adam/cobm
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Asm scaler fixes: restrict RGB15_to_YUV operand to a register, but leave gcc more available regs; fixes scaler problems in MacOSX; Also, properly inform gcc of clobbered condition code reg

Browse Source 
git-svn-id: svn://svn.code.sf.net/p/sc2/code/trunk@3248 8092fc87-c524-0410-9efc-e669fe64eaf9
This commit is contained in:
avolkov
2009-10-25 01:01:45 +00:00
parent 11ffd2ac61
commit a82b17e1c9
1 changed files with 8 additions and 17 deletions
Download Patch File Download Diff File Expand all files Collapse all files
sc2/src/libs/graphics/sdl/scalemmx.h
+8 -17
View File
@@ -489,16 +489,10 @@ SCALE_(Blend_bilinear) (const Uint32* row0, const Uint32* row1,
# define A_REG "rax" # define A_REG "rax"
# define D_REG "rdx" # define D_REG "rdx"
# define CLR_UPPER32(r) "xor " "%%" r "," "%%" r # define CLR_UPPER32(r) "xor " "%%" r "," "%%" r
# define OFFSETABLE_PTR(p) "r" (p)
# define MOVD_WMO(base, ofs, mult, dst) \
"movd " "(" #base "," ofs "," #mult ")," #dst
#else #else
# define A_REG "eax" # define A_REG "eax"
# define D_REG "edx" # define D_REG "edx"
# define CLR_UPPER32(r) # define CLR_UPPER32(r)
# define OFFSETABLE_PTR(p) "o" (*(p))
# define MOVD_WMO(base, ofs, mult, dst) \
"movd " #base "(," ofs "," #mult ")," #dst
#endif #endif
static inline void static inline void
@@ -553,7 +547,7 @@ SCALE_(GetRGBDelta) (Uint32 pix1, Uint32 pix2)
// store result // store result
"movd %%mm1, %0 \n\t" "movd %%mm1, %0 \n\t"
: /*0*/"=r" (res) : /*0*/"=rm" (res)
: /*1*/"rm" (pix1), /*2*/"rm" (pix2) : /*1*/"rm" (pix1), /*2*/"rm" (pix2)
); );
@@ -603,7 +597,7 @@ SCALE_(RGBtoYUV) (Uint32 pix)
"por %%mm2, %%mm1 \n\t" // ditto "por %%mm2, %%mm1 \n\t" // ditto
"movd %%mm1, %0 \n\t" "movd %%mm1, %0 \n\t"
: /*0*/"=r" (i) : /*0*/"=rm" (i)
: /*1*/"rm" (pix), /*2*/"m" (mmx_888to555_mult) : /*1*/"rm" (pix), /*2*/"m" (mmx_888to555_mult)
); );
return RGB15_to_YUV[i]; return RGB15_to_YUV[i];
@@ -641,11 +635,9 @@ SCALE_(CmpYUV) (Uint32 pix1, Uint32 pix2, int toler)
// lookup the YUV vector // lookup the YUV vector
"movd %%mm1, %%eax \n\t" "movd %%mm1, %%eax \n\t"
"movd %%mm3, %%edx \n\t" "movd %%mm3, %%edx \n\t"
//movd (%3, %%eax, 4), %%mm1 "movd (%3, %%" A_REG ", 4), %%mm1 \n\t"
MOVD_WMO (%3, "%%" A_REG, 4, %%mm1) "\n\t"
"movq %%mm1, %%mm4 \n\t" "movq %%mm1, %%mm4 \n\t"
//movd (%3, %%edx, 4), %%mm2 "movd (%3, %%" D_REG ", 4), %%mm2 \n\t"
MOVD_WMO (%3, "%%" D_REG, 4, %%mm2) "\n\t"
// get abs difference between YUV components // get abs difference between YUV components
#ifdef USE_PSADBW #ifdef USE_PSADBW
@@ -670,11 +662,11 @@ SCALE_(CmpYUV) (Uint32 pix1, Uint32 pix2, int toler)
"movd %%mm1, %0 \n\t" "movd %%mm1, %0 \n\t"
"andl $0xff, %0 \n\t" "andl $0xff, %0 \n\t"
#endif /* USE_PSADBW */ #endif /* USE_PSADBW */
: /*0*/"=r" (delta) : /*0*/"=rm" (delta)
: /*1*/"rm" (pix1), /*2*/"rm" (pix2), : /*1*/"rm" (pix1), /*2*/"rm" (pix2),
/*3*/OFFSETABLE_PTR(RGB15_to_YUV), /*3*/ "r" (RGB15_to_YUV),
/*4*/"m" (mmx_888to555_mult) /*4*/"m" (mmx_888to555_mult)
: "%" A_REG, "%" D_REG : "%" A_REG, "%" D_REG, "cc"
); );
return (delta << 1) <= toler; return (delta << 1) <= toler;
@@ -713,7 +705,7 @@ SCALE_(DiffYUV) (Uint32 yuv1, Uint32 yuv2)
: /*0*/"=a" (ret) : /*0*/"=a" (ret)
: /*1*/"rm" (yuv1), /*2*/"rm" (yuv2), : /*1*/"rm" (yuv1), /*2*/"rm" (yuv2),
/*3*/"m" (mmx_YUV_threshold) /*3*/"m" (mmx_YUV_threshold)
: "%" D_REG : "%" D_REG, "cc"
); );
return ret; return ret;
} }
@@ -791,7 +783,6 @@ SCALE_(Blend_bilinear) (const Uint32* row0, const Uint32* row1,
#undef A_REG #undef A_REG
#undef D_REG #undef D_REG
#undef CLR_UPPER32 #undef CLR_UPPER32
#undef OFFSETABLE_PTR
#endif // GCC_ASM #endif // GCC_ASM