1
0
mirror of https://github.com/gryf/wmaker.git synced 2025-12-27 17:02:33 +01:00

Remove assembly/optimized code

Signed-off-by: Carlos R. Mafra <crmafra@gmail.com>
This commit is contained in:
Carlos R. Mafra
2010-09-23 14:32:12 +02:00
parent ce9b375b8d
commit cf2301a048
9 changed files with 3 additions and 768 deletions

View File

@@ -20,8 +20,6 @@ libExtraWINGs_la_SOURCES = \
wtableview.h \ wtableview.h \
wtabledelegates.h wtabledelegates.h
AM_CFLAGS = @NOSTRICTALIASING@
INCLUDES = -I$(top_srcdir)/wrlib -I$(top_srcdir)/WINGs \ INCLUDES = -I$(top_srcdir)/wrlib -I$(top_srcdir)/WINGs \
-DRESOURCE_PATH=\"$(datadir)/WINGs\" @HEADER_SEARCH_PATH@ -DDEBUG -DRESOURCE_PATH=\"$(datadir)/WINGs\" @HEADER_SEARCH_PATH@ -DDEBUG

View File

@@ -83,8 +83,7 @@ libWUtil_la_SOURCES = \
AM_CPPFLAGS = -DLOCALEDIR=\"$(NLSDIR)\" -DRESOURCE_PATH=\"$(datadir)/WINGs\" -DDEBUG AM_CPPFLAGS = -DLOCALEDIR=\"$(NLSDIR)\" -DRESOURCE_PATH=\"$(datadir)/WINGs\" -DDEBUG
AM_CFLAGS =
AM_CFLAGS = @NOSTRICTALIASING@
INCLUDES = -I$(top_srcdir)/WINGs/WINGs -I$(top_srcdir)/wrlib -I$(top_srcdir)/src \ INCLUDES = -I$(top_srcdir)/WINGs/WINGs -I$(top_srcdir)/wrlib -I$(top_srcdir)/src \
@XFTFLAGS@ @HEADER_SEARCH_PATH@ @XFTFLAGS@ @HEADER_SEARCH_PATH@

View File

@@ -43,8 +43,7 @@ WPrefs_SOURCES = \
xmodifier.c xmodifier.c
AM_CPPFLAGS = -DLOCALEDIR=\"$(NLSDIR)\" -DRESOURCE_PATH=\"$(wpdatadir)\" AM_CPPFLAGS = -DLOCALEDIR=\"$(NLSDIR)\" -DRESOURCE_PATH=\"$(wpdatadir)\"
AM_CFLAGS =
AM_CFLAGS = @NOSTRICTALIASING@
INCLUDES = -I$(top_srcdir)/wrlib -I$(top_srcdir)/WINGs @HEADER_SEARCH_PATH@ INCLUDES = -I$(top_srcdir)/wrlib -I$(top_srcdir)/WINGs @HEADER_SEARCH_PATH@

View File

@@ -218,60 +218,6 @@ AC_C_CONST
AC_TYPE_SIGNAL AC_TYPE_SIGNAL
dnl Compiler/architecture specific optimizations
dnl ============================================
dnl GCC/as with MMX support
dnl -----------------------
# until we fix it, leave it disabled
asm_support=no
mmx_support=no
check_for_mmx_support=yes
AC_ARG_ENABLE(mmx, AS_HELP_STRING([--disable-mmx], [disable compilation of MMX inline assembly]),
[if test x$enableval != xyes; then
check_for_mmx_support=no
fi])
if test "$ac_cv_prog_gcc" = yes -a "$check_for_mmx_support" = yes; then
case $host_cpu in
*i?86*)
# gcc-3.3 or newer complains about some of our stuff without this
NOSTRICTALIASING="-fno-strict-aliasing"
AC_CACHE_CHECK(whether gcc supports x86 inline asm,
ac_cv_c_inline_asm,
[AC_TRY_LINK(,[{int x; asm volatile("movl %%eax, %%ebx\n\t pushal\n\t popal"::
"m" (x),"m" (x),"m" (x),"m" (x),"m" (x),"m" (x),
"m" (x),"m" (x),"m" (x),"m" (x),"m" (x),"m" (x));}],
ac_cv_c_inline_asm=yes,
ac_cv_c_inline_asm=no)])
if test "x$ac_cv_c_inline_asm" = xyes; then
AC_DEFINE(ASM_X86, 1, [define if processor is x86 (normally detected by configure)])
asm_support=yes
AC_CACHE_CHECK(whether gcc supports MMX(tm) inline asm,
ac_cv_c_inline_mmx,
[AC_TRY_LINK(,[asm ("movq %mm0, %mm1");],
ac_cv_c_inline_mmx=yes,
ac_cv_c_inline_mmx=no)])
if test "x$ac_cv_c_inline_mmx" = xyes; then
AC_DEFINE(ASM_X86_MMX, 1, [define if processor is x86 with MMX(tm) support (normally autodetected by configure)])
mmx_support=yes
fi
fi
;;
esac
fi
AC_SUBST(NOSTRICTALIASING)
dnl pkg-config dnl pkg-config
dnl ========== dnl ==========
dnl AC_ARG_VAR(PKGCONFIG, [pkg-config command]) dnl AC_ARG_VAR(PKGCONFIG, [pkg-config command])
@@ -1045,8 +991,6 @@ echo "Installation path prefix : $prefix"
echo "Installation path for binaries : $_bindir" echo "Installation path for binaries : $_bindir"
echo "Installation path for WPrefs.app : $wprefs_base_dir" | sed -e 's|\${prefix}|'"$prefix|" echo "Installation path for WPrefs.app : $wprefs_base_dir" | sed -e 's|\${prefix}|'"$prefix|"
echo "Supported graphic format libraries : $supported_gfx" echo "Supported graphic format libraries : $supported_gfx"
echo "Use assembly routines for wrlib : $asm_support"
echo "Use inline MMX(tm) x86 assembly : $mmx_support"
echo "Antialiased text support in WINGs : $xft" echo "Antialiased text support in WINGs : $xft"
echo "Xinerama extension support : $xinerama" echo "Xinerama extension support : $xinerama"
echo "XRandR extension support : $xrandr" echo "XRandR extension support : $xrandr"

View File

@@ -104,9 +104,7 @@ wmaker_SOURCES = \
EXTRA_wmaker_SOURCES = osdep_bsd.c osdep_darwin.c osdep_linux.c osdep_stub.c EXTRA_wmaker_SOURCES = osdep_bsd.c osdep_darwin.c osdep_linux.c osdep_stub.c
AM_CPPFLAGS = $(DFLAGS) -DLOCALEDIR=\"$(NLSDIR)\" AM_CPPFLAGS = $(DFLAGS) -DLOCALEDIR=\"$(NLSDIR)\"
AM_CFLAGS =
AM_CFLAGS = @NOSTRICTALIASING@
INCLUDES = \ INCLUDES = \
-I$(top_srcdir)/wrlib \ -I$(top_srcdir)/wrlib \

View File

@@ -28,7 +28,6 @@ libwraster_la_SOURCES = \
xpixmap.c \ xpixmap.c \
bench.h \ bench.h \
convert.c \ convert.c \
x86_specific.c \
context.c \ context.c \
misc.c \ misc.c \
scale.c \ scale.c \
@@ -47,14 +46,6 @@ libwraster_la_SOURCES = \
LTCOMPILE2=`echo $(LTCOMPILE) | sed -e s/-fomit-frame-pointer//` LTCOMPILE2=`echo $(LTCOMPILE) | sed -e s/-fomit-frame-pointer//`
COMPILE2=`echo $(COMPILE) | sed -e s/-fomit-frame-pointer//` COMPILE2=`echo $(COMPILE) | sed -e s/-fomit-frame-pointer//`
# cant compile asm stuff with optimizations
x86_specific.lo: x86_specific.c
$(LTCOMPILE2) -O0 -c $<
x86_specific.o: x86_specific.c
$(COMPILE2) -O0 -c $<
INCLUDES = $(DFLAGS) @HEADER_SEARCH_PATH@ INCLUDES = $(DFLAGS) @HEADER_SEARCH_PATH@
libwraster_la_LIBADD = @LIBRARY_SEARCH_PATH@ @GFXLIBS@ @XLIBS@ -lm libwraster_la_LIBADD = @LIBRARY_SEARCH_PATH@ @GFXLIBS@ @XLIBS@ -lm

View File

@@ -25,49 +25,19 @@
*/ */
#include <config.h> #include <config.h>
#include <X11/Xlib.h> #include <X11/Xlib.h>
#include <X11/Xutil.h> #include <X11/Xutil.h>
#include <stdlib.h> #include <stdlib.h>
#include <stdio.h> #include <stdio.h>
#include <string.h> #include <string.h>
#include <assert.h> #include <assert.h>
#include "wraster.h" #include "wraster.h"
#ifdef XSHM #ifdef XSHM
extern Pixmap R_CreateXImageMappedPixmap(RContext * context, RXImage * ximage); extern Pixmap R_CreateXImageMappedPixmap(RContext * context, RXImage * ximage);
#endif #endif
#ifdef ASM_X86
extern void x86_PseudoColor_32_to_8(unsigned char *image,
unsigned char *ximage,
char *err, char *nerr,
short *ctable,
int dr, int dg, int db,
unsigned long *pixels,
int cpc, int width, int height, int bytesPerPixel, int line_offset);
#endif /* ASM_X86 */
#ifdef ASM_X86_MMX
extern int x86_check_mmx();
extern void x86_mmx_TrueColor_32_to_16(unsigned char *image,
unsigned short *ximage,
short *err, short *nerr,
const unsigned short *rtable,
const unsigned short *gtable,
const unsigned short *btable,
int dr, int dg, int db,
unsigned int roffs,
unsigned int goffs,
unsigned int boffs, int width, int height, int line_offset);
#endif /* ASM_X86_MMX */
#define NFREE(n) if (n) free(n) #define NFREE(n) if (n) free(n)
#define HAS_ALPHA(I) ((I)->format == RRGBAFormat) #define HAS_ALPHA(I) ((I)->format == RRGBAFormat)
@@ -360,36 +330,6 @@ static RXImage *image2TrueColor(RContext * ctx, RImage * image)
fputs("true color dither\n", stderr); fputs("true color dither\n", stderr);
#endif #endif
#ifdef ASM_X86_MMX
if (ctx->depth == 16 && HAS_ALPHA(image) && x86_check_mmx()) {
short *err;
short *nerr;
err = malloc(8 * (image->width + 3));
nerr = malloc(8 * (image->width + 3));
if (!err || !nerr) {
NFREE(err);
NFREE(nerr);
RErrorCode = RERR_NOMEMORY;
RDestroyXImage(ctx, ximg);
return NULL;
}
memset(err, 0, 8 * (image->width + 3));
memset(nerr, 0, 8 * (image->width + 3));
x86_mmx_TrueColor_32_to_16(image->data,
(unsigned short *)ximg->image->data,
err + 8, nerr + 8,
rtable, gtable, btable,
dr, dg, db,
roffs, goffs, boffs,
image->width, image->height,
ximg->image->bytes_per_line - 2 * image->width);
free(err);
free(nerr);
} else
#endif /* ASM_X86_MMX */
{ {
signed char *err; signed char *err;
signed char *nerr; signed char *nerr;
@@ -575,7 +515,6 @@ static RXImage *image2PseudoColor(RContext * ctx, RImage * image)
memset(err, 0, 4 * (image->width + 3)); memset(err, 0, 4 * (image->width + 3));
memset(nerr, 0, 4 * (image->width + 3)); memset(nerr, 0, 4 * (image->width + 3));
/*#ifdef ASM_X86 */
convertPseudoColor_to_8(ximg, image, err + 4, nerr + 4, convertPseudoColor_to_8(ximg, image, err + 4, nerr + 4,
rtable, gtable, btable, dr, dg, db, ctx->pixels, cpc); rtable, gtable, btable, dr, dg, db, ctx->pixels, cpc);

View File

@@ -93,9 +93,5 @@ LIBWRASTER3
# RSaveXPM # RSaveXPM
# _wraster_change_filter # _wraster_change_filter
# WRasterLibVersion # WRasterLibVersion
# x86_check_mmx
# x86_mmx_TrueColor_24_to_16
# x86_mmx_TrueColor_32_to_16
# x86_PseudoColor_32_to_8
*; *;
}; };

View File

@@ -1,629 +0,0 @@
/* x86_convert.c - convert RImage to XImage with x86 optimizations
*
* Raster graphics library
*
* Copyright (c) 2000-2003 Alfredo K. Kojima
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Library General Public
* License as published by the Free Software Foundation; either
* version 2 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Library General Public License for more details.
*
* You should have received a copy of the GNU Library General Public
* License along with this library; if not, write to the Free
* Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#include <config.h>
#ifdef ASM_X86
#ifdef ASM_X86_MMX
int x86_check_mmx()
{
static int result = -1;
if (result >= 0)
return result;
result = 0;
asm volatile (
"pushal \n\t" /* please dont forget this in any asm */
"pushfl \n\t" /* check whether cpuid supported */
"pop %%eax \n\t"
"movl %%eax, %%ebx \n\t"
"xorl $(1<<21), %%eax \n\t"
"pushl %%eax \n\t"
"popfl \n\t"
"pushfl \n\t"
"popl %%eax \n\t"
"xorl %%ebx, %%eax \n\t"
"andl $(1<<21), %%eax \n\t"
"jz .NotPentium \n\t"
"xorl %%eax, %%eax \n\t" /* no eax effect because of the movl below */
/* except reseting flags. is it needed? */
"movl $1, %%eax \n\t"
"cpuid \n\t"
"test $(1<<23), %%edx \n\t"
"jz .NotMMX \n\t"
"popal \n\t" /* popal needed because the address of */
"movl $1, %0 \n\t" /* variable %0 may be kept in a register */
"jmp .noPop \n\t"
".NotMMX: \n\t"
".NotPentium: \n\t"
"popal \n\t"
".noPop: \n\t"
: "=m" (result)
);
return result;
}
/*
* TODO:
* 32/8 24/8 32/16 24/16 32/24 24/24
* PPlain YES YES
* MMX DONE
*
*
* - try to align stack (local variable space) into quadword boundary
*/
void
x86_mmx_TrueColor_32_to_16(unsigned char *image,
unsigned short *ximage,
short *err,
short *nerr,
unsigned short *rtable,
unsigned short *gtable,
unsigned short *btable,
int dr,
int dg,
int db,
unsigned int roffs,
unsigned int goffs, unsigned int boffs, int width, int height, int line_offset)
{
union {
long long rrggbbaa;
struct {
short int rr, gg, bb, aa;
} words;
} rrggbbaa;
union {
long long pixel;
struct {
short int rr, gg, bb, aa;
} words;
} pixel;
short *tmp_err;
short *tmp_nerr;
int x;
asm volatile (
"pushl %%ebx \n\t"
/* pack dr, dg and db into mm6 */
"movl %7, %%eax \n\t"
"movl %8, %%ebx \n\t"
"movl %9, %%ecx \n\t"
"movw %%ax, %16 \n\t"
"movw %%bx, %17 \n\t"
"movw %%cx, %18 \n\t"
"movw $0, %19 \n\t"
"movq %16, %%mm6 \n\t" /* dr dg db 0 */
/* pack 4|4|4|4 into mm7, for shifting (/16) */
"movl $0x00040004, %16 \n\t"
"movl $0x00040004, %18 \n\t"
"movq %16, %%mm7 \n\t"
/* store constant values for using with mmx when dithering */
"movl $0x00070007, %16 \n\t"
"movl $0x00070007, %18 \n\t"
"movq %16, %%mm5 \n\t"
"movl $0x00050005, %16 \n\t"
"movl $0x00050005, %18 \n\t"
"movq %16, %%mm4 \n\t"
"movl $0x00030003, %16 \n\t"
"movl $0x00030003, %18 \n\t"
"movq %16, %%mm3 \n\t"
/* process 1 pixel / cycle, each component treated as 16bit */
"movl %0, %%esi \n\t" /* esi = image->data */
".LoopYa: \n\t"
"movl %13, %%eax \n\t"
"movl %%eax, %26 \n\t" /* x = width */
"movl %14, %%eax \n\t"
"decl %%eax \n\t" /* y-- */
"movl %%eax, %14 \n\t"
"js .Enda \n\t" /* if y < 0, goto end */
"andl $1, %%eax \n\t"
"jz .LoopY_1a \n\t" /* if (y & 1) goto LoopY_1 */
".LoopY_0a: \n\t"
"movl %2, %%ebx \n\t" /* ebx = err */
"movl %%ebx, %25 \n\t" /* [-36] = err */
"movl %3, %%eax \n\t"
"movl %%eax, %24 \n\t" /* [-32] = nerr */
"jmp .LoopXa \n\t"
".LoopY_1a: \n\t"
"movl %3, %%ebx \n\t" /* ebx = nerr */
"movl %%ebx, %25 \n\t" /* [-36] = nerr */
"movl %2, %%eax \n\t"
"movl %%eax, %24 \n\t" /* [-32] = eerr */
".align 16 \n\t"
".LoopXa: \n\t"
/* calculate errors and pixel components; depend on ebx, esi, mm6 */
"movq (%%ebx), %%mm1 \n\t" /* mm1 = error[0..3] */
"punpcklbw (%%esi), %%mm0 \n\t" /* mm0 = image->data[0..3] */
"psrlw $8, %%mm0 \n\t" /* fixup mm0 */
"paddusb %%mm1, %%mm0 \n\t" /* mm0 = mm0 + mm1 (sat. to 255) */
"movq %%mm0, %20 \n\t" /* save the pixel */
"movzwl %20, %%ecx \n\t" /* ecx = pixel.red */
"movl %4, %%edi \n\t" /* edi = rtable */
/* agi */
"leal (%%edi, %%ecx, 2), %%eax \n\t" /* eax = &rtable[pixel.red] */
/* agi */
"movw (%%eax), %%dx \n\t" /* dx = rtable[pixel.red] */
"movw %%dx, %16 \n\t" /* save rr */
"movzwl %21, %%ecx \n\t" /* ecx = pixel.green */
"movl %5, %%edi \n\t" /* edi = gtable */
/* agi */
"leal (%%edi, %%ecx, 2), %%eax \n\t" /* eax = &gtable[pixel.green] */
/* agi */
"movw (%%eax), %%dx \n\t" /* dx = gtable[pixel.green] */
"movw %%dx, %17 \n\t" /* save gg */
"movzwl %22, %%ecx \n\t" /* ecx = pixel.blue */
"movl %6, %%edi \n\t" /* ebx = btable */
/* agi */
"leal (%%edi, %%ecx, 2), %%eax \n\t" /* eax = &btable[pixel.blue] */
/* agi */
"movw (%%eax), %%dx \n\t" /* dx = btable[pixel.blue] */
"movw %%dx, %18 \n\t" /* save bb */
"movw $0, %19 \n\t" /* save dummy aa */
"movq %16, %%mm1 \n\t" /* load mm1 with rrggbbaa */
"pmullw %%mm6, %%mm1 \n\t" /* mm1 = rr*dr|... */
"psubsw %%mm1, %%mm0 \n\t" /* error = pixel - mm1 */
/* distribute the error; depend on mm0, mm7, mm3, mm4, mm5 */
"movl %25, %%ebx \n\t"
"movq %%mm0, %%mm1 \n\t"
"pmullw %%mm5, %%mm1 \n\t" /* mm1 = mm1*7 */
"psrlw %%mm7, %%mm1 \n\t" /* mm1 = mm1/16 */
"paddw 8(%%ebx), %%mm1 \n\t"
"movq %%mm1, 8(%%ebx) \n\t" /* err[x+1,y] = rer*7/16 */
"movl %24, %%ebx \n\t"
"movq %%mm0, %%mm1 \n\t"
"pmullw %%mm4, %%mm1 \n\t" /* mm1 = mm1*5 */
"psrlw %%mm7, %%mm1 \n\t" /* mm1 = mm1/16 */
"paddw -8(%%ebx), %%mm1 \n\t"
"movq %%mm1, -8(%%ebx) \n\t" /* err[x-1,y+1] += rer*3/16 */
"movq %%mm0, %%mm1 \n\t"
"pmullw %%mm3, %%mm1 \n\t" /* mm1 = mm1*3 */
"psrlw %%mm7, %%mm1 \n\t" /* mm1 = mm1/16 */
"paddw 8(%%ebx), %%mm1 \n\t"
"movq %%mm1, (%%ebx) \n\t" /* err[x,y+1] += rer*5/16 */
"psrlw %%mm7, %%mm0 \n\t" /* mm0 = mm0/16 */
"movq %%mm0, 8(%%ebx) \n\t" /* err[x+1,y+1] = rer/16 */
/* calculate final pixel value and store */
"movl %10, %%ecx \n\t"
"movw %16, %%ax \n\t"
"shlw %%cl, %%ax \n\t" /* NP* ax = r<<roffs */
"movl %11, %%ecx \n\t"
"movw %17, %%bx \n\t"
"shlw %%cl, %%bx \n\t" /* NP* */
"orw %%bx, %%ax \n\t"
"movl %12, %%ecx \n\t"
"movw %18, %%bx \n\t"
"shlw %%cl, %%bx \n\t" /* NP* */
"orw %%bx, %%ax \n\t"
"movl %1, %%edx \n\t"
"movw %%ax, (%%edx) \n\t"
"addl $2, %%edx \n\t" /* increment ximage */
"movl %%edx, %1 \n\t"
/* prepare for next iteration on X */
"addl $8, %24 \n\t" /* nerr += 8 */
"movl %25, %%ebx \n\t"
"addl $8, %%ebx \n\t"
"movl %%ebx, %25 \n\t" /* ebx = err += 8 */
/* Note: in the last pixel, this would cause an invalid memory access
* because, punpcklbw is used (which reads 8 bytes) and the last
* pixel is only 4 bytes. This is no problem because the image data
* was allocated with extra 4 bytes when created. */
"addl $4, %%esi \n\t" /* image->data += 4 */
"decl %26 \n\t" /* x-- */
"jnz .LoopXa \n\t" /* if x>0, goto .LoopX */
/* depend on edx */
"addl %15, %%edx \n\t" /* add extra offset to ximage */
"movl %%edx, %1 \n\t"
"jmp .LoopYa \n\t"
".Enda: \n\t" /* THE END */
"emms \n\t"
"popl %%ebx \n\t"
:
: "m" (image), /* %0 */
"m" (ximage), /* %1 */
"m" (err), /* %2 */
"m" (nerr), /* %3 */
"m" (rtable), /* %4 */
"m" (gtable), /* %5 */
"m" (btable), /* %6 */
"m" (dr), /* %7 */
"m" (dg), /* %8 */
"m" (db), /* %9 */
"m" (roffs), /* %10 */
"m" (goffs), /* %11 */
"m" (boffs), /* %12 */
"m" (width), /* %13 */
"m" (height), /* %14 */
"m" (line_offset), /* %15 */
"m" (rrggbbaa.words.rr), /* %16 (access to rr) */
"m" (rrggbbaa.words.gg), /* %17 (access to gg) */
"m" (rrggbbaa.words.bb), /* %18 (access to bb) */
"m" (rrggbbaa.words.aa), /* %19 (access to aa) */
"m" (pixel.words.rr), /* %20 (access to pixel.r) */
"m" (pixel.words.gg), /* %21 (access to pixel.g) */
"m" (pixel.words.bb), /* %22 (access to pixel.b) */
"m" (pixel.words.aa), /* %23 (access to pixel.a) */
"m" (tmp_err), /* %24 */
"m" (tmp_nerr), /* %25 */
"m" (x) /* %26 */
: "eax",
"ecx",
"edx",
"esi",
"edi"
);
}
void
x86_mmx_TrueColor_24_to_16(unsigned char *image,
unsigned short *ximage,
short *err,
short *nerr,
short *rtable,
short *gtable,
short *btable,
int dr,
int dg,
int db,
unsigned int roffs,
unsigned int goffs, unsigned int boffs, int width, int height, int line_offset)
{
union {
long long rrggbbaa;
struct {
short int rr, gg, bb, aa;
} words;
} rrggbbaa;
union {
long long pixel;
struct {
short int rr, gg, bb, aa;
} words;
} pixel;
short *tmp_err;
short *tmp_nerr;
int x;
int w1;
int w2;
asm volatile (
"pushl %%ebx \n\t"
"movl %13, %%eax \n\t" /* eax = width */
"movl %%eax, %%ebx \n\t"
"shrl $2, %%eax \n\t"
"movl %%eax, %27 \n\t" /* w1 = width / 4 */
"andl $3, %%ebx \n\t"
"movl %%ebx, %28 \n\t" /* w2 = width %% 4 */
".LoopYc: \n\t"
"movl %13, %%eax \n\t"
"movl %%eax, %26 \n\t" /* x = width */
"decl %14 \n\t" /* height-- */
"js .Endc \n\t" /* if height < 0 then end */
"movl %14, %%eax \n\t"
"decl %%eax \n\t" /* y-- */
"movl %%eax, %14 \n\t"
"js .Endc \n\t" /* if y < 0, goto end */
"andl $1, %%eax \n\t"
"jz .LoopY_1c \n\t" /* if (y&1) goto LoopY_1 */
".LoopY_0c: \n\t"
"movl %2, %%ebx \n\t" /* ebx = err */
"movl %%ebx, %25 \n\t" /* [-36] = err */
"movl %3, %%eax \n\t"
"movl %%eax, %24 \n\t" /* [-32] = nerr */
"jmp .LoopX_1c \n\t"
".LoopY_1c: \n\t"
"movl %3, %%ebx \n\t" /* ebx = nerr */
"movl %%ebx, %25 \n\t" /* [-36] = nerr */
"movl %2, %%eax \n\t"
"movl %%eax, %24 \n\t" /* [-32] = eerr */
".align 16 \n\t"
"movl %%eax, %26 \n\t" /* x = w1 */
".LoopX_1c: \n\t"
"decl %26 \n\t" /* x-- */
"js .Xend1_c \n\t" /* if x < 0 then end */
/* do conversion of 4 pixels */
"movq %2, %%mm0 \n\t" /* mm0 = err */
"jmp .LoopX_1c \n\t"
".Xend1_c: \n\t"
"movl %28, %%eax \n\t"
"movl %%eax, %26 \n\t" /* x = w2 */
".LoopX_2c: \n\t"
"decl %26 \n\t" /* x-- */
"js .Xend2_c \n\t"
/* do conversion */
"jmp .LoopX_2c \n\t"
".Xend2_c: \n\t"
"movl %27, %%eax \n\t"
"jmp .LoopYc \n\t"
".Endc: \n\t" /* THE END */
"emms \n\t"
"popl %%ebx \n\t"
:
: "m" (image), /* %0 */
"m" (ximage), /* %1 */
"m" (err), /* %2 */
"m" (nerr), /* %3 */
"m" (rtable), /* %4 */
"m" (gtable), /* %5 */
"m" (btable), /* %6 */
"m" (dr), /* %7 */
"m" (dg), /* %8 */
"m" (db), /* %9 */
"m" (roffs), /* %10 */
"m" (goffs), /* %11 */
"m" (boffs), /* %12 */
"m" (width), /* %13 */
"m" (height), /* %14 */
"m" (line_offset), /* %15 */
"m" (rrggbbaa.words.rr), /* %16 (access to rr) */
"m" (rrggbbaa.words.gg), /* %17 (access to gg) */
"m" (rrggbbaa.words.bb), /* %18 (access to bb) */
"m" (rrggbbaa.words.aa), /* %19 (access to aa) */
"m" (pixel.words.rr), /* %20 (access to pixel.r) */
"m" (pixel.words.gg), /* %21 (access to pixel.g) */
"m" (pixel.words.bb), /* %22 (access to pixel.b) */
"m" (pixel.words.aa), /* %23 (access to pixel.a) */
"m" (tmp_err), /* %24 */
"m" (tmp_nerr), /* %25 */
"m" (x), /* %26 */
"m" (w1), /* %27 */
"m" (w2) /* %28 */
:
"eax",
"ecx",
"edx",
"esi",
"edi"
);
}
#endif /* ASM_X86_MMX */
void
x86_PseudoColor_32_to_8(unsigned char *image,
unsigned char *ximage,
char *err,
char *nerr,
short *ctable,
int dr,
int dg,
int db,
unsigned long *pixels, int cpc, int width, int height, int bytesPerPixel, int line_offset)
{
int x;
int cpcpc;
int rr;
int gg;
int bb;
char *tmp_err;
char *tmp_nerr;
char ndr; // aparently not used
char ndg; // aparently not used
char ndb; // aparently not used
asm volatile (
"pushal \n\t"
"movl %9, %%eax \n\t"
"mulb %9 \n\t"
"movl %%eax, %15 \n\t" /* cpcpc = cpc * cpc */
/* eax will always be <= 0xffff */
/* process 1 pixel / cycle, each component treated as 16bit */
"movl %0, %%esi \n\t" /* esi = image->data */
".LoopYb: \n\t"
"movl %10, %%ecx \n\t"
"movl %%ecx, %14 \n\t" /* x = width */
"movl %11, %%ecx \n\t"
"decl %%ecx \n\t" /* y-- */
"movl %%ecx, %11 \n\t"
"js .Endb \n\t" /* if y < 0, goto end */
"andl $1, %%ecx \n\t"
"jz .LoopY_1b \n\t" /* if (y & 1) goto LoopY_1 */
".LoopY_0b: \n\t"
"movl %2, %%ebx \n\t" /* ebx = err */
/* "movl %%ebx, %20 \n\t" */ /* [-36] = err */ /* useless */
"movl %3, %%ecx \n\t"
"movl %%ecx, %19 \n\t" /* [-32] = nerr */
"movl $0, (%%ecx) \n\t" /* init error of nerr[0] to 0 */
"jmp .LoopXb \n\t"
".LoopY_1b: \n\t"
"movl %3, %%ebx \n\t" /* ebx = nerr */
/* "movl %%ebx, %20 \n\t" */ /* [-36] = nerr */ /* useless */
"movl %2, %%ecx \n\t"
"movl %%ecx, %19 \n\t" /* [-32] = err */
"movl $0, (%%ecx) \n\t" /* init error of nerr[0] to 0 */
".align 16 \n\t"
".LoopXb: \n\t"
"movl %4, %%edi \n\t" /* edi = ctable */
"xorl %%edx, %%edx \n\t" /* zero the upper word on edx */
/* RED; depends on ebx==err, esi==image->data, edi */
"movzbw (%%esi), %%dx \n\t" /* dx = image->data[0] */
"movsbw (%%ebx), %%ax \n\t" /* ax = error[0] */
"addw %%ax, %%dx \n\t" /* pixel.red = data[0] + error[0] */
"testb %%dh, %%dh \n\t" /* test if pixel.red < 0 or > 255 */
"jz .OKRb \n\t" /* 0 <= pixel.red <= 255 */
"js .NEGRb \n\t" /* pixel.red < 0 */
"movw $0xff, %%dx \n\t" /* pixel.red > 255 */
"jmp .OKRb \n\t"
".NEGRb: \n\t"
"xorw %%dx, %%dx \n\t"
".OKRb: \n\t"
/* partial reg */
"leal (%%edi, %%edx, 2), %%ecx \n\t" /* ecx = &ctable[pixel.red] */
/* agi */
"movl (%%ecx), %%eax \n\t" /* ax = ctable[pixel.red] */
"movw %%ax, %16 \n\t" /* save rr */
"mulb %5 \n\t" /* ax = rr*dr */
"subw %%ax, %%dx \n\t" /* rer = dx = dx - rr*dr */
"movswl %%dx, %%eax \n\t" /* save rer */
/* distribute error */
"leal (, %%eax, 8), %%ecx \n\t"
"subw %%dx, %%cx \n\t" /* cx = rer * 7 */
"sarw $4, %%cx \n\t" /* cx = rer * 7 / 16 */
"addb %%cl, 4(%%ebx) \n\t" /* err[x+1] += rer * 7 / 16 */
"movl %19, %%ecx \n\t" /* ecx = nerr */
"leaw (%%eax, %%eax, 4), %%dx \n\t" /* dx = rer * 5 */
"sarw $4, %%dx \n\t" /* dx = rer * 5 / 16 */
"addb %%dl, (%%ecx) \n\t" /* nerr[x] += rer * 5 / 16 */
"leaw (%%eax, %%eax, 2), %%dx \n\t" /* dx = rer * 3 */
"sarw $4, %%dx \n\t" /* dx = rer * 3 / 16 */
"addb %%dl, -4(%%ecx) \n\t" /* nerr[x-1] += rer * 3 / 16 */
"sarw $4, %%ax \n\t" /* ax = rer / 16 */
"movb %%al, 4(%%ecx) \n\t" /* nerr[x+1] = rer / 16 */
/* GREEN; depends on ebx, esi, edi */
"movzbw 1(%%esi), %%dx \n\t" /* dx = image->data[1] */
"movsbw 1(%%ebx), %%ax \n\t" /* ax = error[1] */
"addw %%ax, %%dx \n\t" /* pixel.grn = data[1] + error[1] */
"testb %%dh, %%dh \n\t" /* test if pixel.grn < 0 or > 255 */
"jz .OKGb \n\t" /* 0 <= pixel.grn <= 255 */
"js .NEGGb \n\t" /* pixel.grn < 0 */
"movw $0xff, %%dx \n\t" /* pixel.grn > 255 */
"jmp .OKGb \n\t"
".NEGGb: \n\t"
"xorw %%dx, %%dx \n\t"
".OKGb: \n\t"
/* partial reg */
"leal (%%edi, %%edx, 2), %%ecx \n\t" /* ecx = &ctable[pixel.grn] */
/* agi */
"movw (%%ecx), %%ax \n\t" /* ax = ctable[pixel.grn] */
"movw %%ax, %17 \n\t" /* save gg */
"mulb %6 \n\t" /* ax = gg*dg */
"subw %%ax, %%dx \n\t" /* ger = dx = dx - gg*dg */
"movswl %%dx, %%eax \n\t" /* save ger */
/* distribute error */
"leal (, %%eax, 8), %%ecx \n\t"
"subw %%dx, %%cx \n\t" /* cx = ger * 7 */
"sarw $4, %%cx \n\t" /* cx = ger * 7 / 16 */
"addb %%cl, 5(%%ebx) \n\t" /* err[x+1] += ger * 7 / 16 */
"movl %19, %%ecx \n\t" /* ecx = nerr */
"leaw (%%eax, %%eax, 4), %%dx \n\t" /* dx = ger * 5 */
"sarw $4, %%dx \n\t" /* dx = ger * 5 / 16 */
"addb %%dl, 1(%%ecx) \n\t" /* nerr[x] += ger * 5 / 16 */
"leaw (%%eax, %%eax, 2), %%dx \n\t" /* dx = ger * 3 */
"sarw $4, %%dx \n\t" /* dx = ger * 3 / 16 */
"addb %%dl, -3(%%ecx) \n\t" /* nerr[x-1] += ger * 3 / 16 */
"sarw $4, %%ax \n\t" /* ax = ger / 16 */
"movb %%al, 5(%%ecx) \n\t" /* nerr[x+1] = ger / 16 */
/* BLUE; depends on ebx, esi */
"movzbw 2(%%esi), %%dx \n\t" /* dx = image->data[2] */
"movsbw 2(%%ebx), %%ax \n\t" /* ax = error[2] */
"addw %%ax, %%dx \n\t" /* pixel.grn = data[2] + error[2] */
"testb %%dh, %%dh \n\t" /* test if pixel.blu < 0 or > 255 */
"jz .OKBb \n\t" /* 0 <= pixel.blu <= 255 */
"js .NEGBb \n\t" /* pixel.blu < 0 */
"movw $0xff, %%dx \n\t" /* pixel.blu > 255 */
"jmp .OKBb \n\t"
".NEGBb: \n\t"
"xorw %%dx, %%dx \n\t"
".OKBb: \n\t"
/* partial reg */
"leal (%%edi, %%edx, 2), %%ecx \n\t" /* ecx = &ctable[pixel.blu] */
/* agi */
"movw (%%ecx), %%ax \n\t" /* ax = ctable[pixel.blu] */
"movw %%ax, %18 \n\t" /* save bb */
"mulb %7 \n\t" /* ax = bb*db */
"subw %%ax, %%dx \n\t" /* ber = dx = dx - bb*db */
"movswl %%dx, %%eax \n\t" /* save ber */
/* distribute error */
"leal (, %%eax, 8), %%ecx \n\t"
"subw %%dx, %%cx \n\t" /* cx = ber * 7 */
"sarw $4, %%cx \n\t" /* cx = ber * 7 / 16 */
"addb %%cl, 6(%%ebx) \n\t" /* err[x+1] += ber * 7 / 16 */
"movl %19, %%ecx \n\t" /* ecx = nerr */
"leaw (%%eax, %%eax, 4), %%dx \n\t" /* dx = ber * 5 */
"sarw $4, %%dx \n\t" /* dx = ber * 5 / 16 */
"addb %%dl, 2(%%ecx) \n\t" /* nerr[x] += ber * 5 / 16 */
"leaw (%%eax, %%eax, 2), %%dx \n\t" /* dx = ber * 3 */
"sarw $4, %%dx \n\t" /* dx = ber * 3 / 16 */
"addb %%dl, -4(%%ecx) \n\t" /* nerr[x-1] += ber * 3 / 16 */
"sarw $4, %%ax \n\t" /* ax = ber / 16 */
"movb %%al, 6(%%ecx) \n\t" /* nerr[x+1] = ber / 16 */
"andl $0xffff, %%eax \n\t"
/* depends on eax & 0xffff0000 == 0
* calculate the index of the value of the pixel */
"movw %16, %%ax \n\t" /* ax = rr */
"mulb %15 \n\t" /* ax = cpcpc*rr */
"movw %%ax, %%cx \n\t"
"movw %17, %%ax \n\t" /* ax = gg */
"mulb %9 \n\t" /* ax = cpc*gg */
"addw %%cx, %%ax \n\t" /* ax = cpc*gg + cpcpc*rr */
"addw %18, %%ax \n\t" /* ax = cpcpc*rr + cpc*gg + bb */
"movl %8, %%ecx \n\t"
/* agi */
"leal (%%ecx, %%eax, 4), %%edx \n\t"
/* agi */
"movb (%%edx), %%cl \n\t" /* cl = pixels[ax] */
/* store the pixel */
"movl %1, %%eax \n\t"
"movb %%cl, (%%eax) \n\t" /* *ximage = cl */
"incl %1 \n\t" /* ximage++ */
/* prepare for next iteration on X */
"addl $4, %19 \n\t" /* nerr += 4 */
"addl $4, %%ebx \n\t" /* err += 4 */
"addl %12, %%esi \n\t" /* image->data += bpp */
"decl %14 \n\t" /* x-- */
"jnz .LoopXb \n\t" /* if x>0, goto .LoopX */
"movl %13, %%eax \n\t"
"addl %%eax, %1 \n\t" /* add extra offset to ximage */
"jmp .LoopYb \n\t"
".Endb: \n\t"
"emms \n\t"
"popal \n\t"
:
: "m" (image), /* %0 */
"m" (ximage), /* %1 */
"m" (err), /* %2 */
"m" (nerr), /* %3 */
"m" (ctable), /* %4 */
"m" (dr), /* %5 */
"m" (dg), /* %6 */
"m" (db), /* %7 */
"m" (pixels), /* %8 */
"m" (cpc), /* %9 */
"m" (width), /* %10 */
"m" (height), /* %11 */
"m" (bytesPerPixel), /* %12 */
"m" (line_offset), /* %13 */
"m" (x), /* %14 */
"m" (cpcpc), /* %15 */
"m" (rr), /* %16 */
"m" (gg), /* %17 */
"m" (bb), /* %18 */
"m" (tmp_err), /* %19 */
"m" (tmp_nerr), /* %20 */
"m" (ndr), /* %21 */
"m" (ndg), /* %22 */
"m" (ndb) /* %23 */
);
}
#endif /* ASM_X86 */