annotate src/video/SDL_yuv_mmx.c @ 4045:f420bba13676 SDL-1.2

GCC inline asm for MMX YUV processing no longer has textrels and now works when gcc wants to hog %%ebx for the PIC register. Fixes Bugzilla #418.
author Ryan C. Gordon <icculus@icculus.org>
date Wed, 11 Jul 2007 06:26:22 +0000
parents 40edc79b0926
children 3a9e60224efe
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
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
24 #if (__GNUC__ > 2) && defined(__i386__) && SDL_ASSEMBLY_ROUTINES
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
25
1407
0c6941483cc6 Whoops, forgot to check in this fix
Sam Lantinga <slouken@libsdl.org>
parents: 1402
diff changeset
26 #include "SDL_stdinc.h"
0c6941483cc6 Whoops, forgot to check in this fix
Sam Lantinga <slouken@libsdl.org>
parents: 1402
diff changeset
27
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
28 #include "mmx.h"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
29
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
30 static mmx_t MMX_0080w = { .ud = {0x00800080, 0x00800080} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
31 static mmx_t MMX_00FFw = { .ud = {0x00ff00ff, 0x00ff00ff} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
32 static mmx_t MMX_FF00w = { .ud = {0xff00ff00, 0xff00ff00} };
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
33
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
34 static mmx_t MMX_Ycoeff = { .uw = {0x004a, 0x004a, 0x004a, 0x004a} };
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
35
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
36 static mmx_t MMX_UbluRGB = { .uw = {0x0072, 0x0072, 0x0072, 0x0072} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
37 static mmx_t MMX_VredRGB = { .uw = {0x0059, 0x0059, 0x0059, 0x0059} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
38 static mmx_t MMX_UgrnRGB = { .uw = {0xffea, 0xffea, 0xffea, 0xffea} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
39 static mmx_t MMX_VgrnRGB = { .uw = {0xffd2, 0xffd2, 0xffd2, 0xffd2} };
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
40
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
41 static mmx_t MMX_Ublu5x5 = { .uw = {0x0081, 0x0081, 0x0081, 0x0081} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
42 static mmx_t MMX_Vred5x5 = { .uw = {0x0066, 0x0066, 0x0066, 0x0066} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
43 static mmx_t MMX_Ugrn555 = { .uw = {0xffe7, 0xffe7, 0xffe7, 0xffe7} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
44 static mmx_t MMX_Vgrn555 = { .uw = {0xffcc, 0xffcc, 0xffcc, 0xffcc} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
45 static mmx_t MMX_Ugrn565 = { .uw = {0xffe8, 0xffe8, 0xffe8, 0xffe8} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
46 static mmx_t MMX_Vgrn565 = { .uw = {0xffcd, 0xffcd, 0xffcd, 0xffcd} };
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
47
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
48 static mmx_t MMX_red555 = { .uw = {0x7c00, 0x7c00, 0x7c00, 0x7c00} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
49 static mmx_t MMX_red565 = { .uw = {0xf800, 0xf800, 0xf800, 0xf800} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
50 static mmx_t MMX_grn555 = { .uw = {0x03e0, 0x03e0, 0x03e0, 0x03e0} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
51 static mmx_t MMX_grn565 = { .uw = {0x07e0, 0x07e0, 0x07e0, 0x07e0} };
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
52 static mmx_t MMX_blu5x5 = { .uw = {0x001f, 0x001f, 0x001f, 0x001f} };
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
53
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
54 /**
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
55 This MMX assembler is my first assembler/MMX program ever.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
56 Thus it maybe buggy.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
57 Send patches to:
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
58 mvogt@rhrk.uni-kl.de
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
59
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
60 After it worked fine I have "obfuscated" the code a bit to have
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
61 more parallism in the MMX units. This means I moved
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
62 initilisation around and delayed other instruction.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
63 Performance measurement did not show that this brought any advantage
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
64 but in theory it _should_ be faster this way.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
65
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
66 The overall performanve gain to the C based dither was 30%-40%.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
67 The MMX routine calculates 256bit=8RGB values in each cycle
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
68 (4 for row1 & 4 for row2)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
69
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
70 The red/green/blue.. coefficents are taken from the mpeg_play
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
71 player. They look nice, but I dont know if you can have
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
72 better values, to avoid integer rounding errors.
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
73
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
74
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
75 IMPORTANT:
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
76 ==========
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
77
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
78 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
79 the out are 16byte aligned or you will/may get segfaults
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
80
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 void ColorRGBDitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
84 unsigned char *lum, unsigned char *cr,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
85 unsigned char *cb, unsigned char *out,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
86 int rows, int cols, int mod )
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
87 {
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
88 Uint32 *row1;
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
89 Uint32 *row2;
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
90
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
91 unsigned char* y = lum +cols*rows; // Pointer to the end
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
92 int x = 0;
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
93 row1 = (Uint32 *)out; // 32 bit target
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
94 row2 = (Uint32 *)out+cols+mod; // start of second row
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
95 mod = (mod+cols+mod)*4; // increment for row1 in byte
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
96
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
97 __asm__ __volatile__ (
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
98 // tap dance to workaround the inability to use %%ebx at will...
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
99 // move one thing to the stack...
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
100 "pushl $0\n" // save a slot on the stack.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
101 "pushl %%ebx\n" // save %%ebx.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
102 "movl %0, %%ebx\n" // put the thing in ebx.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
103 "movl %%ebx, 4(%%esp)\n" // put the thing in the stack slot.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
104 "popl %%ebx\n" // get back %%ebx (the PIC register).
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
105
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
106 ".align 8\n"
0
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)
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
110 "pushl %%ebx\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
111 "movl 4(%%esp), %%ebx\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
112 "movd (%%ebx), %%mm1\n" // 0 0 0 0 v3 v2 v1 v0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
113 "popl %%ebx\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
114 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
115 "movd (%2), %%mm2\n" // 0 0 0 0 l3 l2 l1 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
116 "punpcklbw %%mm7,%%mm1\n" // 0 v3 0 v2 00 v1 00 v0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
117 "punpckldq %%mm1,%%mm1\n" // 00 v1 00 v0 00 v1 00 v0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
118 "psubw %9,%%mm1\n" // mm1-128:r1 r1 r0 r0 r1 r1 r0 r0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
119
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
120 // create Cr_g (result in mm0)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
121 "movq %%mm1,%%mm0\n" // r1 r1 r0 r0 r1 r1 r0 r0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
122 "pmullw %10,%%mm0\n" // red*-46dec=0.7136*64
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
123 "pmullw %11,%%mm1\n" // red*89dec=1.4013*64
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
124 "psraw $6, %%mm0\n" // red=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
125 "psraw $6, %%mm1\n" // red=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
126
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
127 // create L1 L2 (result in mm2,mm4)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
128 // L2=lum+cols
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
129 "movq (%2,%4),%%mm3\n" // 0 0 0 0 L3 L2 L1 L0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
130 "punpckldq %%mm3,%%mm2\n" // L3 L2 L1 L0 l3 l2 l1 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
131 "movq %%mm2,%%mm4\n" // L3 L2 L1 L0 l3 l2 l1 l0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
132 "pand %12,%%mm2\n" // L3 0 L1 0 l3 0 l1 0
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
133 "pand %13,%%mm4\n" // 0 L2 0 L0 0 l2 0 l0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
134 "psrlw $8,%%mm2\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
135
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
136 // create R (result in mm6)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
137 "movq %%mm2,%%mm5\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
138 "movq %%mm4,%%mm6\n" // 0 L2 0 L0 0 l2 0 l0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
139 "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
140 "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
141 "packuswb %%mm5,%%mm5\n" // R3 R1 r3 r1 R3 R1 r3 r1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
142 "packuswb %%mm6,%%mm6\n" // R2 R0 r2 r0 R2 R0 r2 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
143 "pxor %%mm7,%%mm7\n" // 00 00 00 00 00 00 00 00
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
144 "punpcklbw %%mm5,%%mm6\n" // R3 R2 R1 R0 r3 r2 r1 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
145
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
146 // create Cb (result in mm1)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
147 "movd (%1), %%mm1\n" // 0 0 0 0 u3 u2 u1 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
148 "punpcklbw %%mm7,%%mm1\n" // 0 u3 0 u2 00 u1 00 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
149 "punpckldq %%mm1,%%mm1\n" // 00 u1 00 u0 00 u1 00 u0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
150 "psubw %9,%%mm1\n" // mm1-128:u1 u1 u0 u0 u1 u1 u0 u0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
151 // create Cb_g (result in mm5)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
152 "movq %%mm1,%%mm5\n" // u1 u1 u0 u0 u1 u1 u0 u0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
153 "pmullw %14,%%mm5\n" // blue*-109dec=1.7129*64
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
154 "pmullw %15,%%mm1\n" // blue*114dec=1.78125*64
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
155 "psraw $6, %%mm5\n" // blue=red/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
156 "psraw $6, %%mm1\n" // blue=blue/64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
157
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
158 // create G (result in mm7)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
159 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
160 "movq %%mm4,%%mm7\n" // 0 L2 0 L0 0 l2 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
161 "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
162 "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
163 "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
164 "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
165 "packuswb %%mm3,%%mm3\n" // G3 G1 g3 g1 G3 G1 g3 g1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
166 "packuswb %%mm7,%%mm7\n" // G2 G0 g2 g0 G2 G0 g2 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
167 "punpcklbw %%mm3,%%mm7\n" // G3 G2 G1 G0 g3 g2 g1 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
168
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
169 // create B (result in mm5)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
170 "movq %%mm2,%%mm3\n" // 0 L3 0 L1 0 l3 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
171 "movq %%mm4,%%mm5\n" // 0 L2 0 L0 0 l2 0 l1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
172 "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
173 "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
174 "packuswb %%mm3,%%mm3\n" // B3 B1 b3 b1 B3 B1 b3 b1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
175 "packuswb %%mm5,%%mm5\n" // B2 B0 b2 b0 B2 B0 b2 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
176 "punpcklbw %%mm3,%%mm5\n" // B3 B2 B1 B0 b3 b2 b1 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
177
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
178 // fill destination row1 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
179
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
180 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
181 "pxor %%mm4,%%mm4\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
182 "movq %%mm6,%%mm1\n" // R3 R2 R1 R0 r3 r2 r1 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
183 "movq %%mm5,%%mm3\n" // B3 B2 B1 B0 b3 b2 b1 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
184 // process lower lum
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
185 "punpcklbw %%mm4,%%mm1\n" // 0 r3 0 r2 0 r1 0 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
186 "punpcklbw %%mm4,%%mm3\n" // 0 b3 0 b2 0 b1 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
187 "movq %%mm1,%%mm2\n" // 0 r3 0 r2 0 r1 0 r0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
188 "movq %%mm3,%%mm0\n" // 0 b3 0 b2 0 b1 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
189 "punpcklwd %%mm1,%%mm3\n" // 0 r1 0 b1 0 r0 0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
190 "punpckhwd %%mm2,%%mm0\n" // 0 r3 0 b3 0 r2 0 b2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
191
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
192 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
193 "movq %%mm7,%%mm1\n" // G3 G2 G1 G0 g3 g2 g1 g0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
194 "punpcklbw %%mm1,%%mm2\n" // g3 0 g2 0 g1 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
195 "punpcklwd %%mm4,%%mm2\n" // 0 0 g1 0 0 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
196 "por %%mm3, %%mm2\n" // 0 r1 g1 b1 0 r0 g0 b0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
197 "movq %%mm2,(%3)\n" // wrote out ! row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
198
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
199 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
200 "punpcklbw %%mm1,%%mm4\n" // g3 0 g2 0 g1 0 g0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
201 "punpckhwd %%mm2,%%mm4\n" // 0 0 g3 0 0 0 g2 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
202 "por %%mm0, %%mm4\n" // 0 r3 g3 b3 0 r2 g2 b2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
203 "movq %%mm4,8(%3)\n" // wrote out ! row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
204
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
205 // fill destination row2 (needed are mm6=Rr,mm7=Gg,mm5=Bb)
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
206 // this can be done "destructive"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
207 "pxor %%mm2,%%mm2\n" // 0 0 0 0 0 0 0 0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
208 "punpckhbw %%mm2,%%mm6\n" // 0 R3 0 R2 0 R1 0 R0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
209 "punpckhbw %%mm1,%%mm5\n" // G3 B3 G2 B2 G1 B1 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
210 "movq %%mm5,%%mm1\n" // G3 B3 G2 B2 G1 B1 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
211 "punpcklwd %%mm6,%%mm1\n" // 0 R1 G1 B1 0 R0 G0 B0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
212 "movq %%mm1,(%5)\n" // wrote out ! row2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
213 "punpckhwd %%mm6,%%mm5\n" // 0 R3 G3 B3 0 R2 G2 B2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
214 "movq %%mm5,8(%5)\n" // wrote out ! row2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
215
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
216 "addl $4,%2\n" // lum+4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
217 "leal 16(%3),%3\n" // row1+16
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
218 "leal 16(%5),%5\n" // row2+16
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
219 "addl $2, (%%esp)\n" // cr+2
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
220 "addl $2, %1\n" // cb+2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
221
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
222 "addl $4,%6\n" // x+4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
223 "cmpl %4,%6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
224
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
225 "jl 1b\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
226 "addl %4, %2\n" // lum += cols
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
227 "addl %8, %3\n" // row1+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
228 "addl %8, %5\n" // row2+= mod
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
229 "movl $0, %6\n" // x=0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
230 "cmpl %7, %2\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
231 "jl 1b\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
232
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
233 "addl $4, %%esp\n" // get rid of the stack slot we reserved.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
234 "emms\n" // reset MMX registers.
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
235 :
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
236 : "m" (cr), "r"(cb),"r"(lum),
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
237 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
238 "m"(MMX_0080w),"m"(MMX_VgrnRGB),"m"(MMX_VredRGB),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
239 "m"(MMX_FF00w),"m"(MMX_00FFw),"m"(MMX_UgrnRGB),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
240 "m"(MMX_UbluRGB)
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
241 );
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
242 }
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
243
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
244 void Color565DitherYV12MMX1X( int *colortab, Uint32 *rgb_2_pix,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
245 unsigned char *lum, unsigned char *cr,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
246 unsigned char *cb, unsigned char *out,
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
247 int rows, int cols, int mod )
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
248 {
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
249 Uint16 *row1;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
250 Uint16 *row2;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
251
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
252 unsigned char* y = lum +cols*rows; /* Pointer to the end */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
253 int x=0;
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
254 row1 = (Uint16 *)out; /* 16 bit target */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
255 row2 = (Uint16 *)out+cols+mod; /* start of second row */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
256 mod = (mod+cols+mod)*2; /* increment for row1 in byte */
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
257
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
258
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
259 __asm__ __volatile__(
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
260 // tap dance to workaround the inability to use %%ebx at will...
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
261 // move one thing to the stack...
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
262 "pushl $0\n" // save a slot on the stack.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
263 "pushl %%ebx\n" // save %%ebx.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
264 "movl %0, %%ebx\n" // put the thing in ebx.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
265 "movl %%ebx, 4(%%esp)\n" // put the thing in the stack slot.
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
266 "popl %%ebx\n" // get back %%ebx (the PIC register).
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
267
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
268 ".align 8\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
269 "1:\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
270 "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
271 "pxor %%mm7, %%mm7\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
272 "pushl %%ebx\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
273 "movl 4(%%esp), %%ebx\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
274 "movd (%%ebx), %%mm1\n" // 4 Cr 0 0 0 0 v3 v2 v1 v0
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
275 "popl %%ebx\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
276
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
277 "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
278 "punpcklbw %%mm7, %%mm1\n" // 4 W cr 0 v3 0 v2 0 v1 0 v0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
279 "psubw %9, %%mm0\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
280 "psubw %9, %%mm1\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
281 "movq %%mm0, %%mm2\n" // Cb 0 u3 0 u2 0 u1 0 u0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
282 "movq %%mm1, %%mm3\n" // Cr
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
283 "pmullw %10, %%mm2\n" // Cb2green 0 R3 0 R2 0 R1 0 R0
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
284 "movq (%2), %%mm6\n" // L1 l7 L6 L5 L4 L3 L2 L1 L0
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
285 "pmullw %11, %%mm0\n" // Cb2blue
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
286 "pand %12, %%mm6\n" // L1 00 L6 00 L4 00 L2 00 L0
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
287 "pmullw %13, %%mm3\n" // Cr2green
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
288 "movq (%2), %%mm7\n" // L2
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
289 "pmullw %14, %%mm1\n" // Cr2red
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
290 "psrlw $8, %%mm7\n" // L2 00 L7 00 L5 00 L3 00 L1
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
291 "pmullw %15, %%mm6\n" // lum1
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
292 "paddw %%mm3, %%mm2\n" // Cb2green + Cr2green == green
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
293 "pmullw %15, %%mm7\n" // lum2
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
294
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
295 "movq %%mm6, %%mm4\n" // lum1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
296 "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
297 "movq %%mm4, %%mm5\n" // lum1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
298 "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
299 "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
300 "psraw $6, %%mm4\n" // R1 0 .. 64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
301 "movq %%mm7, %%mm3\n" // lum2 00 L7 00 L5 00 L3 00 L1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
302 "psraw $6, %%mm5\n" // G1 - .. +
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
303 "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
304 "psraw $6, %%mm6\n" // B1 0 .. 64
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
305 "packuswb %%mm4, %%mm4\n" // R1 R1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
306 "packuswb %%mm5, %%mm5\n" // G1 G1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
307 "packuswb %%mm6, %%mm6\n" // B1 B1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
308 "punpcklbw %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
309 "punpcklbw %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
310
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
311 "pand %16, %%mm4\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
312 "psllw $3, %%mm5\n" // GREEN 1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
313 "punpcklbw %%mm6, %%mm6\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
314 "pand %17, %%mm5\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
315 "pand %16, %%mm6\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
316 "por %%mm5, %%mm4\n" //
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
317 "psrlw $11, %%mm6\n" // BLUE 1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
318 "movq %%mm3, %%mm5\n" // lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
319 "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
320 "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
321 "psraw $6, %%mm3\n" // R2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
322 "por %%mm6, %%mm4\n" // MM4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
323 "psraw $6, %%mm5\n" // G2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
324 "movq (%2, %4), %%mm6\n" // L3 load lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
325 "psraw $6, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
326 "packuswb %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
327 "packuswb %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
328 "packuswb %%mm7, %%mm7\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
329 "pand %12, %%mm6\n" // L3
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
330 "punpcklbw %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
331 "punpcklbw %%mm5, %%mm5\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
332 "pmullw %15, %%mm6\n" // lum3
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
333 "punpcklbw %%mm7, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
334 "psllw $3, %%mm5\n" // GREEN 2
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
335 "pand %16, %%mm7\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
336 "pand %16, %%mm3\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
337 "psrlw $11, %%mm7\n" // BLUE 2
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
338 "pand %17, %%mm5\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
339 "por %%mm7, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
340 "movq (%2,%4), %%mm7\n" // L4 load lum2
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
341 "por %%mm5, %%mm3\n" //
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
342 "psrlw $8, %%mm7\n" // L4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
343 "movq %%mm4, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
344 "punpcklwd %%mm3, %%mm4\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
345 "pmullw %15, %%mm7\n" // lum4
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
346 "punpckhwd %%mm3, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
347
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
348 "movq %%mm4, (%3)\n" // write row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
349 "movq %%mm5, 8(%3)\n" // write row1
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
350
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
351 "movq %%mm6, %%mm4\n" // Lum3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
352 "paddw %%mm0, %%mm6\n" // Lum3 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
353
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
354 "movq %%mm4, %%mm5\n" // Lum3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
355 "paddw %%mm1, %%mm4\n" // Lum3 +red
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
356 "paddw %%mm2, %%mm5\n" // Lum3 +green
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
357 "psraw $6, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
358 "movq %%mm7, %%mm3\n" // Lum4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
359 "psraw $6, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
360 "paddw %%mm0, %%mm7\n" // Lum4 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
361 "psraw $6, %%mm6\n" // Lum3 +blue
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
362 "movq %%mm3, %%mm0\n" // Lum4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
363 "packuswb %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
364 "paddw %%mm1, %%mm3\n" // Lum4 +red
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
365 "packuswb %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
366 "paddw %%mm2, %%mm0\n" // Lum4 +green
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
367 "packuswb %%mm6, %%mm6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
368 "punpcklbw %%mm4, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
369 "punpcklbw %%mm5, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
370 "punpcklbw %%mm6, %%mm6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
371 "psllw $3, %%mm5\n" // GREEN 3
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
372 "pand %16, %%mm4\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
373 "psraw $6, %%mm3\n" // psr 6
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
374 "psraw $6, %%mm0\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
375 "pand %16, %%mm6\n" // BLUE
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
376 "pand %17, %%mm5\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
377 "psrlw $11, %%mm6\n" // BLUE 3
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
378 "por %%mm5, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
379 "psraw $6, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
380 "por %%mm6, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
381 "packuswb %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
382 "packuswb %%mm0, %%mm0\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
383 "packuswb %%mm7, %%mm7\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
384 "punpcklbw %%mm3, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
385 "punpcklbw %%mm0, %%mm0\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
386 "punpcklbw %%mm7, %%mm7\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
387 "pand %16, %%mm3\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
388 "pand %16, %%mm7\n" // BLUE
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
389 "psllw $3, %%mm0\n" // GREEN 4
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
390 "psrlw $11, %%mm7\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
391 "pand %17, %%mm0\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
392 "por %%mm7, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
393 "por %%mm0, %%mm3\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
394
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
395 "movq %%mm4, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
396
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
397 "punpcklwd %%mm3, %%mm4\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
398 "punpckhwd %%mm3, %%mm5\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
399
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
400 "movq %%mm4, (%5)\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
401 "movq %%mm5, 8(%5)\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
402
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
403 "addl $8, %6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
404 "addl $8, %2\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
405 "addl $4, (%%esp)\n"
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
406 "addl $4, %1\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
407 "cmpl %4, %6\n"
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
408 "leal 16(%3), %3\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
409 "leal 16(%5),%5\n" // row2+16
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
410
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
411
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
412 "jl 1b\n"
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
413 "addl %4, %2\n" // lum += cols
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
414 "addl %8, %3\n" // row1+= mod
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
415 "addl %8, %5\n" // row2+= mod
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
416 "movl $0, %6\n" // x=0
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
417 "cmpl %7, %2\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
418 "jl 1b\n"
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
419 "addl $4, %%esp\n" // get rid of the stack slot we reserved.
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
420 "emms\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
421 :
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
422 : "m" (cr), "r"(cb),"r"(lum),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
423 "r"(row1),"r"(cols),"r"(row2),"m"(x),"m"(y),"m"(mod),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
424 "m"(MMX_0080w),"m"(MMX_Ugrn565),"m"(MMX_Ublu5x5),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
425 "m"(MMX_00FFw),"m"(MMX_Vgrn565),"m"(MMX_Vred5x5),
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
426 "m"(MMX_Ycoeff),"m"(MMX_red565),"m"(MMX_grn565));
0
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
427 }
74212992fb08 Initial revision
Sam Lantinga <slouken@lokigames.com>
parents:
diff changeset
428
4045
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
429 #endif /* GCC3 i386 inline assembly */
f420bba13676 GCC inline asm for MMX YUV processing no longer has textrels and now works when
Ryan C. Gordon <icculus@icculus.org>
parents: 1413
diff changeset
430