aboutsummaryrefslogtreecommitdiffstats
path: root/libswscale/ppc
diff options
context:
space:
mode:
authorRonald S. Bultje <rsbultje@gmail.com>2011-10-15 16:00:42 -0700
committerRonald S. Bultje <rsbultje@gmail.com>2011-10-22 10:35:14 -0700
commitf48b12e0a6a14c4f176ae0c9527de4cea456d6a2 (patch)
treee4b68ed78a5a2b1e0fed3f7ade87d79ae3de9292 /libswscale/ppc
parent6cacecdca3b5c9fc769bf404fa19ef3597209e46 (diff)
downloadffmpeg-f48b12e0a6a14c4f176ae0c9527de4cea456d6a2.tar.gz
swscale: update altivec yuv2planeX asm to new per-plane API.
Diffstat (limited to 'libswscale/ppc')
-rw-r--r--libswscale/ppc/swscale_altivec.c94
1 files changed, 13 insertions, 81 deletions
diff --git a/libswscale/ppc/swscale_altivec.c b/libswscale/ppc/swscale_altivec.c
index 3bf9458296..87059d9430 100644
--- a/libswscale/ppc/swscale_altivec.c
+++ b/libswscale/ppc/swscale_altivec.c
@@ -94,34 +94,29 @@ altivec_packIntArrayToCharArray(int *val, uint8_t* dest, int dstW)
//FIXME remove the usage of scratch buffers.
static void
-yuv2yuvX_altivec_real(SwsContext *c,
- const int16_t *lumFilter, const int16_t **lumSrc,
- int lumFilterSize, const int16_t *chrFilter,
- const int16_t **chrUSrc, const int16_t **chrVSrc,
- int chrFilterSize, const int16_t **alpSrc,
- uint8_t *dest[4], int dstW, int chrDstW)
+yuv2planeX_altivec(const int16_t *filter, int filterSize,
+ const int16_t **src, uint8_t *dest, int dstW,
+ const uint8_t *dither, int offset)
{
- uint8_t *yDest = dest[0], *uDest = dest[1], *vDest = dest[2];
- const uint8_t *lumDither = c->lumDither8, *chrDither = c->chrDither8;
register int i, j;
{
DECLARE_ALIGNED(16, int, val)[dstW];
for (i=0; i<dstW; i++)
- val[i] = lumDither[i & 7] << 12;
+ val[i] = dither[(i + offset) & 7] << 12;
- for (j = 0; j < lumFilterSize; j++) {
- vector signed short l1, vLumFilter = vec_ld(j << 1, lumFilter);
- vector unsigned char perm, perm0 = vec_lvsl(j << 1, lumFilter);
+ for (j = 0; j < filterSize; j++) {
+ vector signed short l1, vLumFilter = vec_ld(j << 1, filter);
+ vector unsigned char perm, perm0 = vec_lvsl(j << 1, filter);
vLumFilter = vec_perm(vLumFilter, vLumFilter, perm0);
vLumFilter = vec_splat(vLumFilter, 0); // lumFilter[j] is loaded 8 times in vLumFilter
- perm = vec_lvsl(0, lumSrc[j]);
- l1 = vec_ld(0, lumSrc[j]);
+ perm = vec_lvsl(0, src[j]);
+ l1 = vec_ld(0, src[j]);
for (i = 0; i < (dstW - 7); i+=8) {
int offset = i << 2;
- vector signed short l2 = vec_ld((i << 1) + 16, lumSrc[j]);
+ vector signed short l2 = vec_ld((i << 1) + 16, src[j]);
vector signed int v1 = vec_ld(offset, val);
vector signed int v2 = vec_ld(offset + 16, val);
@@ -143,73 +138,10 @@ yuv2yuvX_altivec_real(SwsContext *c,
l1 = l2;
}
for ( ; i < dstW; i++) {
- val[i] += lumSrc[j][i] * lumFilter[j];
+ val[i] += src[j][i] * filter[j];
}
}
- altivec_packIntArrayToCharArray(val, yDest, dstW);
- }
- if (uDest != 0) {
- DECLARE_ALIGNED(16, int, u)[chrDstW];
- DECLARE_ALIGNED(16, int, v)[chrDstW];
-
- for (i=0; i<chrDstW; i++) {
- u[i] = chrDither[i & 7] << 12;
- v[i] = chrDither[(i + 3) & 7] << 12;
- }
-
- for (j = 0; j < chrFilterSize; j++) {
- vector signed short l1, l1_V, vChrFilter = vec_ld(j << 1, chrFilter);
- vector unsigned char perm, perm0 = vec_lvsl(j << 1, chrFilter);
- vChrFilter = vec_perm(vChrFilter, vChrFilter, perm0);
- vChrFilter = vec_splat(vChrFilter, 0); // chrFilter[j] is loaded 8 times in vChrFilter
-
- perm = vec_lvsl(0, chrUSrc[j]);
- l1 = vec_ld(0, chrUSrc[j]);
- l1_V = vec_ld(0, chrVSrc[j]);
-
- for (i = 0; i < (chrDstW - 7); i+=8) {
- int offset = i << 2;
- vector signed short l2 = vec_ld((i << 1) + 16, chrUSrc[j]);
- vector signed short l2_V = vec_ld((i << 1) + 16, chrVSrc[j]);
-
- vector signed int v1 = vec_ld(offset, u);
- vector signed int v2 = vec_ld(offset + 16, u);
- vector signed int v1_V = vec_ld(offset, v);
- vector signed int v2_V = vec_ld(offset + 16, v);
-
- vector signed short ls = vec_perm(l1, l2, perm); // chrUSrc[j][i] ... chrUSrc[j][i+7]
- vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrVSrc[j][i] ... chrVSrc[j][i]
-
- vector signed int i1 = vec_mule(vChrFilter, ls);
- vector signed int i2 = vec_mulo(vChrFilter, ls);
- vector signed int i1_V = vec_mule(vChrFilter, ls_V);
- vector signed int i2_V = vec_mulo(vChrFilter, ls_V);
-
- vector signed int vf1 = vec_mergeh(i1, i2);
- vector signed int vf2 = vec_mergel(i1, i2); // chrUSrc[j][i] * chrFilter[j] ... chrUSrc[j][i+7] * chrFilter[j]
- vector signed int vf1_V = vec_mergeh(i1_V, i2_V);
- vector signed int vf2_V = vec_mergel(i1_V, i2_V); // chrVSrc[j][i] * chrFilter[j] ... chrVSrc[j][i+7] * chrFilter[j]
-
- vector signed int vo1 = vec_add(v1, vf1);
- vector signed int vo2 = vec_add(v2, vf2);
- vector signed int vo1_V = vec_add(v1_V, vf1_V);
- vector signed int vo2_V = vec_add(v2_V, vf2_V);
-
- vec_st(vo1, offset, u);
- vec_st(vo2, offset + 16, u);
- vec_st(vo1_V, offset, v);
- vec_st(vo2_V, offset + 16, v);
-
- l1 = l2;
- l1_V = l2_V;
- }
- for ( ; i < chrDstW; i++) {
- u[i] += chrUSrc[j][i] * chrFilter[j];
- v[i] += chrVSrc[j][i] * chrFilter[j];
- }
- }
- altivec_packIntArrayToCharArray(u, uDest, chrDstW);
- altivec_packIntArrayToCharArray(v, vDest, chrDstW);
+ altivec_packIntArrayToCharArray(val, dest, dstW);
}
}
@@ -405,7 +337,7 @@ void ff_sws_init_swScale_altivec(SwsContext *c)
if (!is16BPS(dstFormat) && !is9_OR_10BPS(dstFormat) &&
dstFormat != PIX_FMT_NV12 && dstFormat != PIX_FMT_NV21 &&
!c->alpPixBuf) {
-// c->yuv2yuvX = yuv2yuvX_altivec_real;
+ c->yuv2planeX = yuv2planeX_altivec;
}
/* The following list of supported dstFormat values should