annotate src/video/SDL_yuv_mmx.c @ 1452:b7184889c1cb

*** empty log message ***
author Sam Lantinga <slouken@libsdl.org>
date Mon, 27 Feb 2006 18:58:12 +0000
parents 40edc79b0926
children 782fd950bd46 c121d94672cb f420bba13676
rev   line source
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
1 /*
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
2 SDL - Simple DirectMedia Layer
1312
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
3 Copyright (C) 1997-2006 Sam Lantinga
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
5 This library is free software; you can redistribute it and/or
1312
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
6 modify it under the terms of the GNU Lesser General Public
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
7 License as published by the Free Software Foundation; either
1312
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
8 version 2.1 of the License, or (at your option) any later version.
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
9
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
10 This library is distributed in the hope that it will be useful,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
11 but WITHOUT ANY WARRANTY; without even the implied warranty of
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
12 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
1312
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
13 Lesser General Public License for more details.
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
14
1312
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
15 You should have received a copy of the GNU Lesser General Public
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
16 License along with this library; if not, write to the Free Software
c9b51268668f Updated copyright information and removed rcs id lines (problematic in branch merges)
Sam Lantinga <slouken@libsdl.org>
parents: 1148
diff changeset
17 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
18
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
19 Sam Lantinga
252
e8157fcb3114 Updated the source with the correct e-mail address
Sam Lantinga <slouken@libsdl.org>
parents: 0
diff changeset
20 slouken@libsdl.org
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
21 */
1402
d910939febfa Use consistent identifiers for the various platforms we support.
Sam Lantinga <slouken@libsdl.org>
parents: 1361
diff changeset
22 #include "SDL_config.h"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
23
1413
Sam Lantinga <slouken@libsdl.org>
parents: 1407
diff changeset
24 #if 0 /* FIXME: This code needs to be rewritten to reference the static data using relocatable addresses (e.g. http://www.gentoo.org/proj/en/hardened/pic-fix-guide.xml or http://nasm.sourceforge.net/doc/html/nasmdoc8.html#section-8.2) This code currently breaks on systems with readonly text segments (hardened Linux / Intel Mac) */
1402
d910939febfa Use consistent identifiers for the various platforms we support.
Sam Lantinga <slouken@libsdl.org>
parents: 1361
diff changeset
25 #if defined(__GNUC__) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
26
1407
0c6941483cc6 Whoops, forgot to check in this fix
Sam Lantinga <slouken@libsdl.org>
parents: 1402
diff changeset
27 #include "SDL_stdinc.h"
0c6941483cc6 Whoops, forgot to check in this fix
Sam Lantinga <slouken@libsdl.org>
parents: 1402
diff changeset
28
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
29 #define ASM_ARRAY(x) x[] __asm__("_" #x) __attribute__((used))
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
30
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
31 static unsigned int ASM_ARRAY(MMX_0080w) = {0x00800080, 0x00800080};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
32 static unsigned int ASM_ARRAY(MMX_00FFw) = {0x00ff00ff, 0x00ff00ff};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
33 static unsigned int ASM_ARRAY(MMX_FF00w) = {0xff00ff00, 0xff00ff00};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
34
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
35 static unsigned short ASM_ARRAY(MMX_Ycoeff) = {0x004a, 0x004a, 0x004a, 0x004a};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
36
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
37 static unsigned short ASM_ARRAY(MMX_UbluRGB) = {0x0072, 0x0072, 0x0072, 0x0072};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
38 static unsigned short ASM_ARRAY(MMX_VredRGB) = {0x0059, 0x0059, 0x0059, 0x0059};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
39 static unsigned short ASM_ARRAY(MMX_UgrnRGB) = {0xffea, 0xffea, 0xffea, 0xffea};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
40 static unsigned short ASM_ARRAY(MMX_VgrnRGB) = {0xffd2, 0xffd2, 0xffd2, 0xffd2};
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
41
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
42 static unsigned short ASM_ARRAY(MMX_Ublu5x5) = {0x0081, 0x0081, 0x0081, 0x0081};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
43 static unsigned short ASM_ARRAY(MMX_Vred5x5) = {0x0066, 0x0066, 0x0066, 0x0066};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
44 static unsigned short ASM_ARRAY(MMX_Ugrn555) = {0xffe7, 0xffe7, 0xffe7, 0xffe7};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
45 static unsigned short ASM_ARRAY(MMX_Vgrn555) = {0xffcc, 0xffcc, 0xffcc, 0xffcc};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
46 static unsigned short ASM_ARRAY(MMX_Ugrn565) = {0xffe8, 0xffe8, 0xffe8, 0xffe8};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
47 static unsigned short ASM_ARRAY(MMX_Vgrn565) = {0xffcd, 0xffcd, 0xffcd, 0xffcd};
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
48
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
49 static unsigned short ASM_ARRAY(MMX_red555) = {0x7c00, 0x7c00, 0x7c00, 0x7c00};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
50 static unsigned short ASM_ARRAY(MMX_red565) = {0xf800, 0xf800, 0xf800, 0xf800};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
51 static unsigned short ASM_ARRAY(MMX_grn555) = {0x03e0, 0x03e0, 0x03e0, 0x03e0};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
52 static unsigned short ASM_ARRAY(MMX_grn565) = {0x07e0, 0x07e0, 0x07e0, 0x07e0};
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
53 static unsigned short ASM_ARRAY(MMX_blu5x5) = {0x001f, 0x001f, 0x001f, 0x001f};
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
54
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
55 /**
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
56 This MMX assembler is my first assembler/MMX program ever.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
57 Thus it maybe buggy.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
58 Send patches to:
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
59 mvogt@rhrk.uni-kl.de
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
60
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
61 After it worked fine I have "obfuscated" the code a bit to have
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
62 more parallism in the MMX units. This means I moved
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
63 initilisation around and delayed other instruction.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
64 Performance measurement did not show that this brought any advantage
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
65 but in theory it _should_ be faster this way.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
66
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
67 The overall performanve gain to the C based dither was 30%-40%.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
68 The MMX routine calculates 256bit=8RGB values in each cycle
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
69 (4 for row1 & 4 for row2)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
70
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
71 The red/green/blue.. coefficents are taken from the mpeg_play
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
72 player. They look nice, but I dont know if you can have
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
73 better values, to avoid integer rounding errors.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
74
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
75
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
76 IMPORTANT:
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
77 ==========
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
78
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
79 It is a requirement that the cr/cb/lum are 8 byte aligned and
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
80 the out are 16byte aligned or you will/may get segfaults
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
81
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
82 */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
83
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
84 void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
85 unsigned char *lum, unsigned char *cr,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
86 unsigned char *cb, unsigned char *out,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
87 int rows, int cols, int mod )
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
88 {
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
89 Uint32 *row1;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
90 Uint32 *row2;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
91
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
92 unsigned char* y = lum +cols*rows; // Pointer to the end
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
93 int x=0;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
94 row1 = (Uint32 *)out; // 32 bit target
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
95 row2 = (Uint32 *)out+cols+mod; // start of second row
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
96 mod = (mod+cols+mod)*4; // increment for row1 in byte
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
97
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
98 __asm__ __volatile__ (
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
99 /* We don't really care about PIC - the code should be rewritten to use
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
100 relative addressing for the static tables, so right now we take the
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
101 COW hit on the pages this code resides. Big deal.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
102 This spill is just to reduce register pressure in the PIC case. */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
103 "pushl %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
104 "movl %0, %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
105
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
106 ".align 8\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
107 "1:\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
108
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
109 // create Cr (result in mm1)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
110 "movd (%%ebx), %%mm1\n" // 0 0 0 0 v3 v2 v1 v0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
111 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
112 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
113 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
114 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
115 "psubw _MMX_0080w,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
116
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
117 // create Cr_g (result in mm0)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
118 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
119 "pmullw _MMX_VgrnRGB,%%mm0\n"// red*-46dec=0.7136*64
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
120 "pmullw _MMX_VredRGB,%%mm1\n"// red*89dec=1.4013*64
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
121 "psraw $6, %%mm0\n" // red=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
122 "psraw $6, %%mm1\n" // red=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
123
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
124 // create L1 L2 (result in mm2,mm4)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
125 // L2=lum+cols
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
126 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
127 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
128 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
129 "pand _MMX_FF00w,%%mm2\n" // L3 0 L1 0 l3 0 l1 0
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
130 "pand _MMX_00FFw,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
131 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
132
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
133 // create R (result in mm6)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
134 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
135 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
136 "paddsw %%mm1, %%mm5\n" // lum1+red:x R3 x R1 x r3 x r1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
137 "paddsw %%mm1, %%mm6\n" // lum1+red:x R2 x R0 x r2 x r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
138 "packuswb %%mm5,%%mm5\n" // R3 R1 r3 r1 R3 R1 r3 r1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
139 "packuswb %%mm6,%%mm6\n" // R2 R0 r2 r0 R2 R0 r2 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
140 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
141 "punpcklbw %%mm5,%%mm6\n" // R3 R2 R1 R0 r3 r2 r1 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
142
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
143 // create Cb (result in mm1)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
144 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
145 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
146 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
147 "psubw _MMX_0080w,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
148 // create Cb_g (result in mm5)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
149 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
150 "pmullw _MMX_UgrnRGB,%%mm5\n" // blue*-109dec=1.7129*64
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
151 "pmullw _MMX_UbluRGB,%%mm1\n" // blue*114dec=1.78125*64
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
152 "psraw $6, %%mm5\n" // blue=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
153 "psraw $6, %%mm1\n" // blue=blue/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
154
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
155 // create G (result in mm7)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
156 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
157 "movq %%mm4,%%mm7\n" // 0 L2 0 L0 0 l2 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
158 "paddsw %%mm5, %%mm3\n" // lum1+Cb_g:x G3t x G1t x g3t x g1t
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
159 "paddsw %%mm5, %%mm7\n" // lum1+Cb_g:x G2t x G0t x g2t x g0t
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
160 "paddsw %%mm0, %%mm3\n" // lum1+Cr_g:x G3 x G1 x g3 x g1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
161 "paddsw %%mm0, %%mm7\n" // lum1+blue:x G2 x G0 x g2 x g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
162 "packuswb %%mm3,%%mm3\n" // G3 G1 g3 g1 G3 G1 g3 g1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
163 "packuswb %%mm7,%%mm7\n" // G2 G0 g2 g0 G2 G0 g2 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
164 "punpcklbw %%mm3,%%mm7\n" // G3 G2 G1 G0 g3 g2 g1 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
165
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
166 // create B (result in mm5)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
167 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
168 "movq %%mm4,%%mm5\n" // 0 L2 0 L0 0 l2 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
169 "paddsw %%mm1, %%mm3\n" // lum1+blue:x B3 x B1 x b3 x b1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
170 "paddsw %%mm1, %%mm5\n" // lum1+blue:x B2 x B0 x b2 x b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
171 "packuswb %%mm3,%%mm3\n" // B3 B1 b3 b1 B3 B1 b3 b1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
172 "packuswb %%mm5,%%mm5\n" // B2 B0 b2 b0 B2 B0 b2 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
173 "punpcklbw %%mm3,%%mm5\n" // B3 B2 B1 B0 b3 b2 b1 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
174
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
175 // fill destination row1 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
176
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
177 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
178 "pxor %%mm4,%%mm4\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
179 "movq %%mm6,%%mm1\n" // R3 R2 R1 R0 r3 r2 r1 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
180 "movq %%mm5,%%mm3\n" // B3 B2 B1 B0 b3 b2 b1 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
181 // process lower lum
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
182 "punpcklbw %%mm4,%%mm1\n" // 0 r3 0 r2 0 r1 0 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
183 "punpcklbw %%mm4,%%mm3\n" // 0 b3 0 b2 0 b1 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
184 "movq %%mm1,%%mm2\n" // 0 r3 0 r2 0 r1 0 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
185 "movq %%mm3,%%mm0\n" // 0 b3 0 b2 0 b1 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
186 "punpcklwd %%mm1,%%mm3\n" // 0 r1 0 b1 0 r0 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
187 "punpckhwd %%mm2,%%mm0\n" // 0 r3 0 b3 0 r2 0 b2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
188
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
189 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
190 "movq %%mm7,%%mm1\n" // G3 G2 G1 G0 g3 g2 g1 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
191 "punpcklbw %%mm1,%%mm2\n" // g3 0 g2 0 g1 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
192 "punpcklwd %%mm4,%%mm2\n" // 0 0 g1 0 0 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
193 "por %%mm3, %%mm2\n" // 0 r1 g1 b1 0 r0 g0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
194 "movq %%mm2,(%3)\n" // wrote out ! row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
195
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
196 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
197 "punpcklbw %%mm1,%%mm4\n" // g3 0 g2 0 g1 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
198 "punpckhwd %%mm2,%%mm4\n" // 0 0 g3 0 0 0 g2 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
199 "por %%mm0, %%mm4\n" // 0 r3 g3 b3 0 r2 g2 b2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
200 "movq %%mm4,8(%3)\n" // wrote out ! row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
201
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
202 // fill destination row2 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
203 // this can be done "destructive"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
204 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
205 "punpckhbw %%mm2,%%mm6\n" // 0 R3 0 R2 0 R1 0 R0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
206 "punpckhbw %%mm1,%%mm5\n" // G3 B3 G2 B2 G1 B1 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
207 "movq %%mm5,%%mm1\n" // G3 B3 G2 B2 G1 B1 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
208 "punpcklwd %%mm6,%%mm1\n" // 0 R1 G1 B1 0 R0 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
209 "movq %%mm1,(%5)\n" // wrote out ! row2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
210 "punpckhwd %%mm6,%%mm5\n" // 0 R3 G3 B3 0 R2 G2 B2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
211 "movq %%mm5,8(%5)\n" // wrote out ! row2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
212
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
213 "addl $4,%2\n" // lum+4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
214 "leal 16(%3),%3\n" // row1+16
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
215 "leal 16(%5),%5\n" // row2+16
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
216 "addl $2, %%ebx\n" // cr+2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
217 "addl $2, %1\n" // cb+2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
218
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
219 "addl $4,%6\n" // x+4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
220 "cmpl %4,%6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
221
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
222 "jl 1b\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
223 "addl %4, %2\n" // lum += cols
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
224 "addl %8, %3\n" // row1+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
225 "addl %8, %5\n" // row2+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
226 "movl $0, %6\n" // x=0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
227 "cmpl %7, %2\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
228 "jl 1b\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
229 "emms\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
230 "popl %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
231 :
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
232 : "m" (cr), "r"(cb),"r"(lum),
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
233 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod));
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
234 }
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
235
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
236 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
237 unsigned char *lum, unsigned char *cr,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
238 unsigned char *cb, unsigned char *out,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
239 int rows, int cols, int mod )
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
240 {
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
241 Uint16 *row1;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
242 Uint16 *row2;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
243
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
244 unsigned char* y = lum +cols*rows; /* Pointer to the end */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
245 int x=0;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
246 row1 = (Uint16 *)out; /* 16 bit target */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
247 row2 = (Uint16 *)out+cols+mod; /* start of second row */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
248 mod = (mod+cols+mod)*2; /* increment for row1 in byte */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
249
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
250
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
251 __asm__ __volatile__(
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
252 "pushl %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
253 "movl %0, %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
254
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
255 ".align 8\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
256 "1:\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
257 "movd (%1), %%mm0\n" // 4 Cb 0 0 0 0 u3 u2 u1 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
258 "pxor %%mm7, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
259 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
260 "punpcklbw %%mm7, %%mm0\n" // 4 W cb 0 u3 0 u2 0 u1 0 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
261 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
262 "psubw _MMX_0080w, %%mm0\n"
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
263 "psubw _MMX_0080w, %%mm1\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
264 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
265 "movq %%mm1, %%mm3\n" // Cr
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
266 "pmullw _MMX_Ugrn565, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
267 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
268 "pmullw _MMX_Ublu5x5, %%mm0\n" // Cb2blue
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
269 "pand _MMX_00FFw, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
270 "pmullw _MMX_Vgrn565, %%mm3\n" // Cr2green
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
271 "movq (%2), %%mm7\n" // L2
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
272 "pmullw _MMX_Vred5x5, %%mm1\n" // Cr2red
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
273 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
274 "pmullw _MMX_Ycoeff, %%mm6\n" // lum1
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
275 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
276 "pmullw _MMX_Ycoeff, %%mm7\n" // lum2
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
277
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
278 "movq %%mm6, %%mm4\n" // lum1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
279 "paddw %%mm0, %%mm6\n" // lum1 +blue 00 B6 00 B4 00 B2 00 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
280 "movq %%mm4, %%mm5\n" // lum1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
281 "paddw %%mm1, %%mm4\n" // lum1 +red 00 R6 00 R4 00 R2 00 R0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
282 "paddw %%mm2, %%mm5\n" // lum1 +green 00 G6 00 G4 00 G2 00 G0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
283 "psraw $6, %%mm4\n" // R1 0 .. 64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
284 "movq %%mm7, %%mm3\n" // lum2 00 L7 00 L5 00 L3 00 L1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
285 "psraw $6, %%mm5\n" // G1 - .. +
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
286 "paddw %%mm0, %%mm7\n" // Lum2 +blue 00 B7 00 B5 00 B3 00 B1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
287 "psraw $6, %%mm6\n" // B1 0 .. 64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
288 "packuswb %%mm4, %%mm4\n" // R1 R1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
289 "packuswb %%mm5, %%mm5\n" // G1 G1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
290 "packuswb %%mm6, %%mm6\n" // B1 B1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
291 "punpcklbw %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
292 "punpcklbw %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
293
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
294 "pand _MMX_red565, %%mm4\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
295 "psllw $3, %%mm5\n" // GREEN 1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
296 "punpcklbw %%mm6, %%mm6\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
297 "pand _MMX_grn565, %%mm5\n"
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
298 "pand _MMX_red565, %%mm6\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
299 "por %%mm5, %%mm4\n" //
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
300 "psrlw $11, %%mm6\n" // BLUE 1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
301 "movq %%mm3, %%mm5\n" // lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
302 "paddw %%mm1, %%mm3\n" // lum2 +red 00 R7 00 R5 00 R3 00 R1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
303 "paddw %%mm2, %%mm5\n" // lum2 +green 00 G7 00 G5 00 G3 00 G1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
304 "psraw $6, %%mm3\n" // R2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
305 "por %%mm6, %%mm4\n" // MM4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
306 "psraw $6, %%mm5\n" // G2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
307 "movq (%2, %4), %%mm6\n" // L3 load lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
308 "psraw $6, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
309 "packuswb %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
310 "packuswb %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
311 "packuswb %%mm7, %%mm7\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
312 "pand _MMX_00FFw, %%mm6\n" // L3
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
313 "punpcklbw %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
314 "punpcklbw %%mm5, %%mm5\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
315 "pmullw _MMX_Ycoeff, %%mm6\n" // lum3
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
316 "punpcklbw %%mm7, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
317 "psllw $3, %%mm5\n" // GREEN 2
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
318 "pand _MMX_red565, %%mm7\n"
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
319 "pand _MMX_red565, %%mm3\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
320 "psrlw $11, %%mm7\n" // BLUE 2
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
321 "pand _MMX_grn565, %%mm5\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
322 "por %%mm7, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
323 "movq (%2,%4), %%mm7\n" // L4 load lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
324 "por %%mm5, %%mm3\n" //
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
325 "psrlw $8, %%mm7\n" // L4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
326 "movq %%mm4, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
327 "punpcklwd %%mm3, %%mm4\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
328 "pmullw _MMX_Ycoeff, %%mm7\n" // lum4
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
329 "punpckhwd %%mm3, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
330
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
331 "movq %%mm4, (%3)\n" // write row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
332 "movq %%mm5, 8(%3)\n" // write row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
333
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
334 "movq %%mm6, %%mm4\n" // Lum3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
335 "paddw %%mm0, %%mm6\n" // Lum3 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
336
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
337 "movq %%mm4, %%mm5\n" // Lum3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
338 "paddw %%mm1, %%mm4\n" // Lum3 +red
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
339 "paddw %%mm2, %%mm5\n" // Lum3 +green
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
340 "psraw $6, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
341 "movq %%mm7, %%mm3\n" // Lum4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
342 "psraw $6, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
343 "paddw %%mm0, %%mm7\n" // Lum4 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
344 "psraw $6, %%mm6\n" // Lum3 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
345 "movq %%mm3, %%mm0\n" // Lum4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
346 "packuswb %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
347 "paddw %%mm1, %%mm3\n" // Lum4 +red
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
348 "packuswb %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
349 "paddw %%mm2, %%mm0\n" // Lum4 +green
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
350 "packuswb %%mm6, %%mm6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
351 "punpcklbw %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
352 "punpcklbw %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
353 "punpcklbw %%mm6, %%mm6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
354 "psllw $3, %%mm5\n" // GREEN 3
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
355 "pand _MMX_red565, %%mm4\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
356 "psraw $6, %%mm3\n" // psr 6
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
357 "psraw $6, %%mm0\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
358 "pand _MMX_red565, %%mm6\n" // BLUE
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
359 "pand _MMX_grn565, %%mm5\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
360 "psrlw $11, %%mm6\n" // BLUE 3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
361 "por %%mm5, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
362 "psraw $6, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
363 "por %%mm6, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
364 "packuswb %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
365 "packuswb %%mm0, %%mm0\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
366 "packuswb %%mm7, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
367 "punpcklbw %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
368 "punpcklbw %%mm0, %%mm0\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
369 "punpcklbw %%mm7, %%mm7\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
370 "pand _MMX_red565, %%mm3\n"
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
371 "pand _MMX_red565, %%mm7\n" // BLUE
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
372 "psllw $3, %%mm0\n" // GREEN 4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
373 "psrlw $11, %%mm7\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
374 "pand _MMX_grn565, %%mm0\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
375 "por %%mm7, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
376 "por %%mm0, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
377
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
378 "movq %%mm4, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
379
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
380 "punpcklwd %%mm3, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
381 "punpckhwd %%mm3, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
382
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
383 "movq %%mm4, (%5)\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
384 "movq %%mm5, 8(%5)\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
385
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
386 "addl $8, %6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
387 "addl $8, %2\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
388 "addl $4, %%ebx\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
389 "addl $4, %1\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
390 "cmpl %4, %6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
391 "leal 16(%3), %3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
392 "leal 16(%5),%5\n" // row2+16
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
393
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
394
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
395 "jl 1b\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
396 "addl %4, %2\n" // lum += cols
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
397 "addl %8, %3\n" // row1+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
398 "addl %8, %5\n" // row2+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
399 "movl $0, %6\n" // x=0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
400 "cmpl %7, %2\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
401 "jl 1b\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
402 "emms\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
403 "popl %%ebx\n"
1038
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
404 :
29d7db09776e Ugly hack to make this work with gcc 2.x and 3.x
Sam Lantinga <slouken@libsdl.org>
parents: 949
diff changeset
405 :"m" (cr), "r"(cb),"r"(lum),
1148
63fb2da89a4b Patched inline assembly to compile on gcc 4.0.1. Details are here:
Ryan C. Gordon <icculus@icculus.org>
parents: 1038
diff changeset
406 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod));
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
407 }
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
408
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
409 #endif /* GCC i386 inline assembly */
1413
Sam Lantinga <slouken@libsdl.org>
parents: 1407
diff changeset
410 #endif /* 0 */