Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(268)

Side by Side Diff: source/row_gcc.cc

Issue 2387713002: HalfFloat_SSE2 for Visual C (Closed)
Patch Set: planar functions sse2 only Created 4 years, 2 months ago
Use n/p to move between diff chunks; N/P to move between comments. Draft comments are only viewable by you.
Jump to:
View unified diff | Download patch
« no previous file with comments | « source/planar_functions.cc ('k') | source/row_win.cc » ('j') | no next file with comments »
Toggle Intra-line Diffs ('i') | Expand Comments ('e') | Collapse Comments ('c') | Show Comments Hide Comments ('s')
OLDNEW
1 // VERSION 2 1 // VERSION 2
2 /* 2 /*
3 * Copyright 2011 The LibYuv Project Authors. All rights reserved. 3 * Copyright 2011 The LibYuv Project Authors. All rights reserved.
4 * 4 *
5 * Use of this source code is governed by a BSD-style license 5 * Use of this source code is governed by a BSD-style license
6 * that can be found in the LICENSE file in the root of the source 6 * that can be found in the LICENSE file in the root of the source
7 * tree. An additional intellectual property rights grant can be found 7 * tree. An additional intellectual property rights grant can be found
8 * in the file PATENTS. All contributing project authors may 8 * in the file PATENTS. All contributing project authors may
9 * be found in the AUTHORS file in the root of the source tree. 9 * be found in the AUTHORS file in the root of the source tree.
10 */ 10 */
(...skipping 5349 matching lines...) Expand 10 before | Expand all | Expand 10 after
5360 "+r"(dst_argb), // %1 5360 "+r"(dst_argb), // %1
5361 "+r"(width) // %2 5361 "+r"(width) // %2
5362 : "r"(poly) // %3 5362 : "r"(poly) // %3
5363 : "memory", "cc", 5363 : "memory", "cc",
5364 "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7" 5364 "xmm0", "xmm1", "xmm2", "xmm3", "xmm4", "xmm5", "xmm6", "xmm7"
5365 ); 5365 );
5366 } 5366 }
5367 #endif // HAS_ARGBPOLYNOMIALROW_AVX2 5367 #endif // HAS_ARGBPOLYNOMIALROW_AVX2
5368 5368
5369 #ifdef HAS_HALFFLOATROW_SSE2 5369 #ifdef HAS_HALFFLOATROW_SSE2
5370 static float kScaleBias = 1.9259299444e-34f;
5370 void HalfFloatRow_SSE2(const uint16* src, uint16* dst, float scale, int width) { 5371 void HalfFloatRow_SSE2(const uint16* src, uint16* dst, float scale, int width) {
5371 float mult = 1.9259299444e-34f * scale;
5372 asm volatile ( 5372 asm volatile (
5373 "movd %3,%%xmm4 \n" 5373 "pshufd $0x0,%3,%%xmm4 \n"
5374 "pshufd $0x0,%%xmm4,%%xmm4 \n"
5375 "pxor %%xmm5,%%xmm5 \n" 5374 "pxor %%xmm5,%%xmm5 \n"
5376 5375
5377 // 16 pixel loop. 5376 // 16 pixel loop.
5378 LABELALIGN 5377 LABELALIGN
5379 "1: \n" 5378 "1: \n"
5380 "movdqu " MEMACCESS(0) ",%%xmm0 \n" // 8 shorts 5379 "movdqu " MEMACCESS(0) ",%%xmm2 \n" // 8 shorts
5381 "lea " MEMLEA(0x10,0) ",%0 \n" 5380 "lea " MEMLEA(0x10,0) ",%0 \n"
5382 "movdqa %%xmm0,%%xmm1 \n" 5381 "movdqa %%xmm2,%%xmm3 \n"
5383 "punpcklwd %%xmm5,%%xmm0 \n" // 8 ints in xmm0/1 5382 "punpcklwd %%xmm5,%%xmm2 \n" // 8 ints in xmm2/1
5384 "cvtdq2ps %%xmm0,%%xmm0 \n" // 8 floats 5383 "cvtdq2ps %%xmm2,%%xmm2 \n" // 8 floats
5385 "punpckhwd %%xmm5,%%xmm1 \n" 5384 "punpckhwd %%xmm5,%%xmm3 \n"
5386 "cvtdq2ps %%xmm1,%%xmm1 \n" 5385 "cvtdq2ps %%xmm3,%%xmm3 \n"
5387 "mulps %%xmm4,%%xmm0 \n" 5386 "mulps %%xmm4,%%xmm2 \n"
5388 "mulps %%xmm4,%%xmm1 \n" 5387 "mulps %%xmm4,%%xmm3 \n"
5389 "psrld $0xd,%%xmm0 \n" 5388 "psrld $0xd,%%xmm2 \n"
5390 "psrld $0xd,%%xmm1 \n" 5389 "psrld $0xd,%%xmm3 \n"
5391 "packssdw %%xmm1,%%xmm0 \n" 5390 "packssdw %%xmm3,%%xmm2 \n"
5392 "movdqu %%xmm0," MEMACCESS(1) " \n" 5391 "movdqu %%xmm2," MEMACCESS(1) " \n"
5393 "lea " MEMLEA(0x10,1) ",%1 \n" 5392 "lea " MEMLEA(0x10,1) ",%1 \n"
5394 "sub $0x8,%2 \n" 5393 "sub $0x8,%2 \n"
5395 "jg 1b \n" 5394 "jg 1b \n"
5396 : "+r"(src), // %0 5395 : "+r"(src), // %0
5397 "+r"(dst), // %1 5396 "+r"(dst), // %1
5398 "+r"(width) // %2 5397 "+r"(width) // %2
5399 : "rm"(mult) // %3 5398 : "x"(scale * kScaleBias) // %3
5400 : "memory", "cc", 5399 : "memory", "cc",
5401 "xmm0", "xmm1", "xmm4", "xmm5" 5400 "xmm2", "xmm3", "xmm4", "xmm5"
5402 ); 5401 );
5403 } 5402 }
5404 #endif // HAS_HALFFLOATROW_SSE2 5403 #endif // HAS_HALFFLOATROW_SSE2
5405 5404
5406 #ifdef HAS_HALFFLOATROW_AVX2 5405 #ifdef HAS_HALFFLOATROW_AVX2
5407 void HalfFloatRow_AVX2(const uint16* src, uint16* dst, float scale, int width) { 5406 void HalfFloatRow_AVX2(const uint16* src, uint16* dst, float scale, int width) {
5408 asm volatile ( 5407 asm volatile (
5409 "vbroadcastss %3, %%ymm4 \n" 5408 "vbroadcastss %3, %%ymm4 \n"
5410 5409
5411 // 16 pixel loop. 5410 // 16 pixel loop.
5412 LABELALIGN 5411 LABELALIGN
5413 "1: \n" 5412 "1: \n"
5414 "vpmovzxwd " MEMACCESS(0) ",%%ymm0 \n" // 8 shorts -> 8 ints 5413 "vpmovzxwd " MEMACCESS(0) ",%%ymm2 \n" // 8 shorts -> 8 ints
5415 "vpmovzxwd " MEMACCESS2(0x10,0) ",%%ymm1 \n" // 8 more 5414 "vpmovzxwd " MEMACCESS2(0x10,0) ",%%ymm3 \n" // 8 more
5416 "lea " MEMLEA(0x20,0) ",%0 \n" 5415 "lea " MEMLEA(0x20,0) ",%0 \n"
5417 "vcvtdq2ps %%ymm0,%%ymm0 \n" 5416 "vcvtdq2ps %%ymm2,%%ymm2 \n"
5418 "vcvtdq2ps %%ymm1,%%ymm1 \n" 5417 "vcvtdq2ps %%ymm3,%%ymm3 \n"
5419 "vmulps %%ymm0,%%ymm4,%%ymm0 \n" 5418 "vmulps %%ymm2,%%ymm4,%%ymm2 \n"
5420 "vmulps %%ymm1,%%ymm4,%%ymm1 \n" 5419 "vmulps %%ymm3,%%ymm4,%%ymm3 \n"
5421 "vcvtps2ph $3, %%ymm0, %%xmm0 \n" 5420 "vcvtps2ph $3, %%ymm2, %%xmm2 \n"
5422 "vcvtps2ph $3, %%ymm1, %%xmm1 \n" 5421 "vcvtps2ph $3, %%ymm3, %%xmm3 \n"
5423 "vmovdqu %%xmm0," MEMACCESS(1) " \n" 5422 "vmovdqu %%xmm2," MEMACCESS(1) " \n"
5424 "vmovdqu %%xmm1," MEMACCESS2(0x10,1) " \n" 5423 "vmovdqu %%xmm3," MEMACCESS2(0x10,1) " \n"
5425 "lea " MEMLEA(0x20,1) ",%1 \n" 5424 "lea " MEMLEA(0x20,1) ",%1 \n"
5426 "sub $0x10,%2 \n" 5425 "sub $0x10,%2 \n"
5427 "jg 1b \n" 5426 "jg 1b \n"
5428 "vzeroupper \n" 5427 "vzeroupper \n"
5429 : "+r"(src), // %0 5428 : "+r"(src), // %0
5430 "+r"(dst), // %1 5429 "+r"(dst), // %1
5431 "+r"(width) // %2 5430 "+r"(width) // %2
5432 : "x"(scale) // %3 5431 : "x"(scale) // %3
5433 : "memory", "cc", 5432 : "memory", "cc",
5434 "xmm0", "xmm1", "xmm4" 5433 "xmm2", "xmm3", "xmm4"
5435 ); 5434 );
5436 } 5435 }
5437 #endif // HAS_HALFFLOATROW_AVX2 5436 #endif // HAS_HALFFLOATROW_AVX2
5438 5437
5439 #ifdef HAS_ARGBCOLORTABLEROW_X86 5438 #ifdef HAS_ARGBCOLORTABLEROW_X86
5440 // Tranform ARGB pixels with color table. 5439 // Tranform ARGB pixels with color table.
5441 void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb, 5440 void ARGBColorTableRow_X86(uint8* dst_argb, const uint8* table_argb,
5442 int width) { 5441 int width) {
5443 uintptr_t pixel_temp; 5442 uintptr_t pixel_temp;
5444 asm volatile ( 5443 asm volatile (
(...skipping 150 matching lines...) Expand 10 before | Expand all | Expand 10 after
5595 ); 5594 );
5596 } 5595 }
5597 #endif // HAS_ARGBLUMACOLORTABLEROW_SSSE3 5596 #endif // HAS_ARGBLUMACOLORTABLEROW_SSSE3
5598 5597
5599 #endif // defined(__x86_64__) || defined(__i386__) 5598 #endif // defined(__x86_64__) || defined(__i386__)
5600 5599
5601 #ifdef __cplusplus 5600 #ifdef __cplusplus
5602 } // extern "C" 5601 } // extern "C"
5603 } // namespace libyuv 5602 } // namespace libyuv
5604 #endif 5603 #endif
OLDNEW
« no previous file with comments | « source/planar_functions.cc ('k') | source/row_win.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698