media/libvpx/vp9/encoder/x86/vp9_subtract_sse2.asm@6474c204b198 (annotated)
media/libvpx/vp9/encoder/x86/vp9_subtract_sse2.asm
Wed, 31 Dec 2014 06:09:35 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Wed, 31 Dec 2014 06:09:35 +0100
- changeset 0
- 6474c204b198
- permissions
- -rw-r--r--
Cloned upstream origin tor-browser at tor-browser-31.3.0esr-4.5-1-build1
revision ID fc1c9ff7c1b2defdbc039f12214767608f46423f for hacking purpose.
| michael@0 | 1 | ; |
| michael@0 | 2 | ; Copyright (c) 2010 The WebM project authors. All Rights Reserved. |
| michael@0 | 3 | ; |
| michael@0 | 4 | ; Use of this source code is governed by a BSD-style license |
| michael@0 | 5 | ; that can be found in the LICENSE file in the root of the source |
| michael@0 | 6 | ; tree. An additional intellectual property rights grant can be found |
| michael@0 | 7 | ; in the file PATENTS. All contributing project authors may |
| michael@0 | 8 | ; be found in the AUTHORS file in the root of the source tree. |
| michael@0 | 9 | ; |
| michael@0 | 10 | |
| michael@0 | 11 | %include "third_party/x86inc/x86inc.asm" |
| michael@0 | 12 | |
| michael@0 | 13 | SECTION .text |
| michael@0 | 14 | |
| michael@0 | 15 | ; void vp9_subtract_block(int rows, int cols, |
| michael@0 | 16 | ; int16_t *diff, ptrdiff_t diff_stride, |
| michael@0 | 17 | ; const uint8_t *src, ptrdiff_t src_stride, |
| michael@0 | 18 | ; const uint8_t *pred, ptrdiff_t pred_stride) |
| michael@0 | 19 | |
| michael@0 | 20 | INIT_XMM sse2 |
| michael@0 | 21 | cglobal subtract_block, 7, 7, 8, \ |
| michael@0 | 22 | rows, cols, diff, diff_stride, src, src_stride, \ |
| michael@0 | 23 | pred, pred_stride |
| michael@0 | 24 | %define pred_str colsq |
| michael@0 | 25 | pxor m7, m7 ; dedicated zero register |
| michael@0 | 26 | cmp colsd, 4 |
| michael@0 | 27 | je .case_4 |
| michael@0 | 28 | cmp colsd, 8 |
| michael@0 | 29 | je .case_8 |
| michael@0 | 30 | cmp colsd, 16 |
| michael@0 | 31 | je .case_16 |
| michael@0 | 32 | cmp colsd, 32 |
| michael@0 | 33 | je .case_32 |
| michael@0 | 34 | |
| michael@0 | 35 | %macro loop16 6 |
| michael@0 | 36 | mova m0, [srcq+%1] |
| michael@0 | 37 | mova m4, [srcq+%2] |
| michael@0 | 38 | mova m1, [predq+%3] |
| michael@0 | 39 | mova m5, [predq+%4] |
| michael@0 | 40 | punpckhbw m2, m0, m7 |
| michael@0 | 41 | punpckhbw m3, m1, m7 |
| michael@0 | 42 | punpcklbw m0, m7 |
| michael@0 | 43 | punpcklbw m1, m7 |
| michael@0 | 44 | psubw m2, m3 |
| michael@0 | 45 | psubw m0, m1 |
| michael@0 | 46 | punpckhbw m1, m4, m7 |
| michael@0 | 47 | punpckhbw m3, m5, m7 |
| michael@0 | 48 | punpcklbw m4, m7 |
| michael@0 | 49 | punpcklbw m5, m7 |
| michael@0 | 50 | psubw m1, m3 |
| michael@0 | 51 | psubw m4, m5 |
| michael@0 | 52 | mova [diffq+mmsize*0+%5], m0 |
| michael@0 | 53 | mova [diffq+mmsize*1+%5], m2 |
| michael@0 | 54 | mova [diffq+mmsize*0+%6], m4 |
| michael@0 | 55 | mova [diffq+mmsize*1+%6], m1 |
| michael@0 | 56 | %endmacro |
| michael@0 | 57 | |
| michael@0 | 58 | mov pred_str, pred_stridemp |
| michael@0 | 59 | .loop_64: |
| michael@0 | 60 | loop16 0*mmsize, 1*mmsize, 0*mmsize, 1*mmsize, 0*mmsize, 2*mmsize |
| michael@0 | 61 | loop16 2*mmsize, 3*mmsize, 2*mmsize, 3*mmsize, 4*mmsize, 6*mmsize |
| michael@0 | 62 | lea diffq, [diffq+diff_strideq*2] |
| michael@0 | 63 | add predq, pred_str |
| michael@0 | 64 | add srcq, src_strideq |
| michael@0 | 65 | dec rowsd |
| michael@0 | 66 | jg .loop_64 |
| michael@0 | 67 | RET |
| michael@0 | 68 | |
| michael@0 | 69 | .case_32: |
| michael@0 | 70 | mov pred_str, pred_stridemp |
| michael@0 | 71 | .loop_32: |
| michael@0 | 72 | loop16 0, mmsize, 0, mmsize, 0, 2*mmsize |
| michael@0 | 73 | lea diffq, [diffq+diff_strideq*2] |
| michael@0 | 74 | add predq, pred_str |
| michael@0 | 75 | add srcq, src_strideq |
| michael@0 | 76 | dec rowsd |
| michael@0 | 77 | jg .loop_32 |
| michael@0 | 78 | RET |
| michael@0 | 79 | |
| michael@0 | 80 | .case_16: |
| michael@0 | 81 | mov pred_str, pred_stridemp |
| michael@0 | 82 | .loop_16: |
| michael@0 | 83 | loop16 0, src_strideq, 0, pred_str, 0, diff_strideq*2 |
| michael@0 | 84 | lea diffq, [diffq+diff_strideq*4] |
| michael@0 | 85 | lea predq, [predq+pred_str*2] |
| michael@0 | 86 | lea srcq, [srcq+src_strideq*2] |
| michael@0 | 87 | sub rowsd, 2 |
| michael@0 | 88 | jg .loop_16 |
| michael@0 | 89 | RET |
| michael@0 | 90 | |
| michael@0 | 91 | %macro loop_h 0 |
| michael@0 | 92 | movh m0, [srcq] |
| michael@0 | 93 | movh m2, [srcq+src_strideq] |
| michael@0 | 94 | movh m1, [predq] |
| michael@0 | 95 | movh m3, [predq+pred_str] |
| michael@0 | 96 | punpcklbw m0, m7 |
| michael@0 | 97 | punpcklbw m1, m7 |
| michael@0 | 98 | punpcklbw m2, m7 |
| michael@0 | 99 | punpcklbw m3, m7 |
| michael@0 | 100 | psubw m0, m1 |
| michael@0 | 101 | psubw m2, m3 |
| michael@0 | 102 | mova [diffq], m0 |
| michael@0 | 103 | mova [diffq+diff_strideq*2], m2 |
| michael@0 | 104 | %endmacro |
| michael@0 | 105 | |
| michael@0 | 106 | .case_8: |
| michael@0 | 107 | mov pred_str, pred_stridemp |
| michael@0 | 108 | .loop_8: |
| michael@0 | 109 | loop_h |
| michael@0 | 110 | lea diffq, [diffq+diff_strideq*4] |
| michael@0 | 111 | lea srcq, [srcq+src_strideq*2] |
| michael@0 | 112 | lea predq, [predq+pred_str*2] |
| michael@0 | 113 | sub rowsd, 2 |
| michael@0 | 114 | jg .loop_8 |
| michael@0 | 115 | RET |
| michael@0 | 116 | |
| michael@0 | 117 | INIT_MMX |
| michael@0 | 118 | .case_4: |
| michael@0 | 119 | mov pred_str, pred_stridemp |
| michael@0 | 120 | .loop_4: |
| michael@0 | 121 | loop_h |
| michael@0 | 122 | lea diffq, [diffq+diff_strideq*4] |
| michael@0 | 123 | lea srcq, [srcq+src_strideq*2] |
| michael@0 | 124 | lea predq, [predq+pred_str*2] |
| michael@0 | 125 | sub rowsd, 2 |
| michael@0 | 126 | jg .loop_4 |
| michael@0 | 127 | RET |
