RepoMirrors
/
ffmpeg
mirror of https://git.ffmpeg.org/ffmpeg.git
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

avcodec/ppc/lossless_audiodsp_altivec: POWER LE support for scalarproduct_and_madd_int16_altivec() 

adds macro GET_T()

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Rong Yan 2014-12-02 07:37:11 +00:00 committed by Michael Niedermayer
parent d7716961a8
commit b4d41beebe
1 changed files with 21 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

31
libavcodec/ppc/lossless_audiodsp_altivec.c
View File

@ -29,6 +29,20 @@
#include "libavutil/ppc/types_altivec.h" #include "libavutil/ppc/types_altivec.h"
#include "libavcodec/lossless_audiodsp.h" #include "libavcodec/lossless_audiodsp.h"
#if HAVE_BIGENDIAN
#define GET_T(tt0,tt1,src,a,b){ \
a = vec_ld(16, src); \
tt0 = vec_perm(b, a, align); \
b = vec_ld(32, src); \
tt1 = vec_perm(a, b, align); \
}
#else
#define GET_T(tt0,tt1,src,a,b){ \
tt0 = vec_vsx_ld(0, src); \
tt1 = vec_vsx_ld(16, src); \
}
#endif
#if HAVE_ALTIVEC #if HAVE_ALTIVEC
static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1, static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1,
const int16_t *v2, const int16_t *v2,
@ -38,26 +52,23 @@ static int32_t scalarproduct_and_madd_int16_altivec(int16_t *v1,
LOAD_ZERO; LOAD_ZERO;
vec_s16 *pv1 = (vec_s16 *) v1; vec_s16 *pv1 = (vec_s16 *) v1;
register vec_s16 muls = { mul, mul, mul, mul, mul, mul, mul, mul }; register vec_s16 muls = { mul, mul, mul, mul, mul, mul, mul, mul };
register vec_s16 t0, t1, i0, i1, i4; register vec_s16 t0, t1, i0, i1, i4, i2, i3;
register vec_s16 i2 = vec_ld(0, v2), i3 = vec_ld(0, v3);
register vec_s32 res = zero_s32v; register vec_s32 res = zero_s32v;
#if HAVE_BIGENDIAN
register vec_u8 align = vec_lvsl(0, v2); register vec_u8 align = vec_lvsl(0, v2);
i2 = vec_ld(0, v2);
i3 = vec_ld(0, v3);
#endif
int32_t ires; int32_t ires;
order >>= 4; order >>= 4;
do { do {
i1 = vec_ld(16, v2); GET_T(t0,t1,v2,i1,i2);
t0 = vec_perm(i2, i1, align);
i2 = vec_ld(32, v2);
t1 = vec_perm(i1, i2, align);
i0 = pv1[0]; i0 = pv1[0];
i1 = pv1[1]; i1 = pv1[1];
res = vec_msum(t0, i0, res); res = vec_msum(t0, i0, res);
res = vec_msum(t1, i1, res); res = vec_msum(t1, i1, res);
i4 = vec_ld(16, v3); GET_T(t0,t1,v3,i4,i3);
t0 = vec_perm(i3, i4, align);
i3 = vec_ld(32, v3);
t1 = vec_perm(i4, i3, align);
pv1[0] = vec_mladd(t0, muls, i0); pv1[0] = vec_mladd(t0, muls, i0);
pv1[1] = vec_mladd(t1, muls, i1); pv1[1] = vec_mladd(t1, muls, i1);
pv1 += 2; pv1 += 2;