/* * Copyright (c) 2023. * * This software is free software; * * You can redistribute it or modify it under terms of the MIT, Apache License or Zlib license */ #![cfg(any(target_arch = "x86", target_arch = "x86_64"))] //! AVX optimised IDCT. //! //! Okay not thaat optimised. //! //! //! # The implementation //! The implementation is neatly broken down into two operations. //! //! 1. Test for zeroes //! > There is a shortcut method for idct where when all AC values are zero, we can get the answer really quickly. //! by scaling the 1/8th of the DCT coefficient of the block to the whole block and level shifting. //! //! 2. If above fails, we proceed to carry out IDCT as a two pass one dimensional algorithm. //! IT does two whole scans where it carries out IDCT on all items //! After each successive scan, data is transposed in register(thank you x86 SIMD powers). and the second //! pass is carried out. //! //! The code is not super optimized, it produces bit identical results with scalar code hence it's //! `mm256_add_epi16` //! and it also has the advantage of making this implementation easy to maintain. #![cfg(feature = "x86")] #![allow(dead_code)] #[cfg(target_arch = "x86")] use core::arch::x86::*; #[cfg(target_arch = "x86_64")] use core::arch::x86_64::*; use crate::unsafe_utils::{transpose, YmmRegister}; const SCALE_BITS: i32 = 512 + 65536 + (128 << 17); // Pack i32 to i16's, // clamp them to be between 0-255 // Undo shuffling // Store back to array macro_rules! permute_store { ($x:tt,$y:tt,$index:tt,$out:tt,$stride:tt) => { let a = _mm256_packs_epi32($x, $y); // Clamp the values after packing, we can clamp more values at once let b = clamp_avx(a); // /Undo shuffling let c = _mm256_permute4x64_epi64(b, shuffle(3, 1, 2, 0)); // store first vector _mm_storeu_si128( ($out) .get_mut($index..$index + 8) .unwrap() .as_mut_ptr() .cast(), _mm256_extractf128_si256::<0>(c), ); $index += $stride; // second vector _mm_storeu_si128( ($out) .get_mut($index..$index + 8) .unwrap() .as_mut_ptr() .cast(), _mm256_extractf128_si256::<1>(c), ); $index += $stride; }; } #[target_feature(enable = "avx2")] #[allow( clippy::too_many_lines, clippy::cast_possible_truncation, clippy::similar_names, clippy::op_ref, unused_assignments, clippy::zero_prefixed_literal )] pub unsafe fn idct_avx2( in_vector: &mut [i32; 64], out_vector: &mut [i16], stride: usize, ) { let mut pos = 0; // load into registers // // We sign extend i16's to i32's and calculate them with extended precision and // later reduce them to i16's when we are done carrying out IDCT let rw0 = _mm256_loadu_si256(in_vector[00..].as_ptr().cast()); let rw1 = _mm256_loadu_si256(in_vector[08..].as_ptr().cast()); let rw2 = _mm256_loadu_si256(in_vector[16..].as_ptr().cast()); let rw3 = _mm256_loadu_si256(in_vector[24..].as_ptr().cast()); let rw4 = _mm256_loadu_si256(in_vector[32..].as_ptr().cast()); let rw5 = _mm256_loadu_si256(in_vector[40..].as_ptr().cast()); let rw6 = _mm256_loadu_si256(in_vector[48..].as_ptr().cast()); let rw7 = _mm256_loadu_si256(in_vector[56..].as_ptr().cast()); // Forward DCT and quantization may cause all the AC terms to be zero, for such // cases we can try to accelerate it // Basically the poop is that whenever the array has 63 zeroes, its idct is // (arr[0]>>3)or (arr[0]/8) propagated to all the elements. // We first test to see if the array contains zero elements and if it does, we go the // short way. // // This reduces IDCT overhead from about 39% to 18 %, almost half // Do another load for the first row, we don't want to check DC value, because // we only care about AC terms let rw8 = _mm256_loadu_si256(in_vector[1..].as_ptr().cast()); let mut bitmap = _mm256_or_si256(rw1, rw2); bitmap = _mm256_or_si256(bitmap, rw3); bitmap = _mm256_or_si256(bitmap, rw4); bitmap = _mm256_or_si256(bitmap, rw5); bitmap = _mm256_or_si256(bitmap, rw6); bitmap = _mm256_or_si256(bitmap, rw7); bitmap = _mm256_or_si256(bitmap, rw8); if _mm256_testz_si256(bitmap, bitmap) == 1 { // AC terms all zero, idct of the block is ( coeff[0] * qt[0] )/8 + 128 (bias) // (and clamped to 255) // Round by adding 0.5 * (1 << 3) and offset by adding (128 << 3) before scaling let coeff = ((in_vector[0] + 4 + 1024) >> 3).clamp(0, 255) as i16; let idct_value = _mm_set1_epi16(coeff); macro_rules! store { ($pos:tt,$value:tt) => { // store _mm_storeu_si128( out_vector .get_mut($pos..$pos + 8) .unwrap() .as_mut_ptr() .cast(), $value, ); $pos += stride; }; } store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); store!(pos, idct_value); return; } let mut row0 = YmmRegister { mm256: rw0 }; let mut row1 = YmmRegister { mm256: rw1 }; let mut row2 = YmmRegister { mm256: rw2 }; let mut row3 = YmmRegister { mm256: rw3 }; let mut row4 = YmmRegister { mm256: rw4 }; let mut row5 = YmmRegister { mm256: rw5 }; let mut row6 = YmmRegister { mm256: rw6 }; let mut row7 = YmmRegister { mm256: rw7 }; macro_rules! dct_pass { ($SCALE_BITS:tt,$scale:tt) => { // There are a lot of ways to do this // but to keep it simple(and beautiful), ill make a direct translation of the // scalar code to also make this code fully transparent(this version and the non // avx one should produce identical code.) // even part let p1 = (row2 + row6) * 2217; let mut t2 = p1 + row6 * -7567; let mut t3 = p1 + row2 * 3135; let mut t0 = YmmRegister { mm256: _mm256_slli_epi32((row0 + row4).mm256, 12), }; let mut t1 = YmmRegister { mm256: _mm256_slli_epi32((row0 - row4).mm256, 12), }; let x0 = t0 + t3 + $SCALE_BITS; let x3 = t0 - t3 + $SCALE_BITS; let x1 = t1 + t2 + $SCALE_BITS; let x2 = t1 - t2 + $SCALE_BITS; let p3 = row7 + row3; let p4 = row5 + row1; let p1 = row7 + row1; let p2 = row5 + row3; let p5 = (p3 + p4) * 4816; t0 = row7 * 1223; t1 = row5 * 8410; t2 = row3 * 12586; t3 = row1 * 6149; let p1 = p5 + p1 * -3685; let p2 = p5 + (p2 * -10497); let p3 = p3 * -8034; let p4 = p4 * -1597; t3 += p1 + p4; t2 += p2 + p3; t1 += p2 + p4; t0 += p1 + p3; row0.mm256 = _mm256_srai_epi32((x0 + t3).mm256, $scale); row1.mm256 = _mm256_srai_epi32((x1 + t2).mm256, $scale); row2.mm256 = _mm256_srai_epi32((x2 + t1).mm256, $scale); row3.mm256 = _mm256_srai_epi32((x3 + t0).mm256, $scale); row4.mm256 = _mm256_srai_epi32((x3 - t0).mm256, $scale); row5.mm256 = _mm256_srai_epi32((x2 - t1).mm256, $scale); row6.mm256 = _mm256_srai_epi32((x1 - t2).mm256, $scale); row7.mm256 = _mm256_srai_epi32((x0 - t3).mm256, $scale); }; } // Process rows dct_pass!(512, 10); transpose( &mut row0, &mut row1, &mut row2, &mut row3, &mut row4, &mut row5, &mut row6, &mut row7, ); // process columns dct_pass!(SCALE_BITS, 17); transpose( &mut row0, &mut row1, &mut row2, &mut row3, &mut row4, &mut row5, &mut row6, &mut row7, ); // Pack and write the values back to the array permute_store!((row0.mm256), (row1.mm256), pos, out_vector, stride); permute_store!((row2.mm256), (row3.mm256), pos, out_vector, stride); permute_store!((row4.mm256), (row5.mm256), pos, out_vector, stride); permute_store!((row6.mm256), (row7.mm256), pos, out_vector, stride); } #[target_feature(enable = "avx2")] #[allow( clippy::too_many_lines, clippy::cast_possible_truncation, clippy::similar_names, clippy::op_ref, unused_assignments, clippy::zero_prefixed_literal )] pub unsafe fn idct_avx2_4x4( in_vector: &mut [i32; 64], out_vector: &mut [i16], stride: usize, ) { let rw0 = _mm256_loadu_si256(in_vector[00..].as_ptr().cast()); let rw1 = _mm256_loadu_si256(in_vector[08..].as_ptr().cast()); let rw2 = _mm256_loadu_si256(in_vector[16..].as_ptr().cast()); let rw3 = _mm256_loadu_si256(in_vector[24..].as_ptr().cast()); let mut row0 = YmmRegister { mm256: rw0 }; let mut row1 = YmmRegister { mm256: rw1 }; let mut row2 = YmmRegister { mm256: rw2 }; let mut row3 = YmmRegister { mm256: rw3 }; let mut row4 = YmmRegister { mm256: rw0 }; let mut row5 = YmmRegister { mm256: rw0 }; let mut row6 = YmmRegister { mm256: rw0 }; let mut row7 = YmmRegister { mm256: rw0 }; { row0.mm256 = _mm256_slli_epi32(row0.mm256, 12); row0 += 512; let i2 = row2; let p1 = i2 * 2217; let p3 = i2 * 5352; let x0 = row0 + p3; let x1 = row0 + p1; let x2 = row0 - p1; let x3 = row0 - p3; // odd part let i4 = row3; let i3 = row1; let p5 = (i4 + i3) * 4816; let p1 = p5 + i3 * -3685; let p2 = p5 + i4 * -10497; let t3 = p5 + i3 * 867; let t2 = p5 + i4 * -5945; let t1 = p2 + i3 * -1597; let t0 = p1 + i4 * -8034; row0.mm256 = _mm256_srai_epi32((x0 + t3).mm256, 10); row1.mm256 = _mm256_srai_epi32((x1 + t2).mm256, 10); row2.mm256 = _mm256_srai_epi32((x2 + t1).mm256, 10); row3.mm256 = _mm256_srai_epi32((x3 + t0).mm256, 10); row4.mm256 = _mm256_srai_epi32((x3 - t0).mm256, 10); row5.mm256 = _mm256_srai_epi32((x2 - t1).mm256, 10); row6.mm256 = _mm256_srai_epi32((x1 - t2).mm256, 10); row7.mm256 = _mm256_srai_epi32((x0 - t3).mm256, 10); } transpose( &mut row0, &mut row1, &mut row2, &mut row3, &mut row4, &mut row5, &mut row6, &mut row7, ); { let i2 = row2; let i0 = row0; row0.mm256 = _mm256_slli_epi32(i0.mm256, 12); let t0 = row0 + SCALE_BITS; let t2 = i2 * 2217; let t3 = i2 * 5352; // constants scaled things up by 1<<12, plus we had 1<<2 from first // loop, plus horizontal and vertical each scale by sqrt(8) so together // we've got an extra 1<<3, so 1<<17 total we need to remove. // so we want to round that, which means adding 0.5 * 1<<17, // aka 65536. Also, we'll end up with -128 to 127 that we want // to encode as 0..255 by adding 128, so we'll add that before the shift // Rounding constant is already added into `t0` let x0 = t0 + t3; let x3 = t0 - t3; let x1 = t0 + t2; let x2 = t0 - t2; // odd part let i3 = row3; let i1 = row1; let p5 = (i3 + i1) * 4816; let p1 = p5 + i1 * -3685; let p2 = p5 + i3 * -10497; let t3 = p5 + i1 * 867; let t2 = p5 + i3 * -5945; let t1 = p2 + i1 * -1597; let t0 = p1 + i3 * -8034; row0.mm256 = _mm256_srai_epi32((x0 + t3).mm256, 17); row1.mm256 = _mm256_srai_epi32((x1 + t2).mm256, 17); row2.mm256 = _mm256_srai_epi32((x2 + t1).mm256, 17); row3.mm256 = _mm256_srai_epi32((x3 + t0).mm256, 17); row4.mm256 = _mm256_srai_epi32((x3 - t0).mm256, 17); row5.mm256 = _mm256_srai_epi32((x2 - t1).mm256, 17); row6.mm256 = _mm256_srai_epi32((x1 - t2).mm256, 17); row7.mm256 = _mm256_srai_epi32((x0 - t3).mm256, 17); } transpose( &mut row0, &mut row1, &mut row2, &mut row3, &mut row4, &mut row5, &mut row6, &mut row7, ); let mut pos = 0; // Pack and write the values back to the array permute_store!((row0.mm256), (row1.mm256), pos, out_vector, stride); permute_store!((row2.mm256), (row3.mm256), pos, out_vector, stride); permute_store!((row4.mm256), (row5.mm256), pos, out_vector, stride); permute_store!((row6.mm256), (row7.mm256), pos, out_vector, stride); } #[inline] #[target_feature(enable = "avx2")] unsafe fn clamp_avx(reg: __m256i) -> __m256i { let min_s = _mm256_set1_epi16(0); let max_s = _mm256_set1_epi16(255); let max_v = _mm256_max_epi16(reg, min_s); //max(a,0) let min_v = _mm256_min_epi16(max_v, max_s); //min(max(a,0),255) return min_v; } /// A copy of `_MM_SHUFFLE()` that doesn't require /// a nightly compiler #[inline] const fn shuffle(z: i32, y: i32, x: i32, w: i32) -> i32 { ((z << 6) | (y << 4) | (x << 2) | w) }