fdo-mirrors/cairo
1
0
Fork 0
mirror of https://gitlab.freedesktop.org/cairo/cairo.git synced 2026-05-05 07:38:22 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions

Finish up merging the optimization work from jaymz.

Browse source
This commit is contained in:
Jeff Muizelaar 2005-07-30 10:23:21 +00:00
parent bdfb7f4db0
commit 05d84a0a1e
4 changed files with 170 additions and 173 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
pixman/ChangeLog
View file

@ -1,3 +1,13 @@
2005-07-30 Jeff Muizelaar <jeff@infidigm.net>
* src/ic.c: (fbCompositeSolidMask_nx8x0888),
(fbCompositeSolidMask_nx8x0565), (fbCompositeSolidMask_nx1xn),
(fbCompositeTrans_0565xnx0565), (fbCompositeTrans_0888xnx0888),
(fbCompositeSrcSrc_nxn), (pixman_composite):
* src/icblt.c: (fbBlt):
* src/iccompose.c: (fbFetch_r8g8b8):
Finish up merging the optimization work from jaymz.
2005-07-30 Jeff Muizelaar <jeff@infidigm.net>
* src/iccompose.c: (fbCombineMaskAlphaU),

316
pixman/src/ic.c
View file

@ -152,21 +152,25 @@ fbIn24 (CARD32 x, CARD8 y)
#define genericCombine24(a,b,c,d) (((a)*(c)+(b)*(d)))
#define fastcombine32(alpha, source, destval, destptr, dstrb, dstag, drb, dag) \
dstrb=destval&0xFF00FF; dstag=(destval>>8)&0xFF00FF; \
drb=((source&0xFF00FF)-dstrb)*alpha; dag=(((source>>8)&0xFF00FF)-dstag)*alpha; \
#define fastcombine32(alpha, source, destval, destptr) { \
CARD32 dstrb=destval&0xFF00FF; CARD32 dstag=(destval>>8)&0xFF00FF; \
CARD32 drb=((source&0xFF00FF)-dstrb)*alpha; CARD32 dag=(((source>>8)&0xFF00FF)-dstag)*alpha; \
*destptr++=((((drb>>8) + dstrb) & 0x00FF00FF) | ((((dag>>8) + dstag) << 8) & 0xFF00FF00)); \
}
#define fastcombine32(alpha, source, destval, destptr, dstrb, dstag, drb, dag) \
dstrb=destval&0xFF00FF; dstag=(destval>>8)&0xFF00FF; \
drb=((source&0xFF00FF)-dstrb)*alpha; dag=(((source>>8)&0xFF00FF)-dstag)*alpha; \
*destptr++=((((drb>>8) + dstrb) & 0x00FF00FF) | ((((dag>>8) + dstag) << 8) & 0xFF00FF00)); \
// Note: this macro expects 6 bits of alpha, not 8!
/* Note: this macro expects 5 bits of alpha, not 8!
* (single op can actually be changed to take 6 bits, but double op can't)
*/
#define fastCombine0565(alpha, source, destval, destptr) { \
CARD16 dstrb = destval & 0xf81f; CARD16 dstg = destval & 0x7e0; \
CARD32 drb = ((source&0xf81f)-dstrb)*alpha; CARD32 dg=((source & 0x7e0)-dstg)*alpha; \
destptr= ((((drb>>6) + dstrb)&0xf81f) | (((dg>>6) + dstg) & 0x7e0)); \
destptr= ((((drb>>5) + dstrb)&0xf81f) | (((dg>>5) + dstg) & 0x7e0)); \
}
#define fastCombine2x0565(alpha, source, destval, destptr) { \
CARD32 dstrb = destval & 0x07e0f81f; CARD32 dstg = (destval & 0xf81f07e0)>>5; \
CARD32 drb = ((source&0x07e0f81f)-dstrb)*alpha; CARD32 dg=(((source & 0xf81f07e0)>>5)-dstg)*alpha; \
destptr= ((((drb>>5) + dstrb)&0x07e0f81f) | ((((dg>>5) + dstg)<<5) & 0xf81f07e0)); \
}
#if IMAGE_BYTE_ORDER == LSBFirst
@ -371,7 +375,7 @@ fbCompositeSolidMask_nx8x0888 (pixman_operator_t op,
while (height--)
{
// fixme: cleanup unused
/* fixme: cleanup unused */
unsigned int wt,wd;
CARD32 workingiDest;
CARD32 *widst;
@ -454,7 +458,7 @@ fbCompositeSolidMask_nx8x0565 (pixman_operator_t op,
src16 = cvt8888to0565(src);
rsrca = src >> 24;
srca=rsrca>>2;
srca=rsrca>>3;
if (src == 0)
return;
@ -486,7 +490,7 @@ fbCompositeSolidMask_nx8x0565 (pixman_operator_t op,
}
else if (m)
{
na=(rsrca*(int)m)>>10;
na=(rsrca*(int)m)>>11;
d = *dst;
fastCombine0565(na, src16, d, *dst++);
}
@ -924,7 +928,7 @@ fbCompositeSolidMask_nx1xn (pixman_operator_t op,
int dstXoff, dstYoff;
int maskXoff, maskYoff;
FbBits src;
fbComposeGetSolid(pSrc, src);
if ((src & 0xff000000) != 0xff000000)
@ -995,7 +999,7 @@ fbCompositeTrans_0565xnx0565(pixman_operator_t op,
CARD32 s_32, d_32, i_32, r_32;
fbComposeGetSolid (pMask, mask);
maskAlpha = mask >> 26;
maskAlpha = mask >> 27;
if (!maskAlpha)
return;
@ -1012,7 +1016,7 @@ fbCompositeTrans_0565xnx0565(pixman_operator_t op,
while (height--)
{
CARD32 *isrc;
CARD32 *isrc, *idst;
dst = dstLine;
dstLine += dstStride;
src = srcLine;
@ -1027,24 +1031,38 @@ fbCompositeTrans_0565xnx0565(pixman_operator_t op,
w--;
}
isrc=(CARD32 *)src;
while (w>1)
if(((int)dst&1)==0)
{
s_32=*isrc++;
idst=(CARD32 *)dst;
while (w>1)
{
s_32 = *isrc++;
d_32 = *idst;
fastCombine2x0565(maskAlpha,s_32,d_32,*idst++);
w-=2;
}
dst=(CARD16 *)idst;
}
else
{
if(w>1)
{
#if IMAGE_BYTE_ORDER == LSBFirst
s_16=s_32&0xffff;
s_16=s_32&0xffff;
#else
s_16=s_32>>16;
s_16=s_32>>16;
#endif
d_16 = *dst;
fastCombine0565(maskAlpha, s_16, d_16, *dst++);
d_16 = *dst;
fastCombine0565(maskAlpha, s_16, d_16, *dst++);
#if IMAGE_BYTE_ORDER == LSBFirst
s_16=s_32>>16;
s_16=s_32>>16;
#else
s_16=s_32&0xffff;
s_16=s_32&0xffff;
#endif
d_16 = *dst;
fastCombine0565(maskAlpha, s_16, d_16, *dst++);
w-=2;
d_16 = *dst;
fastCombine0565(maskAlpha, s_16, d_16, *dst++);
w-=2;
}
}
src=(CARD16 *)isrc;
if(w!=0)
@ -1058,7 +1076,7 @@ fbCompositeTrans_0565xnx0565(pixman_operator_t op,
// macros for "i can't believe it's not fast" packed pixel handling
/* macros for "i can't believe it's not fast" packed pixel handling */
#define alphamaskCombine24(a,b) genericCombine24(a,b,maskAlpha,maskiAlpha)
static void
fbCompositeTrans_0888xnx0888(pixman_operator_t op,
@ -1087,140 +1105,104 @@ fbCompositeTrans_0888xnx0888(pixman_operator_t op,
if (!maskAlpha)
return;
//if (maskAlpha == 0xff)
//{
//fbCompositeSrc_0888x0888 (op, pSrc, pMask, pDst,
// xSrc, ySrc, xMask, yMask, xDst, yDst,
// width, height);
//return;
//}
/*
if (maskAlpha == 0xff)
{
fbCompositeSrc_0888x0888 (op, pSrc, pMask, pDst,
xSrc, ySrc, xMask, yMask, xDst, yDst,
width, height);
return;
}
*/
fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 3);
fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3);
{
unsigned int ws,wt,wd,ww;
unsigned int ws,wt,ww;
CARD32 workingSource;
CARD32 *wsrc;
CARD32 rs,gs,bs;
CARD32 rd,gd,bd;
CARD32 *wsrc, *wdst, *widst;
CARD32 rs, rd, nd;
CARD8 *isrc;
CARD32 workingiDest,workingoDest;
CARD32 *widst,*wodst;
// are xSrc and xDst at the same alignment? if not, we need to be complicated :)
//if(0==0)
if( (((xSrc*3)&3)!=((xDst*3)&3)) || (srcStride&3)!=0 || (dstStride&3)!=0)
/* are xSrc and xDst at the same alignment? if not, we need to be complicated :)*/
/* if(0==0) */
if( (((xSrc*3)&3)!=((xDst*3)&3)) || ((srcStride&3)!=(dstStride&3)))
{
while (height--)
{
idst=dst = dstLine;
dst = dstLine;
dstLine += dstStride;
src = srcLine;
isrc = src = srcLine;
srcLine += srcStride;
w = width*3;
setupPackedReader(wd,wt,idst,widst,workingiDest);
ww=(int)dst;
wt=ww&3;
dst-=wt;
wodst=(CARD32 *)dst;
workingoDest=*wodst;
ww=4-wt;
#if IMAGE_BYTE_ORDER == LSBFirst
workingoDest<<=(8*(ww+1));
#else
workingoDest>>=(8*(ww+1));
#endif
setupPackedReader(ws,wt,isrc,wsrc,workingSource);
// get to word aligned
switch(!(int)src&3)
/* get to word aligned */
switch(!(int)dst&3)
{
case 1:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
readPackedSource(rs);
/* *dst++=alphamaskCombine24(rs, *dst)>>8; */
rd=*dst; /* make gcc happy. hope it doens't cost us too much performance*/
*dst++=alphamaskCombine24(rs, rd)>>8;
w--; if(w==0) break;
case 2:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
readPackedSource(rs);
rd=*dst;
*dst++=alphamaskCombine24(rs, rd)>>8;
w--; if(w==0) break;
case 3:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
readPackedSource(rs);
rd=*dst;
*dst++=alphamaskCombine24(rs, rd)>>8;
w--; if(w==0) break;
}
wsrc=(CARD32 *)src;
wdst=(CARD32 *)dst;
while (w>3)
{
rs=*wsrc++;
// FIXME: write a version of readPackedDest() which
// can collect 4 bytes at once if we're on a boundry (which we're
// actually guarenteed not to be in this version, but do it anyhow), and can
// collect as 2 16bit words on a 2byte boundry, and then use the 32bit combine here
/* FIXME: write a special readPackedWord macro, which knows how to
* halfword combine
*/
#if IMAGE_BYTE_ORDER == LSBFirst
readPackedDest(rd);
rd=alphamaskCombine24(rs&0xff, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24((rs>>8)&0xff, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24((rs>>16)&0xff, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24(rs>>24, rd)>>8;
writePacked(rd);
rd=*wdst;
readPackedSource(nd);
readPackedSource(rs);
nd|=rs<<8;
readPackedSource(rs);
nd|=rs<<16;
readPackedSource(rs);
nd|=rs<<24;
#else
readPackedDest(rd);
rd=alphamaskCombine24(rs>>24, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24((rs>>16)&0xff, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24((rs>>8)&0xff, rd)>>8;
writePacked(rd);
readPackedDest(rd);
rd=alphamaskCombine24(rs&0xff, rd)>>8;
writePacked(rd);
readPackedSource(nd);
nd<<=24;
readPackedSource(rs);
nd|=rs<<16;
readPackedSource(rs);
nd|=rs<<8;
readPackedSource(rs);
nd|=rs;
#endif
fastcombine32(maskAlpha, nd, rd, wdst)
w-=4;
}
src=(CARD8 *)wsrc;
dst=(CARD8 *)wdst;
switch(w)
{
case 3:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
readPackedSource(rs);
rd=*dst;
*dst++=alphamaskCombine24(rs, rd)>>8;
case 2:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
readPackedSource(rs);
rd=*dst;
*dst++=alphamaskCombine24(rs, rd)>>8;
case 1:
readPackedDest(rd);
rd=alphamaskCombine24(*src++, rd)>>8;
writePacked(rd);
}
dst=(CARD8 *)wodst;
switch(ww)
{
case 1:
dst[2]=(workingoDest>>8)&0xff;
case 2:
dst[1]=(workingoDest>>16)&0xff;
case 3:
dst[0]=workingoDest>>24;
readPackedSource(rs);
rd=*dst;
*dst++=alphamaskCombine24(rs, rd)>>8;
}
}
}
@ -1233,7 +1215,7 @@ fbCompositeTrans_0888xnx0888(pixman_operator_t op,
src = srcLine;
srcLine += srcStride;
w = width*3;
// get to word aligned
/* get to word aligned */
switch(!(int)src&3)
{
case 1:
@ -1252,12 +1234,11 @@ fbCompositeTrans_0888xnx0888(pixman_operator_t op,
wsrc=(CARD32 *)src;
widst=(CARD32 *)dst;
register CARD32 t1, t2, t3, t4;
while(w>3)
{
rs = *wsrc++;
rd = *widst;
fastcombine32(maskAlpha, rs, rd, widst, t1, t2, t3, t4);
fastcombine32(maskAlpha, rs, rd, widst);
w-=4;
}
src=(CARD8 *)wsrc;
@ -1299,33 +1280,41 @@ fbCompositeSrcSrc_nxn (pixman_operator_t op,
{
FbBits *dst;
FbBits *src;
FbStride dstStride, srcStride;
int srcXoff, srcYoff;
int dstXoff, dstYoff;
int srcBpp;
int dstBpp;
// these need to be signed now!
int iwidth=width;
int iheight=height;
Bool reverse = FALSE;
Bool upsidedown = FALSE;
CARD8 *isrc;
CARD8 *idst;
int dstStride, srcStride;
int srcXoff, srcYoff;
int dstXoff, dstYoff;
int srcBpp;
int dstBpp;
/* these need to be signed now! */
int iwidth=width;
int iheight=height;
int bytesPerPixel;
int initialWidth=width;
int initialX=xDst;
// FIXME: this is possibly the worst piece of code I've ever written.
// My main objection to it, is that it is incrfedibly slow in a few cases, due to the
// call-per-repeat structure of it - the *correct* solution is to implement
// repeat into fbBlt(), but that's a nontrivial job, and it's far more
// important to get the "requireRepeat" stuff implented functionally
// first, *then* make it fast.
// -- jj
/* FIXME: this is possibly the second worst piece of code I've ever written.
* (the worst being the previous imlementation of this)
* -- jj
*/
Bool srcRepeat=pSrc->repeat;
CARD32 srcHeight=pSrc->pDrawable->height;
CARD32 srcWidth=pSrc->pDrawable->width;
FbGetPixels(pSrc->pixels,src,srcStride,srcBpp,srcXoff,srcYoff);
FbGetPixels(pDst->pixels,dst,dstStride,dstBpp,dstXoff,dstYoff);
src+=(srcXoff+(srcYoff*srcStride));
dst+=(dstXoff+(dstYoff*dstStride));
isrc=(CARD8 *)src;
idst=(CARD8 *)dst;
bytesPerPixel=(srcBpp>>3);
srcStride*=sizeof(FbBits);
dstStride*=sizeof(FbBits);
if(srcRepeat)
{
xSrc%=srcWidth;
@ -1342,26 +1331,16 @@ fbCompositeSrcSrc_nxn (pixman_operator_t op,
while(iwidth>0)
{
int wwidth=iwidth;
int j;
if(wwidth>(srcWidth-xSrc))
wwidth=(srcWidth-xSrc);
fbBlt (src + (ySrc + srcYoff) * srcStride,
srcStride,
(xSrc + srcXoff) * srcBpp,
dst + (yDst + dstYoff) * dstStride,
dstStride,
(xDst + dstXoff) * dstBpp,
(wwidth) * dstBpp,
(wheight),
GXcopy,
FB_ALLONES,
dstBpp,
reverse,
upsidedown);
for(j=0;j<wheight;j++)
memcpy( idst + (yDst + j) * dstStride + ((xDst * bytesPerPixel)),
isrc + (ySrc + j) * srcStride + ((xSrc * bytesPerPixel)),
(wwidth*bytesPerPixel));
/* reset the xSrc pointer. D'oh. */
xSrc=0;
if(!srcRepeat)
iwidth=0;
else
@ -1370,6 +1349,8 @@ fbCompositeSrcSrc_nxn (pixman_operator_t op,
iwidth-=wwidth;
}
}
ySrc=0;
/* reset the ySrc pointer. D'oh. */
if(!srcRepeat)
iheight=0;
else
@ -1660,9 +1641,10 @@ pixman_composite (pixman_operator_t op,
n = pixman_region_num_rects (region);
pbox = pixman_region_rects (region);
// FIXME: this is bascially a "white list" of composites that work
// with repeat until they are all implented. Once that's done, we
// remove the checks below entirely
/* FIXME: this is bascially a "white list" of composites that work
* with repeat until they are all implented. Once that's done, we
* remove the checks below entirely
*/
if(func==fbCompositeSrcSrc_nxn)
{
srcRepeat=maskRepeat=FALSE;

4
pixman/src/icblt.c
View file

@ -70,8 +70,8 @@ fbBlt (FbBits *srcLine,
*/
if((pm==FB_ALLONES) && (alu==GXcopy) && !reverse && (srcX&7)==0 && (dstX&7)==0 && (width&7)==0)
{
uint8_t *isrc=(uint8_t *)srcLine;
uint8_t *idst=(uint8_t *)dstLine;
CARD8 *isrc=(CARD8 *)srcLine;
CARD8 *idst=(CARD8 *)dstLine;
int sstride=srcStride*sizeof(FbBits);
int dstride=dstStride*sizeof(FbBits);
int j;

13
pixman/src/iccompose.c
View file

@ -1555,15 +1555,20 @@ fbFetch_r8g8b8 (FbCompositeOperand *op)
FbBits *line = op->u.drawable.line; CARD32 offset = op->u.drawable.offset;
CARD8 *pixel = ((CARD8 *) line) + (offset >> 3);
#if IMAGE_BYTE_ORDER == MSBFirst
/* FIXME: implent WORKING_UNALIGNED_INT for this endian :) */
return (0xff000000 |
(pixel[0] << 16) |
(pixel[1] << 8) |
(pixel[2]));
#else
return (0xff000000 |
(pixel[2] << 16) |
(pixel[1] << 8) |
(pixel[0]));
#ifdef WORKING_UNALIGNED_INT
return *(CARD32 *)pixel|0xff000000;
#else
return (0xff000000 |
(pixel[2] << 16) |
(pixel[1] << 8) |
(pixel[0]));
#endif
#endif
}