| 1 |
/* |
| 2 |
* Copyright (C) 2020 Linux Studio Plugins Project <https://lsp-plug.in/> |
| 3 |
* (C) 2020 Vladimir Sadovnikov <[email protected]> |
| 4 |
* |
| 5 |
* This file is part of lsp-dsp-lib |
| 6 |
* Created on: 10 окт. 2020 г. |
| 7 |
* |
| 8 |
* lsp-dsp-lib is free software: you can redistribute it and/or modify |
| 9 |
* it under the terms of the GNU Lesser General Public License as published by |
| 10 |
* the Free Software Foundation, either version 3 of the License, or |
| 11 |
* any later version. |
| 12 |
* |
| 13 |
* lsp-dsp-lib is distributed in the hope that it will be useful, |
| 14 |
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 15 |
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 16 |
* GNU Lesser General Public License for more details. |
| 17 |
* |
| 18 |
* You should have received a copy of the GNU Lesser General Public License |
| 19 |
* along with lsp-dsp-lib. If not, see <https://www.gnu.org/licenses/>. |
| 20 |
*/ |
| 21 |
|
| 22 |
#ifndef PRIVATE_DSP_ARCH_X86_SSE2_PIXELFMT_H_ |
| 23 |
#define PRIVATE_DSP_ARCH_X86_SSE2_PIXELFMT_H_ |
| 24 |
|
| 25 |
namespace lsp |
| 26 |
{ |
| 27 |
namespace sse2 |
| 28 |
{ |
| 29 |
IF_ARCH_X86( |
| 30 |
static const uint32_t rgba_const[] __lsp_aligned16 = |
| 31 |
{ |
| 32 |
LSP_DSP_VEC4(0x00ff00ff) |
| 33 |
}; |
| 34 |
) |
| 35 |
|
| 36 |
void rgba32_to_bgra32(void *dst, const void *src, size_t count) |
| 37 |
{ |
| 38 |
size_t off; |
| 39 |
|
| 40 |
ARCH_X86_ASM |
| 41 |
( |
| 42 |
__ASM_EMIT("movdqa %[MASK], %%xmm6") // xmm6 = 00 ff 00 ff |
| 43 |
__ASM_EMIT("xor %[off], %[off]") // off = 0 |
| 44 |
__ASM_EMIT("movdqa %%xmm6, %%xmm7") // xmm7 = 00 ff 00 ff |
| 45 |
__ASM_EMIT("pslld $8, %%xmm6") // xmm6 = ff 00 ff 00 |
| 46 |
|
| 47 |
// 8-element blocks |
| 48 |
__ASM_EMIT("sub $8, %[count]") |
| 49 |
__ASM_EMIT("jb 2f") |
| 50 |
__ASM_EMIT("1:") |
| 51 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = A1 R1 G1 B1 |
| 52 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm1") // xmm1 = A2 R2 G2 B2 |
| 53 |
__ASM_EMIT("movdqa %%xmm0, %%xmm2") // xmm2 = A1 R1 G1 B1 |
| 54 |
__ASM_EMIT("movdqa %%xmm1, %%xmm3") // xmm3 = A2 R2 G2 B2 |
| 55 |
__ASM_EMIT("pand %%xmm7, %%xmm0") // xmm0 = 00 R1 00 B1 |
| 56 |
__ASM_EMIT("pand %%xmm6, %%xmm2") // xmm2 = A1 00 G1 00 |
| 57 |
__ASM_EMIT("pand %%xmm7, %%xmm1") // xmm1 = 00 R2 00 B2 |
| 58 |
__ASM_EMIT("pand %%xmm6, %%xmm3") // xmm3 = A2 00 G2 00 |
| 59 |
__ASM_EMIT("movdqa %%xmm0, %%xmm4") // xmm4 = A1 00 G1 00 |
| 60 |
__ASM_EMIT("movdqa %%xmm1, %%xmm5") // xmm5 = A2 00 G2 00 |
| 61 |
__ASM_EMIT("pslld $16, %%xmm0") // xmm0 = 00 B1 00 00 |
| 62 |
__ASM_EMIT("pslld $16, %%xmm1") // xmm1 = 00 B2 00 00 |
| 63 |
__ASM_EMIT("psrld $16, %%xmm4") // xmm4 = 00 00 00 R1 |
| 64 |
__ASM_EMIT("psrld $16, %%xmm5") // xmm5 = 00 00 00 R2 |
| 65 |
__ASM_EMIT("por %%xmm2, %%xmm0") // xmm0 = A1 B1 G1 00 |
| 66 |
__ASM_EMIT("por %%xmm3, %%xmm1") // xmm1 = A2 B2 G2 00 |
| 67 |
__ASM_EMIT("por %%xmm4, %%xmm0") // xmm0 = A1 B1 G1 R1 |
| 68 |
__ASM_EMIT("por %%xmm5, %%xmm1") // xmm1 = A2 B2 G2 R2 |
| 69 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 70 |
__ASM_EMIT("movdqu %%xmm1, 0x10(%[dst], %[off])") |
| 71 |
__ASM_EMIT("add $0x20, %[off]") |
| 72 |
__ASM_EMIT("sub $8, %[count]") |
| 73 |
__ASM_EMIT("jae 1b") |
| 74 |
|
| 75 |
// 4-element block |
| 76 |
__ASM_EMIT("2:") |
| 77 |
__ASM_EMIT("add $4, %[count]") |
| 78 |
__ASM_EMIT("jl 4f") |
| 79 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = A1 R1 G1 B1 |
| 80 |
__ASM_EMIT("movdqa %%xmm0, %%xmm2") // xmm2 = A1 R1 G1 B1 |
| 81 |
__ASM_EMIT("pand %%xmm7, %%xmm0") // xmm0 = 00 R1 00 B1 |
| 82 |
__ASM_EMIT("pand %%xmm6, %%xmm2") // xmm2 = A1 00 G1 00 |
| 83 |
__ASM_EMIT("movdqa %%xmm0, %%xmm4") // xmm4 = A1 00 G1 00 |
| 84 |
__ASM_EMIT("pslld $16, %%xmm0") // xmm0 = 00 B1 00 00 |
| 85 |
__ASM_EMIT("psrld $16, %%xmm4") // xmm4 = 00 00 00 R1 |
| 86 |
__ASM_EMIT("por %%xmm2, %%xmm0") // xmm0 = A1 B1 G1 00 |
| 87 |
__ASM_EMIT("por %%xmm4, %%xmm0") // xmm0 = A1 B1 G1 R1 |
| 88 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 89 |
__ASM_EMIT("add $0x10, %[off]") |
| 90 |
__ASM_EMIT("sub $4, %[count]") |
| 91 |
|
| 92 |
// Tail |
| 93 |
__ASM_EMIT("4:") |
| 94 |
__ASM_EMIT("add $3, %[count]") |
| 95 |
__ASM_EMIT("jl 6f") |
| 96 |
__ASM_EMIT("5:") |
| 97 |
__ASM_EMIT("movd 0x00(%[src], %[off]), %%xmm0") // xmm0 = AA RR GG BB |
| 98 |
__ASM_EMIT("movdqa %%xmm0, %%xmm1") // xmm1 = AA RR GG BB |
| 99 |
__ASM_EMIT("pand %%xmm7, %%xmm0") // xmm0 = 00 RR 00 BB |
| 100 |
__ASM_EMIT("pand %%xmm6, %%xmm1") // xmm1 = AA 00 GG 00 |
| 101 |
__ASM_EMIT("movdqa %%xmm0, %%xmm2") // xmm2 = 00 RR 00 BB |
| 102 |
__ASM_EMIT("pslld $16, %%xmm0") // xmm0 = 00 BB 00 00 |
| 103 |
__ASM_EMIT("psrld $16, %%xmm2") // xmm2 = 00 00 00 RR |
| 104 |
__ASM_EMIT("por %%xmm1, %%xmm0") // xmm0 = AA 00 GG RR |
| 105 |
__ASM_EMIT("por %%xmm2, %%xmm0") // xmm0 = AA BB GG RR |
| 106 |
__ASM_EMIT("movd %%xmm0, 0x00(%[dst], %[off])") |
| 107 |
__ASM_EMIT("add $4, %[off]") |
| 108 |
__ASM_EMIT("dec %[count]") |
| 109 |
__ASM_EMIT("jge 5b") |
| 110 |
|
| 111 |
// End |
| 112 |
__ASM_EMIT("6:") |
| 113 |
|
| 114 |
: [dst] "+r"(dst), [src] "+r"(src), [count] "+r" (count), |
| 115 |
[off] "=&r" (off) |
| 116 |
: [MASK] "m" (rgba_const) |
| 117 |
: "cc", "memory", |
| 118 |
"%xmm0", "%xmm1", "%xmm2", "%xmm3", |
| 119 |
"%xmm4", "%xmm5", "%xmm6", "%xmm7" |
| 120 |
); |
| 121 |
} |
| 122 |
|
| 123 |
IF_ARCH_X86( |
| 124 |
static const uint32_t abgr32_to_bgrff32_const[] __lsp_aligned16 = |
| 125 |
{ |
| 126 |
LSP_DSP_VEC4(0xff000000) |
| 127 |
}; |
| 128 |
) |
| 129 |
|
| 130 |
void abgr32_to_bgrff32(void *dst, const void *src, size_t count) |
| 131 |
{ |
| 132 |
size_t off; |
| 133 |
|
| 134 |
ARCH_X86_ASM |
| 135 |
( |
| 136 |
__ASM_EMIT("movdqa %[MASK], %%xmm6") // xmm6 = ff 00 00 00 |
| 137 |
__ASM_EMIT("xor %[off], %[off]") // off = 0 |
| 138 |
__ASM_EMIT("movdqa %%xmm6, %%xmm7") // xmm7 = ff 00 00 00 |
| 139 |
|
| 140 |
// 24-element blocks |
| 141 |
__ASM_EMIT("sub $24, %[count]") |
| 142 |
__ASM_EMIT("jb 2f") |
| 143 |
__ASM_EMIT("1:") |
| 144 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = R1 G1 B1 A1 |
| 145 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm1") |
| 146 |
__ASM_EMIT("movdqu 0x20(%[src], %[off]), %%xmm2") |
| 147 |
__ASM_EMIT("movdqu 0x30(%[src], %[off]), %%xmm3") |
| 148 |
__ASM_EMIT("movdqu 0x40(%[src], %[off]), %%xmm4") |
| 149 |
__ASM_EMIT("movdqu 0x50(%[src], %[off]), %%xmm5") |
| 150 |
__ASM_EMIT("psrld $8, %%xmm0") // xmm0 = 00 R1 G1 B1 |
| 151 |
__ASM_EMIT("psrld $8, %%xmm1") |
| 152 |
__ASM_EMIT("psrld $8, %%xmm2") |
| 153 |
__ASM_EMIT("psrld $8, %%xmm3") |
| 154 |
__ASM_EMIT("psrld $8, %%xmm4") |
| 155 |
__ASM_EMIT("psrld $8, %%xmm5") |
| 156 |
__ASM_EMIT("por %%xmm6, %%xmm0") // xmm6 = FF R1 G1 B1 |
| 157 |
__ASM_EMIT("por %%xmm7, %%xmm1") |
| 158 |
__ASM_EMIT("por %%xmm6, %%xmm2") |
| 159 |
__ASM_EMIT("por %%xmm7, %%xmm3") |
| 160 |
__ASM_EMIT("por %%xmm6, %%xmm4") |
| 161 |
__ASM_EMIT("por %%xmm7, %%xmm5") |
| 162 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 163 |
__ASM_EMIT("movdqu %%xmm1, 0x10(%[dst], %[off])") |
| 164 |
__ASM_EMIT("movdqu %%xmm2, 0x20(%[dst], %[off])") |
| 165 |
__ASM_EMIT("movdqu %%xmm3, 0x30(%[dst], %[off])") |
| 166 |
__ASM_EMIT("movdqu %%xmm4, 0x40(%[dst], %[off])") |
| 167 |
__ASM_EMIT("movdqu %%xmm5, 0x50(%[dst], %[off])") |
| 168 |
__ASM_EMIT("add $0x60, %[off]") |
| 169 |
__ASM_EMIT("sub $24, %[count]") |
| 170 |
__ASM_EMIT("jae 1b") |
| 171 |
|
| 172 |
// 16-element block |
| 173 |
__ASM_EMIT("2:") |
| 174 |
__ASM_EMIT("add $8, %[count]") |
| 175 |
__ASM_EMIT("jl 4f") |
| 176 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = R1 G1 B1 A1 |
| 177 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm1") |
| 178 |
__ASM_EMIT("movdqu 0x20(%[src], %[off]), %%xmm2") |
| 179 |
__ASM_EMIT("movdqu 0x30(%[src], %[off]), %%xmm3") |
| 180 |
__ASM_EMIT("psrld $8, %%xmm0") // xmm0 = 00 R1 G1 B1 |
| 181 |
__ASM_EMIT("psrld $8, %%xmm1") |
| 182 |
__ASM_EMIT("psrld $8, %%xmm2") |
| 183 |
__ASM_EMIT("psrld $8, %%xmm3") |
| 184 |
__ASM_EMIT("por %%xmm6, %%xmm0") // xmm6 = FF R1 G1 B1 |
| 185 |
__ASM_EMIT("por %%xmm7, %%xmm1") |
| 186 |
__ASM_EMIT("por %%xmm6, %%xmm2") |
| 187 |
__ASM_EMIT("por %%xmm7, %%xmm3") |
| 188 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 189 |
__ASM_EMIT("movdqu %%xmm1, 0x10(%[dst], %[off])") |
| 190 |
__ASM_EMIT("movdqu %%xmm2, 0x20(%[dst], %[off])") |
| 191 |
__ASM_EMIT("movdqu %%xmm3, 0x30(%[dst], %[off])") |
| 192 |
__ASM_EMIT("add $0x40, %[off]") |
| 193 |
__ASM_EMIT("sub $16, %[count]") |
| 194 |
|
| 195 |
// 8-element block |
| 196 |
__ASM_EMIT("4:") |
| 197 |
__ASM_EMIT("add $8, %[count]") |
| 198 |
__ASM_EMIT("jl 6f") |
| 199 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = R1 G1 B1 A1 |
| 200 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm1") |
| 201 |
__ASM_EMIT("psrld $8, %%xmm0") // xmm0 = 00 R1 G1 B1 |
| 202 |
__ASM_EMIT("psrld $8, %%xmm1") |
| 203 |
__ASM_EMIT("por %%xmm6, %%xmm0") // xmm6 = FF R1 G1 B1 |
| 204 |
__ASM_EMIT("por %%xmm7, %%xmm1") |
| 205 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 206 |
__ASM_EMIT("movdqu %%xmm1, 0x10(%[dst], %[off])") |
| 207 |
__ASM_EMIT("add $0x20, %[off]") |
| 208 |
__ASM_EMIT("sub $8, %[count]") |
| 209 |
|
| 210 |
// 4-element block |
| 211 |
__ASM_EMIT("6:") |
| 212 |
__ASM_EMIT("add $4, %[count]") |
| 213 |
__ASM_EMIT("jl 8f") |
| 214 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm0") // xmm0 = R1 G1 B1 A1 |
| 215 |
__ASM_EMIT("psrld $8, %%xmm0") // xmm0 = 00 R1 G1 B1 |
| 216 |
__ASM_EMIT("por %%xmm6, %%xmm0") // xmm6 = FF R1 G1 B1 |
| 217 |
__ASM_EMIT("movdqu %%xmm0, 0x00(%[dst], %[off])") |
| 218 |
__ASM_EMIT("add $0x10, %[off]") |
| 219 |
__ASM_EMIT("sub $4, %[count]") |
| 220 |
|
| 221 |
// Tail |
| 222 |
__ASM_EMIT("8:") |
| 223 |
__ASM_EMIT("add $3, %[count]") |
| 224 |
__ASM_EMIT("jl 10f") |
| 225 |
__ASM_EMIT("9:") |
| 226 |
__ASM_EMIT("movd 0x00(%[src], %[off]), %%xmm0") // xmm0 = R1 G1 B1 A1 |
| 227 |
__ASM_EMIT("psrld $8, %%xmm0") // xmm0 = 00 R1 G1 B1 |
| 228 |
__ASM_EMIT("por %%xmm6, %%xmm0") // xmm6 = FF R1 G1 B1 |
| 229 |
__ASM_EMIT("movd %%xmm0, 0x00(%[dst], %[off])") |
| 230 |
__ASM_EMIT("add $4, %[off]") |
| 231 |
__ASM_EMIT("dec %[count]") |
| 232 |
__ASM_EMIT("jge 9b") |
| 233 |
|
| 234 |
// End |
| 235 |
__ASM_EMIT("10:") |
| 236 |
|
| 237 |
: [dst] "+r"(dst), [src] "+r"(src), [count] "+r" (count), |
| 238 |
[off] "=&r" (off) |
| 239 |
: [MASK] "m" (abgr32_to_bgrff32_const) |
| 240 |
: "cc", "memory", |
| 241 |
"%xmm0", "%xmm1", "%xmm2", "%xmm3", |
| 242 |
"%xmm4", "%xmm5", "%xmm6", "%xmm7" |
| 243 |
); |
| 244 |
} |
| 245 |
|
| 246 |
IF_ARCH_X86( |
| 247 |
static const uint32_t pabc32_set_alpha_const[] __lsp_aligned16 = |
| 248 |
{ |
| 249 |
LSP_DSP_VEC4(0x00ffffff) |
| 250 |
}; |
| 251 |
) |
| 252 |
|
| 253 |
void pabc32_set_alpha(void *dst, const void *src, uint8_t alpha, size_t count) |
| 254 |
{ |
| 255 |
IF_ARCH_X86( |
| 256 |
size_t off; |
| 257 |
uint32_t value = uint32_t(alpha) << 24; |
| 258 |
); |
| 259 |
|
| 260 |
ARCH_X86_ASM |
| 261 |
( |
| 262 |
__ASM_EMIT("movdqa %[MASK], %%xmm1") // xmm1 = 00 ff ff ff |
| 263 |
__ASM_EMIT("pshufd $0x00, %%xmm0, %%xmm0") // xmm0 = vv vv vv vv |
| 264 |
__ASM_EMIT("xor %[off], %[off]") // off = 0 |
| 265 |
|
| 266 |
// 24-element blocks |
| 267 |
__ASM_EMIT("sub $24, %[count]") |
| 268 |
__ASM_EMIT("jb 2f") |
| 269 |
__ASM_EMIT("1:") |
| 270 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm2") // xmm2 = AA XX YY ZZ |
| 271 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm3") |
| 272 |
__ASM_EMIT("movdqu 0x20(%[src], %[off]), %%xmm4") |
| 273 |
__ASM_EMIT("pand %%xmm1, %%xmm2") // xmm2 = 00 XX YY ZZ |
| 274 |
__ASM_EMIT("movdqu 0x30(%[src], %[off]), %%xmm5") |
| 275 |
__ASM_EMIT("pand %%xmm1, %%xmm3") |
| 276 |
__ASM_EMIT("movdqu 0x40(%[src], %[off]), %%xmm6") |
| 277 |
__ASM_EMIT("pand %%xmm1, %%xmm4") |
| 278 |
__ASM_EMIT("por %%xmm0, %%xmm2") // xmm2 = VV XX YY ZZ |
| 279 |
__ASM_EMIT("movdqu 0x50(%[src], %[off]), %%xmm7") |
| 280 |
__ASM_EMIT("por %%xmm0, %%xmm3") |
| 281 |
__ASM_EMIT("pand %%xmm1, %%xmm5") |
| 282 |
__ASM_EMIT("por %%xmm0, %%xmm4") |
| 283 |
__ASM_EMIT("movdqu %%xmm2, 0x00(%[dst], %[off])") |
| 284 |
__ASM_EMIT("pand %%xmm1, %%xmm6") |
| 285 |
__ASM_EMIT("por %%xmm0, %%xmm5") |
| 286 |
__ASM_EMIT("movdqu %%xmm3, 0x10(%[dst], %[off])") |
| 287 |
__ASM_EMIT("pand %%xmm1, %%xmm7") |
| 288 |
__ASM_EMIT("movdqu %%xmm4, 0x20(%[dst], %[off])") |
| 289 |
__ASM_EMIT("por %%xmm0, %%xmm6") |
| 290 |
__ASM_EMIT("movdqu %%xmm5, 0x30(%[dst], %[off])") |
| 291 |
__ASM_EMIT("por %%xmm0, %%xmm7") |
| 292 |
__ASM_EMIT("movdqu %%xmm6, 0x40(%[dst], %[off])") |
| 293 |
__ASM_EMIT("movdqu %%xmm7, 0x50(%[dst], %[off])") |
| 294 |
__ASM_EMIT("add $0x60, %[off]") |
| 295 |
__ASM_EMIT("sub $24, %[count]") |
| 296 |
__ASM_EMIT("jae 1b") |
| 297 |
|
| 298 |
// 16-element block |
| 299 |
__ASM_EMIT("2:") |
| 300 |
__ASM_EMIT("add $8, %[count]") |
| 301 |
__ASM_EMIT("jl 4f") |
| 302 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm2") // xmm2 = AA XX YY ZZ |
| 303 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm3") |
| 304 |
__ASM_EMIT("movdqu 0x20(%[src], %[off]), %%xmm4") |
| 305 |
__ASM_EMIT("movdqu 0x30(%[src], %[off]), %%xmm5") |
| 306 |
__ASM_EMIT("pand %%xmm1, %%xmm2") // xmm2 = 00 XX YY ZZ |
| 307 |
__ASM_EMIT("pand %%xmm1, %%xmm3") |
| 308 |
__ASM_EMIT("pand %%xmm1, %%xmm4") |
| 309 |
__ASM_EMIT("pand %%xmm1, %%xmm5") |
| 310 |
__ASM_EMIT("por %%xmm0, %%xmm2") // xmm2 = VV XX YY ZZ |
| 311 |
__ASM_EMIT("por %%xmm0, %%xmm3") |
| 312 |
__ASM_EMIT("por %%xmm0, %%xmm4") |
| 313 |
__ASM_EMIT("por %%xmm0, %%xmm5") |
| 314 |
__ASM_EMIT("movdqu %%xmm2, 0x00(%[dst], %[off])") |
| 315 |
__ASM_EMIT("movdqu %%xmm3, 0x10(%[dst], %[off])") |
| 316 |
__ASM_EMIT("movdqu %%xmm4, 0x20(%[dst], %[off])") |
| 317 |
__ASM_EMIT("movdqu %%xmm5, 0x30(%[dst], %[off])") |
| 318 |
__ASM_EMIT("add $0x40, %[off]") |
| 319 |
__ASM_EMIT("sub $16, %[count]") |
| 320 |
|
| 321 |
// 8-element block |
| 322 |
__ASM_EMIT("4:") |
| 323 |
__ASM_EMIT("add $8, %[count]") |
| 324 |
__ASM_EMIT("jl 6f") |
| 325 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm2") // xmm2 = AA XX YY ZZ |
| 326 |
__ASM_EMIT("movdqu 0x10(%[src], %[off]), %%xmm3") |
| 327 |
__ASM_EMIT("pand %%xmm1, %%xmm2") // xmm2 = 00 XX YY ZZ |
| 328 |
__ASM_EMIT("pand %%xmm1, %%xmm3") |
| 329 |
__ASM_EMIT("por %%xmm0, %%xmm2") // xmm2 = VV XX YY ZZ |
| 330 |
__ASM_EMIT("por %%xmm0, %%xmm3") |
| 331 |
__ASM_EMIT("movdqu %%xmm2, 0x00(%[dst], %[off])") |
| 332 |
__ASM_EMIT("movdqu %%xmm3, 0x10(%[dst], %[off])") |
| 333 |
__ASM_EMIT("add $0x20, %[off]") |
| 334 |
__ASM_EMIT("sub $8, %[count]") |
| 335 |
|
| 336 |
// 4-element block |
| 337 |
__ASM_EMIT("6:") |
| 338 |
__ASM_EMIT("add $4, %[count]") |
| 339 |
__ASM_EMIT("jl 8f") |
| 340 |
__ASM_EMIT("movdqu 0x00(%[src], %[off]), %%xmm2") // xmm2 = AA XX YY ZZ |
| 341 |
__ASM_EMIT("pand %%xmm1, %%xmm2") // xmm2 = 00 XX YY ZZ |
| 342 |
__ASM_EMIT("por %%xmm0, %%xmm2") // xmm2 = VV XX YY ZZ |
| 343 |
__ASM_EMIT("movdqu %%xmm2, 0x00(%[dst], %[off])") |
| 344 |
__ASM_EMIT("add $0x10, %[off]") |
| 345 |
__ASM_EMIT("sub $4, %[count]") |
| 346 |
|
| 347 |
// Tail |
| 348 |
__ASM_EMIT("8:") |
| 349 |
__ASM_EMIT("add $3, %[count]") |
| 350 |
__ASM_EMIT("jl 10f") |
| 351 |
__ASM_EMIT("9:") |
| 352 |
__ASM_EMIT("movd 0x00(%[src], %[off]), %%xmm2") // xmm2 = AA XX YY ZZ |
| 353 |
__ASM_EMIT("pand %%xmm1, %%xmm2") // xmm2 = 00 XX YY ZZ |
| 354 |
__ASM_EMIT("por %%xmm0, %%xmm2") // xmm2 = VV XX YY ZZ |
| 355 |
__ASM_EMIT("movd %%xmm2, 0x00(%[dst], %[off])") |
| 356 |
__ASM_EMIT("add $4, %[off]") |
| 357 |
__ASM_EMIT("dec %[count]") |
| 358 |
__ASM_EMIT("jge 9b") |
| 359 |
|
| 360 |
// End |
| 361 |
__ASM_EMIT("10:") |
| 362 |
|
| 363 |
: [dst] "+r"(dst), [src] "+r"(src), [count] "+r" (count), |
| 364 |
[off] "=&r" (off), [value] "+Yz" (value) |
| 365 |
: [MASK] "m" (pabc32_set_alpha_const) |
| 366 |
: "cc", "memory", |
| 367 |
"%xmm1", "%xmm2", "%xmm3", |
| 368 |
"%xmm4", "%xmm5", "%xmm6", "%xmm7" |
| 369 |
); |
| 370 |
} |
| 371 |
|
| 372 |
} /* namespace sse2 */ |
| 373 |
} /* namespace lsp */ |
| 374 |
|
| 375 |
|
| 376 |
#endif /* PRIVATE_DSP_ARCH_X86_SSE2_PIXELFMT_H_ */ |