generic-library/ffmpeg
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

sws-PPC: fix after VOFW change.

Browse Source 
It seems sws-PPC did hardcode 2048 at various places instead of using VOFW.
This also means that all past VOFW benchmarks on PPC are meaningless
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This commit is contained in:
Michael Niedermayer 2011-05-12 02:26:22 +02:00
parent 83630f79ec
commit 1cafc23288
2 changed files with 6 additions and 6 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
libswscale/ppc/swscale_altivec_template.c
View File

@ -161,12 +161,12 @@ yuv2yuvX_altivec_real(const int16_t *lumFilter, const int16_t **lumSrc, int lumF
perm = vec_lvsl(0, chrSrc[j]); perm = vec_lvsl(0, chrSrc[j]);
l1 = vec_ld(0, chrSrc[j]); l1 = vec_ld(0, chrSrc[j]);
l1_V = vec_ld(2048 << 1, chrSrc[j]); l1_V = vec_ld(VOFW << 1, chrSrc[j]);
for (i = 0; i < (chrDstW - 7); i+=8) { for (i = 0; i < (chrDstW - 7); i+=8) {
int offset = i << 2; int offset = i << 2;
vector signed short l2 = vec_ld((i << 1) + 16, chrSrc[j]); vector signed short l2 = vec_ld((i << 1) + 16, chrSrc[j]);
vector signed short l2_V = vec_ld(((i + 2048) << 1) + 16, chrSrc[j]); vector signed short l2_V = vec_ld(((i + VOFW) << 1) + 16, chrSrc[j]);
vector signed int v1 = vec_ld(offset, u); vector signed int v1 = vec_ld(offset, u);
vector signed int v2 = vec_ld(offset + 16, u); vector signed int v2 = vec_ld(offset + 16, u);
@ -174,7 +174,7 @@ yuv2yuvX_altivec_real(const int16_t *lumFilter, const int16_t **lumSrc, int lumF
vector signed int v2_V = vec_ld(offset + 16, v); vector signed int v2_V = vec_ld(offset + 16, v);
vector signed short ls = vec_perm(l1, l2, perm); // chrSrc[j][i] ... chrSrc[j][i+7] vector signed short ls = vec_perm(l1, l2, perm); // chrSrc[j][i] ... chrSrc[j][i+7]
vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrSrc[j][i+2048] ... chrSrc[j][i+2055] vector signed short ls_V = vec_perm(l1_V, l2_V, perm); // chrSrc[j][i+VOFW] ... chrSrc[j][i+2055]
vector signed int i1 = vec_mule(vChrFilter, ls); vector signed int i1 = vec_mule(vChrFilter, ls);
vector signed int i2 = vec_mulo(vChrFilter, ls); vector signed int i2 = vec_mulo(vChrFilter, ls);
@ -201,7 +201,7 @@ yuv2yuvX_altivec_real(const int16_t *lumFilter, const int16_t **lumSrc, int lumF
} }
for ( ; i < chrDstW; i++) { for ( ; i < chrDstW; i++) {
u[i] += chrSrc[j][i] * chrFilter[j]; u[i] += chrSrc[j][i] * chrFilter[j];
v[i] += chrSrc[j][i + 2048] * chrFilter[j]; v[i] += chrSrc[j][i + VOFW] * chrFilter[j];
} }
} }
altivec_packIntArrayToCharArray(u, uDest, chrDstW); altivec_packIntArrayToCharArray(u, uDest, chrDstW);

4
libswscale/ppc/yuv2rgb_altivec.c
View File

@ -817,7 +817,7 @@ ff_yuv2packedX_altivec(SwsContext *c,
for (j=0; j<chrFilterSize; j++) { for (j=0; j<chrFilterSize; j++) {
X = vec_ld (0, &chrSrc[j][i/2]); X = vec_ld (0, &chrSrc[j][i/2]);
U = vec_mradds (X, CCoeffs[j], U); U = vec_mradds (X, CCoeffs[j], U);
X = vec_ld (0, &chrSrc[j][i/2+2048]); X = vec_ld (0, &chrSrc[j][i/2+VOFW]);
V = vec_mradds (X, CCoeffs[j], V); V = vec_mradds (X, CCoeffs[j], V);
} }
@ -895,7 +895,7 @@ ff_yuv2packedX_altivec(SwsContext *c,
for (j=0; j<chrFilterSize; j++) { for (j=0; j<chrFilterSize; j++) {
X = vec_ld (0, &chrSrc[j][i/2]); X = vec_ld (0, &chrSrc[j][i/2]);
U = vec_mradds (X, CCoeffs[j], U); U = vec_mradds (X, CCoeffs[j], U);
X = vec_ld (0, &chrSrc[j][i/2+2048]); X = vec_ld (0, &chrSrc[j][i/2+VOFW]);
V = vec_mradds (X, CCoeffs[j], V); V = vec_mradds (X, CCoeffs[j], V);
} }