
let t_b2_12 = f64x2_mul(self.twiddle7im, x12m19); |
| 7475 |
let t_b2_13 = f64x2_mul(self.twiddle5im, x13m18); |
| 7476 |
let t_b2_14 = f64x2_mul(self.twiddle3im, x14m17); |
| 7477 |
let t_b2_15 = f64x2_mul(self.twiddle1im, x15m16); |
| 7478 |
let t_b3_1 = f64x2_mul(self.twiddle3im, x1m30); |
| 7479 |
let t_b3_2 = f64x2_mul(self.twiddle6im, x2m29); |
| 7480 |
let t_b3_3 = f64x2_mul(self.twiddle9im, x3m28); |
| 7481 |
let t_b3_4 = f64x2_mul(self.twiddle12im, x4m27); |
| 7482 |
let t_b3_5 = f64x2_mul(self.twiddle15im, x5m26); |
| 7483 |
let t_b3_6 = f64x2_mul(self.twiddle13im, x6m25); |
| 7484 |
let t_b3_7 = f64x2_mul(self.twiddle10im, x7m24); |
| 7485 |
let t_b3_8 = f64x2_mul(self.twiddle7im, x8m23); |
| 7486 |
let t_b3_9 = f64x2_mul(self.twiddle4im, x9m22); |
| 7487 |
let t_b3_10 = f64x2_mul(self.twiddle1im, x10m21); |
| 7488 |
let t_b3_11 = f64x2_mul(self.twiddle2im, x11m20); |
| 7489 |
let t_b3_12 = f64x2_mul(self.twiddle5im, x12m19); |
| 7490 |
let t_b3_13 = f64x2_mul(self.twiddle8im, x13m18); |
| 7491 |
let t_b3_14 = f64x2_mul(self.twiddle11im, x14m17); |
| 7492 |
let t_b3_15 = f64x2_mul(self.twiddle14im, x15m16); |
| 7493 |
let t_b4_1 = f64x2_mul(self.twiddle4im, x1m30); |
| 7494 |
let t_b4_2 = f64x2_mul(self.twiddle8im, x2m29); |
| 7495 |
let t_b4_3 = f64x2_mul(self.twiddle12im, x3m28); |
| 7496 |
let t_b4_4 = f64x2_mul(self.twiddle15im, x4m27); |
| 7497 |
let t_b4_5 = f64x2_mul(self.twiddle11im, x5m26); |
| 7498 |
let t_b4_6 = f64x2_mul(self.twiddle7im, x6m25); |
| 7499 |
let t_b4_7 = f64x2_mul(self.twiddle3im, x7m24); |
| 7500 |
let t_b4_8 = f64x2_mul(self.twiddle1im, x8m23); |
| 7501 |
let t_b4_9 = f64x2_mul(self.twiddle5im, x9m22); |
| 7502 |
let t_b4_10 = f64x2_mul(self.twiddle9im, x10m21); |
| 7503 |
let t_b4_11 = f64x2_mul(self.twiddle13im, x11m20); |
| 7504 |
let t_b4_12 = f64x2_mul(self.twiddle14im, x12m19); |
| 7505 |
let t_b4_13 = f64x2_mul(self.twiddle10im, x13m18); |
| 7506 |
let t_b4_14 = f64x2_mul(self.twiddle6im, x14m17); |
| 7507 |
let t_b4_15 = f64x2_mul(self.twiddle2im, x15m16); |
| 7508 |
let t_b5_1 = f64x2_mul(self.twiddle5im, x1m30); |
| 7509 |
let t_b5_2 = f64x2_mul(self.twiddle10im, x2m29); |
| 7510 |
let t_b5_3 = f64x2_mul(self.twiddle15im, x3m28); |
| 7511 |
let t_b5_4 = f64x2_mul(self.twiddle11im, x4m27); |
| 7512 |
let t_b5_5 = f64x2_mul(self.twiddle6im, x5m26); |
| 7513 |
let t_b5_6 = f64x2_mul(self.twiddle1im, x6m25); |
| 7514 |
let t_b5_7 = f64x2_mul(self.twiddle4im, x7m24); |
| 7515 |
let t_b5_8 = f64x2_mul(self.twiddle9im, x8m23); |
| 7516 |
let t_b5_9 = f64x2_mul(self.twiddle14im, x9m22); |
| 7517 |
let t_b5_10 = f64x2_mul(self.twiddle12im, x10m21); |
| 7518 |
let t_b5_11 = f64x2_mul(self.twiddle7im, x11m20); |
| 7519 |
let t_b5_12 = f64x2_mul(self.twiddle2im, x12m19); |
| 7520 |
let t_b5_13 = f64x2_mul(self.twiddle3im, x13m18); |
| 7521 |
let t_b5_14 = f64x2_mul(self.twiddle8im, x14m17); |
| 7522 |
let t_b5_15 = f64x2_mul(self.twiddle13im, x15m16); |
| 7523 |
let t_b6_1 = f64x2_mul(self.twiddle6im, x1m30); |
| 7524 |
let t_b6_2 = f64x2_mul(self.twiddle12im, x2m29); |
| 7525 |
let t_b6_3 = f64x2_mul(self.twiddle13im, x3m28); |
| 7526 |
let t_b6_4 = f64x2_mul(self.twiddle7im, x4m27); |
| 7527 |
let t_b6_5 = f64x2_mul(self.twiddle1im, x5m26); |
| 7528 |
let t_b6_6 = f64x2_mul(self.twiddle5im, x6m25); |
| 7529 |
let t_b6_7 = f64x2_mul(self.twiddle11im, x7m24); |
| 7530 |
let t_b6_8 = f64x2_mul(self.twiddle14im, x8m23); |
| 7531 |
let t_b6_9 = f64x2_mul(self.twiddle8im, x9m22); |
| 7532 |
let t_b6_10 = f64x2_mul(self.twiddle2im, x10m21); |
| 7533 |
let t_b6_11 = f64x2_mul(self.twiddle4im, x11m20); |
| 7534 |
let t_b6_12 = f64x2_mul(self.twiddle10im, x12m19); |
| 7535 |
let t_b6_13 = f64x2_mul(self.twiddle15im, x13m18); |
| 7536 |
let t_b6_14 = f64x2_mul(self.twiddle9im, x14m17); |
| 7537 |
let t_b6_15 = f64x2_mul(self.twiddle3im, x15m16); |
| 7538 |
let t_b7_1 = f64x2_mul(self.twiddle7im, x1m30); |
| 7539 |
let t_b7_2 = f64x2_mul(self.twiddle14im, x2m29); |
| 7540 |
let t_b7_3 = f64x2_mul(self.twiddle10im, x3m28); |
| 7541 |
let t_b7_4 = f64x2_mul(self.twiddle3im, x4m27); |
| 7542 |
let t_b7_5 = f64x2_mul(self.twiddle4im, x5m26); |
| 7543 |
let t_b7_6 = f64x2_mul(self.twiddle11im, x6m25); |
| 7544 |
let t_b7_7 = f64x2_mul(self.twiddle13im, x7m24); |
| 7545 |
let t_b7_8 = f64x2_mul(self.twiddle6im, x8m23); |
| 7546 |
let t_b7_9 = f64x2_mul(self.twiddle1im, x9m22); |
| 7547 |
let t_b7_10 = f64x2_mul(self.twiddle8im, x10m21); |
| 7548 |
let t_b7_11 = f64x2_mul(self.twiddle15im, x11m20); |
| 7549 |
let t_b7_12 = f64x2_mul(self.twiddle9im, x12m19); |
| 7550 |
let t_b7_13 = f64x2_mul(self.twiddle2im, x13m18); |
| 7551 |
let t_b7_14 = f64x2_mul(self.twiddle5im, x14m17); |
| 7552 |
let t_b7_15 = f64x2_mul(self.twiddle12im, x15m16); |
| 7553 |
let t_b8_1 = f64x2_mul(self.twiddle8im, x1m30); |
| 7554 |
let t_b8_2 = f64x2_mul(self.twiddle15im, x2m29); |
| 7555 |
let t_b8_3 = f64x2_mul(self.twiddle7im, x3m28); |
| 7556 |
let t_b8_4 = f64x2_mul(self.twiddle1im, x4m27); |
| 7557 |
let t_b8_5 = f64x2_mul(self.twiddle9im, x5m26); |
| 7558 |
let t_b8_6 = f64x2_mul(self.twiddle14im, x6m25); |
| 7559 |
let t_b8_7 = f64x2_mul(self.twiddle6im, x7m24); |
| 7560 |
let t_b8_8 = f64x2_mul(self.twiddle2im, x8m23); |
| 7561 |
let t_b8_9 = f64x2_mul(self.twiddle10im, x9m22); |
| 7562 |
let t_b8_10 = f64x2_mul(self.twiddle13im, x10m21); |
| 7563 |
let t_b8_11 = f64x2_mul(self.twiddle5im, x11m20); |
| 7564 |
let t_b8_12 = f64x2_mul(self.twiddle3im, x12m19); |
| 7565 |
let t_b8_13 = f64x2_mul(self.twiddle11im, x13m18); |
| 7566 |
let t_b8_14 = f64x2_mul(self.twiddle12im, x14m17); |
| 7567 |
let t_b8_15 = f64x2_mul(self.twiddle4im, x15m16); |
| 7568 |
let t_b9_1 = f64x2_mul(self.twiddle9im, x1m30); |
| 7569 |
let t_b9_2 = f64x2_mul(self.twiddle13im, x2m29); |
| 7570 |
let t_b9_3 = f64x2_mul(self.twiddle4im, x3m28); |
| 7571 |
let t_b9_4 = f64x2_mul(self.twiddle5im, x4m27); |
| 7572 |
let t_b9_5 = f64x2_mul(self.twiddle14im, x5m26); |
| 7573 |
let t_b9_6 = f64x2_mul(self.twiddle8im, x6m25); |
| 7574 |
let t_b9_7 = f64x2_mul(self.twiddle1im, x7m24); |
| 7575 |
let t_b9_8 = f64x2_mul(self.twiddle10im, x8m23); |
| 7576 |
let t_b9_9 = f64x2_mul(self.twiddle12im, x9m22); |
| 7577 |
let t_b9_10 = f64x2_mul(self.twiddle3im, x10m21); |
| 7578 |
let t_b9_11 = f64x2_mul(self.twiddle6im, x11m20); |
| 7579 |
let t_b9_12 = f64x2_mul(self.twiddle15im, x12m19); |
| 7580 |
let t_b9_13 = f64x2_mul(self.twiddle7im, x13m18); |
| 7581 |
let t_b9_14 = f64x2_mul(self.twiddle2im, x14m17); |
| 7582 |
let t_b9_15 = f64x2_mul(self.twiddle11im, x15m16); |
| 7583 |
let t_b10_1 = f64x2_mul(self.twiddle10im, x1m30); |
| 7584 |
let t_b10_2 = f64x2_mul(self.twiddle11im, x2m29); |
| 7585 |
let t_b10_3 = f64x2_mul(self.twiddle1im, x3m28); |
| 7586 |
let t_b10_4 = f64x2_mul(self.twiddle9im, x4m27); |
| 7587 |
let t_b10_5 = f64x2_mul(self.twiddle12im, x5m26); |
| 7588 |
let t_b10_6 = f64x2_mul(self.twiddle2im, x6m25); |
| 7589 |
let t_b10_7 = f64x2_mul(self.twiddle8im, x7m24); |
| 7590 |
let t_b10_8 = f64x2_mul(self.twiddle13im, x8m23); |
| 7591 |
let t_b10_9 = f64x2_mul(self.twiddle3im, x9m22); |
| 7592 |
let t_b10_10 = f64x2_mul(self.twiddle7im, x10m21); |
| 7593 |
let t_b10_11 = f64x2_mul(self.twiddle14im, x11m20); |
| 7594 |
let t_b10_12 = f64x2_mul(self.twiddle4im, x12m19); |
| 7595 |
let t_b10_13 = f64x2_mul(self.twiddle6im, x13m18); |
| 7596 |
let t_b10_14 = f64x2_mul(self.twiddle15im, x14m17); |
| 7597 |
let t_b10_15 = f64x2_mul(self.twiddle5im, x15m16); |
| 7598 |
let t_b11_1 = f64x2_mul(self.twiddle11im, x1m30); |
| 7599 |
let t_b11_2 = f64x2_mul(self.twiddle9im, x2m29); |
| 7600 |
let t_b11_3 = f64x2_mul(self.twiddle2im, x3m28); |
| 7601 |
let t_b11_4 = f64x2_mul(self.twiddle13im, x4m27); |
| 7602 |
let t_b11_5 = f64x2_mul(self.twiddle7im, x5m26); |
| 7603 |
let t_b11_6 = f64x2_mul(self.twiddle4im, x6m25); |
| 7604 |
let t_b11_7 = f64x2_mul(self.twiddle15im, x7m24); |
| 7605 |
let t_b11_8 = f64x2_mul(self.twiddle5im, x8m23); |
| 7606 |
let t_b11_9 = f64x2_mul(self.twiddle6im, x9m22); |
| 7607 |
let t_b11_10 = f64x2_mul(self.twiddle14im, x10m21); |
| 7608 |
let t_b11_11 = f64x2_mul(self.twiddle3im, x11m20); |
| 7609 |
let t_b11_12 = f64x2_mul(self.twiddle8im, x12m19); |
| 7610 |
let t_b11_13 = f64x2_mul(self.twiddle12im, x13m18); |
| 7611 |
let t_b11_14 = f64x2_mul(self.twiddle1im, x14m17); |
| 7612 |
let t_b11_15 = f64x2_mul(self.twiddle10im, x15m16); |
| 7613 |
let t_b12_1 = f64x2_mul(self.twiddle12im, x1m30); |
| 7614 |
let t_b12_2 = f64x2_mul(self.twiddle7im, x2m29); |
| 7615 |
let t_b12_3 = f64x2_mul(self.twiddle5im, x3m28); |
| 7616 |
let t_b12_4 = f64x2_mul(self.twiddle14im, x4m27); |
| 7617 |
let t_b12_5 = f64x2_mul(self.twiddle2im, x5m26); |
| 7618 |
let t_b12_6 = f64x2_mul(self.twiddle10im, x6m25); |
| 7619 |
let t_b12_7 = f64x2_mul(self.twiddle9im, x7m24); |
| 7620 |
let t_b12_8 = f64x2_mul(self.twiddle3im, x8m23); |
| 7621 |
let t_b12_9 = f64x2_mul(self.twiddle15im, x9m22); |
| 7622 |
let t_b12_10 = f64x2_mul(self.twiddle4im, x10m21); |
| 7623 |
let t_b12_11 = f64x2_mul(self.twiddle8im, x11m20); |
| 7624 |
let t_b12_12 = f64x2_mul(self.twiddle11im, x12m19); |
| 7625 |
let t_b12_13 = f64x2_mul(self.twiddle1im, x13m18); |
| 7626 |
let t_b12_14 = f64x2_mul(self.twiddle13im, x14m17); |
| 7627 |
let t_b12_15 = f64x2_mul(self.twiddle6im, x15m16); |
| 7628 |
let t_b13_1 = f64x2_mul(self.twiddle13im, x1m30); |
| 7629 |
let t_b13_2 = f64x2_mul(self.twiddle5im, x2m29); |
| 7630 |
let t_b13_3 = f64x2_mul(self.twiddle8im, x3m28); |
| 7631 |
let t_b13_4 = f64x2_mul(self.twiddle10im, x4m27); |
| 7632 |
let t_b13_5 = f64x2_mul(self.twiddle3im, x5m26); |
| 7633 |
let t_b13_6 = f64x2_mul(self.twiddle15im, x6m25); |
| 7634 |
let t_b13_7 = f64x2_mul(self.twiddle2im, x7m24); |
| 7635 |
let t_b13_8 = f64x2_mul(self.twiddle11im, x8m23); |
| 7636 |
let t_b13_9 = f64x2_mul(self.twiddle7im, x9m22); |
| 7637 |
let t_b13_10 = f64x2_mul(self.twiddle6im, x10m21); |
| 7638 |
let t_b13_11 = f64x2_mul(self.twiddle12im, x11m20); |
| 7639 |
let t_b13_12 = f64x2_mul(self.twiddle1im, x12m19); |
| 7640 |
let t_b13_13 = f64x2_mul(self.twiddle14im, x13m18); |
| 7641 |
let t_b13_14 = f64x2_mul(self.twiddle4im, x14m17); |
| 7642 |
let t_b13_15 = f64x2_mul(self.twiddle9im, x15m16); |
| 7643 |
let t_b14_1 = f64x2_mul(self.twiddle14im, x1m30); |
| 7644 |
let t_b14_2 = f64x2_mul(self.twiddle3im, x2m29); |
| 7645 |
let t_b14_3 = f64x2_mul(self.twiddle11im, x3m28); |
| 7646 |
let t_b14_4 = f64x2_mul(self.twiddle6im, x4m27); |
| 7647 |
let t_b14_5 = f64x2_mul(self.twiddle8im, x5m26); |
| 7648 |
let t_b14_6 = f64x2_mul(self.twiddle9im, x6m25); |
| 7649 |
let t_b14_7 = f64x2_mul(self.twiddle5im, x7m24); |
| 7650 |
let t_b14_8 = f64x2_mul(self.twiddle12im, x8m23); |
| 7651 |
let t_b14_9 = f64x2_mul(self.twiddle2im, x9m22); |
| 7652 |
let t_b14_10 = f64x2_mul(self.twiddle15im, x10m21); |
| 7653 |
let t_b14_11 = f64x2_mul(self.twiddle1im, x11m20); |
| 7654 |
let t_b14_12 = f64x2_mul(self.twiddle13im, x12m19); |
| 7655 |
let t_b14_13 = f64x2_mul(self.twiddle4im, x13m18); |
| 7656 |
let t_b14_14 = f64x2_mul(self.twiddle10im, x14m17); |
| 7657 |
let t_b14_15 = f64x2_mul(self.twiddle7im, x15m16); |
| 7658 |
let t_b15_1 = f64x2_mul(self.twiddle15im, x1m30); |
| 7659 |
let t_b15_2 = f64x2_mul(self.twiddle1im, x2m29); |
| 7660 |
let t_b15_3 = f64x2_mul(self.twiddle14im, x3m28); |
| 7661 |
let t_b15_4 = f64x2_mul(self.twiddle2im, x4m27); |
| 7662 |
let t_b15_5 = f64x2_mul(self.twiddle13im, x5m26); |
| 7663 |
let t_b15_6 = f64x2_mul(self.twiddle3im, x6m25); |
| 7664 |
let t_b15_7 = f64x2_mul(self.twiddle12im, x7m24); |
| 7665 |
let t_b15_8 = f64x2_mul(self.twiddle4im, x8m23); |
| 7666 |
let t_b15_9 = f64x2_mul(self.twiddle11im, x9m22); |
| 7667 |
let t_b15_10 = f64x2_mul(self.twiddle5im, x10m21); |
| 7668 |
let t_b15_11 = f64x2_mul(self.twiddle10im, x11m20); |
| 7669 |
let t_b15_12 = f64x2_mul(self.twiddle6im, x12m19); |
| 7670 |
let t_b15_13 = f64x2_mul(self.twiddle9im, x13m18); |
| 7671 |
let t_b15_14 = f64x2_mul(self.twiddle7im, x14m17); |
| 7672 |
let t_b15_15 = f64x2_mul(self.twiddle8im, x15m16); |
| 7673 |
|
| 7674 |
let x0 = values[0]; |
| 7675 |
let t_a1 = calc_f64!( |
| 7676 |
x0 + t_a1_1 |
| 7677 |
+ t_a1_2 |
| 7678 |
+ t_a1_3 |
| 7679 |
+ t_a1_4 |
| 7680 |
+ t_a1_5 |
| 7681 |
+ t_a1_6 |
| 7682 |
+ t_a1_7 |
| 7683 |
+ t_a1_8 |
| 7684 |
+ t_a1_9 |
| 7685 |
+ t_a1_10 |
| 7686 |
+ t_a1_11 |
| 7687 |
+ t_a1_12 |
| 7688 |
+ t_a1_13 |
| 7689 |
+ t_a1_14 |
| 7690 |
+ t_a1_15 |
| 7691 |
); |
| 7692 |
let t_a2 = calc_f64!( |
| 7693 |
x0 + t_a2_1 |
| 7694 |
+ t_a2_2 |
| 7695 |
+ t_a2_3 |
| 7696 |
+ t_a2_4 |
| 7697 |
+ t_a2_5 |
| 7698 |
+ t_a2_6 |
| 7699 |
+ t_a2_7 |
| 7700 |
+ t_a2_8 |
| 7701 |
+ t_a2_9 |
| 7702 |
+ t_a2_10 |
| 7703 |
+ t_a2_11 |
| 7704 |
+ t_a2_12 |
| 7705 |
+ t_a2_13 |
| 7706 |
+ t_a2_14 |
| 7707 |
+ t_a2_15 |
| 7708 |
); |
| 7709 |
let t_a3 = calc_f64!( |
| 7710 |
x0 + t_a3_1 |
| 7711 |
+ t_a3_2 |
| 7712 |
+ t_a3_3 |
| 7713 |
+ t_a3_4 |
| 7714 |
+ t_a3_5 |
| 7715 |
+ t_a3_6 |
| 7716 |
+ t_a3_7 |
| 7717 |
+ t_a3_8 |
| 7718 |
+ t_a3_9 |
| 7719 |
+ t_a3_10 |
| 7720 |
+ t_a3_11 |
| 7721 |
+ t_a3_12 |
| 7722 |
+ t_a3_13 |
| 7723 |
+ t_a3_14 |
| 7724 |
+ t_a3_15 |
| 7725 |
); |
| 7726 |
let t_a4 = calc_f64!( |
| 7727 |
x0 + t_a4_1 |
| 7728 |
+ t_a4_2 |
| 7729 |
+ t_a4_3 |
| 7730 |
+ t_a4_4 |
| 7731 |
+ t_a4_5 |
| 7732 |
+ t_a4_6 |
| 7733 |
+ t_a4_7 |
| 7734 |
+ t_a4_8 |
| 7735 |
+ t_a4_9 |
| 7736 |
+ t_a4_10 |
| 7737 |
+ t_a4_11 |
| 7738 |
+ t_a4_12 |
| 7739 |
+ t_a4_13 |
| 7740 |
+ t_a4_14 |
| 7741 |
+ t_a4_15 |
| 7742 |
); |
| 7743 |
let t_a5 = calc_f64!( |
| 7744 |
x0 + t_a5_1 |
| 7745 |
+ t_a5_2 |
| 7746 |
+ t_a5_3 |
| 7747 |
+ t_a5_4 |
| 7748 |
+ t_a5_5 |
| 7749 |
+ t_a5_6 |
| 7750 |
+ t_a5_7 |
| 7751 |
+ t_a5_8 |
| 7752 |
+ t_a5_9 |
| 7753 |
+ t_a5_10 |
| 7754 |
+ t_a5_11 |
| 7755 |
+ t_a5_12 |
| 7756 |
+ t_a5_13 |
| 7757 |
+ t_a5_14 |
| 7758 |
+ t_a5_15 |
| 7759 |
); |
| 7760 |
let t_a6 = calc_f64!( |
| 7761 |
x0 + t_a6_1 |
| 7762 |
+ t_a6_2 |
| 7763 |
+ t_a6_3 |
| 7764 |
+ t_a6_4 |
| 7765 |
+ t_a6_5 |
| 7766 |
+ t_a6_6 |
| 7767 |
+ t_a6_7 |
| 7768 |
+ t_a6_8 |
| 7769 |
+ t_a6_9 |
| 7770 |
+ t_a6_10 |
| 7771 |
+ t_a6_11 |
| 7772 |
+ t_a6_12 |
| 7773 |
+ t_a6_13 |
| 7774 |
+ t_a6_14 |
| 7775 |
+ t_a6_15 |
| 7776 |
); |
| 7777 |
let t_a7 = calc_f64!( |
| 7778 |
x0 + t_a7_1 |
| 7779 |
+ t_a7_2 |
| 7780 |
+ t_a7_3 |
| 7781 |
+ t_a7_4 |
| 7782 |
+ t_a7_5 |
| 7783 |
+ t_a7_6 |
| 7784 |
+ t_a7_7 |
| 7785 |
+ t_a7_8 |
| 7786 |
+ t_a7_9 |
| 7787 |
+ t_a7_10 |
| 7788 |
+ t_a7_11 |
| 7789 |
+ t_a7_12 |
| 7790 |
+ t_a7_13 |
| 7791 |
+ t_a7_14 |
| 7792 |
+ t_a7_15 |
| 7793 |
); |
| 7794 |
let t_a8 = calc_f64!( |
| 7795 |
x0 + t_a8_1 |
| 7796 |
+ t_a8_2 |
| 7797 |
+ t_a8_3 |
| 7798 |
+ t_a8_4 |
| 7799 |
+ t_a8_5 |
| 7800 |
+ t_a8_6 |
| 7801 |
+ t_a8_7 |
| 7802 |
+ t_a8_8 |
| 7803 |
+ t_a8_9 |
| 7804 |
+ t_a8_10 |
| 7805 |
+ t_a8_11 |
| 7806 |
+ t_a8_12 |
| 7807 |
+ t_a8_13 |
| 7808 |
+ t_a8_14 |
| 7809 |
+ t_a8_15 |
| 7810 |
); |
| 7811 |
let t_a9 = calc_f64!( |
| 7812 |
x0 + t_a9_1 |
| 7813 |
+ t_a9_2 |
| 7814 |
+ t_a9_3 |
| 7815 |
+ t_a9_4 |
| 7816 |
+ t_a9_5 |
| 7817 |
+ t_a9_6 |
| 7818 |
+ t_a9_7 |
| 7819 |
+ t_a9_8 |
| 7820 |
+ t_a9_9 |
| 7821 |
+ t_a9_10 |
| 7822 |
+ t_a9_11 |
| 7823 |
+ t_a9_12 |
| 7824 |
+ t_a9_13 |
| 7825 |
+ t_a9_14 |
| 7826 |
+ t_a9_15 |
| 7827 |
); |
| 7828 |
let t_a10 = calc_f64!( |
| 7829 |
x0 + t_a10_1 |
| 7830 |
+ t_a10_2 |
| 7831 |
+ t_a10_3 |
| 7832 |
+ t_a10_4 |
| 7833 |
+ t_a10_5 |
| 7834 |
+ t_a10_6 |
| 7835 |
+ t_a10_7 |
| 7836 |
+ t_a10_8 |
| 7837 |
+ t_a10_9 |
| 7838 |
+ t_a10_10 |
| 7839 |
+ t_a10_11 |
| 7840 |
+ t_a10_12 |
| 7841 |
+ t_a10_13 |
| 7842 |
+ t_a10_14 |
| 7843 |
+ t_a10_15 |
| 7844 |
); |
| 7845 |
let t_a11 = calc_f64!( |
| 7846 |
x0 + t_a11_1 |
| 7847 |
+ t_a11_2 |
| 7848 |
+ t_a11_3 |
| 7849 |
+ t_a11_4 |
| 7850 |
+ t_a11_5 |
| 7851 |
+ t_a11_6 |
| 7852 |
+ t_a11_7 |
| 7853 |
+ t_a11_8 |
| 7854 |
+ t_a11_9 |
| 7855 |
+ t_a11_10 |
| 7856 |
+ t_a11_11 |
| 7857 |
+ t_a11_12 |
| 7858 |
+ t_a11_13 |
| 7859 |
+ t_a11_14 |
| 7860 |
+ t_a11_15 |
| 7861 |
); |
| 7862 |
let t_a12 = calc_f64!( |
| 7863 |
x0 + t_a12_1 |
| 7864 |
+ t_a12_2 |
| 7865 |
+ t_a12_3 |
| 7866 |
+ t_a12_4 |
| 7867 |
+ t_a12_5 |
| 7868 |
+ t_a12_6 |
| 7869 |
+ t_a12_7 |
| 7870 |
+ t_a12_8 |
| 7871 |
+ t_a12_9 |
| 7872 |
+ t_a12_10 |
| 7873 |
+ t_a12_11 |
| 7874 |
+ t_a12_12 |
| 7875 |
+ t_a12_13 |
| 7876 |
+ t_a12_14 |
| 7877 |
+ t_a12_15 |
| 7878 |
); |
| 7879 |
let t_a13 = calc_f64!( |
| 7880 |
x0 + t_a13_1 |
| 7881 |
+ t_a13_2 |
| 7882 |
+ t_a13_3 |
| 7883 |
+ t_a13_4 |
| 7884 |
+ t_a13_5 |
| 7885 |
+ t_a13_6 |
| 7886 |
+ t_a13_7 |
| 7887 |
+ t_a13_8 |
| 7888 |
+ t_a13_9 |
| 7889 |
+ t_a13_10 |
| 7890 |
+ t_a13_11 |
| 7891 |
+ t_a13_12 |
| 7892 |
+ t_a13_13 |
| 7893 |
+ t_a13_14 |
| 7894 |
+ t_a13_15 |
| 7895 |
); |
| 7896 |
let t_a14 = calc_f64!( |
| 7897 |
x0 + t_a14_1 |
| 7898 |
+ t_a14_2 |
| 7899 |
+ t_a14_3 |
| 7900 |
+ t_a14_4 |
| 7901 |
+ t_a14_5 |
| 7902 |
+ t_a14_6 |
| 7903 |
+ t_a14_7 |
| 7904 |
+ t_a14_8 |
| 7905 |
+ t_a14_9 |
| 7906 |
+ t_a14_10 |
| 7907 |
+ t_a14_11 |
| 7908 |
+ t_a14_12 |
| 7909 |
+ t_a14_13 |
| 7910 |
+ t_a14_14 |
| 7911 |
+ t_a14_15 |
| 7912 |
); |
| 7913 |
let t_a15 = calc_f64!( |
| 7914 |
x0 + t_a15_1 |
| 7915 |
+ t_a15_2 |
| 7916 |
+ t_a15_3 |
| 7917 |
+ t_a15_4 |
| 7918 |
+ t_a15_5 |
| 7919 |
+ t_a15_6 |
| 7920 |
+ t_a15_7 |
| 7921 |
+ t_a15_8 |
| 7922 |
+ t_a15_9 |
| 7923 |
+ t_a15_10 |
| 7924 |
+ t_a15_11 |
| 7925 |
+ t_a15_12 |
| 7926 |
+ t_a15_13 |
| 7927 |
+ t_a15_14 |
| 7928 |
+ t_a15_15 |
| 7929 |
); |
| 7930 |
|
| 7931 |
let t_b1 = calc_f64!( |
| 7932 |
t_b1_1 |
| 7933 |
+ t_b1_2 |
| 7934 |
+ t_b1_3 |
| 7935 |
+ t_b1_4 |
| 7936 |
+ t_b1_5 |
| 7937 |
+ t_b1_6 |
| 7938 |
+ t_b1_7 |
| 7939 |
+ t_b1_8 |
| 7940 |
+ t_b1_9 |
| 7941 |
+ t_b1_10 |
| 7942 |
+ t_b1_11 |
| 7943 |
+ t_b1_12 |
| 7944 |
+ t_b1_13 |
| 7945 |
+ t_b1_14 |
| 7946 |
+ t_b1_15 |
| 7947 |
); |
| 7948 |
let t_b2 = calc_f64!( |
| 7949 |
t_b2_1 + t_b2_2 + t_b2_3 + t_b2_4 + t_b2_5 + t_b2_6 + t_b2_7 |
| 7950 |
- t_b2_8 |
| 7951 |
- t_b2_9 |
| 7952 |
- t_b2_10 |
| 7953 |
- t_b2_11 |
| 7954 |
- t_b2_12 |
| 7955 |
- t_b2_13 |
| 7956 |
- t_b2_14 |
| 7957 |
- t_b2_15 |
| 7958 |
); |
| 7959 |
let t_b3 = calc_f64!( |
| 7960 |
t_b3_1 + t_b3_2 + t_b3_3 + t_b3_4 + t_b3_5 |
| 7961 |
- t_b3_6 |
| 7962 |
- t_b3_7 |
| 7963 |
- t_b3_8 |
| 7964 |
- t_b3_9 |
| 7965 |
- t_b3_10 |
| 7966 |
+ t_b3_11 |
| 7967 |
+ t_b3_12 |
| 7968 |
+ t_b3_13 |
| 7969 |
+ t_b3_14 |
| 7970 |
+ t_b3_15 |
| 7971 |
); |
| 7972 |
let t_b4 = calc_f64!( |
| 7973 |
t_b4_1 + t_b4_2 + t_b4_3 - t_b4_4 - t_b4_5 - t_b4_6 - t_b4_7 |
| 7974 |
+ t_b4_8 |
| 7975 |
+ t_b4_9 |
| 7976 |
+ t_b4_10 |
| 7977 |
+ t_b4_11 |
| 7978 |
- t_b4_12 |
| 7979 |
- t_b4_13 |
| 7980 |
- t_b4_14 |
| 7981 |
- t_b4_15 |
| 7982 |
); |
| 7983 |
let t_b5 = calc_f64!( |
| 7984 |
t_b5_1 + t_b5_2 + t_b5_3 - t_b5_4 - t_b5_5 - t_b5_6 + t_b5_7 + t_b5_8 + t_b5_9 |
| 7985 |
- t_b5_10 |
| 7986 |
- t_b5_11 |
| 7987 |
- t_b5_12 |
| 7988 |
+ t_b5_13 |
| 7989 |
+ t_b5_14 |
| 7990 |
+ t_b5_15 |
| 7991 |
); |
| 7992 |
let t_b6 = calc_f64!( |
| 7993 |
t_b6_1 + t_b6_2 - t_b6_3 - t_b6_4 - t_b6_5 + t_b6_6 + t_b6_7 |
| 7994 |
- t_b6_8 |
| 7995 |
- t_b6_9 |
| 7996 |
- t_b6_10 |
| 7997 |
+ t_b6_11 |
| 7998 |
+ t_b6_12 |
| 7999 |
- t_b6_13 |
| 8000 |
- t_b6_14 |
| 8001 |
- t_b6_15 |
| 8002 |
); |
| 8003 |
let t_b7 = calc_f64!( |
| 8004 |
t_b7_1 + t_b7_2 - t_b7_3 - t_b7_4 + t_b7_5 + t_b7_6 - t_b7_7 - t_b7_8 |
| 8005 |
+ t_b7_9 |
| 8006 |
+ t_b7_10 |
| 8007 |
+ t_b7_11 |
| 8008 |
- t_b7_12 |
| 8009 |
- t_b7_13 |
| 8010 |
+ t_b7_14 |
| 8011 |
+ t_b7_15 |
| 8012 |
); |
| 8013 |
let t_b8 = calc_f64!( |
| 8014 |
t_b8_1 - t_b8_2 - t_b8_3 + t_b8_4 + t_b8_5 - t_b8_6 - t_b8_7 + t_b8_8 + t_b8_9 |
| 8015 |
- t_b8_10 |
| 8016 |
- t_b8_11 |
| 8017 |
+ t_b8_12 |
| 8018 |
+ t_b8_13 |
| 8019 |
- t_b8_14 |
| 8020 |
- t_b8_15 |
| 8021 |
); |
| 8022 |
let t_b9 = calc_f64!( |
| 8023 |
t_b9_1 - t_b9_2 - t_b9_3 + t_b9_4 + t_b9_5 - t_b9_6 + t_b9_7 + t_b9_8 |
| 8024 |
- t_b9_9 |
| 8025 |
- t_b9_10 |
| 8026 |
+ t_b9_11 |
| 8027 |
+ t_b9_12 |
| 8028 |
- t_b9_13 |
| 8029 |
+ t_b9_14 |
| 8030 |
+ t_b9_15 |
| 8031 |
); |
| 8032 |
let t_b10 = calc_f64!( |
| 8033 |
t_b10_1 - t_b10_2 - t_b10_3 + t_b10_4 - t_b10_5 - t_b10_6 + t_b10_7 - t_b10_8 - t_b10_9 |
| 8034 |
+ t_b10_10 |
| 8035 |
- t_b10_11 |
| 8036 |
- t_b10_12 |
| 8037 |
+ t_b10_13 |
| 8038 |
- t_b10_14 |
| 8039 |
- t_b10_15 |
| 8040 |
); |
| 8041 |
let t_b11 = calc_f64!( |
| 8042 |
t_b11_1 - t_b11_2 + t_b11_3 + t_b11_4 - t_b11_5 + t_b11_6 + t_b11_7 - t_b11_8 + t_b11_9 |
| 8043 |
- t_b11_10 |
| 8044 |
- t_b11_11 |
| 8045 |
+ t_b11_12 |
| 8046 |
- t_b11_13 |
| 8047 |
- t_b11_14 |
| 8048 |
+ t_b11_15 |
| 8049 |
); |
| 8050 |
let t_b12 = calc_f64!( |
| 8051 |
t_b12_1 - t_b12_2 + t_b12_3 - t_b12_4 - t_b12_5 + t_b12_6 - t_b12_7 + t_b12_8 + t_b12_9 |
| 8052 |
- t_b12_10 |
| 8053 |
+ t_b12_11 |
| 8054 |
- t_b12_12 |
| 8055 |
+ t_b12_13 |
| 8056 |
+ t_b12_14 |
| 8057 |
- t_b12_15 |
| 8058 |
); |
| 8059 |
let t_b13 = calc_f64!( |
| 8060 |
t_b13_1 - t_b13_2 + t_b13_3 - t_b13_4 + t_b13_5 - t_b13_6 - t_b13_7 + t_b13_8 - t_b13_9 |
| 8061 |
+ t_b13_10 |
| 8062 |
- t_b13_11 |
| 8063 |
+ t_b13_12 |
| 8064 |
+ t_b13_13 |
| 8065 |
- t_b13_14 |
| 8066 |
+ t_b13_15 |
| 8067 |
); |
| 8068 |
let t_b14 = calc_f64!( |
| 8069 |
t_b14_1 - t_b14_2 + t_b14_3 - t_b14_4 + t_b14_5 - t_b14_6 + t_b14_7 - t_b14_8 + t_b14_9 |
| 8070 |
- t_b14_10 |
| 8071 |
- t_b14_11 |
| 8072 |
+ t_b14_12 |
| 8073 |
- t_b14_13 |
| 8074 |
+ t_b14_14 |
| 8075 |
- t_b14_15 |
| 8076 |
); |
| 8077 |
let t_b15 = calc_f64!( |
| 8078 |
t_b15_1 - t_b15_2 + t_b15_3 - t_b15_4 + t_b15_5 - t_b15_6 + t_b15_7 - t_b15_8 + t_b15_9 |
| 8079 |
- t_b15_10 |
| 8080 |
+ t_b15_11 |
| 8081 |
- t_b15_12 |
| 8082 |
+ t_b15_13 |
| 8083 |
- t_b15_14 |
| 8084 |
+ t_b15_15 |
| 8085 |
); |
| 8086 |
|
| 8087 |
let t_b1_rot = self.rotate.rotate(t_b1); |
| 8088 |
let t_b2_rot = self.rotate.rotate(t_b2); |
| 8089 |
let t_b3_rot = self.rotate.rotate(t_b3); |
| 8090 |
let t_b4_rot = self.rotate.rotate(t_b4); |
| 8091 |
let t_b5_rot = self.rotate.rotate(t_b5); |
| 8092 |
let t_b6_rot = self.rotate.rotate(t_b6); |
| 8093 |
let t_b7_rot = self.rotate.rotate(t_b7); |
| 8094 |
let t_b8_rot = self.rotate.rotate(t_b8); |
| 8095 |
let t_b9_rot = self.rotate.rotate(t_b9); |
| 8096 |
let t_b10_rot = self.rotate.rotate(t_b10); |
| 8097 |
let t_b11_rot = self.rotate.rotate(t_b11); |
| 8098 |
let t_b12_rot = self.rotate.rotate(t_b12); |
| 8099 |
let t_b13_rot = self.rotate.rotate(t_b13); |
| 8100 |
let t_b14_rot = self.rotate.rotate(t_b14); |
| 8101 |
let t_b15_rot = self.rotate.rotate(t_b15); |
| 8102 |
|
| 8103 |
let y0 = calc_f64!( |
| 8104 |
x0 + x1p30 |
| 8105 |
+ x2p29 |
| 8106 |
+ x3p28 |
| 8107 |
+ x4p27 |
| 8108 |
+ x5p26 |
| 8109 |
+ x6p25 |
| 8110 |
+ x7p24 |
| 8111 |
+ x8p23 |
| 8112 |
+ x9p22 |
| 8113 |
+ x10p21 |
| 8114 |
+ x11p20 |
| 8115 |
+ x12p19 |
| 8116 |
+ x13p18 |
| 8117 |
+ x14p17 |
| 8118 |
+ x15p16 |
| 8119 |
); |
| 8120 |
let [y1, y30] = solo_fft2_f64(t_a1, t_b1_rot); |
| 8121 |
let [y2, y29] = solo_fft2_f64(t_a2, t_b2_rot); |
| 8122 |
let [y3, y28] = solo_fft2_f64(t_a3, t_b3_rot); |
| 8123 |
let [y4, y27] = solo_fft2_f64(t_a4, t_b4_rot); |
| 8124 |
let [y5, y26] = solo_fft2_f64(t_a5, t_b5_rot); |
| 8125 |
let [y6, y25] = solo_fft2_f64(t_a6, t_b6_rot); |
| 8126 |
let [y7, y24] = solo_fft2_f64(t_a7, t_b7_rot); |
| 8127 |
let [y8, y23] = solo_fft2_f64(t_a8, t_b8_rot); |
| 8128 |
let [y9, y22] = solo_fft2_f64(t_a9, t_b9_rot); |
| 8129 |
let [y10, y21] = solo_fft2_f64(t_a10, t_b10_rot); |
| 8130 |
let [y11, y20] = solo_fft2_f64(t_a11, t_b11_rot); |
| 8131 |
let [y12, y19] = solo_fft2_f64(t_a12, t_b12_rot); |
| 8132 |
let [y13, y18] = solo_fft2_f64(t_a13, t_b13_rot); |
| 8133 |
let [y14, y17] = solo_fft2_f64(t_a14, t_b14_rot); |
| 8134 |
let [y15, y16] = solo_fft2_f64(t_a15, t_b15_rot); |
| 8135 |
[ |
| 8136 |
y0, y1, y2, y3, y4, y5, y6, y7, y8, y9, y10, y11, y12, y13, y14, y15, y16, y17, y18, |
| 8137 |
y19, y20, y21, y22, y23, y24, y25, y26, y27, y28, y29, y30, |
| 8138 |
] |
| 8139 |
} |
| 8140 |
} |
| 8141 |
|
| 8142 |
// _____ _____ ____ _____ ____ |
| 8143 |
// |_ _| ____/ ___|_ _/ ___| |
| 8144 |
// | | | _| \___ \ | | \___ \ |
| 8145 |
// | | | |___ ___) || | ___) | |
| 8146 |
// |_| |_____|____/ |_| |____/ |
| 8147 |
// |
| 8148 |
|
| 8149 |
#[cfg(test)] |
| 8150 |
mod unit_tests { |
| 8151 |
use super::*; |
| 8152 |
use crate::test_utils::check_fft_algorithm; |
| 8153 |
use wasm_bindgen_test::wasm_bindgen_test; |
| 8154 |
|
| 8155 |
//the tests for all butterflies will be identical except for the identifiers used and size |
| 8156 |
//so it's ideal for a macro |
| 8157 |
macro_rules! test_butterfly_32_func { |
| 8158 |
($test_name:ident, $struct_name:ident, $size:expr) => { |
| 8159 |
#[wasm_bindgen_test] |
| 8160 |
fn $test_name() { |
| 8161 |
let butterfly = $struct_name::new(FftDirection::Forward); |
| 8162 |
check_fft_algorithm::<f32>(&butterfly, $size, FftDirection::Forward); |
| 8163 |
|
| 8164 |
let butterfly_direction = $struct_name::new(FftDirection::Inverse); |
| 8165 |
check_fft_algorithm::<f32>(&butterfly_direction, $size, FftDirection::Inverse); |
| 8166 |
} |
| 8167 |
}; |
| 8168 |
} |
| 8169 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly7, WasmSimdF32Butterfly7, 7); |
| 8170 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly11, WasmSimdF32Butterfly11, 11); |
| 8171 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly13, WasmSimdF32Butterfly13, 13); |
| 8172 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly17, WasmSimdF32Butterfly17, 17); |
| 8173 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly19, WasmSimdF32Butterfly19, 19); |
| 8174 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly23, WasmSimdF32Butterfly23, 23); |
| 8175 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly29, WasmSimdF32Butterfly29, 29); |
| 8176 |
test_butterfly_32_func!(test_wasm_simdf32_butterfly31, WasmSimdF32Butterfly31, 31); |
| 8177 |
|
| 8178 |
//the tests for all butterflies will be identical except for the identifiers used and size |
| 8179 |
//so it's ideal for a macro |
| 8180 |
macro_rules! test_butterfly_64_func { |
| 8181 |
($test_name:ident, $struct_name:ident, $size:expr) => { |
| 8182 |
#[test] |
| 8183 |
fn $test_name() { |
| 8184 |
let butterfly = $struct_name::new(FftDirection::Forward); |
| 8185 |
check_fft_algorithm::<f64>(&butterfly, $size, FftDirection::Forward); |
| 8186 |
|
| 8187 |
let butterfly_direction = $struct_name::new(FftDirection::Inverse); |
| 8188 |
check_fft_algorithm::<f64>(&butterfly_direction, $size, FftDirection::Inverse); |
| 8189 |
} |
| 8190 |
}; |
| 8191 |
} |
| 8192 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly7, WasmSimdF64Butterfly7, 7); |
| 8193 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly11, WasmSimdF64Butterfly11, 11); |
| 8194 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly13, WasmSimdF64Butterfly13, 13); |
| 8195 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly17, WasmSimdF64Butterfly17, 17); |
| 8196 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly19, WasmSimdF64Butterfly19, 19); |
| 8197 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly23, WasmSimdF64Butterfly23, 23); |
| 8198 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly29, WasmSimdF64Butterfly29, 29); |
| 8199 |
test_butterfly_64_func!(test_wasm_simdf64_butterfly31, WasmSimdF64Butterfly31, 31); |
| 8200 |
} |