| 1 |
/* { dg-do compile } */ |
| 2 |
/* { dg-skip-if "" { powerpc*-*-darwin* } { "*" } { "" } } */ |
| 3 |
/* { dg-options "-O3 -mvsx" } */ |
| 4 |
/* { dg-additional-options "-mdejagnu-cpu=power8" { target { ! has_arch_pwr8 } } } */ |
| 5 |
/* { dg-require-effective-target p8vector_hw } */ |
| 6 |
|
| 7 |
#define NO_WARN_X86_INTRINSICS 1 |
| 8 |
|
| 9 |
#include <mmintrin.h> |
| 10 |
#include <stddef.h> |
| 11 |
#include <stdint.h> |
| 12 |
|
| 13 |
#if 0 |
| 14 |
extern const uint64_t ff_bone; |
| 15 |
#endif |
| 16 |
|
| 17 |
static inline void transpose4x4(uint8_t *dst, uint8_t *src, ptrdiff_t dst_stride, ptrdiff_t src_stride) { |
| 18 |
__m64 row0 = _mm_cvtsi32_si64(*(unsigned*)(src + (0 * src_stride))); |
| 19 |
__m64 row1 = _mm_cvtsi32_si64(*(unsigned*)(src + (1 * src_stride))); |
| 20 |
__m64 row2 = _mm_cvtsi32_si64(*(unsigned*)(src + (2 * src_stride))); |
| 21 |
__m64 row3 = _mm_cvtsi32_si64(*(unsigned*)(src + (3 * src_stride))); |
| 22 |
__m64 tmp0 = _mm_unpacklo_pi8(row0, row1); |
| 23 |
__m64 tmp1 = _mm_unpacklo_pi8(row2, row3); |
| 24 |
__m64 row01 = _mm_unpacklo_pi16(tmp0, tmp1); |
| 25 |
__m64 row23 = _mm_unpackhi_pi16(tmp0, tmp1); |
| 26 |
*((unsigned*)(dst + (0 * dst_stride))) = _mm_cvtsi64_si32(row01); |
| 27 |
*((unsigned*)(dst + (1 * dst_stride))) = _mm_cvtsi64_si32(_mm_unpackhi_pi32(row01, row01)); |
| 28 |
*((unsigned*)(dst + (2 * dst_stride))) = _mm_cvtsi64_si32(row23); |
| 29 |
*((unsigned*)(dst + (3 * dst_stride))) = _mm_cvtsi64_si32(_mm_unpackhi_pi32(row23, row23)); |
| 30 |
} |
| 31 |
|
| 32 |
#if 0 |
| 33 |
static inline void h264_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha1, int beta1) |
| 34 |
{ |
| 35 |
asm volatile( |
| 36 |
"" |
| 37 |
:: "r"(pix-2*stride), "r"(pix), "r"((long)stride), |
| 38 |
"m"(alpha1), "m"(beta1), "m"(ff_bone) |
| 39 |
); |
| 40 |
} |
| 41 |
#endif |
| 42 |
|
| 43 |
void h264_h_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha, int beta) |
| 44 |
{ |
| 45 |
uint8_t trans[8*4] __attribute__ ((aligned (8))); |
| 46 |
transpose4x4(trans, pix-2, 8, stride); |
| 47 |
transpose4x4(trans+4, pix-2+4*stride, 8, stride); |
| 48 |
// h264_loop_filter_chroma_intra_mmx2(trans+2*8, 8, alpha-1, beta-1); |
| 49 |
transpose4x4(pix-2, trans, stride, 8); |
| 50 |
transpose4x4(pix-2+4*stride, trans+4, stride, 8); |
| 51 |
} |