WARPER88_1616(hadamard8_diff_mmx, hadamard8_diff16_mmx)
+#define put_no_rnd_pixels8_mmx(a,b,c,d) put_pixels8_mmx(a,b,c,d)
+#define put_no_rnd_pixels16_mmx(a,b,c,d) put_pixels16_mmx(a,b,c,d)
+
#define QPEL_V_LOW(m3,m4,m5,m6, pw_20, pw_3, rnd, in0, in1, in2, in7, out, OP)\
"paddw " #m4 ", " #m3 " \n\t" /* x1 */\
"movq " #pw_20 ", %%mm4 \n\t" /* 20 */\
"packuswb %%mm5, %%mm5 \n\t"\
OP(%%mm5, out, %%mm7, d)
-#define QPEL_BASE(OPNAME, ROUNDER, RND, OP)\
+#define QPEL_BASE(OPNAME, ROUNDER, RND, OP_MMX2, OP_3DNOW)\
void OPNAME ## mpeg4_qpel16_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
uint64_t temp;\
\
"psraw $5, %%mm3 \n\t"\
"movq %7, %%mm1 \n\t"\
"packuswb %%mm3, %%mm1 \n\t"\
- OP(%%mm1, (%1),%%mm4, q)\
+ OP_MMX2(%%mm1, (%1),%%mm4, q)\
/* mm0= GHIJ, mm2=FGHI, mm5=HIJK, mm6=IJKL, mm7=0 */\
\
"movq 9(%0), %%mm1 \n\t" /* JKLMNOPQ */\
"paddw %%mm3, %%mm4 \n\t" /* 20a - 6b + 3c - d */\
"psraw $5, %%mm4 \n\t"\
"packuswb %%mm4, %%mm0 \n\t"\
- OP(%%mm0, 8(%1), %%mm4, q)\
+ OP_MMX2(%%mm0, 8(%1), %%mm4, q)\
\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"psraw $5, %%mm0 \n\t"\
"psraw $5, %%mm1 \n\t"\
"packuswb %%mm1, %%mm0 \n\t"\
- OP(%%mm0, (%1), %%mm1, q)\
+ OP_3DNOW(%%mm0, (%1), %%mm1, q)\
"movq 16(%0), %%mm0 \n\t"\
"movq 24(%0), %%mm1 \n\t"\
"paddw %2, %%mm0 \n\t"\
"psraw $5, %%mm0 \n\t"\
"psraw $5, %%mm1 \n\t"\
"packuswb %%mm1, %%mm0 \n\t"\
- OP(%%mm0, 8(%1), %%mm1, q)\
+ OP_3DNOW(%%mm0, 8(%1), %%mm1, q)\
:: "r"(temp), "r"(dst), "m"(ROUNDER)\
);\
dst+=dstStride;\
}\
}\
\
-void OPNAME ## mpeg4_qpel16_v_lowpass_mmx(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
- uint64_t temp[17*4];\
- uint64_t *temp_ptr= temp;\
- int count= 17;\
-\
- /*FIXME unroll */\
- asm volatile(\
- "pxor %%mm7, %%mm7 \n\t"\
- "1: \n\t"\
- "movq (%0), %%mm0 \n\t"\
- "movq (%0), %%mm1 \n\t"\
- "movq 8(%0), %%mm2 \n\t"\
- "movq 8(%0), %%mm3 \n\t"\
- "punpcklbw %%mm7, %%mm0 \n\t"\
- "punpckhbw %%mm7, %%mm1 \n\t"\
- "punpcklbw %%mm7, %%mm2 \n\t"\
- "punpckhbw %%mm7, %%mm3 \n\t"\
- "movq %%mm0, (%1) \n\t"\
- "movq %%mm1, 17*8(%1) \n\t"\
- "movq %%mm2, (%1, %4) \n\t"\
- "movq %%mm3, (%1, %5) \n\t"\
- "addl $8, %1 \n\t"\
- "addl %3, %0 \n\t"\
- "decl %2 \n\t"\
- " jnz 1b \n\t"\
- : "+r" (src), "+r" (temp_ptr), "+r"(count)\
- : "r" (srcStride), "r"(2*8*17), "r"(3*8*17)\
- );\
- \
- temp_ptr= temp;\
- count=4;\
- \
-/*FIXME reorder for speed */\
-/*FIXME remove push/pop gcc 2.95 bug workaround here and in the other 3 lowpass filters */\
- asm volatile(\
- /*"pxor %%mm7, %%mm7 \n\t"*/\
- "pushl %0 \n\t"\
- "pushl %1 \n\t"\
- "pushl %2 \n\t"\
- "1: \n\t"\
- "movq (%0), %%mm0 \n\t"\
- "movq 8(%0), %%mm1 \n\t"\
- "movq 16(%0), %%mm2 \n\t"\
- "movq 24(%0), %%mm3 \n\t"\
- QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
- QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
- \
- QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
- QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
- QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
- QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
- QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t"\
- QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
- \
- QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
- "addl %4, %1 \n\t" \
- QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
- QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
- \
- "addl $136, %0 \n\t"\
- "addl %8, %1 \n\t"\
- "decl %2 \n\t"\
- " jnz 1b \n\t"\
- "popl %2 \n\t"\
- "popl %1 \n\t"\
- "popl %0 \n\t"\
- \
- :: "r"(temp_ptr), "r"(dst), "r"(count),\
- "r"(dstStride), "r"(2*dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-14*dstStride)\
- );\
-}\
void OPNAME ## mpeg4_qpel8_h_lowpass_mmx2(uint8_t *dst, uint8_t *src, int dstStride, int srcStride, int h){\
uint64_t temp;\
\
"paddw %%mm1, %%mm3 \n\t" /* 20a - 6b + 3c - d */\
"psraw $5, %%mm3 \n\t"\
"packuswb %%mm3, %%mm0 \n\t"\
- OP(%%mm0, (%1), %%mm4, q)\
+ OP_MMX2(%%mm0, (%1), %%mm4, q)\
\
"addl %3, %0 \n\t"\
"addl %4, %1 \n\t"\
"psraw $5, %%mm0 \n\t"\
"psraw $5, %%mm1 \n\t"\
"packuswb %%mm1, %%mm0 \n\t"\
- OP(%%mm0, (%1), %%mm1, q)\
+ OP_3DNOW(%%mm0, (%1), %%mm1, q)\
:: "r"(temp), "r"(dst), "m"(ROUNDER)\
);\
dst+=dstStride;\
src+=srcStride;\
}\
+}
+
+#define QPEL_OP(OPNAME, ROUNDER, RND, OP, MMX)\
+\
+static void OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+ uint64_t temp[17*4];\
+ uint64_t *temp_ptr= temp;\
+ int count= 17;\
+\
+ /*FIXME unroll */\
+ asm volatile(\
+ "pxor %%mm7, %%mm7 \n\t"\
+ "1: \n\t"\
+ "movq (%0), %%mm0 \n\t"\
+ "movq (%0), %%mm1 \n\t"\
+ "movq 8(%0), %%mm2 \n\t"\
+ "movq 8(%0), %%mm3 \n\t"\
+ "punpcklbw %%mm7, %%mm0 \n\t"\
+ "punpckhbw %%mm7, %%mm1 \n\t"\
+ "punpcklbw %%mm7, %%mm2 \n\t"\
+ "punpckhbw %%mm7, %%mm3 \n\t"\
+ "movq %%mm0, (%1) \n\t"\
+ "movq %%mm1, 17*8(%1) \n\t"\
+ "movq %%mm2, (%1, %4) \n\t"\
+ "movq %%mm3, (%1, %5) \n\t"\
+ "addl $8, %1 \n\t"\
+ "addl %3, %0 \n\t"\
+ "decl %2 \n\t"\
+ " jnz 1b \n\t"\
+ : "+r" (src), "+r" (temp_ptr), "+r"(count)\
+ : "r" (srcStride), "r"(2*8*17), "r"(3*8*17)\
+ );\
+ \
+ temp_ptr= temp;\
+ count=4;\
+ \
+/*FIXME reorder for speed */\
+/*FIXME remove push/pop gcc 2.95 bug workaround here and in the other 3 lowpass filters */\
+ asm volatile(\
+ /*"pxor %%mm7, %%mm7 \n\t"*/\
+ "pushl %0 \n\t"\
+ "pushl %1 \n\t"\
+ "pushl %2 \n\t"\
+ "1: \n\t"\
+ "movq (%0), %%mm0 \n\t"\
+ "movq 8(%0), %%mm1 \n\t"\
+ "movq 16(%0), %%mm2 \n\t"\
+ "movq 24(%0), %%mm3 \n\t"\
+ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 16(%0), 8(%0), (%0), 32(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 8(%0), (%0), (%0), 40(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, (%0), (%0), 8(%0), 48(%0), (%1), OP)\
+ \
+ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, (%0), 8(%0), 16(%0), 56(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 8(%0), 16(%0), 24(%0), 64(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 16(%0), 24(%0), 32(%0), 72(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 24(%0), 32(%0), 40(%0), 80(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 32(%0), 40(%0), 48(%0), 88(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 40(%0), 48(%0), 56(%0), 96(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 48(%0), 56(%0), 64(%0),104(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 56(%0), 64(%0), 72(%0),112(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 64(%0), 72(%0), 80(%0),120(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t"\
+ QPEL_V_LOW(%%mm0, %%mm1, %%mm2, %%mm3, %5, %6, %7, 72(%0), 80(%0), 88(%0),128(%0), (%1), OP)\
+ \
+ QPEL_V_LOW(%%mm1, %%mm2, %%mm3, %%mm0, %5, %6, %7, 80(%0), 88(%0), 96(%0),128(%0), (%1, %3), OP)\
+ "addl %4, %1 \n\t" \
+ QPEL_V_LOW(%%mm2, %%mm3, %%mm0, %%mm1, %5, %6, %7, 88(%0), 96(%0),104(%0),120(%0), (%1), OP)\
+ QPEL_V_LOW(%%mm3, %%mm0, %%mm1, %%mm2, %5, %6, %7, 96(%0),104(%0),112(%0),112(%0), (%1, %3), OP)\
+ \
+ "addl $136, %0 \n\t"\
+ "addl %8, %1 \n\t"\
+ "decl %2 \n\t"\
+ " jnz 1b \n\t"\
+ "popl %2 \n\t"\
+ "popl %1 \n\t"\
+ "popl %0 \n\t"\
+ \
+ :: "r"(temp_ptr), "r"(dst), "r"(count),\
+ "r"(dstStride), "r"(2*dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-14*dstStride)\
+ );\
}\
\
-void OPNAME ## mpeg4_qpel8_v_lowpass_mmx(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
+void OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(uint8_t *dst, uint8_t *src, int dstStride, int srcStride){\
uint64_t temp[9*4];\
uint64_t *temp_ptr= temp;\
int count= 9;\
:: "r"(temp_ptr), "r"(dst), "r"(count),\
"r"(dstStride), "r"(2*dstStride), "m"(ff_pw_20), "m"(ff_pw_3), "m"(ROUNDER), "g"(4-6*dstStride)\
);\
-}
-
-#define QPEL_OP(OPNAME, ROUNDER, RND, OP, MMX)\
+}\
\
static void OPNAME ## qpel8_mc00_ ## MMX (UINT8 *dst, UINT8 *src, int stride){\
- put_pixels8_mmx(dst, src, stride, 8);\
+ OPNAME ## pixels8_mmx(dst, src, stride, 8);\
}\
\
static void OPNAME ## qpel8_mc10_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
static void OPNAME ## qpel8_mc01_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t temp[32];\
uint8_t * const half= (uint8_t*)temp;\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(half, src, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);\
OPNAME ## pixels8_l2_mmx(dst, src, half, stride, stride, 8);\
}\
\
static void OPNAME ## qpel8_mc02_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
- OPNAME ## mpeg4_qpel8_v_lowpass_mmx(dst, src, stride, stride);\
+ OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, src, stride, stride);\
}\
\
static void OPNAME ## qpel8_mc03_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t temp[32];\
uint8_t * const half= (uint8_t*)temp;\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(half, src, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(half, src, 8, stride);\
OPNAME ## pixels8_l2_mmx(dst, src+stride, half, stride, stride, 8);\
}\
static void OPNAME ## qpel8_mc11_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l4_mmx(dst, src, (uint8_t*)half, stride, 8);\
}\
static void OPNAME ## qpel8_mc31_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src+1, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src+1, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l4_mmx(dst, src+1, (uint8_t*)half, stride, 8);\
}\
static void OPNAME ## qpel8_mc13_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l4_mmx(dst, src+stride, (uint8_t*)half, stride, 8);\
}\
static void OPNAME ## qpel8_mc33_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src , 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src+1, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src+1, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l4_mmx(dst, src+stride+1, (uint8_t*)half, stride, 8);\
}\
static void OPNAME ## qpel8_mc21_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfH= ((uint8_t*)half) + 64;\
uint8_t * const halfHV= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l2_mmx(dst, halfH, halfHV, stride, 8, 8);\
}\
static void OPNAME ## qpel8_mc23_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfH= ((uint8_t*)half) + 64;\
uint8_t * const halfHV= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l2_mmx(dst, halfH+8, halfHV, stride, 8, 8);\
}\
static void OPNAME ## qpel8_mc12_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l2_mmx(dst, halfV, halfHV, stride, 8, 8);\
}\
static void OPNAME ## qpel8_mc32_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 64;\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfV, src+1, 8, stride);\
- put ## RND ## mpeg4_qpel8_v_lowpass_mmx(halfHV, halfH, 8, 8);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfV, src+1, 8, stride);\
+ put ## RND ## mpeg4_qpel8_v_lowpass_ ## MMX(halfHV, halfH, 8, 8);\
OPNAME ## pixels8_l2_mmx(dst, halfV, halfHV, stride, 8, 8);\
}\
static void OPNAME ## qpel8_mc22_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t half[9*2];\
uint8_t * const halfH= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel8_h_lowpass_ ## MMX(halfH, src, 8, stride, 9);\
- OPNAME ## mpeg4_qpel8_v_lowpass_mmx(dst, halfH, stride, 8);\
+ OPNAME ## mpeg4_qpel8_v_lowpass_ ## MMX(dst, halfH, stride, 8);\
}\
static void OPNAME ## qpel16_mc00_ ## MMX (UINT8 *dst, UINT8 *src, int stride){\
- put_pixels16_mmx(dst, src, stride, 16);\
+ OPNAME ## pixels16_mmx(dst, src, stride, 16);\
}\
\
static void OPNAME ## qpel16_mc10_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
static void OPNAME ## qpel16_mc01_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t temp[32];\
uint8_t * const half= (uint8_t*)temp;\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(half, src, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, stride);\
OPNAME ## pixels16_l2_mmx(dst, src, half, stride, stride, 16);\
}\
\
static void OPNAME ## qpel16_mc02_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
- OPNAME ## mpeg4_qpel16_v_lowpass_mmx(dst, src, stride, stride);\
+ OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, src, stride, stride);\
}\
\
static void OPNAME ## qpel16_mc03_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t temp[32];\
uint8_t * const half= (uint8_t*)temp;\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(half, src, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(half, src, 16, stride);\
OPNAME ## pixels16_l2_mmx(dst, src+stride, half, stride, stride, 16);\
}\
static void OPNAME ## qpel16_mc11_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l4_mmx(dst, src, (uint8_t*)half, stride, 16);\
}\
static void OPNAME ## qpel16_mc31_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src+1, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src+1, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l4_mmx(dst, src+1, (uint8_t*)half, stride, 16);\
}\
static void OPNAME ## qpel16_mc13_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l4_mmx(dst, src+stride, (uint8_t*)half, stride, 16);\
}\
static void OPNAME ## qpel16_mc33_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src , 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src+1, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src+1, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l4_mmx(dst, src+stride+1, (uint8_t*)half, stride, 16);\
}\
static void OPNAME ## qpel16_mc21_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfH= ((uint8_t*)half) + 256;\
uint8_t * const halfHV= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l2_mmx(dst, halfH, halfHV, stride, 16, 16);\
}\
static void OPNAME ## qpel16_mc23_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfH= ((uint8_t*)half) + 256;\
uint8_t * const halfHV= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l2_mmx(dst, halfH+16, halfHV, stride, 16, 16);\
}\
static void OPNAME ## qpel16_mc12_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l2_mmx(dst, halfV, halfHV, stride, 16, 16);\
}\
static void OPNAME ## qpel16_mc32_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint8_t * const halfV= ((uint8_t*)half);\
uint8_t * const halfHV= ((uint8_t*)half) + 256;\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfV, src+1, 16, stride);\
- put ## RND ## mpeg4_qpel16_v_lowpass_mmx(halfHV, halfH, 16, 16);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfV, src+1, 16, stride);\
+ put ## RND ## mpeg4_qpel16_v_lowpass_ ## MMX(halfHV, halfH, 16, 16);\
OPNAME ## pixels16_l2_mmx(dst, halfV, halfHV, stride, 16, 16);\
}\
static void OPNAME ## qpel16_mc22_ ## MMX(UINT8 *dst, UINT8 *src, int stride){\
uint64_t half[17*2];\
uint8_t * const halfH= ((uint8_t*)half);\
put ## RND ## mpeg4_qpel16_h_lowpass_ ## MMX(halfH, src, 16, stride, 17);\
- OPNAME ## mpeg4_qpel16_v_lowpass_mmx(dst, halfH, stride, 16);\
+ OPNAME ## mpeg4_qpel16_v_lowpass_ ## MMX(dst, halfH, stride, 16);\
}
#define PUT_OP(a,b,temp, size) "mov" #size " " #a ", " #b " \n\t"
-#define AVG_OP(a,b,temp, size) \
+#define AVG_3DNOW_OP(a,b,temp, size) \
"mov" #size " " #b ", " #temp " \n\t"\
"pavgusb " #temp ", " #a " \n\t"\
"mov" #size " " #a ", " #b " \n\t"
-
-QPEL_BASE(put_ , ff_pw_16, _ , PUT_OP)
-QPEL_BASE(avg_ , ff_pw_16, _ , AVG_OP)
-QPEL_BASE(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP)
-QPEL_OP(put_ , ff_pw_16, _ , PUT_OP, 3dnow)
-QPEL_OP(avg_ , ff_pw_16, _ , AVG_OP, 3dnow)
-QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, 3dnow)
-
-#undef AVG_OP
-#define AVG_OP(a,b,temp, size) \
+#define AVG_MMX2_OP(a,b,temp, size) \
"mov" #size " " #b ", " #temp " \n\t"\
"pavgb " #temp ", " #a " \n\t"\
"mov" #size " " #a ", " #b " \n\t"
+
+QPEL_BASE(put_ , ff_pw_16, _ , PUT_OP, PUT_OP)
+QPEL_BASE(avg_ , ff_pw_16, _ , AVG_MMX2_OP, AVG_3DNOW_OP)
+QPEL_BASE(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, PUT_OP)
+QPEL_OP(put_ , ff_pw_16, _ , PUT_OP, 3dnow)
+QPEL_OP(avg_ , ff_pw_16, _ , AVG_3DNOW_OP, 3dnow)
+QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, 3dnow)
QPEL_OP(put_ , ff_pw_16, _ , PUT_OP, mmx2)
-QPEL_OP(avg_ , ff_pw_16, _ , AVG_OP, mmx2)
+QPEL_OP(avg_ , ff_pw_16, _ , AVG_MMX2_OP, mmx2)
QPEL_OP(put_no_rnd_, ff_pw_15, _no_rnd_, PUT_OP, mmx2)
#if 0
c->avg_pixels_tab[1][1] = avg_pixels8_x2_mmx2;
c->avg_pixels_tab[1][2] = avg_pixels8_y2_mmx2;
c->avg_pixels_tab[1][3] = avg_pixels8_xy2_mmx2;
+
SET_QPEL_FUNC(qpel_pixels_tab[0][ 0], qpel16_mc00_mmx2)
SET_QPEL_FUNC(qpel_pixels_tab[0][ 1], qpel16_mc10_mmx2)
SET_QPEL_FUNC(qpel_pixels_tab[0][ 2], qpel16_mc20_mmx2)
projects / platform / upstream / libav.git / commitdiff