1/* { dg-do compile } */
2/* { dg-options "-O1 -mmmx" } */
3/* { dg-skip-if "no stdint" { vxworks_kernel } } */
4
5#include <mmintrin.h>
6#include <stddef.h>
7#include <stdint.h>
8
9extern const uint64_t ff_bone;
10
11static inline void transpose4x4(uint8_t *dst, uint8_t *src, ptrdiff_t dst_stride, ptrdiff_t src_stride) {
12    __m64 row0 = _mm_cvtsi32_si64(*(unsigned*)(src + (0 * src_stride)));
13    __m64 row1 = _mm_cvtsi32_si64(*(unsigned*)(src + (1 * src_stride)));
14    __m64 row2 = _mm_cvtsi32_si64(*(unsigned*)(src + (2 * src_stride)));
15    __m64 row3 = _mm_cvtsi32_si64(*(unsigned*)(src + (3 * src_stride)));
16    __m64 tmp0 = _mm_unpacklo_pi8(row0, row1);
17    __m64 tmp1 = _mm_unpacklo_pi8(row2, row3);
18    __m64 row01 = _mm_unpacklo_pi16(tmp0, tmp1);
19    __m64 row23 = _mm_unpackhi_pi16(tmp0, tmp1);
20    *((unsigned*)(dst + (0 * dst_stride))) = _mm_cvtsi64_si32(row01);
21    *((unsigned*)(dst + (1 * dst_stride))) = _mm_cvtsi64_si32(_mm_unpackhi_pi32(row01, row01));
22    *((unsigned*)(dst + (2 * dst_stride))) = _mm_cvtsi64_si32(row23);
23    *((unsigned*)(dst + (3 * dst_stride))) = _mm_cvtsi64_si32(_mm_unpackhi_pi32(row23, row23));
24}
25
26static inline void h264_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha1, int beta1)
27{
28    asm volatile(
29        ""
30        :: "r"(pix-2*stride), "r"(pix), "r"((long)stride),
31           "m"(alpha1), "m"(beta1), "m"(ff_bone)
32    );
33}
34
35void h264_h_loop_filter_chroma_intra_mmx2(uint8_t *pix, int stride, int alpha, int beta)
36{
37
38    uint8_t trans[8*4] __attribute__ ((aligned (8)));
39    transpose4x4(trans, pix-2, 8, stride);
40    transpose4x4(trans+4, pix-2+4*stride, 8, stride);
41    h264_loop_filter_chroma_intra_mmx2(trans+2*8, 8, alpha-1, beta-1);
42    transpose4x4(pix-2, trans, stride, 8);
43    transpose4x4(pix-2+4*stride, trans+4, stride, 8);
44}
45