diff options
| -rw-r--r-- | pixman/pixman/config.h | 6 | ||||
| -rw-r--r-- | pixman/pixman/makefile | 32 | ||||
| -rw-r--r-- | pixman/pixman/pixman-mmx.c | 34 |
3 files changed, 25 insertions, 47 deletions
diff --git a/pixman/pixman/config.h b/pixman/pixman/config.h index 50adacc11..b16bc8da4 100644 --- a/pixman/pixman/config.h +++ b/pixman/pixman/config.h @@ -122,7 +122,7 @@ /* #undef TRANS_REOPEN */ /* use MMX compiler intrinsics */ -#define USE_MMX 1 +#define USE_X86_MMX 1 /* Support UNIX socket connections */ #define UNIXCONN 1 @@ -134,7 +134,7 @@ #define USE_DYNAMIC_XCURSOR 1 /* use SSE2 compiler intrinsics */ -#undef USE_SSE2 +#define USE_SSE2 1 /* use VMX compiler intrinsics */ #undef USE_VMX @@ -225,3 +225,5 @@ /* Support gzip for bitmap fonts */ #define X_GZIP_FONT_COMPRESSION 1 + +#define __inline__ __inline
\ No newline at end of file diff --git a/pixman/pixman/makefile b/pixman/pixman/makefile index 869040492..d85478e98 100644 --- a/pixman/pixman/makefile +++ b/pixman/pixman/makefile @@ -42,36 +42,8 @@ $(OBJDIR)\pixman-combine64.h : pixman-combine.h.template make-combine.pl perl make-combine.pl 16 < pixman-combine.h.template > $@ ## mmx code -#if USE_MMX -#noinst_LTLIBRARIES += libpixman-mmx.la -CSRCS += \ - pixman-mmx.c -# pixman-mmx.h -#libpixman_mmx_la_CFLAGS = $(DEP_CFLAGS) $(MMX_CFLAGS) -#libpixman_mmx_la_LIBADD = $(DEP_LIBS) -#libpixman_1_la_LIBADD += libpixman-mmx.la -#endif - -## vmx code -#if USE_VMX -#noinst_LTLIBRARIES += libpixman-vmx.la -#libpixman_vmx_la_SOURCES = \ -# pixman-vmx.c \ -# pixman-vmx.h \ -# pixman-combine32.h -#libpixman_vmx_la_CFLAGS = $(DEP_CFLAGS) $(VMX_CFLAGS) -#libpixman_vmx_la_LIBADD = $(DEP_LIBS) -#libpixman_1_la_LIBADD += libpixman-vmx.la -#endif +CSRCS += pixman-mmx.c # sse2 code -#if USE_SSE2 -#noinst_LTLIBRARIES += libpixman-sse2.la -#libpixman_sse2_la_SOURCES = \ -# pixman-sse2.c \ -# pixman-sse2.h -#libpixman_sse2_la_CFLAGS = $(DEP_CFLAGS) $(SSE2_CFLAGS) -#libpixman_sse2_la_LIBADD = $(DEP_LIBS) -#libpixman_1_la_LIBADD += libpixman-sse2.la -#endif +CSRCS += pixman-sse2.c diff --git a/pixman/pixman/pixman-mmx.c b/pixman/pixman/pixman-mmx.c index 9c428a3de..ea244f67e 100644 --- a/pixman/pixman/pixman-mmx.c +++ b/pixman/pixman/pixman-mmx.c @@ -313,24 +313,28 @@ in_over (__m64 src, __m64 srca, __m64 mask, __m64 dest) /* Elemental unaligned loads */ -static __inline__ uint64_t ldq_u(uint64_t *p) +#ifdef _MSC_VER +#define ldq_u(p) *((__m64*)(p)) +#else +static __inline__ __m64 ldq_u(uint64_t *p) { #ifdef USE_X86_MMX /* x86's alignment restrictions are very relaxed. */ - return *p; + return (__m64)*p; #elif defined USE_ARM_IWMMXT int align = (uintptr_t)p & 7; __m64 *aligned_p; if (align == 0) - return *p; + return (__m64)*p; aligned_p = (__m64 *)((uintptr_t)p & ~7); return _mm_align_si64 (aligned_p[0], aligned_p[1], align); #else struct __una_u64 { uint64_t x __attribute__((packed)); }; const struct __una_u64 *ptr = (const struct __una_u64 *) p; - return ptr->x; + return (__m64)ptr->x; #endif } +#endif static __inline__ uint32_t ldl_u(uint32_t *p) { @@ -1424,7 +1428,7 @@ mmx_composite_over_8888_n_8888 (pixman_implementation_t *imp, while (w >= 2) { - __m64 vs = (__m64)ldq_u((uint64_t *)src); + __m64 vs = ldq_u((uint64_t *)src); __m64 vd = *(__m64 *)dst; __m64 vsrc0 = expand8888 (vs, 0); __m64 vsrc1 = expand8888 (vs, 1); @@ -1505,14 +1509,14 @@ mmx_composite_over_x888_n_8888 (pixman_implementation_t *imp, __m64 vd6 = *(__m64 *)(dst + 12); __m64 vd7 = *(__m64 *)(dst + 14); - __m64 vs0 = (__m64)ldq_u((uint64_t *)(src + 0)); - __m64 vs1 = (__m64)ldq_u((uint64_t *)(src + 2)); - __m64 vs2 = (__m64)ldq_u((uint64_t *)(src + 4)); - __m64 vs3 = (__m64)ldq_u((uint64_t *)(src + 6)); - __m64 vs4 = (__m64)ldq_u((uint64_t *)(src + 8)); - __m64 vs5 = (__m64)ldq_u((uint64_t *)(src + 10)); - __m64 vs6 = (__m64)ldq_u((uint64_t *)(src + 12)); - __m64 vs7 = (__m64)ldq_u((uint64_t *)(src + 14)); + __m64 vs0 = ldq_u((uint64_t *)(src + 0)); + __m64 vs1 = ldq_u((uint64_t *)(src + 2)); + __m64 vs2 = ldq_u((uint64_t *)(src + 4)); + __m64 vs3 = ldq_u((uint64_t *)(src + 6)); + __m64 vs4 = ldq_u((uint64_t *)(src + 8)); + __m64 vs5 = ldq_u((uint64_t *)(src + 10)); + __m64 vs6 = ldq_u((uint64_t *)(src + 12)); + __m64 vs7 = ldq_u((uint64_t *)(src + 14)); vd0 = pack8888 ( in_over (expandx888 (vs0, 0), srca, vmask, expand8888 (vd0, 0)), @@ -2791,7 +2795,7 @@ mmx_composite_add_8_8 (pixman_implementation_t *imp, while (w >= 8) { - *(__m64*)dst = _mm_adds_pu8 ((__m64)ldq_u((uint64_t *)src), *(__m64*)dst); + *(__m64*)dst = _mm_adds_pu8 (ldq_u((uint64_t *)src), *(__m64*)dst); dst += 8; src += 8; w -= 8; @@ -2849,7 +2853,7 @@ mmx_composite_add_8888_8888 (pixman_implementation_t *imp, while (w >= 2) { - dst64 = _mm_adds_pu8 ((__m64)ldq_u((uint64_t *)src), *(__m64*)dst); + dst64 = _mm_adds_pu8 (ldq_u((uint64_t *)src), *(__m64*)dst); *(uint64_t*)dst = to_uint64 (dst64); dst += 2; src += 2; |