clean up an ugliness introduced in r11826. this syntax will require fewer changes when adding future sse2 code.

Originally committed as revision 11868 to svn://svn.ffmpeg.org/ffmpeg/trunk
pull/126/head
Loren Merritt 17 years ago
parent 37b74c8651
commit fa9b873e08
  1. 6
      libavcodec/i386/dsputil_h264_template_mmx.c
  2. 2
      libavcodec/i386/dsputil_mmx.c
  3. 4
      libavcodec/i386/dsputil_mmx.h
  4. 8
      libavcodec/i386/h264dsp_mmx.c

@ -98,7 +98,7 @@ static void H264_CHROMA_MC8_TMPL(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*
}
/* general case, bilinear */
rnd_reg = rnd ? ff_pw_32 : &ff_pw_28;
rnd_reg = rnd ? &ff_pw_32.a : &ff_pw_28;
asm volatile("movd %2, %%mm4\n\t"
"movd %3, %%mm6\n\t"
"punpcklwd %%mm4, %%mm4\n\t"
@ -250,7 +250,7 @@ static void H264_CHROMA_MC4_TMPL(uint8_t *dst/*align 4*/, uint8_t *src/*align 1*
"sub $2, %2 \n\t"
"jnz 1b \n\t"
: "+r"(dst), "+r"(src), "+r"(h)
: "r"((long)stride), "m"(*ff_pw_32), "m"(x), "m"(y)
: "r"((long)stride), "m"(ff_pw_32), "m"(x), "m"(y)
);
}
@ -301,7 +301,7 @@ static void H264_CHROMA_MC2_TMPL(uint8_t *dst/*align 2*/, uint8_t *src/*align 1*
"sub $1, %2\n\t"
"jnz 1b\n\t"
: "+r" (dst), "+r"(src), "+r"(h)
: "m" (*ff_pw_32), "r"((long)stride)
: "m" (ff_pw_32), "r"((long)stride)
: "%esi");
}

@ -54,7 +54,7 @@ DECLARE_ALIGNED_8 (const uint64_t, ff_pw_8 ) = 0x0008000800080008ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_15 ) = 0x000F000F000F000FULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_16 ) = 0x0010001000100010ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_20 ) = 0x0014001400140014ULL;
DECLARE_ALIGNED_16(const uint64_t, ff_pw_32[2]) = {0x0020002000200020ULL, 0x0020002000200020ULL};
DECLARE_ALIGNED_16(const xmm_t, ff_pw_32 ) = {0x0020002000200020ULL, 0x0020002000200020ULL};
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_42 ) = 0x002A002A002A002AULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_64 ) = 0x0040004000400040ULL;
DECLARE_ALIGNED_8 (const uint64_t, ff_pw_96 ) = 0x0060006000600060ULL;

@ -24,6 +24,8 @@
#include <stdint.h>
typedef struct { uint64_t a, b; } xmm_t;
extern const uint64_t ff_bone;
extern const uint64_t ff_wtwo;
@ -36,7 +38,7 @@ extern const uint64_t ff_pw_8;
extern const uint64_t ff_pw_15;
extern const uint64_t ff_pw_16;
extern const uint64_t ff_pw_20;
extern const uint64_t ff_pw_32[2];
extern const xmm_t ff_pw_32;
extern const uint64_t ff_pw_42;
extern const uint64_t ff_pw_64;
extern const uint64_t ff_pw_96;

@ -75,7 +75,7 @@ static void ff_h264_idct_add_mmx(uint8_t *dst, int16_t *block, int stride)
IDCT4_1D( %%mm4, %%mm2, %%mm3, %%mm0, %%mm1 )
"pxor %%mm7, %%mm7 \n\t"
:: "m"(*ff_pw_32));
:: "m"(ff_pw_32));
asm volatile(
STORE_DIFF_4P( %%mm0, %%mm1, %%mm7)
@ -294,7 +294,7 @@ static void ff_h264_idct8_add_sse2(uint8_t *dst, int16_t *block, int stride)
STORE_DIFF_8P(%%xmm0, (%0,%2,2), %%xmm6, %%xmm7)
STORE_DIFF_8P(%%xmm1, (%0,%3), %%xmm6, %%xmm7)
:"+r"(dst)
:"r"(block), "r"((long)stride), "r"(3L*stride), "m"(*ff_pw_32)
:"r"(block), "r"((long)stride), "r"(3L*stride), "m"(ff_pw_32)
);
}
@ -926,7 +926,7 @@ static av_noinline void OPNAME ## h264_qpel4_hv_lowpass_ ## MMX(uint8_t *dst, in
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(tmp), "+c"(dst), "+m"(h)\
: "S"((long)dstStride), "m"(*ff_pw_32)\
: "S"((long)dstStride), "m"(ff_pw_32)\
: "memory"\
);\
}\
@ -1200,7 +1200,7 @@ static av_noinline void OPNAME ## h264_qpel8or16_hv_lowpass_ ## MMX(uint8_t *dst
"decl %2 \n\t"\
" jnz 1b \n\t"\
: "+a"(tmp), "+c"(dst), "+m"(h)\
: "S"((long)dstStride), "m"(*ff_pw_32)\
: "S"((long)dstStride), "m"(ff_pw_32)\
: "memory"\
);\
tmp += 8 - size*24;\

Loading…
Cancel
Save