Remove unmaintained MMX optimized function and runtime detection.

10 files changed, 9 insertions, 564 deletions

diff --git a/Makefile b/Makefile
index 1cc963c5..e0545614 100644
--- a/Makefile
+++ b/Makefile
@@ -74,7 +74,6 @@ $(CMAPDUMP_EXE): $(OBJDIR)/cmapdump.o
 
 FITZ_HDR := fitz/fitz.h fitz/fitz_base.h fitz/fitz_draw.h fitz/fitz_stream.h
 FITZ_SRC := $(addprefix fitz/, \
-	base_cpudep.c \
 	base_error.c base_memory.c base_string.c base_unicode.c \
 	base_hash.c base_matrix.c base_rect.c \
 	crypt_aes.c crypt_arc4.c crypt_md5.c \
diff --git a/Makerules b/Makerules
index 1d828e6f..d584ca8b 100644
--- a/Makerules
+++ b/Makerules
@@ -24,13 +24,6 @@ SYS_FREETYPE_LIB := `pkg-config --libs freetype2`
 X11LIBS := -lX11 -lXext
 PDFVIEW_EXE = $(X11VIEW_EXE)
 
-ifeq "$(build)" "release"
-ifeq "$(shell uname -m)" "i686"
-CFLAGS += -ffast-math -mmmx -msse -msse2 -march=k8 -DARCH_X86
-DRAW_ARCH_SRC := archx86.c
-endif
-endif
-
 endif
 
 ifeq "$(OS)" "Darwin"
@@ -49,15 +42,6 @@ CFLAGS += -m32
 LDFLAGS += -m32
 endif
 
-ifeq "$(build)" "release"
-ifeq "$(arch)" "amd64"
-CFLAGS += -ffast-math -mmmx -msse -msse2 -DARCH_X86_64
-else
-CFLAGS += -ffast-math -mmmx -msse -msse2 -DARCH_X86
-endif
-DRAW_ARCH_SRC := archx86.c
-endif
-
 endif
 
 # MinGW build depends on thirdparty
diff --git a/apps/pdfdraw.c b/apps/pdfdraw.c
index 45079ffb..c3b93f6e 100644
--- a/apps/pdfdraw.c
+++ b/apps/pdfdraw.c
@@ -466,7 +466,6 @@ int main(int argc, char **argv)
 	int c;
 	enum { NO_FILE_OPENED, NO_PAGES_DRAWN, DREW_PAGES } state;
 
-	fz_cpudetect();
 	fz_accelerate();
 
 	while ((c = fz_getopt(argc, argv, "b:p:o:r:gtxms")) != -1)
diff --git a/apps/win_main.c b/apps/win_main.c
index 90aef1d2..8ab8b371 100644
--- a/apps/win_main.c
+++ b/apps/win_main.c
@@ -773,7 +773,6 @@ WinMain(HINSTANCE hInstance, HINSTANCE hPrevInstance, LPSTR lpCmdLine, int nShow
 	int fd;
 	int code;
 
-	fz_cpudetect();
 	fz_accelerate();
 
 	pdfapp_init(&gapp);
diff --git a/apps/x11_main.c b/apps/x11_main.c
index 5dd9efb5..8172b651 100644
--- a/apps/x11_main.c
+++ b/apps/x11_main.c
@@ -590,7 +590,6 @@ int main(int argc, char **argv)
 	if (argc - fz_optind == 1)
 		pageno = atoi(argv[fz_optind++]);
 
-	fz_cpudetect();
 	fz_accelerate();
 
 	winopen();
diff --git a/draw/archx86.c b/draw/archx86.c
deleted file mode 100644
index 6bdb1f6b..00000000
--- a/draw/archx86.c
+++ /dev/null
@@ -1,228 +0,0 @@
-/*
- * x86 specific render optims live here
- */
-
-#include "fitz.h"
-
-typedef unsigned char byte;
-
-/* always surround cpu specific code with HAVE_XXX */
-#ifdef HAVE_MMX
-
-/* -mmmx for gcc >= 3.4 enables the mmx intrinsic functions, icc and VC
-shouldn't require anything */
-#include <mmintrin.h>
-
-static void duff_4i1o4mmx(byte *sp0, int sw, byte *mp0, int mw, byte *dp0, int dw, int w0, int h)
-{
-	__m64 mzero = _mm_setzero_si64();
-	while (h--)
-	{
-		byte *sp = sp0;
-		byte *mp = mp0;
-		byte *dp = dp0;
-
-		unsigned *s = (unsigned *)sp;
-		unsigned *d = (unsigned *)dp;
-
-		int w = w0;
-
-		/* TODO: unroll and process two pixels/iteration */
-		while (w--)
-		{
-			int ts = *s++;
-			int ma = *mp++ + 1;
-			int sa = (((ts>>24) & 0xff) * ma) >> 8;
-			int ssa = 255 - sa;
-
-			__m64 d0 = _mm_cvtsi32_si64(*d);
-			__m64 s0 = _mm_cvtsi32_si64(ts);
-
-			/* 4 x 9 bit alpha value */
-			__m64 mma = _mm_set1_pi16(ma);
-			__m64 mssa = _mm_set1_pi16(ssa);
-
-			/* unpack 0000rgba => r0g0b0a0 */
-			__m64 d1 = _mm_unpacklo_pi8(d0, mzero);
-			__m64 s1 = _mm_unpacklo_pi8(s0, mzero);
-
-			/* s1 * ma => r0g0b0a0 */
-			__m64 msma = _mm_mullo_pi16(s1, mma);
-			/* d1 * mssa */
-			__m64 mdssa = _mm_mullo_pi16(d1, mssa);
-
-			__m64 res0 = _mm_add_pi16(msma, mdssa);
-			/* TODO: is it possible to get rid of the shift? */
-			__m64 res1 = _mm_srli_pi16(res0, 8);
-
-			/* pack */
-			__m64 res2 = _mm_packs_pu16(res1, mzero);
-
-			*d++ = _mm_cvtsi64_si32(res2);
-		}
-
-		sp0 += sw;
-		mp0 += mw;
-		dp0 += dw;
-	}
-
-	_mm_empty();
-}
-
-#if 0 /* TODO */
-
-/* Needs to be rgba, not bgra, as well as needing finishing */
-
-static inline unsigned
-getargb(unsigned *s, int w, int h, int u, int v)
-{
-	if ((u < 0) | (u >= w) | (v < 0) | (v >= h)) return 0;
-	return s[w * v + u];
-}
-
-static void img_4o4mmx(FZ_PSRC, FZ_PDST, FZ_PCTM)
-{
-	/* since mmx does not have an unsigned multiply instruction we use
-	17.15 fixed point */
-	u0 >>= 1;	v0 >>= 1;
-	fa >>= 1;	fb >>= 1;
-	fc >>= 1;	fd >>= 1;
-
-	while (h--)
-	{
-		unsigned *s = (unsigned *)src;
-		unsigned *d = (unsigned *)dst0;
-		int u = u0;
-		int v = v0;
-		int w = w0;
-
-		__m64 mzero = _mm_setzero_si64();
-		__m64 m256 = _mm_set1_pi16(256);
-		__m64 malphamask = _mm_cvtsi32_si64(0xff);
-
-		while (w--)
-		{
-			int iu = u >> 15;
-			int iv = v >> 15;
-
-			int fu = u & 0x7fff;
-			int fv = v & 0x7fff;
-
-			int atedge =
-			(iu < 0) | (iu >= (srcw - 1)) |
-			(iv < 0) | (iv >= (srch - 1));
-
-			__m64 ms0s1;
-			__m64 ms2s3;
-
-			if (atedge)
-			{
-				unsigned s0, s1, s2, s3;
-
-				/* edge cases use scalar loads */
-				s0 = getargb(s, srcw, srch, iu + 0, iv + 0);
-				s1 = getargb(s, srcw, srch, iu + 1, iv + 0);
-				s2 = getargb(s, srcw, srch, iu + 0, iv + 1);
-				s3 = getargb(s, srcw, srch, iu + 1, iv + 1);
-
-				/* move to mmx registers */
-				ms0s1 = _mm_set_pi32(s1, s0);
-				ms2s3 = _mm_set_pi32(s3, s2);
-			}
-			else
-			{
-				__m64 *m0s = (__m64*)(s + srcw * (iv + 0) + iu);
-				__m64 *m2s = (__m64*)(s + srcw * (iv + 1) + iu);
-
-				/* faster vector loads for interior */
-				ms0s1 = *m0s;
-				ms2s3 = *m2s;
-			}
-
-			/* unpack src into 4x16bit vectors */
-			__m64 ms0 = _mm_unpacklo_pi8(ms0s1, mzero);
-			__m64 ms1 = _mm_unpackhi_pi8(ms0s1, mzero);
-			__m64 ms2 = _mm_unpacklo_pi8(ms2s3, mzero);
-			__m64 ms3 = _mm_unpackhi_pi8(ms2s3, mzero);
-
-			/* lerp fu */
-
-			__m64 mfu = _mm_set1_pi16(fu);
-
-			/* t2 = (s1 - s0) * fu + s0 */
-			__m64 t0 = _mm_sub_pi16(ms1, ms0);
-			__m64 t1 = _mm_mulhi_pi16(t0, mfu);
-			t1 = _mm_adds_pi16(t1, t1);
-			__m64 t2 = _mm_add_pi16(t1, ms0);
-
-			/* t3 = (s3 - s2) * fu + s2 */
-			__m64 t3 = _mm_sub_pi16(ms3, ms2);
-			__m64 t4 = _mm_mulhi_pi16(t3, mfu);
-			t4 = _mm_adds_pi16(t4, t4);
-			__m64 t5 = _mm_add_pi16(t4, ms2);
-
-			/* lerp fv */
-
-			__m64 mfv = _mm_set1_pi16(fv);
-
-			/* t8 = (t5 - t2) * fv + t2 */
-			__m64 t6 = _mm_sub_pi16(t5, t2);
-			__m64 t7 = _mm_mulhi_pi16(t6, mfv);
-			t7 = _mm_adds_pi16(t7, t7);
-			__m64 t8 = _mm_add_pi16(t7, t2);
-
-			/* load and prepare dst */
-			__m64 d0 = _mm_cvtsi32_si64(*d);
-
-			__m64 d1 = _mm_unpacklo_pi8(d0, mzero);
-
-			/* get src alpha */
-
-			/* splat alpha */
-			__m64 a0001 = _mm_and_si64(malphamask, t8);
-			__m64 a0011 = _mm_unpacklo_pi16(a0001, a0001);
-			__m64 a1111 = _mm_unpacklo_pi16(a0011, a0011);
-
-			/* 255+1 - sa */
-			__m64 sna = _mm_sub_pi16(m256, a1111);
-
-			/* blend src with dst */
-			__m64 d2 = _mm_mullo_pi16(d1, sna);
-			__m64 d3 = _mm_srli_pi16(d2, 8);
-			__m64 d4 = _mm_add_pi16(t8, d3);
-
-			/* pack and store new dst */
-			__m64 d5 = _mm_packs_pu16(d4, mzero);
-
-			*d++ = _mm_cvtsi64_si32(d5);
-
-			u += fa;
-			v += fb;
-		}
-
-		dst0 += dstw;
-		u0 += fc;
-		v0 += fd;
-	}
-
-	_mm_empty();
-}
-
-#endif
-
-#endif /* HAVE_MMX */
-
-#if defined (ARCH_X86) || defined(ARCH_X86_64)
-void
-fz_acceleratearch(void)
-{
-#ifdef HAVE_MMX
-	if (fz_cpuflags & HAVE_MMX)
-	{
-		fz_duff_4i1o4 = duff_4i1o4mmx;
-// TODO		fz_img_4o4 = img_4o4mmx;
-	}
-#endif
-}
-#endif
-
diff --git a/fitz/base_cpudep.c b/fitz/base_cpudep.c
deleted file mode 100644
index 586a478a..00000000
--- a/fitz/base_cpudep.c
+++ /dev/null
@@ -1,283 +0,0 @@
-/*
-run-time cpu feature detection code
-mm, alphabet soup...
-
-Glenn Kennard <d98gk@efd.lth.se>
-*/
-
-#include "fitz.h"
-
-/* global run-time constant */
-unsigned fz_cpuflags = 0;
-
-int fz_isbigendian(void)
-{
-	static const int one = 1;
-	return *(char*)&one == 0;
-}
-
-#ifndef HAVE_CPUDEP
-
-void fz_cpudetect(void)
-{
-}
-
-#else
-
-#ifndef _WIN32
-#include <signal.h> /* signal/sigaction */
-#include <setjmp.h> /* sigsetjmp/siglongjmp */
-#endif
-
-/*
-#ifdef _WIN32
-#define sigjmp_buf jmp_buf
-#define sigsetjmp(a,b) setjmp(a)
-#define siglongjmp longjmp
-#endif
-*/
-
-typedef struct {
-	void (*test)(void);
-	const unsigned flag;
-	const char *name;
-} featuretest;
-
-
-#if defined(ARCH_X86) || defined(ARCH_X86_64)
-
-#ifdef __GNUC__
-static void mmx(void)
-{ __asm__ ("pand %mm0, %mm0\n\t"); }
-
-static void m3dnow(void)
-{ __asm__ ("pavgusb %mm0, %mm0\n\t"); }
-
-static void mmxext(void) /* aka Extended 3DNow! */
-{ __asm__ ("pmaxsw %mm0, %mm0\n\t"); }
-
-static void sse(void)
-{ __asm__ ("andps %xmm0, %xmm0\n\t"); }
-
-static void sse2(void)
-{ __asm__ ("andpd %xmm0, %xmm0\n\t"); }
-
-/*
-static void sse3(void)
-{ __asm__ ("haddps %%xmm0, %%xmm0\n\t" : : : "%xmm0"); }
-*/
-#else
-static void mmx(void)
-{ __asm pand mm0, mm0; }
-
-static void m3dnow(void)
-{ __asm pavgusb mm0, mm0; }
-
-static void mmxext(void) /* aka Extended 3DNow! */
-{ __asm pmaxsw mm0, mm0; }
-
-static void sse(void)
-{ __asm andps xmm0, xmm0; }
-
-static void sse2(void)
-{ __asm andpd xmm0, xmm0; }
-#endif
-
-
-#ifdef ARCH_X86_64
-static void amd64(void)
-#ifdef __GNUC__
-{ __asm__ ("and %rax, %rax\n\t"); }
-#else
-{ __asm and rax, rax; }
-#endif
-#endif
-
-
-static const featuretest features[] = {
-	{ mmx, HAVE_MMX, "mmx" },
-	{ m3dnow, HAVE_3DNOW, "3dnow" },
-	{ mmxext, HAVE_MMXEXT, "mmxext" },
-	{ sse, HAVE_SSE, "sse" },
-	{ sse2, HAVE_SSE2, "sse2" },
-	/*	{ sse3, HAVE_SSE3, "sse3" }, */
-#ifdef ARCH_X86_64
-	{ amd64, HAVE_AMD64, "amd64" },
-#endif
-};
-
-#endif
-
-static int
-enabled(char *env, const char *ext)
-{
-	int len;
-	char *s;
-	if (!env)
-		return 1;
-	len = strlen(ext);
-	s = env;
-	while ((s = strstr(s, ext)))
-	{
-		int atstart = s == env || *(s-1) == ',' || *(s-1) == ' ';
-		s += len;
-		if (atstart && (*s == ' ' || *s == ',' || *s == '\0')) {
-			return 1;
-		}
-	}
-	return 0;
-}
-
-static void
-dumpflags(void)
-{
-	unsigned f = fz_cpuflags;
-	int i, n;
-
-	fputs("detected cpu features:", stdout);
-	n = 0;
-	for (i = 0; i < sizeof(features) / sizeof(featuretest); i++)
-	{
-		if (f & features[i].flag)
-		{
-			fputc(' ', stdout);
-			fputs(features[i].name, stdout);
-			n ++;
-		}
-	}
-	if (!n)
-		fputs(" none", stdout);
-	fputc('\n', stdout);
-}
-
-#ifndef _WIN32
-
-static sigjmp_buf jmpbuf;
-static volatile sig_atomic_t canjump;
-
-static void
-sigillhandler(int sig)
-{
-	if (!canjump) {
-		signal(sig, SIG_DFL);
-		raise(sig);
-	}
-
-	canjump = 0;
-	siglongjmp(jmpbuf, 1);
-}
-
-void fz_cpudetect(void)
-{
-	static int hasrun = 0;
-
-	unsigned flags = 0;
-	int i;
-	void (*oldhandler)(int) = NULL;
-	void (*tmphandler)(int);
-	char *env;
-
-	if (hasrun)
-		return;
-	hasrun = 1;
-
-	env = getenv("CPUACCEL");
-
-	for (i = 0; i < sizeof(features) / sizeof(featuretest); i++)
-	{
-		canjump = 0;
-
-		tmphandler = signal(SIGILL, sigillhandler);
-		if (!oldhandler)
-			oldhandler = tmphandler;
-
-		if (sigsetjmp(jmpbuf, 1))
-		{
-			/* test failed - disable feature */
-			flags &= ~features[i].flag;
-			continue;
-		}
-
-		canjump = 1;
-
-		features[i].test();
-
-#if defined(ARCH_X86) || defined(ARCH_X86_64)
-		/* reset mmx/x87 pipeline state */
-		if (features[i].flag & (HAVE_MMX | HAVE_3DNOW | HAVE_MMXEXT)) {
-			__asm__ __volatile__ ("emms\n\t");
-		}
-#endif
-
-		/* if we got here the test succeeded */
-		if (enabled(env, features[i].name))
-			flags |= features[i].flag;
-		else
-			flags &= ~features[i].flag;
-	}
-
-	/* restore previous signal handler */
-	signal(SIGILL, oldhandler);
-
-	fz_cpuflags = flags;
-
-#if defined(ARCH_X86) || defined(ARCH_X86_64)
-	__asm__ __volatile__ ("emms\n\t");
-#endif
-
-	dumpflags();
-}
-
-#else /* _WIN32 */
-
-void fz_cpudetect(void)
-{
-	static int hasrun = 0;
-
-	unsigned flags = 0;
-	int i;
-	char *env;
-
-	if (hasrun)
-		return;
-	hasrun = 1;
-
-	env = getenv("CPUACCEL");
-
-	for (i = 0; i < sizeof(features) / sizeof(featuretest); i++)
-	{
-		__try
-		{
-			features[i].test();
-		}
-		__except(EXCEPTION_EXECUTE_HANDLER)
-		{
-			/* test failed - disable feature */
-			flags &= ~features[i].flag;
-			continue;
-		}
-
-#if defined(ARCH_X86) || defined(ARCH_X86_64)
-		if (features[i].flag & (HAVE_MMX | HAVE_3DNOW | HAVE_MMXEXT)) {
-			/* reset mmx/x87 pipeline state */
-			__asm emms;
-		}
-#endif
-
-		/* if we got here the test succeeded */
-		if (enabled(env, features[i].name))
-			flags |= features[i].flag;
-		else
-			flags &= ~features[i].flag;
-	}
-
-	fz_cpuflags = flags;
-
-	dumpflags();
-}
-
-
-#endif
-
-#endif
-
diff --git a/fitz/base_string.c b/fitz/base_string.c
index e18d8720..b6b85865 100644
--- a/fitz/base_string.c
+++ b/fitz/base_string.c
@@ -2,6 +2,12 @@
 
 #include <string.h>
 
+int fz_isbigendian(void)
+{
+	static const int one = 1;
+	return *(char*)&one == 0;
+}
+
 char *fz_strsep(char **stringp, const char *delim)
 {
 	char *ret = *stringp;
diff --git a/fitz/fitz_base.h b/fitz/fitz_base.h
index ec1eb9fc..9563b0c9 100644
--- a/fitz/fitz_base.h
+++ b/fitz/fitz_base.h
@@ -69,33 +69,8 @@ extern int gettimeofday(struct timeval *tv, struct timezone *tz);
 #endif
 #endif
 
-/*
- * CPU detection and flags
- */
-
-#if defined(ARCH_X86) || defined(ARCH_X86_64)
-#define HAVE_CPUDEP
-#define HAVE_MMX        (1<<0)
-#define HAVE_MMXEXT     (1<<1)
-#define HAVE_SSE        (1<<2)
-#define HAVE_SSE2       (1<<3)
-#define HAVE_SSE3       (1<<4)
-#define HAVE_3DNOW      (1<<5)
-#define HAVE_AMD64      (1<<6)
 #endif
 
-#ifdef ARCH_ARM
-#define HAVE_CPUDEP
-#endif
-
-/* call this before using fitz */
-extern void fz_cpudetect(void);
-
-/* treat as constant! */
-extern unsigned fz_cpuflags;
-
-int fz_isbigendian(void);
-
 /*
  * Base Fitz runtime.
  */
@@ -137,6 +112,9 @@ int fz_isbigendian(void);
 #define CLAMP(x,a,b) ( (x) > (b) ? (b) : ( (x) < (a) ? (a) : (x) ) )
 #endif
 
+/* runtime (hah!) test for endian-ness */
+int fz_isbigendian(void);
+
 /* utf-8 encoding and decoding */
 int chartorune(int *rune, char *str);
 int runetochar(char *str, int *rune);
diff --git a/win32/mupdf/mupdf.vcproj b/win32/mupdf/mupdf.vcproj
index 2816ca8b..4d7ac6c0 100644
--- a/win32/mupdf/mupdf.vcproj
+++ b/win32/mupdf/mupdf.vcproj
@@ -289,10 +289,6 @@
 			Name="fitz"

 			>

 			<File

-				RelativePath="..\..\fitz\base_cpudep.c"

-				>

-			</File>

-			<File

 				RelativePath="..\..\fitz\base_error.c"

 				>

 			</File>

@@ -505,10 +501,6 @@
 				>

 			</File>

 			<File

-				RelativePath="..\..\draw\archx86.c"

-				>

-			</File>

-			<File

 				RelativePath="..\..\draw\blendmodes.c"

 				>

 			</File>