diff options
Diffstat (limited to 'libavcodec/mips/vp8dsp_mmi.c')
| -rw-r--r-- | libavcodec/mips/vp8dsp_mmi.c | 263 |
1 files changed, 199 insertions, 64 deletions
diff --git a/libavcodec/mips/vp8dsp_mmi.c b/libavcodec/mips/vp8dsp_mmi.c index b352906f5b..327eaf561e 100644 --- a/libavcodec/mips/vp8dsp_mmi.c +++ b/libavcodec/mips/vp8dsp_mmi.c @@ -1128,12 +1128,14 @@ void ff_vp8_luma_dc_wht_dc_mmi(int16_t block[4][4][16], int16_t dc[16]) void ff_vp8_idct_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride) { #if 1 - DECLARE_ALIGNED(8, const uint64_t, ff_ph_4e7b) = {0x4e7b4e7b4e7b4e7bULL}; - DECLARE_ALIGNED(8, const uint64_t, ff_ph_22a3) = {0x22a322a322a322a3ULL}; double ftmp[12]; uint32_t tmp[1]; + union av_intfloat64 ff_ph_4e7b_u; + union av_intfloat64 ff_ph_22a3_u; DECLARE_VAR_LOW32; DECLARE_VAR_ALL64; + ff_ph_4e7b_u.i = 0x4e7b4e7b4e7b4e7bULL; + ff_ph_22a3_u.i = 0x22a322a322a322a3ULL; __asm__ volatile ( "pxor %[ftmp0], %[ftmp0], %[ftmp0] \n\t" @@ -1253,8 +1255,8 @@ void ff_vp8_idct_add_mmi(uint8_t *dst, int16_t block[16], ptrdiff_t stride) [tmp0]"=&r"(tmp[0]) : [dst0]"r"(dst), [dst1]"r"(dst+stride), [dst2]"r"(dst+2*stride), [dst3]"r"(dst+3*stride), - [block]"r"(block), [ff_pw_4]"f"(ff_pw_4), - [ff_ph_4e7b]"f"(ff_ph_4e7b), [ff_ph_22a3]"f"(ff_ph_22a3) + [block]"r"(block), [ff_pw_4]"f"(ff_pw_4.f), + [ff_ph_4e7b]"f"(ff_ph_4e7b_u.f), [ff_ph_22a3]"f"(ff_ph_22a3_u.f) : "memory" ); #else @@ -1595,8 +1597,16 @@ void ff_put_vp8_epel16_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, const uint64_t *filter = fourtap_subpel_filters[mx - 1]; double ftmp[9]; uint32_t tmp[1]; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; mips_reg src1, dst1; DECLARE_VAR_ALL64; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7]; @@ -1644,11 +1654,11 @@ void ff_put_vp8_epel16_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [dst1]"=&r"(dst1), [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -1672,7 +1682,16 @@ void ff_put_vp8_epel8_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, const uint64_t *filter = fourtap_subpel_filters[mx - 1]; double ftmp[9]; uint32_t tmp[1]; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; DECLARE_VAR_ALL64; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7]; @@ -1705,11 +1724,11 @@ void ff_put_vp8_epel8_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, RESTRICT_ASM_ALL64 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -1733,7 +1752,15 @@ void ff_put_vp8_epel4_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, const uint64_t *filter = fourtap_subpel_filters[mx - 1]; double ftmp[6]; uint32_t tmp[1]; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; DECLARE_VAR_LOW32; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[-1] + filter[3] * src[1] - filter[4] * src[2] + 64) >> 7]; @@ -1760,11 +1787,11 @@ void ff_put_vp8_epel4_h4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, RESTRICT_ASM_LOW32 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -1789,7 +1816,19 @@ void ff_put_vp8_epel16_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[9]; uint32_t tmp[1]; mips_reg src1, dst1; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_ALL64; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[ 0] = cm[(filter[2]*src[ 0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[ 1] - filter[4]*src[ 2] + filter[5]*src[ 3] + 64) >> 7]; @@ -1837,12 +1876,12 @@ void ff_put_vp8_epel16_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [dst1]"=&r"(dst1), [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -1866,7 +1905,19 @@ void ff_put_vp8_epel8_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, const uint64_t *filter = fourtap_subpel_filters[mx - 1]; double ftmp[9]; uint32_t tmp[1]; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_ALL64; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[0] = cm[(filter[2]*src[0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[1] - filter[4]*src[2] + filter[5]*src[ 3] + 64) >> 7]; @@ -1899,12 +1950,12 @@ void ff_put_vp8_epel8_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, RESTRICT_ASM_ALL64 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -1928,7 +1979,19 @@ void ff_put_vp8_epel4_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, const uint64_t *filter = fourtap_subpel_filters[mx - 1]; double ftmp[6]; uint32_t tmp[1]; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_LOW32; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[0] = cm[(filter[2]*src[0] - filter[1]*src[-1] + filter[0]*src[-2] + filter[3]*src[1] - filter[4]*src[2] + filter[5]*src[ 3] + 64) >> 7]; @@ -1955,12 +2018,12 @@ void ff_put_vp8_epel4_h6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, RESTRICT_ASM_LOW32 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -1985,7 +2048,15 @@ void ff_put_vp8_epel16_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[9]; uint32_t tmp[1]; mips_reg src0, src1, dst0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; DECLARE_VAR_ALL64; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7]; @@ -2034,11 +2105,11 @@ void ff_put_vp8_epel16_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -2063,7 +2134,15 @@ void ff_put_vp8_epel8_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[9]; uint32_t tmp[1]; mips_reg src1; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; DECLARE_VAR_ALL64; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7]; @@ -2097,11 +2176,11 @@ void ff_put_vp8_epel8_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -2126,7 +2205,15 @@ void ff_put_vp8_epel4_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[6]; uint32_t tmp[1]; mips_reg src1; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; DECLARE_VAR_LOW32; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; /* dst[0] = cm[(filter[2] * src[0] - filter[1] * src[ -srcstride] + filter[3] * src[ srcstride] - filter[4] * src[ 2*srcstride] + 64) >> 7]; @@ -2154,11 +2241,11 @@ void ff_put_vp8_epel4_v4_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter1]"f"(filter[1]), [filter2]"f"(filter[2]), - [filter3]"f"(filter[3]), [filter4]"f"(filter[4]) + [filter1]"f"(filter1.f), [filter2]"f"(filter2.f), + [filter3]"f"(filter3.f), [filter4]"f"(filter4.f) : "memory" ); #else @@ -2183,7 +2270,19 @@ void ff_put_vp8_epel16_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[9]; uint32_t tmp[1]; mips_reg src0, src1, dst0; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_ALL64; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7]; @@ -2232,12 +2331,12 @@ void ff_put_vp8_epel16_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -2262,7 +2361,19 @@ void ff_put_vp8_epel8_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[9]; uint32_t tmp[1]; mips_reg src1; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_ALL64; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7]; @@ -2296,12 +2407,12 @@ void ff_put_vp8_epel8_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -2326,7 +2437,19 @@ void ff_put_vp8_epel4_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, double ftmp[6]; uint32_t tmp[1]; mips_reg src1; + union av_intfloat64 filter0; + union av_intfloat64 filter1; + union av_intfloat64 filter2; + union av_intfloat64 filter3; + union av_intfloat64 filter4; + union av_intfloat64 filter5; DECLARE_VAR_LOW32; + filter0.i = filter[0]; + filter1.i = filter[1]; + filter2.i = filter[2]; + filter3.i = filter[3]; + filter4.i = filter[4]; + filter5.i = filter[5]; /* dst[0] = cm[(filter[2]*src[0] - filter[1]*src[0-srcstride] + filter[0]*src[0-2*srcstride] + filter[3]*src[0+srcstride] - filter[4]*src[0+2*srcstride] + filter[5]*src[0+3*srcstride] + 64) >> 7]; @@ -2354,12 +2477,12 @@ void ff_put_vp8_epel4_v6_mmi(uint8_t *dst, ptrdiff_t dststride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src) - : [ff_pw_64]"f"(ff_pw_64), + : [ff_pw_64]"f"(ff_pw_64.f), [srcstride]"r"((mips_reg)srcstride), [dststride]"r"((mips_reg)dststride), - [filter0]"f"(filter[0]), [filter1]"f"(filter[1]), - [filter2]"f"(filter[2]), [filter3]"f"(filter[3]), - [filter4]"f"(filter[4]), [filter5]"f"(filter[5]) + [filter0]"f"(filter0.f), [filter1]"f"(filter1.f), + [filter2]"f"(filter2.f), [filter3]"f"(filter3.f), + [filter4]"f"(filter4.f), [filter5]"f"(filter5.f) : "memory" ); #else @@ -2847,11 +2970,13 @@ void ff_put_vp8_bilinear16_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int a = 8 - mx, b = mx; + union mmi_intfloat64 a, b; double ftmp[7]; uint32_t tmp[1]; mips_reg dst0, src0; DECLARE_VAR_ALL64; + a.i = 8 - mx; + b.i = mx; /* dst[0] = (a * src[0] + b * src[1] + 4) >> 3; @@ -2900,10 +3025,10 @@ void ff_put_vp8_bilinear16_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, [dst0]"=&r"(dst0), [src0]"=&r"(src0), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [a]"+&f"(a), [b]"+&f"(b) + [a]"+&f"(a.f), [b]"+&f"(b.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else @@ -2923,11 +3048,13 @@ void ff_put_vp8_bilinear16_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int c = 8 - my, d = my; + union mmi_intfloat64 c, d; double ftmp[7]; uint32_t tmp[1]; mips_reg src0, src1, dst0; DECLARE_VAR_ALL64; + c.i = 8 - my; + d.i = my; /* dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3; @@ -2968,10 +3095,10 @@ void ff_put_vp8_bilinear16_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [c]"+&f"(c), [d]"+&f"(d) + [c]"+&f"(c.f), [d]"+&f"(d.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else @@ -3025,10 +3152,12 @@ void ff_put_vp8_bilinear8_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int a = 8 - mx, b = mx; + union mmi_intfloat64 a, b; double ftmp[7]; uint32_t tmp[1]; DECLARE_VAR_ALL64; + a.i = 8 - mx; + b.i = mx; /* dst[0] = (a * src[0] + b * src[1] + 4) >> 3; @@ -3062,10 +3191,10 @@ void ff_put_vp8_bilinear8_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, RESTRICT_ASM_ALL64 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [a]"+&f"(a), [b]"+&f"(b) + [a]"+&f"(a.f), [b]"+&f"(b.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else @@ -3085,11 +3214,13 @@ void ff_put_vp8_bilinear8_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int c = 8 - my, d = my; + union mmi_intfloat64 c, d; double ftmp[7]; uint32_t tmp[1]; mips_reg src1; DECLARE_VAR_ALL64; + c.i = 8 - my; + d.i = my; /* dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3; @@ -3124,10 +3255,10 @@ void ff_put_vp8_bilinear8_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [c]"+&f"(c), [d]"+&f"(d) + [c]"+&f"(c.f), [d]"+&f"(d.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else @@ -3181,11 +3312,13 @@ void ff_put_vp8_bilinear4_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int a = 8 - mx, b = mx; + union mmi_intfloat64 a, b; double ftmp[5]; uint32_t tmp[1]; DECLARE_VAR_LOW32; DECLARE_VAR_ALL64; + a.i = 8 - mx; + b.i = mx; /* dst[0] = (a * src[0] + b * src[1] + 4) >> 3; @@ -3215,10 +3348,10 @@ void ff_put_vp8_bilinear4_h_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, RESTRICT_ASM_ALL64 [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [a]"+&f"(a), [b]"+&f"(b) + [a]"+&f"(a.f), [b]"+&f"(b.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else @@ -3238,12 +3371,14 @@ void ff_put_vp8_bilinear4_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, ptrdiff_t sstride, int h, int mx, int my) { #if 1 - int c = 8 - my, d = my; + union mmi_intfloat64 c, d; double ftmp[7]; uint32_t tmp[1]; mips_reg src1; DECLARE_VAR_LOW32; DECLARE_VAR_ALL64; + c.i = 8 - my; + d.i = my; /* dst[0] = (c * src[0] + d * src[ sstride] + 4) >> 3; @@ -3274,10 +3409,10 @@ void ff_put_vp8_bilinear4_v_mmi(uint8_t *dst, ptrdiff_t dstride, uint8_t *src, [src1]"=&r"(src1), [h]"+&r"(h), [dst]"+&r"(dst), [src]"+&r"(src), - [c]"+&f"(c), [d]"+&f"(d) + [c]"+&f"(c.f), [d]"+&f"(d.f) : [sstride]"r"((mips_reg)sstride), [dstride]"r"((mips_reg)dstride), - [ff_pw_4]"f"(ff_pw_4) + [ff_pw_4]"f"(ff_pw_4.f) : "memory" ); #else |