OLD | NEW |
1 /* | 1 /* |
2 * Copyright 2013 The LibYuv Project Authors. All rights reserved. | 2 * Copyright 2013 The LibYuv Project Authors. All rights reserved. |
3 * | 3 * |
4 * Use of this source code is governed by a BSD-style license | 4 * Use of this source code is governed by a BSD-style license |
5 * that can be found in the LICENSE file in the root of the source | 5 * that can be found in the LICENSE file in the root of the source |
6 * tree. An additional intellectual property rights grant can be found | 6 * tree. An additional intellectual property rights grant can be found |
7 * in the file PATENTS. All contributing project authors may | 7 * in the file PATENTS. All contributing project authors may |
8 * be found in the AUTHORS file in the root of the source tree. | 8 * be found in the AUTHORS file in the root of the source tree. |
9 */ | 9 */ |
10 | 10 |
(...skipping 601 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
612 "vpxor %%ymm5,%%ymm5,%%ymm5 \n" | 612 "vpxor %%ymm5,%%ymm5,%%ymm5 \n" |
613 | 613 |
614 LABELALIGN | 614 LABELALIGN |
615 "1: \n" | 615 "1: \n" |
616 "vmovdqu " MEMACCESS(0) ",%%ymm3 \n" | 616 "vmovdqu " MEMACCESS(0) ",%%ymm3 \n" |
617 "lea " MEMLEA(0x20,0) ",%0 \n" // src_ptr += 32 | 617 "lea " MEMLEA(0x20,0) ",%0 \n" // src_ptr += 32 |
618 "vpermq $0xd8,%%ymm3,%%ymm3 \n" | 618 "vpermq $0xd8,%%ymm3,%%ymm3 \n" |
619 "vpunpcklbw %%ymm5,%%ymm3,%%ymm2 \n" | 619 "vpunpcklbw %%ymm5,%%ymm3,%%ymm2 \n" |
620 "vpunpckhbw %%ymm5,%%ymm3,%%ymm3 \n" | 620 "vpunpckhbw %%ymm5,%%ymm3,%%ymm3 \n" |
621 "vpaddusw " MEMACCESS(1) ",%%ymm2,%%ymm0 \n" | 621 "vpaddusw " MEMACCESS(1) ",%%ymm2,%%ymm0 \n" |
622 "vpaddusw " MEMACCESS2(0x10,1) ",%%ymm3,%%ymm1 \n" | 622 "vpaddusw " MEMACCESS2(0x20,1) ",%%ymm3,%%ymm1 \n" |
623 "vmovdqu %%ymm0," MEMACCESS(1) " \n" | 623 "vmovdqu %%ymm0," MEMACCESS(1) " \n" |
624 "vmovdqu %%ymm1," MEMACCESS2(0x10,1) " \n" | 624 "vmovdqu %%ymm1," MEMACCESS2(0x20,1) " \n" |
625 "lea " MEMLEA(0x40,1) ",%1 \n" | 625 "lea " MEMLEA(0x40,1) ",%1 \n" |
626 "sub $0x20,%2 \n" | 626 "sub $0x20,%2 \n" |
627 "jg 1b \n" | 627 "jg 1b \n" |
628 "vzeroupper \n" | 628 "vzeroupper \n" |
629 : "+r"(src_ptr), // %0 | 629 : "+r"(src_ptr), // %0 |
630 "+r"(dst_ptr), // %1 | 630 "+r"(dst_ptr), // %1 |
631 "+r"(src_width) // %2 | 631 "+r"(src_width) // %2 |
632 : | 632 : |
633 : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5" | 633 : "memory", "cc", "xmm0", "xmm1", "xmm2", "xmm3", "xmm5" |
634 ); | 634 ); |
(...skipping 462 matching lines...) Expand 10 before | Expand all | Expand 10 after Loading... |
1097 ); | 1097 ); |
1098 return num; | 1098 return num; |
1099 } | 1099 } |
1100 | 1100 |
1101 #endif // defined(__x86_64__) || defined(__i386__) | 1101 #endif // defined(__x86_64__) || defined(__i386__) |
1102 | 1102 |
1103 #ifdef __cplusplus | 1103 #ifdef __cplusplus |
1104 } // extern "C" | 1104 } // extern "C" |
1105 } // namespace libyuv | 1105 } // namespace libyuv |
1106 #endif | 1106 #endif |
OLD | NEW |