Commit 5edb653b authored by Alan Curry's avatar Alan Curry Committed by Diego Biurrun

AltiVec operations need to have memory aligned on 16-byte boundaries.

patch by Alan Curry, pacman at world dot std dot com

Originally committed as revision 17559 to svn://svn.mplayerhq.hu/mplayer/trunk/postproc
parent be9d060d
......@@ -1166,7 +1166,8 @@ static inline void initFilter(int16_t **outFilter, int16_t **filterPos, int *out
}
// Note the +1 is for the MMXscaler which reads over the end
*outFilter= (int16_t*)memalign(8, *outFilterSize*(dstW+1)*sizeof(int16_t));
/* align at 16 for AltiVec (needed by hScale_altivec_real) */
*outFilter= (int16_t*)memalign(16, *outFilterSize*(dstW+1)*sizeof(int16_t));
memset(*outFilter, 0, *outFilterSize*(dstW+1)*sizeof(int16_t));
/* Normalize & Store in outFilter */
......@@ -2132,10 +2133,11 @@ SwsContext *sws_getContext(int srcW, int srcH, int origSrcFormat, int dstW, int
c->lumPixBuf= (int16_t**)memalign(4, c->vLumBufSize*2*sizeof(int16_t*));
c->chrPixBuf= (int16_t**)memalign(4, c->vChrBufSize*2*sizeof(int16_t*));
//Note we need at least one pixel more at the end because of the mmx code (just in case someone wanna replace the 4000/8000)
/* align at 16 bytes for AltiVec */
for(i=0; i<c->vLumBufSize; i++)
c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= (uint16_t*)memalign(8, 4000);
c->lumPixBuf[i]= c->lumPixBuf[i+c->vLumBufSize]= (uint16_t*)memalign(16, 4000);
for(i=0; i<c->vChrBufSize; i++)
c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= (uint16_t*)memalign(8, 8000);
c->chrPixBuf[i]= c->chrPixBuf[i+c->vChrBufSize]= (uint16_t*)memalign(16, 8000);
//try to avoid drawing green stuff between the right end and the stride end
for(i=0; i<c->vLumBufSize; i++) memset(c->lumPixBuf[i], 0, 4000);
......
......@@ -68,6 +68,9 @@
#include <inttypes.h>
#include <assert.h>
#include "config.h"
#ifdef HAVE_MALLOC_H
#include <malloc.h>
#endif
#include "rgb2rgb.h"
#include "swscale.h"
#include "swscale_internal.h"
......@@ -788,8 +791,8 @@ altivec_yuv2packedX (SwsContext *c,
vector signed short *YCoeffs, *CCoeffs;
vYCoeffsBank = malloc (sizeof (vector signed short)*lumFilterSize*c->dstH);
vCCoeffsBank = malloc (sizeof (vector signed short)*chrFilterSize*c->dstH);
vYCoeffsBank = memalign (16, sizeof (vector signed short)*lumFilterSize*c->dstH);
vCCoeffsBank = memalign (16, sizeof (vector signed short)*chrFilterSize*c->dstH);
for (i=0;i<lumFilterSize*c->dstH;i++) {
tmp = c->vLumFilter[i];
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment