Chromium Code Reviews
chromiumcodereview-hr@appspot.gserviceaccount.com (chromiumcodereview-hr) | Please choose your nickname with Settings | Help | Chromium Project | Gerrit Changes | Sign out
(118)

Issues Search
    My Issues | Starred     Open | Closed | All

Side by Side Diff: media/base/simd/convert_yuva_to_argb_mmx.inc

Issue 271443006: Fix crash and off-by-one error in MMX YUVA->ARGB conversion. (Closed) Base URL: https://chromium.googlesource.com/chromium/src.git@yuv_tables
Patch Set: Make off-by-one fix 32-bit-friendly, fix Android unused stuff Created 6 years, 7 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 | « no previous file | media/base/yuv_convert_unittest.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 ; Copyright (c) 2011 The Chromium Authors. All rights reserved. 1 ; Copyright (c) 2011 The Chromium Authors. All rights reserved.
2 ; Use of this source code is governed by a BSD-style license that can be 2 ; Use of this source code is governed by a BSD-style license that can be
3 ; found in the LICENSE file. 3 ; found in the LICENSE file.
4 4
5 %include "media/base/simd/media_export.asm" 5 %include "media/base/simd/media_export.asm"
6 6
7 EXPORT SYMBOL 7 EXPORT SYMBOL
8 align function_align 8 align function_align
9 9
10 mangle(SYMBOL): 10 mangle(SYMBOL):
11 %assign stack_offset 0 11 %assign stack_offset 0
12 PROLOGUE 7, 7, 3, Y, U, V, A, ARGB, WIDTH, TABLE, TEMP 12 PROLOGUE 7, 7, 3, Y, U, V, A, ARGB, WIDTH, TABLE
13 PUSH WIDTHq 13 PUSH WIDTHq
14 DEFINE_ARGS Y, U, V, A, ARGB, TABLE, TEMP 14 DEFINE_ARGS Y, U, V, A, ARGB, TABLE, TEMP
15 mov TABLEq, TEMPq
15 jmp .convertend 16 jmp .convertend
16 17
17 .convertloop: 18 .convertloop:
18 movzx TEMPd, BYTE [Uq] 19 movzx TEMPd, BYTE [Uq]
19 movq mm0, [TABLEq + 2048 + 8 * TEMPq] 20 movq mm0, [TABLEq + 2048 + 8 * TEMPq]
20 add Uq, 1 21 add Uq, 1
21 22
22 movzx TEMPd, BYTE [Vq] 23 movzx TEMPd, BYTE [Vq]
23 paddsw mm0, [TABLEq + 4096 + 8 * TEMPq] 24 paddsw mm0, [TABLEq + 4096 + 8 * TEMPq]
24 add Vq, 1 25 add Vq, 1
25 26
26 movzx TEMPd, BYTE [Yq] 27 movzx TEMPd, BYTE [Yq]
27 movq mm1, [TABLEq + 8 * TEMPq] 28 movq mm1, [TABLEq + 8 * TEMPq]
28 29
29 movzx TEMPd, BYTE [Yq + 1] 30 movzx TEMPd, BYTE [Yq + 1]
30 movq mm2, [TABLEq + 8 * TEMPq] 31 movq mm2, [TABLEq + 8 * TEMPq]
31 add Yq, 2 32 add Yq, 2
32 33
33 ; Add UV components to Y component. 34 ; Add UV components to Y component.
34 paddsw mm1, mm0 35 paddsw mm1, mm0
35 paddsw mm2, mm0 36 paddsw mm2, mm0
36 37
37 ; Down shift and then pack. 38 ; Down shift and then pack.
38 psraw mm1, 6 39 psraw mm1, 6
39 psraw mm2, 6 40 psraw mm2, 6
40 packuswb mm1, mm2 41 packuswb mm1, mm2
41 42
42 ; Unpack and multiply by alpha value, then repack high bytes of words. 43 ; Unpack
43 movq mm0, mm1 44 movq mm0, mm1
44 pxor mm2, mm2 45 pxor mm2, mm2
45 punpcklbw mm0, mm2 46 punpcklbw mm0, mm2
46 punpckhbw mm1, mm2 47 punpckhbw mm1, mm2
48
49 ; Add one to our alpha values, this is a somewhat unfortunate hack; while
50 ; the pack/unpack above handle saturating any negative numbers to 0, they also
51 ; truncate the alpha value to 255. The math ahead wants to produce the same
52 ; ARGB alpha value as the source pixel in YUVA, but this depends on the alpha
53 ; value in |mm0| and |mm1| being 256, (let A be the source image alpha,
54 ; 256 * A >> 8 == A, whereas 255 * A >> 8 is off by one except at 0).
55 mov TEMPq, 0x00010000
56 movd mm2, TEMPd
57 psllq mm2, 32
58 paddsw mm0, mm2
59 paddsw mm1, mm2
60
61 ; Multiply by alpha value, then repack high bytes of words.
47 movzx TEMPd, BYTE [Aq] 62 movzx TEMPd, BYTE [Aq]
48 movq mm2, [TABLEq + 6144 + 8 * TEMPq] 63 movq mm2, [TABLEq + 6144 + 8 * TEMPq]
49 pmullw mm0, mm2 64 pmullw mm0, mm2
50 psrlw mm0, 8 65 psrlw mm0, 8
51 movzx TEMPd, BYTE [Aq + 1] 66 movzx TEMPd, BYTE [Aq + 1]
52 movq mm2, [TABLEq + 6144 + 8 * TEMPq] 67 movq mm2, [TABLEq + 6144 + 8 * TEMPq]
53 add Aq, 2 68 add Aq, 2
54 pmullw mm1, mm2 69 pmullw mm1, mm2
55 psrlw mm1, 8 70 psrlw mm1, 8
56 packuswb mm0, mm1 71 packuswb mm0, mm1
(...skipping 15 matching lines...) Expand all
72 paddsw mm0, [TABLEq + 4096 + 8 * TEMPq] 87 paddsw mm0, [TABLEq + 4096 + 8 * TEMPq]
73 movzx TEMPd, BYTE [Yq] 88 movzx TEMPd, BYTE [Yq]
74 movq mm1, [TABLEq + 8 * TEMPq] 89 movq mm1, [TABLEq + 8 * TEMPq]
75 paddsw mm1, mm0 90 paddsw mm1, mm0
76 psraw mm1, 6 91 psraw mm1, 6
77 packuswb mm1, mm1 92 packuswb mm1, mm1
78 93
79 ; Multiply ARGB by alpha value. 94 ; Multiply ARGB by alpha value.
80 pxor mm0, mm0 95 pxor mm0, mm0
81 punpcklbw mm1, mm0 96 punpcklbw mm1, mm0
97
98 ; See above note about this hack.
99 mov TEMPq, 0x00010000
100 movd mm0, TEMPd
101 psllq mm0, 32
102 paddsw mm1, mm0
103
82 movzx TEMPd, BYTE [Aq] 104 movzx TEMPd, BYTE [Aq]
83 movq mm0, [TABLEq + 6144 + 8 * TEMPq] 105 movq mm0, [TABLEq + 6144 + 8 * TEMPq]
84 pmullw mm1, mm0 106 pmullw mm1, mm0
85 psrlw mm1, 8 107 psrlw mm1, 8
86 packuswb mm1, mm1 108 packuswb mm1, mm1
87 109
88 movd [ARGBq], mm1 110 movd [ARGBq], mm1
89 111
90 .convertdone: 112 .convertdone:
91 POP TABLEq 113 POP TABLEq
92 RET 114 RET
OLDNEW
« no previous file with comments | « no previous file | media/base/yuv_convert_unittest.cc » ('j') | no next file with comments »

Powered by Google App Engine
This is Rietveld 408576698