xserver: Branch 'master' - 6 commits
Søren Sandmann Pedersen
sandmann at kemper.freedesktop.org
Wed Apr 25 02:05:55 EEST 2007
fb/fbpict.c | 755 ++++++++++++++++++++++++++++++++++++++++++++++++++++--------
fb/fbpict.h | 12
2 files changed, 671 insertions(+), 96 deletions(-)
New commits:
diff-tree 0a2fe443d25b1ca25349aba3f748df986952e20f (from 7e16da7b78c422f96387502b9cc29eaa1741543f)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 19:02:44 2007 -0400
Use READ/WRITE macros for new functions introduced in previous commits.
diff --git a/fb/fbpict.c b/fb/fbpict.c
index 8ca7336..44bee1b 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -100,7 +100,7 @@ fbIn (CARD32 x, CARD8 y)
#define inOver0888(alpha, source, destval, dest) { \
CARD32 dstrb=destval&0xFF00FF; CARD32 dstag=(destval>>8)&0xFF00FF; \
CARD32 drb=((source&0xFF00FF)-dstrb)*alpha; CARD32 dag=(((source>>8)&0xFF00FF)-dstag)*alpha; \
- dest =((((drb>>8) + dstrb) & 0x00FF00FF) | ((((dag>>8) + dstag) << 8) & 0xFF00FF00)); \
+ WRITE(dest, ((((drb>>8) + dstrb) & 0x00FF00FF) | ((((dag>>8) + dstag) << 8) & 0xFF00FF00))); \
}
/*
@@ -111,42 +111,42 @@ fbIn (CARD32 x, CARD8 y)
#define inOver0565(alpha, source, destval, dest) { \
CARD16 dstrb = destval & 0xf81f; CARD16 dstg = destval & 0x7e0; \
CARD32 drb = ((source&0xf81f)-dstrb)*alpha; CARD32 dg=((source & 0x7e0)-dstg)*alpha; \
- dest = ((((drb>>5) + dstrb)&0xf81f) | (((dg>>5) + dstg) & 0x7e0)); \
+ WRITE(dest, ((((drb>>5) + dstrb)&0xf81f) | (((dg>>5) + dstg) & 0x7e0))); \
}
#define inOver2x0565(alpha, source, destval, dest) { \
CARD32 dstrb = destval & 0x07e0f81f; CARD32 dstg = (destval & 0xf81f07e0)>>5; \
CARD32 drb = ((source&0x07e0f81f)-dstrb)*alpha; CARD32 dg=(((source & 0xf81f07e0)>>5)-dstg)*alpha; \
- dest = ((((drb>>5) + dstrb)&0x07e0f81f) | ((((dg>>5) + dstg)<<5) & 0xf81f07e0)); \
+ WRITE(dest, ((((drb>>5) + dstrb)&0x07e0f81f) | ((((dg>>5) + dstg)<<5) & 0xf81f07e0))); \
}
#if IMAGE_BYTE_ORDER == LSBFirst
- #define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
+#define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
temp=count&3; \
where-=temp; \
workingWhere=(CARD32 *)where; \
- workingVal=*workingWhere++; \
+ workingVal=READ(workingWhere++); \
count=4-temp; \
workingVal>>=(8*temp)
- #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y=*z++; } where=(y)&0xff; (y)>>=8; (x)--;}
+ #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y = READ(z++); } where=(y)&0xff; (y)>>=8; (x)--;}
#define readPackedSource(where) readPacked(where,ws,workingSource,wsrc)
#define readPackedDest(where) readPacked(where,wd,workingiDest,widst)
- #define writePacked(what) workingoDest>>=8; workingoDest|=(what<<24); ww--; if(!ww) { ww=4; *wodst++=workingoDest; }
+ #define writePacked(what) workingoDest>>=8; workingoDest|=(what<<24); ww--; if(!ww) { ww=4; WRITE (wodst++, workingoDest); }
#else
#warning "I havn't tested fbCompositeTrans_0888xnx0888() on big endian yet!"
#define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
temp=count&3; \
where-=temp; \
workingWhere=(CARD32 *)where; \
- workingVal=*workingWhere++; \
+ workingVal=READ(workingWhere)++; \
count=4-temp; \
workingVal<<=(8*temp)
- #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y=*z++; } where=(y)>>24; (y)<<=8; (x)--;}
+ #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y = READ(z++); } where=(y)>>24; (y)<<=8; (x)--;}
#define readPackedSource(where) readPacked(where,ws,workingSource,wsrc)
#define readPackedDest(where) readPacked(where,wd,workingiDest,widst)
- #define writePacked(what) workingoDest<<=8; workingoDest|=what; ww--; if(!ww) { ww=4; *wodst++=workingoDest; }
+ #define writePacked(what) workingoDest<<=8; workingoDest|=what; ww--; if(!ww) { ww=4; WRITE(wodst++, workingoDest); }
#endif
/*
@@ -328,65 +328,65 @@ fbCompositeSolidMask_nx8x0888 (CARD8
fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3);
fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
- while (height--)
- {
- /* fixme: cleanup unused */
- unsigned int wt,wd;
- CARD32 workingiDest;
- CARD32 *widst;
-
- edst=dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
+ while (height--)
+ {
+ /* fixme: cleanup unused */
+ unsigned int wt, wd;
+ CARD32 workingiDest;
+ CARD32 *widst;
+
+ edst = dst = dstLine;
+ dstLine += dstStride;
+ mask = maskLine;
+ maskLine += maskStride;
+ w = width;
+
#ifndef NO_MASKED_PACKED_READ
- setupPackedReader(wd,wt,edst,widst,workingiDest);
+ setupPackedReader(wd,wt,edst,widst,workingiDest);
#endif
-
- while (w--)
- {
+
+ while (w--)
+ {
#ifndef NO_MASKED_PACKED_READ
- readPackedDest(rd);
- readPackedDest(gd);
- readPackedDest(bd);
+ readPackedDest(rd);
+ readPackedDest(gd);
+ readPackedDest(bd);
#else
- rd= *edst++;
- gd= *edst++;
- bd= *edst++;
-#endif
- m = *mask++;
- if (m == 0xff)
- {
- if (srca == 0xff)
- {
- *dst++=rs;
- *dst++=gs;
- *dst++=bs;
- }
- else
- {
- *dst++=(srcAlphaCombine24(rs, rd)>>8);
- *dst++=(srcAlphaCombine24(gs, gd)>>8);
- *dst++=(srcAlphaCombine24(bs, bd)>>8);
- }
- }
- else if (m)
- {
- int na=(srca*(int)m)>>8;
- int nia=255-na;
- *dst++=(genericCombine24(rs, rd, na, nia)>>8);
- *dst++=(genericCombine24(gs, gd, na, nia)>>8);
- *dst++=(genericCombine24(bs, bd, na, nia)>>8);
- }
- else
- {
- dst+=3;
- }
+ rd = READ(edst++);
+ gd = READ(edst++);
+ bd = READ(edst++);
+#endif
+ m = READ(mask++);
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ {
+ WRITE(dst++, rs);
+ WRITE(dst++, gs);
+ WRITE(dst++, bs);
}
+ else
+ {
+ WRITE(dst++, (srcAlphaCombine24(rs, rd)>>8));
+ WRITE(dst++, (srcAlphaCombine24(gs, gd)>>8));
+ WRITE(dst++, (srcAlphaCombine24(bs, bd)>>8));
+ }
+ }
+ else if (m)
+ {
+ int na=(srca*(int)m)>>8;
+ int nia=255-na;
+ WRITE(dst++, (genericCombine24(rs, rd, na, nia)>>8));
+ WRITE(dst++, (genericCombine24(gs, gd, na, nia)>>8));
+ WRITE(dst++, (genericCombine24(bs, bd, na, nia)>>8));
+ }
+ else
+ {
+ dst+=3;
+ }
}
-
+ }
+
fbFinishAccess (pMask->pDrawable);
fbFinishAccess (pDst->pDrawable);
}
@@ -414,8 +414,6 @@ fbCompositeSolidMask_nx8x0565 (CARD8
CARD16 w,src16;
fbComposeGetSolid(pSrc, src, pDst->format);
-
-
if (src == 0)
return;
@@ -437,23 +435,23 @@ fbCompositeSolidMask_nx8x0565 (CARD8
while (w--)
{
- m = *mask++;
+ m = READ(mask++);
if (m == 0)
dst++;
else if (srca5 == (0xff >> 3))
{
if (m == 0xff)
- *dst++ = src16;
+ WRITE(dst++, src16);
else
{
- d = *dst;
+ d = READ(dst);
m >>= 3;
- inOver0565 (m, src16, d, *dst++);
+ inOver0565 (m, src16, d, dst++);
}
}
else
{
- d = *dst;
+ d = READ(dst);
if (m == 0xff)
{
t = fbOver24 (src, cvt0565to0888 (d));
@@ -463,7 +461,7 @@ fbCompositeSolidMask_nx8x0565 (CARD8
t = fbIn (src, m);
t = fbOver (t, cvt0565to0888 (d));
}
- *dst++ = cvt8888to0565 (t);
+ WRITE(dst++, cvt8888to0565 (t));
}
}
}
@@ -517,34 +515,34 @@ fbCompositeSolidMask_nx8888x0565 (CARD8
while (w--)
{
- m = *mask++ >> 24;
+ m = READ(mask++) >> 24;
if (m == 0)
dst++;
else if (srca5 == (0xff >> 3))
{
if (m == 0xff)
- *dst++ = src16;
+ WRITE(dst++, src16);
else
{
- d = *dst;
+ d = READ(dst);
m >>= 3;
- inOver0565 (m, src16, d, *dst++);
+ inOver0565 (m, src16, d, dst++);
}
}
else
{
if (m == 0xff)
{
- d = *dst;
+ d = READ(dst);
t = fbOver24 (src, cvt0565to0888 (d));
- *dst++ = cvt8888to0565 (t);
+ WRITE(dst++, cvt8888to0565 (t));
}
else
{
- d = *dst;
+ d = READ(dst);
t = fbIn (src, m);
t = fbOver (t, cvt0565to0888 (d));
- *dst++ = cvt8888to0565 (t);
+ WRITE(dst++, cvt8888to0565 (t));
}
}
}
@@ -1078,9 +1076,9 @@ fbCompositeTrans_0565xnx0565(CARD8
if(((int)src&1)==1)
{
- s_16 = *src++;
- d_16 = *dst;
- inOver0565(maskAlpha, s_16, d_16, *dst++);
+ s_16 = READ(src++);
+ d_16 = READ(dst);
+ inOver0565(maskAlpha, s_16, d_16, dst++);
w--;
}
isrc=(CARD32 *)src;
@@ -1089,9 +1087,9 @@ fbCompositeTrans_0565xnx0565(CARD8
idst=(CARD32 *)dst;
while (w>1)
{
- s_32 = *isrc++;
- d_32 = *idst;
- inOver2x0565(maskAlpha, s_32, d_32, *idst++);
+ s_32 = READ(isrc++);
+ d_32 = READ(idst);
+ inOver2x0565(maskAlpha, s_32, d_32, idst++);
w-=2;
}
dst=(CARD16 *)idst;
@@ -1100,30 +1098,30 @@ fbCompositeTrans_0565xnx0565(CARD8
{
while (w > 1)
{
- s_32 = *isrc++;
+ s_32 = READ(isrc++);
#if IMAGE_BYTE_ORDER == LSBFirst
s_16=s_32&0xffff;
#else
s_16=s_32>>16;
#endif
- d_16 = *dst;
- inOver0565 (maskAlpha, s_16, d_16, *dst++);
+ d_16 = READ(dst);
+ inOver0565 (maskAlpha, s_16, d_16, dst++);
#if IMAGE_BYTE_ORDER == LSBFirst
s_16=s_32>>16;
#else
s_16=s_32&0xffff;
#endif
- d_16 = *dst;
- inOver0565(maskAlpha, s_16, d_16, *dst++);
+ d_16 = READ(dst);
+ inOver0565(maskAlpha, s_16, d_16, dst++);
w-=2;
}
}
src=(CARD16 *)isrc;
if(w!=0)
{
- s_16 = *src;
- d_16 = *dst;
- inOver0565(maskAlpha, s_16, d_16, *dst);
+ s_16 = READ(src);
+ d_16 = READ(dst);
+ inOver0565(maskAlpha, s_16, d_16, dst);
}
}
@@ -1133,6 +1131,7 @@ fbCompositeTrans_0565xnx0565(CARD8
/* 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(CARD8 op,
PicturePtr pSrc,
@@ -1146,175 +1145,175 @@ fbCompositeTrans_0888xnx0888(CARD8
INT16 yDst,
CARD16 width,
CARD16 height)
- {
- CARD8 *dstLine, *dst,*idst;
- CARD8 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
- FbBits mask;
- CARD16 maskAlpha,maskiAlpha;
-
- fbComposeGetSolid (pMask, mask, pDst->format);
- maskAlpha = mask >> 24;
- maskiAlpha= 255-maskAlpha;
-
- if (!maskAlpha)
- return;
- /*
- if (maskAlpha == 0xff)
- {
- fbCompositeSrc_0888x0888 (op, pSrc, pMask, pDst,
- xSrc, ySrc, xMask, yMask, xDst, yDst,
- width, height);
+{
+ CARD8 *dstLine, *dst,*idst;
+ CARD8 *srcLine, *src;
+ FbStride dstStride, srcStride;
+ CARD16 w;
+ FbBits mask;
+ CARD16 maskAlpha,maskiAlpha;
+
+ fbComposeGetSolid (pMask, mask, pDst->format);
+ maskAlpha = mask >> 24;
+ maskiAlpha= 255-maskAlpha;
+
+ if (!maskAlpha)
return;
- }
- */
-
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 3);
- fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3);
-
- {
- unsigned int ws,wt;
- CARD32 workingSource;
- CARD32 *wsrc, *wdst, *widst;
- CARD32 rs, rd, nd;
- CARD8 *isrc;
-
-
- /* 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--)
- {
- dst = dstLine;
- dstLine += dstStride;
- isrc = src = srcLine;
- srcLine += srcStride;
- w = width*3;
-
- setupPackedReader(ws,wt,isrc,wsrc,workingSource);
-
- /* get to word aligned */
- switch(!(int)src&3)
- {
- case 1:
- 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:
- readPackedSource(rs);
- rd=*dst;
- *dst++=alphamaskCombine24(rs, rd)>>8;
- w--; if(w==0) break;
- case 3:
- readPackedSource(rs);
- rd=*dst;
- *dst++=alphamaskCombine24(rs, rd)>>8;
- w--; if(w==0) break;
- }
- wdst=(CARD32 *)dst;
- while (w>3)
- {
- rs=*wsrc++;
- /* FIXME: write a special readPackedWord macro, which knows how to
- * halfword combine
- */
- #if IMAGE_BYTE_ORDER == LSBFirst
- rd=*wdst;
- readPackedSource(nd);
- readPackedSource(rs);
- nd|=rs<<8;
- readPackedSource(rs);
- nd|=rs<<16;
- readPackedSource(rs);
- nd|=rs<<24;
+ /*
+ 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;
+ CARD32 workingSource;
+ CARD32 *wsrc, *wdst, *widst;
+ CARD32 rs, rd, nd;
+ CARD8 *isrc;
+
+
+ /* 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--)
+ {
+ dst = dstLine;
+ dstLine += dstStride;
+ isrc = src = srcLine;
+ srcLine += srcStride;
+ w = width*3;
+
+ setupPackedReader(ws,wt,isrc,wsrc,workingSource);
+
+ /* get to word aligned */
+ switch(!(int)src&3)
+ {
+ case 1:
+ readPackedSource(rs);
+ /* *dst++=alphamaskCombine24(rs, *dst)>>8; */
+ rd = READ(dst); /* make gcc happy. hope it doens't cost us too much performance*/
+ WRITE(dst++, alphamaskCombine24(rs, rd) >> 8);
+ w--; if(w==0) break;
+ case 2:
+ readPackedSource(rs);
+ rd = READ(dst);
+ WRITE(dst++, alphamaskCombine24(rs, rd) >> 8);
+ w--; if(w==0) break;
+ case 3:
+ readPackedSource(rs);
+ rd = READ(dst);
+ WRITE(dst++,alphamaskCombine24(rs, rd) >> 8);
+ w--; if(w==0) break;
+ }
+ wdst=(CARD32 *)dst;
+ while (w>3)
+ {
+ rs=READ(wsrc++);
+ /* FIXME: write a special readPackedWord macro, which knows how to
+ * halfword combine
+ */
+#if IMAGE_BYTE_ORDER == LSBFirst
+ rd=READ(wdst);
+ readPackedSource(nd);
+ readPackedSource(rs);
+ nd|=rs<<8;
+ readPackedSource(rs);
+ nd|=rs<<16;
+ readPackedSource(rs);
+ nd|=rs<<24;
#else
- readPackedSource(nd);
- nd<<=24;
- readPackedSource(rs);
- nd|=rs<<16;
- readPackedSource(rs);
- nd|=rs<<8;
- readPackedSource(rs);
- nd|=rs;
-#endif
- inOver0888(maskAlpha, nd, rd, *wdst++);
- w-=4;
- }
- src=(CARD8 *)wdst;
- switch(w)
- {
- case 3:
- readPackedSource(rs);
- rd=*dst;
- *dst++=alphamaskCombine24(rs, rd)>>8;
- case 2:
- readPackedSource(rs);
- rd=*dst;
- *dst++=alphamaskCombine24(rs, rd)>>8;
- case 1:
- readPackedSource(rs);
- rd=*dst;
- *dst++=alphamaskCombine24(rs, rd)>>8;
- }
- }
- }
- else
- {
- while (height--)
- {
- idst=dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width*3;
- /* get to word aligned */
- switch(!(int)src&3)
- {
- case 1:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- w--; if(w==0) break;
- case 2:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- w--; if(w==0) break;
- case 3:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- w--; if(w==0) break;
- }
- wsrc=(CARD32 *)src;
- widst=(CARD32 *)dst;
- while(w>3)
- {
- rs = *wsrc++;
- rd = *widst;
- inOver0888 (maskAlpha, rs, rd, *widst++);
- w-=4;
- }
- src=(CARD8 *)wsrc;
- dst=(CARD8 *)widst;
- switch(w)
- {
- case 3:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- case 2:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- case 1:
- rd=alphamaskCombine24(*src++, *dst)>>8;
- *dst++=rd;
- }
- }
- }
- }
- }
-
+ readPackedSource(nd);
+ nd<<=24;
+ readPackedSource(rs);
+ nd|=rs<<16;
+ readPackedSource(rs);
+ nd|=rs<<8;
+ readPackedSource(rs);
+ nd|=rs;
+#endif
+ inOver0888(maskAlpha, nd, rd, wdst++);
+ w-=4;
+ }
+ src=(CARD8 *)wdst;
+ switch(w)
+ {
+ case 3:
+ readPackedSource(rs);
+ rd=READ(dst);
+ WRITE(dst++,alphamaskCombine24(rs, rd)>>8);
+ case 2:
+ readPackedSource(rs);
+ rd = READ(dst);
+ WRITE(dst++, alphamaskCombine24(rs, rd)>>8);
+ case 1:
+ readPackedSource(rs);
+ rd = READ(dst);
+ WRITE(dst++, alphamaskCombine24(rs, rd)>>8);
+ }
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ idst=dst = dstLine;
+ dstLine += dstStride;
+ src = srcLine;
+ srcLine += srcStride;
+ w = width*3;
+ /* get to word aligned */
+ switch(!(int)src&3)
+ {
+ case 1:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ w--; if(w==0) break;
+ case 2:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ w--; if(w==0) break;
+ case 3:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ w--; if(w==0) break;
+ }
+ wsrc=(CARD32 *)src;
+ widst=(CARD32 *)dst;
+ while(w>3)
+ {
+ rs = READ(wsrc++);
+ rd = READ(widst);
+ inOver0888 (maskAlpha, rs, rd, widst++);
+ w-=4;
+ }
+ src=(CARD8 *)wsrc;
+ dst=(CARD8 *)widst;
+ switch(w)
+ {
+ case 3:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ case 2:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ case 1:
+ rd=alphamaskCombine24(READ(src++), READ(dst))>>8;
+ WRITE(dst++, rd);
+ }
+ }
+ }
+ }
+}
/*
* Simple bitblt
diff-tree 7e16da7b78c422f96387502b9cc29eaa1741543f (from 2d9a7a768747ca39a800475f12c424c298018dc6)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 18:15:34 2007 -0400
Remove #if 0'ed blocks
diff --git a/fb/fbpict.c b/fb/fbpict.c
index d027e35..8ca7336 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -783,47 +783,6 @@ fbCompositeSrc_8888x0565 (CARD8 op,
fbFinishAccess (pSrc->pDrawable);
}
-#if 0
-void
-fbCompositeSrc_0565x0565 (CARD8 op,
- PicturePtr pSrc,
- PicturePtr pMask,
- PicturePtr pDst,
- INT16 xSrc,
- INT16 ySrc,
- INT16 xMask,
- INT16 yMask,
- INT16 xDst,
- INT16 yDst,
- CARD16 width,
- CARD16 height)
-{
- CARD16 *dstLine, *dst;
- CARD16 *srcLine, *src;
- FbStride dstStride, srcStride;
- CARD16 w;
-
- fbComposeGetStart (pSrc, xSrc, ySrc, CARD16, srcStride, srcLine, 1);
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w--)
- WRITE(dst, READ(src++));
- }
-
- fbFinishAccess (pDst->pDrawable);
- fbFinishAccess (pSrc->pDrawable);
-}
-#endif
-
void
fbCompositeSrcAdd_8000x8000 (CARD8 op,
PicturePtr pSrc,
diff-tree 2d9a7a768747ca39a800475f12c424c298018dc6 (from fde4a5adf02d3067a064ebf6bdd666aa5784cfe9)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 14:46:59 2007 -0400
From pixman (Jeff Muizelaar)
Fix up the fast-path compositing operators; those are useful for
sources without alpha, but can't be used for sources with
alpha. Also, replaced fbCompositeSrcSrc_nxn with call to fbBlt as
this function must handle 1, 4, 8, 16, 24, 32 bpp objects. Would
be nice to optimize fbBlt for common cases involving 8, 16, 24 and
32bpp.
From Keith Packard.
diff --git a/fb/fbpict.c b/fb/fbpict.c
index 660112c..d027e35 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -64,20 +64,6 @@ fbOver (CARD32 x, CARD32 y)
return m|n|o|p;
}
-static CARD32
-fbIn24 (CARD32 x, CARD8 y)
-{
- CARD16 a = y;
- CARD16 t;
- CARD32 m,n,o,p;
-
- m = FbInU(x,0,a,t);
- n = FbInU(x,8,a,t);
- o = FbInU(x,16,a,t);
- p = (y << 24);
- return m|n|o|p;
-}
-
CARD32
fbOver24 (CARD32 x, CARD32 y)
{
@@ -107,22 +93,34 @@ fbIn (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; \
- *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!
-#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)); \
- }
+/*
+ * This macro does src IN mask OVER dst when src and dst are 0888.
+ * If src has alpha, this will not work
+ */
+#define inOver0888(alpha, source, destval, dest) { \
+ CARD32 dstrb=destval&0xFF00FF; CARD32 dstag=(destval>>8)&0xFF00FF; \
+ CARD32 drb=((source&0xFF00FF)-dstrb)*alpha; CARD32 dag=(((source>>8)&0xFF00FF)-dstag)*alpha; \
+ dest =((((drb>>8) + dstrb) & 0x00FF00FF) | ((((dag>>8) + dstag) << 8) & 0xFF00FF00)); \
+ }
+
+/*
+ * This macro does src IN mask OVER dst when src and dst are 0565 and
+ * mask is a 5-bit alpha value. Again, if src has alpha, this will not
+ * work.
+ */
+#define inOver0565(alpha, source, destval, dest) { \
+ CARD16 dstrb = destval & 0xf81f; CARD16 dstg = destval & 0x7e0; \
+ CARD32 drb = ((source&0xf81f)-dstrb)*alpha; CARD32 dg=((source & 0x7e0)-dstg)*alpha; \
+ dest = ((((drb>>5) + dstrb)&0xf81f) | (((dg>>5) + dstg) & 0x7e0)); \
+ }
+
+
+#define inOver2x0565(alpha, source, destval, dest) { \
+ CARD32 dstrb = destval & 0x07e0f81f; CARD32 dstg = (destval & 0xf81f07e0)>>5; \
+ CARD32 drb = ((source&0x07e0f81f)-dstrb)*alpha; CARD32 dg=(((source & 0xf81f07e0)>>5)-dstg)*alpha; \
+ dest = ((((drb>>5) + dstrb)&0x07e0f81f) | ((((dg>>5) + dstg)<<5) & 0xf81f07e0)); \
+ }
+
#if IMAGE_BYTE_ORDER == LSBFirst
#define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
@@ -150,6 +148,7 @@ fbIn (CARD32 x, CARD8 y)
#define readPackedDest(where) readPacked(where,wd,workingiDest,widst)
#define writePacked(what) workingoDest<<=8; workingoDest|=what; ww--; if(!ww) { ww=4; *wodst++=workingoDest; }
#endif
+
/*
* Naming convention:
*
@@ -310,7 +309,6 @@ fbCompositeSolidMask_nx8x0888 (CARD8
{
CARD32 src, srca, srcia;
CARD8 *dstLine, *dst, *edst;
- CARD32 d;
CARD8 *maskLine, *mask, m;
FbStride dstStride, maskStride;
CARD16 w;
@@ -332,7 +330,7 @@ fbCompositeSolidMask_nx8x0888 (CARD8
while (height--)
{
- // fixme: cleanup unused
+ /* fixme: cleanup unused */
unsigned int wt,wd;
CARD32 workingiDest;
CARD32 *widst;
@@ -407,61 +405,152 @@ fbCompositeSolidMask_nx8x0565 (CARD8
CARD16 width,
CARD16 height)
{
- CARD32 src, srca,na, rsrca;
- CARD16 *dstLine, *dst;
- CARD16 d;
- CARD8 *maskLine, *mask, m;
- FbStride dstStride, maskStride;
- CARD16 w,src16;
-
- fbComposeGetSolid(pSrc, src, pDst->format);
- src16 = cvt8888to0565(src);
-
- rsrca = src >> 24;
- srca=rsrca>>2;
- if (src == 0)
- return;
+ CARD32 src, srca8, srca5;
+ CARD16 *dstLine, *dst;
+ CARD16 d;
+ CARD32 t;
+ CARD8 *maskLine, *mask, m;
+ FbStride dstStride, maskStride;
+ CARD16 w,src16;
+
+ fbComposeGetSolid(pSrc, src, pDst->format);
+
+
+
+ if (src == 0)
+ return;
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
+ srca8 = (src >> 24);
+ srca5 = (srca8 >> 3);
+ src16 = cvt8888to0565(src);
+
+ fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
+ fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
+ while (height--)
+ {
+ dst = dstLine;
+ dstLine += dstStride;
+ mask = maskLine;
+ maskLine += maskStride;
+ w = width;
- while (w--)
- {
- m = *mask++;
- if (m == 0xff)
- {
- if (srca == 0xff)
- {
- *dst=src16;
- }
- else
- {
- d = *dst;
- fastCombine0565(srca, src16, d, *dst++);
- }
- }
- else if (m)
- {
- na=(rsrca*(int)m)>>10;
- d = *dst;
- fastCombine0565(na, src16, d, *dst++);
- }
- else
- dst++;
- }
- }
+ while (w--)
+ {
+ m = *mask++;
+ if (m == 0)
+ dst++;
+ else if (srca5 == (0xff >> 3))
+ {
+ if (m == 0xff)
+ *dst++ = src16;
+ else
+ {
+ d = *dst;
+ m >>= 3;
+ inOver0565 (m, src16, d, *dst++);
+ }
+ }
+ else
+ {
+ d = *dst;
+ if (m == 0xff)
+ {
+ t = fbOver24 (src, cvt0565to0888 (d));
+ }
+ else
+ {
+ t = fbIn (src, m);
+ t = fbOver (t, cvt0565to0888 (d));
+ }
+ *dst++ = cvt8888to0565 (t);
+ }
+ }
+ }
+
fbFinishAccess (pMask->pDrawable);
fbFinishAccess (pDst->pDrawable);
}
+static void
+fbCompositeSolidMask_nx8888x0565 (CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height)
+{
+ CARD32 src, srca8, srca5;
+ CARD16 *dstLine, *dst;
+ CARD16 d;
+ CARD32 *maskLine, *mask;
+ CARD32 t;
+ CARD8 m;
+ FbStride dstStride, maskStride;
+ CARD16 w, src16;
+
+ fbComposeGetSolid(pSrc, src, pDst->format);
+
+ if (src == 0)
+ return;
+
+ srca8 = src >> 24;
+ srca5 = srca8 >> 3;
+ src16 = cvt8888to0565(src);
+
+ fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
+ fbComposeGetStart (pMask, xMask, yMask, CARD32, maskStride, maskLine, 1);
+
+ while (height--)
+ {
+ dst = dstLine;
+ dstLine += dstStride;
+ mask = maskLine;
+ maskLine += maskStride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++ >> 24;
+ if (m == 0)
+ dst++;
+ else if (srca5 == (0xff >> 3))
+ {
+ if (m == 0xff)
+ *dst++ = src16;
+ else
+ {
+ d = *dst;
+ m >>= 3;
+ inOver0565 (m, src16, d, *dst++);
+ }
+ }
+ else
+ {
+ if (m == 0xff)
+ {
+ d = *dst;
+ t = fbOver24 (src, cvt0565to0888 (d));
+ *dst++ = cvt8888to0565 (t);
+ }
+ else
+ {
+ d = *dst;
+ t = fbIn (src, m);
+ t = fbOver (t, cvt0565to0888 (d));
+ *dst++ = cvt8888to0565 (t);
+ }
+ }
+ }
+ }
+}
+
void
fbCompositeSolidMask_nx8888x0565C (CARD8 op,
PicturePtr pSrc,
@@ -516,14 +605,14 @@ fbCompositeSolidMask_nx8888x0565C (CARD8
else
{
d = READ(dst);
- d = fbOver24 (src, cvt0565to8888(d));
+ d = fbOver24 (src, cvt0565to0888(d));
WRITE(dst, cvt8888to0565(d));
}
}
else if (ma)
{
d = READ(dst);
- d = cvt0565to8888(d);
+ d = cvt0565to0888(d);
FbInOverC (src, srca, ma, d, 0, m);
FbInOverC (src, srca, ma, d, 8, n);
FbInOverC (src, srca, ma, d, 16, o);
@@ -682,7 +771,7 @@ fbCompositeSrc_8888x0565 (CARD8 op,
else
{
d = READ(dst);
- d = fbOver24 (s, cvt0565to8888(d));
+ d = fbOver24 (s, cvt0565to0888(d));
}
WRITE(dst, cvt8888to0565(d));
}
@@ -694,6 +783,7 @@ fbCompositeSrc_8888x0565 (CARD8 op,
fbFinishAccess (pSrc->pDrawable);
}
+#if 0
void
fbCompositeSrc_0565x0565 (CARD8 op,
PicturePtr pSrc,
@@ -732,6 +822,7 @@ fbCompositeSrc_0565x0565 (CARD8 op,
fbFinishAccess (pDst->pDrawable);
fbFinishAccess (pSrc->pDrawable);
}
+#endif
void
fbCompositeSrcAdd_8000x8000 (CARD8 op,
@@ -964,6 +1055,19 @@ fbCompositeSolidMask_nx1xn (CARD8 o
/*
* Apply a constant alpha value in an over computation
*/
+static void
+fbCompositeSrcSrc_nxn (CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height);
static void
fbCompositeTrans_0565xnx0565(CARD8 op,
@@ -985,19 +1089,19 @@ fbCompositeTrans_0565xnx0565(CARD8
CARD16 w;
FbBits mask;
CARD8 maskAlpha;
- CARD16 s_16, d_16, r_16;
- CARD32 s_32, d_32, i_32, r_32;
+ CARD16 s_16, d_16;
+ CARD32 s_32, d_32;
fbComposeGetSolid (pMask, mask, pDst->format);
- maskAlpha = mask >> 26;
+ maskAlpha = mask >> 27;
if (!maskAlpha)
return;
if (maskAlpha == 0xff)
{
- fbCompositeSrc_0565x0565 (op, pSrc, pMask, pDst,
- xSrc, ySrc, xMask, yMask, xDst, yDst,
- width, height);
+ fbCompositeSrcSrc_nxn (PictOpSrc, pSrc, pMask, pDst,
+ xSrc, ySrc, xMask, yMask, xDst, yDst,
+ width, height);
return;
}
@@ -1005,55 +1109,70 @@ fbCompositeTrans_0565xnx0565(CARD8
fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
while (height--)
- {
- CARD32 *isrc;
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- if(((int)src&1)==1)
- {
- s_16 = *src++;
- d_16 = *dst;
- fastCombine0565(maskAlpha, s_16, d_16, *dst++);
- w--;
- }
- isrc=(CARD32 *)src;
- while (w>1)
- {
- s_32=*isrc++;
+ {
+ CARD32 *isrc, *idst;
+ dst = dstLine;
+ dstLine += dstStride;
+ src = srcLine;
+ srcLine += srcStride;
+ w = width;
+
+ if(((int)src&1)==1)
+ {
+ s_16 = *src++;
+ d_16 = *dst;
+ inOver0565(maskAlpha, s_16, d_16, *dst++);
+ w--;
+ }
+ isrc=(CARD32 *)src;
+ if(((int)dst&1)==0)
+ {
+ idst=(CARD32 *)dst;
+ while (w>1)
+ {
+ s_32 = *isrc++;
+ d_32 = *idst;
+ inOver2x0565(maskAlpha, s_32, d_32, *idst++);
+ w-=2;
+ }
+ dst=(CARD16 *)idst;
+ }
+ else
+ {
+ while (w > 1)
+ {
+ s_32 = *isrc++;
#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++);
- #if IMAGE_BYTE_ORDER == LSBFirst
- s_16=s_32>>16;
- #else
- s_16=s_32&0xffff;
- #endif
- d_16 = *dst;
- fastCombine0565(maskAlpha, s_16, d_16, *dst++);
- w-=2;
- }
- src=(CARD16 *)isrc;
- if(w!=0)
- {
- s_16 = *src;
- d_16 = *dst;
- fastCombine0565(maskAlpha, s_16, d_16, *dst);
- }
- }
-
+ d_16 = *dst;
+ inOver0565 (maskAlpha, s_16, d_16, *dst++);
+#if IMAGE_BYTE_ORDER == LSBFirst
+ s_16=s_32>>16;
+#else
+ s_16=s_32&0xffff;
+#endif
+ d_16 = *dst;
+ inOver0565(maskAlpha, s_16, d_16, *dst++);
+ w-=2;
+ }
+ }
+ src=(CARD16 *)isrc;
+ if(w!=0)
+ {
+ s_16 = *src;
+ d_16 = *dst;
+ inOver0565(maskAlpha, s_16, d_16, *dst);
+ }
+ }
+
fbFinishAccess (pSrc->pDrawable);
fbFinishAccess (pDst->pDrawable);
}
-// 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(CARD8 op,
@@ -1082,140 +1201,105 @@ fbCompositeTrans_0888xnx0888(CARD8
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;
CARD32 workingSource;
- CARD32 *wsrc;
- CARD32 rs,gs,bs;
- CARD32 rd,gd,bd;
-
- CARD32 workingiDest,workingoDest;
- CARD32 *widst,*wodst;
-
+ CARD32 *wsrc, *wdst, *widst;
+ CARD32 rs, rd, nd;
+ CARD8 *isrc;
+
- // 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
-
- // get to word aligned
+ setupPackedReader(ws,wt,isrc,wsrc,workingSource);
+
+ /* get to word aligned */
switch(!(int)src&3)
{
case 1:
- readPackedDest(rd);
- rd=alphamaskCombine24(*src++, rd)>>8;
- writePacked(rd);
- w--; if(w==0) break;
+ 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);
- #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);
- #endif
+ rd=*wdst;
+ readPackedSource(nd);
+ readPackedSource(rs);
+ nd|=rs<<8;
+ readPackedSource(rs);
+ nd|=rs<<16;
+ readPackedSource(rs);
+ nd|=rs<<24;
+#else
+ readPackedSource(nd);
+ nd<<=24;
+ readPackedSource(rs);
+ nd|=rs<<16;
+ readPackedSource(rs);
+ nd|=rs<<8;
+ readPackedSource(rs);
+ nd|=rs;
+#endif
+ inOver0888(maskAlpha, nd, rd, *wdst++);
w-=4;
}
- src=(CARD8 *)wsrc;
+ src=(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;
}
}
}
@@ -1228,7 +1312,7 @@ fbCompositeTrans_0888xnx0888(CARD8
src = srcLine;
srcLine += srcStride;
w = width*3;
- // get to word aligned
+ /* get to word aligned */
switch(!(int)src&3)
{
case 1:
@@ -1246,13 +1330,11 @@ fbCompositeTrans_0888xnx0888(CARD8
}
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);
+ inOver0888 (maskAlpha, rs, rd, *widst++);
w-=4;
}
src=(CARD8 *)wsrc;
@@ -1300,80 +1382,29 @@ fbCompositeSrcSrc_nxn (CARD8 op,
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;
- 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
- Bool srcRepeat=pSrc->repeat;
- CARD32 srcHeight=pSrc->pDrawable->height;
- CARD32 srcWidth=pSrc->pDrawable->width;
+
+ fbGetDrawable(pSrc->pDrawable,src,srcStride,srcBpp,srcXoff,srcYoff);
+ fbGetDrawable(pDst->pDrawable,dst,dstStride,dstBpp,dstXoff,dstYoff);
+
+ fbBlt (src + (ySrc + srcYoff) * srcStride,
+ srcStride,
+ (xSrc + srcXoff) * srcBpp,
- fbGetDrawable(pSrc->pDrawable,src,srcStride,srcBpp,srcXoff,srcYoff);
- fbGetDrawable(pDst->pDrawable,dst,dstStride,dstBpp,dstXoff,dstYoff);
+ dst + (yDst + dstYoff) * dstStride,
+ dstStride,
+ (xDst + dstXoff) * dstBpp,
- if(srcRepeat)
- {
- xSrc%=srcWidth;
- ySrc%=srcHeight;
- }
-
- while(iheight>0)
- {
- int wheight=iheight;
- if(wheight>(srcHeight-ySrc))
- wheight=(srcHeight-ySrc);
- iwidth=initialWidth;
- xDst=initialX;
- while(iwidth>0)
- {
- int wwidth=iwidth;
- 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);
- if(!srcRepeat)
- iwidth=0;
- else
- {
- xDst+=wwidth;
- iwidth-=wwidth;
- }
- }
- if(!srcRepeat)
- iheight=0;
- else
- {
- yDst+=wheight;
- iheight-=wheight;
- }
- }
+ (width) * dstBpp,
+ (height),
+
+ GXcopy,
+ FB_ALLONES,
+ dstBpp,
+
+ reverse,
+ upsidedown);
fbFinishAccess(pSrc->pDrawable);
fbFinishAccess(pDst->pDrawable);
@@ -1530,6 +1561,16 @@ fbComposite (CARD8 op,
break;
}
}
+ else
+ {
+ switch (pDst->format) {
+ case PICT_r5g6b5:
+ func = fbCompositeSolidMask_nx8888x0565;
+ break;
+ default:
+ break;
+ }
+ }
break;
case PICT_a8b8g8r8:
if (pMask->componentAlpha) {
@@ -1555,6 +1596,16 @@ fbComposite (CARD8 op,
break;
}
}
+ else
+ {
+ switch (pDst->format) {
+ case PICT_b5g6r5:
+ func = fbCompositeSolidMask_nx8888x0565;
+ break;
+ default:
+ break;
+ }
+ }
break;
case PICT_a1:
switch (pDst->format) {
@@ -1722,7 +1773,20 @@ fbComposite (CARD8 op,
}
else if (! srcRepeat)
{
- switch (pSrc->format) {
+ /*
+ * Formats without alpha bits are just Copy with Over
+ */
+ if (pSrc->format == pDst->format && !PICT_FORMAT_A(pSrc->format))
+ {
+#ifdef USE_MMX
+ if (fbHaveMMX() &&
+ (pSrc->format == PICT_x8r8g8b8 || pSrc->format == PICT_x8b8g8r8))
+ func = fbCompositeCopyAreammx;
+ else
+#endif
+ func = fbCompositeSrcSrc_nxn;
+ }
+ else switch (pSrc->format) {
case PICT_a8r8g8b8:
switch (pDst->format) {
case PICT_a8r8g8b8:
@@ -1800,24 +1864,6 @@ fbComposite (CARD8 op,
break;
}
break;
- case PICT_r5g6b5:
- switch (pDst->format) {
- case PICT_r5g6b5:
- func = fbCompositeSrc_0565x0565;
- break;
- default:
- break;
- }
- break;
- case PICT_b5g6r5:
- switch (pDst->format) {
- case PICT_b5g6r5:
- func = fbCompositeSrc_0565x0565;
- break;
- default:
- break;
- }
- break;
default:
break;
}
@@ -1908,14 +1954,6 @@ fbComposite (CARD8 op,
n = REGION_NUM_RECTS (®ion);
pbox = REGION_RECTS (®ion);
- // 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;
- }
-
while (n--)
{
h = pbox->y2 - pbox->y1;
diff --git a/fb/fbpict.h b/fb/fbpict.h
index 76cab5b..5246cd5 100644
--- a/fb/fbpict.h
+++ b/fb/fbpict.h
@@ -121,7 +121,7 @@ fbCanGetSolid(PicturePtr pict)
break; \
case 16: \
(bits) = READ((CARD16 *) __bits__); \
- (bits) = cvt0565to8888(bits); \
+ (bits) = cvt0565to0888(bits); \
break; \
case 8: \
(bits) = READ((CARD8 *) __bits__); \
@@ -161,7 +161,7 @@ fbCanGetSolid(PicturePtr pict)
#define cvt8888to0565(s) ((((s) >> 3) & 0x001f) | \
(((s) >> 5) & 0x07e0) | \
(((s) >> 8) & 0xf800))
-#define cvt0565to8888(s) (((((s) << 3) & 0xf8) | (((s) >> 2) & 0x7)) | \
+#define cvt0565to0888(s) (((((s) << 3) & 0xf8) | (((s) >> 2) & 0x7)) | \
((((s) << 5) & 0xfc00) | (((s) >> 1) & 0x300)) | \
((((s) << 8) & 0xf80000) | (((s) << 3) & 0x70000)))
diff-tree fde4a5adf02d3067a064ebf6bdd666aa5784cfe9 (from 13e1d5ea55b0a3b7729316c8e37d3d8fca2075b5)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 13:30:43 2007 -0400
From xserver via pixman (Jeff Muizelaar)
Add some optimizations from jaymz. Also adds some compile
warnings that will hopefully go awa y as we continue merging.
diff --git a/fb/fbpict.c b/fb/fbpict.c
index 072f8d8..660112c 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -105,6 +105,51 @@ fbIn (CARD32 x, CARD8 y)
return m|n|o|p;
}
+#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; \
+ *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!
+#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)); \
+ }
+
+#if IMAGE_BYTE_ORDER == LSBFirst
+ #define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
+ temp=count&3; \
+ where-=temp; \
+ workingWhere=(CARD32 *)where; \
+ workingVal=*workingWhere++; \
+ count=4-temp; \
+ workingVal>>=(8*temp)
+ #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y=*z++; } where=(y)&0xff; (y)>>=8; (x)--;}
+ #define readPackedSource(where) readPacked(where,ws,workingSource,wsrc)
+ #define readPackedDest(where) readPacked(where,wd,workingiDest,widst)
+ #define writePacked(what) workingoDest>>=8; workingoDest|=(what<<24); ww--; if(!ww) { ww=4; *wodst++=workingoDest; }
+#else
+ #warning "I havn't tested fbCompositeTrans_0888xnx0888() on big endian yet!"
+ #define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \
+ temp=count&3; \
+ where-=temp; \
+ workingWhere=(CARD32 *)where; \
+ workingVal=*workingWhere++; \
+ count=4-temp; \
+ workingVal<<=(8*temp)
+ #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y=*z++; } where=(y)>>24; (y)<<=8; (x)--;}
+ #define readPackedSource(where) readPacked(where,ws,workingSource,wsrc)
+ #define readPackedDest(where) readPacked(where,wd,workingiDest,widst)
+ #define writePacked(what) workingoDest<<=8; workingoDest|=what; ww--; if(!ww) { ww=4; *wodst++=workingoDest; }
+#endif
/*
* Naming convention:
*
@@ -248,6 +293,7 @@ fbCompositeSolidMask_nx8888x8888C (CARD8
fbFinishAccess (pDst->pDrawable);
}
+#define srcAlphaCombine24(a,b) genericCombine24(a,b,srca,srcia)
void
fbCompositeSolidMask_nx8x0888 (CARD8 op,
PicturePtr pSrc,
@@ -262,52 +308,86 @@ fbCompositeSolidMask_nx8x0888 (CARD8
CARD16 width,
CARD16 height)
{
- CARD32 src, srca;
- CARD8 *dstLine, *dst;
+ CARD32 src, srca, srcia;
+ CARD8 *dstLine, *dst, *edst;
CARD32 d;
CARD8 *maskLine, *mask, m;
FbStride dstStride, maskStride;
CARD16 w;
+ CARD32 rs,gs,bs,rd,gd,bd;
fbComposeGetSolid(pSrc, src, pDst->format);
srca = src >> 24;
+ srcia = 255-srca;
if (src == 0)
return;
+ rs=src&0xff;
+ gs=(src>>8)&0xff;
+ bs=(src>>16)&0xff;
+
fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3);
fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- while (w--)
- {
- m = READ(mask++);
- if (m == 0xff)
- {
- if (srca == 0xff)
- d = src;
- else
- {
- d = Fetch24(dst);
- d = fbOver24 (src, d);
+ while (height--)
+ {
+ // fixme: cleanup unused
+ unsigned int wt,wd;
+ CARD32 workingiDest;
+ CARD32 *widst;
+
+ edst=dst = dstLine;
+ dstLine += dstStride;
+ mask = maskLine;
+ maskLine += maskStride;
+ w = width;
+
+#ifndef NO_MASKED_PACKED_READ
+ setupPackedReader(wd,wt,edst,widst,workingiDest);
+#endif
+
+ while (w--)
+ {
+#ifndef NO_MASKED_PACKED_READ
+ readPackedDest(rd);
+ readPackedDest(gd);
+ readPackedDest(bd);
+#else
+ rd= *edst++;
+ gd= *edst++;
+ bd= *edst++;
+#endif
+ m = *mask++;
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ {
+ *dst++=rs;
+ *dst++=gs;
+ *dst++=bs;
+ }
+ else
+ {
+ *dst++=(srcAlphaCombine24(rs, rd)>>8);
+ *dst++=(srcAlphaCombine24(gs, gd)>>8);
+ *dst++=(srcAlphaCombine24(bs, bd)>>8);
+ }
+ }
+ else if (m)
+ {
+ int na=(srca*(int)m)>>8;
+ int nia=255-na;
+ *dst++=(genericCombine24(rs, rd, na, nia)>>8);
+ *dst++=(genericCombine24(gs, gd, na, nia)>>8);
+ *dst++=(genericCombine24(bs, bd, na, nia)>>8);
+ }
+ else
+ {
+ dst+=3;
+ }
}
- Store24(dst,d);
- }
- else if (m)
- {
- d = fbOver24 (fbIn(src,m), Fetch24(dst));
- Store24(dst,d);
- }
- dst += 3;
}
- }
fbFinishAccess (pMask->pDrawable);
fbFinishAccess (pDst->pDrawable);
@@ -327,54 +407,57 @@ fbCompositeSolidMask_nx8x0565 (CARD8
CARD16 width,
CARD16 height)
{
- CARD32 src, srca;
- CARD16 *dstLine, *dst;
- CARD32 d;
- CARD8 *maskLine, *mask, m;
- FbStride dstStride, maskStride;
- CARD16 w;
-
- fbComposeGetSolid(pSrc, src, pDst->format);
-
- srca = src >> 24;
- if (src == 0)
- return;
-
- fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
- fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
-
- while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- mask = maskLine;
- maskLine += maskStride;
- w = width;
-
- while (w--)
- {
- m = READ(mask++);
- if (m == 0xff)
- {
- if (srca == 0xff)
- d = src;
- else
- {
- d = READ(dst);
- d = fbOver24 (src, cvt0565to8888(d));
- }
- WRITE(dst, cvt8888to0565(d));
- }
- else if (m)
- {
- d = READ(dst);
- d = fbOver24 (fbIn(src,m), cvt0565to8888(d));
- WRITE(dst, cvt8888to0565(d));
- }
- dst++;
- }
- }
-
+ CARD32 src, srca,na, rsrca;
+ CARD16 *dstLine, *dst;
+ CARD16 d;
+ CARD8 *maskLine, *mask, m;
+ FbStride dstStride, maskStride;
+ CARD16 w,src16;
+
+ fbComposeGetSolid(pSrc, src, pDst->format);
+ src16 = cvt8888to0565(src);
+
+ rsrca = src >> 24;
+ srca=rsrca>>2;
+ if (src == 0)
+ return;
+
+ fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
+ fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1);
+
+ while (height--)
+ {
+ dst = dstLine;
+ dstLine += dstStride;
+ mask = maskLine;
+ maskLine += maskStride;
+ w = width;
+
+ while (w--)
+ {
+ m = *mask++;
+ if (m == 0xff)
+ {
+ if (srca == 0xff)
+ {
+ *dst=src16;
+ }
+ else
+ {
+ d = *dst;
+ fastCombine0565(srca, src16, d, *dst++);
+ }
+ }
+ else if (m)
+ {
+ na=(rsrca*(int)m)>>10;
+ d = *dst;
+ fastCombine0565(na, src16, d, *dst++);
+ }
+ else
+ dst++;
+ }
+ }
fbFinishAccess (pMask->pDrawable);
fbFinishAccess (pDst->pDrawable);
}
@@ -906,7 +989,7 @@ fbCompositeTrans_0565xnx0565(CARD8
CARD32 s_32, d_32, i_32, r_32;
fbComposeGetSolid (pMask, mask, pDst->format);
- maskAlpha = mask >> 24;
+ maskAlpha = mask >> 26;
if (!maskAlpha)
return;
@@ -922,31 +1005,276 @@ fbCompositeTrans_0565xnx0565(CARD8
fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
while (height--)
- {
- dst = dstLine;
- dstLine += dstStride;
- src = srcLine;
- srcLine += srcStride;
- w = width;
-
- while (w--)
- {
- s_16 = READ(src++);
- s_32 = cvt0565to8888(s_16);
- d_16 = READ(dst);
- d_32 = cvt0565to8888(d_16);
-
- i_32 = fbIn24 (s_32, maskAlpha);
- r_32 = fbOver24 (i_32, d_32);
- r_16 = cvt8888to0565(r_32);
- WRITE(dst++, r_16);
- }
- }
+ {
+ CARD32 *isrc;
+ dst = dstLine;
+ dstLine += dstStride;
+ src = srcLine;
+ srcLine += srcStride;
+ w = width;
+
+ if(((int)src&1)==1)
+ {
+ s_16 = *src++;
+ d_16 = *dst;
+ fastCombine0565(maskAlpha, s_16, d_16, *dst++);
+ w--;
+ }
+ isrc=(CARD32 *)src;
+ while (w>1)
+ {
+ s_32=*isrc++;
+#if IMAGE_BYTE_ORDER == LSBFirst
+ s_16=s_32&0xffff;
+#else
+ s_16=s_32>>16;
+#endif
+ d_16 = *dst;
+ fastCombine0565(maskAlpha, s_16, d_16, *dst++);
+ #if IMAGE_BYTE_ORDER == LSBFirst
+ s_16=s_32>>16;
+ #else
+ s_16=s_32&0xffff;
+ #endif
+ d_16 = *dst;
+ fastCombine0565(maskAlpha, s_16, d_16, *dst++);
+ w-=2;
+ }
+ src=(CARD16 *)isrc;
+ if(w!=0)
+ {
+ s_16 = *src;
+ d_16 = *dst;
+ fastCombine0565(maskAlpha, s_16, d_16, *dst);
+ }
+ }
fbFinishAccess (pSrc->pDrawable);
fbFinishAccess (pDst->pDrawable);
}
+// 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(CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height)
+ {
+ CARD8 *dstLine, *dst,*idst;
+ CARD8 *srcLine, *src;
+ FbStride dstStride, srcStride;
+ CARD16 w;
+ FbBits mask;
+ CARD16 maskAlpha,maskiAlpha;
+
+ fbComposeGetSolid (pMask, mask, pDst->format);
+ maskAlpha = mask >> 24;
+ maskiAlpha= 255-maskAlpha;
+
+ if (!maskAlpha)
+ 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;
+ CARD32 workingSource;
+ CARD32 *wsrc;
+ CARD32 rs,gs,bs;
+ CARD32 rd,gd,bd;
+
+ 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)
+ {
+ while (height--)
+ {
+ idst=dst = dstLine;
+ dstLine += dstStride;
+ 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
+
+ // get to word aligned
+ switch(!(int)src&3)
+ {
+ case 1:
+ readPackedDest(rd);
+ rd=alphamaskCombine24(*src++, rd)>>8;
+ writePacked(rd);
+ w--; if(w==0) break;
+ case 2:
+ readPackedDest(rd);
+ rd=alphamaskCombine24(*src++, rd)>>8;
+ writePacked(rd);
+ w--; if(w==0) break;
+ case 3:
+ readPackedDest(rd);
+ rd=alphamaskCombine24(*src++, rd)>>8;
+ writePacked(rd);
+ w--; if(w==0) break;
+ }
+ wsrc=(CARD32 *)src;
+ 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
+ #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);
+ #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);
+ #endif
+ w-=4;
+ }
+ src=(CARD8 *)wsrc;
+ switch(w)
+ {
+ case 3:
+ readPackedDest(rd);
+ rd=alphamaskCombine24(*src++, rd)>>8;
+ writePacked(rd);
+ case 2:
+ readPackedDest(rd);
+ rd=alphamaskCombine24(*src++, rd)>>8;
+ writePacked(rd);
+ 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;
+ }
+ }
+ }
+ else
+ {
+ while (height--)
+ {
+ idst=dst = dstLine;
+ dstLine += dstStride;
+ src = srcLine;
+ srcLine += srcStride;
+ w = width*3;
+ // get to word aligned
+ switch(!(int)src&3)
+ {
+ case 1:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ w--; if(w==0) break;
+ case 2:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ w--; if(w==0) break;
+ case 3:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ w--; if(w==0) break;
+ }
+ 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);
+ w-=4;
+ }
+ src=(CARD8 *)wsrc;
+ dst=(CARD8 *)widst;
+ switch(w)
+ {
+ case 3:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ case 2:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ case 1:
+ rd=alphamaskCombine24(*src++, *dst)>>8;
+ *dst++=rd;
+ }
+ }
+ }
+ }
+ }
+
+
/*
* Simple bitblt
*/
@@ -972,30 +1300,81 @@ fbCompositeSrcSrc_nxn (CARD8 op,
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;
+ 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
+ Bool srcRepeat=pSrc->repeat;
+ CARD32 srcHeight=pSrc->pDrawable->height;
+ CARD32 srcWidth=pSrc->pDrawable->width;
+
+ fbGetDrawable(pSrc->pDrawable,src,srcStride,srcBpp,srcXoff,srcYoff);
+ fbGetDrawable(pDst->pDrawable,dst,dstStride,dstBpp,dstXoff,dstYoff);
+
+ if(srcRepeat)
+ {
+ xSrc%=srcWidth;
+ ySrc%=srcHeight;
+ }
+
+ while(iheight>0)
+ {
+ int wheight=iheight;
+ if(wheight>(srcHeight-ySrc))
+ wheight=(srcHeight-ySrc);
+ iwidth=initialWidth;
+ xDst=initialX;
+ while(iwidth>0)
+ {
+ int wwidth=iwidth;
+ 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);
+ if(!srcRepeat)
+ iwidth=0;
+ else
+ {
+ xDst+=wwidth;
+ iwidth-=wwidth;
+ }
+ }
+ if(!srcRepeat)
+ iheight=0;
+ else
+ {
+ yDst+=wheight;
+ iheight-=wheight;
+ }
+ }
- fbGetDrawable(pSrc->pDrawable,src,srcStride,srcBpp,srcXoff,srcYoff);
- fbGetDrawable(pDst->pDrawable,dst,dstStride,dstBpp,dstXoff,dstYoff);
-
- fbBlt (src + (ySrc + srcYoff) * srcStride,
- srcStride,
- (xSrc + srcXoff) * srcBpp,
-
- dst + (yDst + dstYoff) * dstStride,
- dstStride,
- (xDst + dstXoff) * dstBpp,
-
- (width) * dstBpp,
- (height),
-
- GXcopy,
- FB_ALLONES,
- dstBpp,
-
- reverse,
- upsidedown);
-
fbFinishAccess(pSrc->pDrawable);
fbFinishAccess(pDst->pDrawable);
}
@@ -1294,6 +1673,11 @@ fbComposite (CARD8 op,
if (pDst->format == pSrc->format)
func = fbCompositeTrans_0565xnx0565;
break;
+ case PICT_r8g8b8:
+ case PICT_b8g8r8:
+ if (pDst->format == pSrc->format)
+ func = fbCompositeTrans_0888xnx0888;
+ break;
default:
break;
}
@@ -1524,6 +1908,14 @@ fbComposite (CARD8 op,
n = REGION_NUM_RECTS (®ion);
pbox = REGION_RECTS (®ion);
+ // 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;
+ }
+
while (n--)
{
h = pbox->y2 - pbox->y1;
diff-tree 13e1d5ea55b0a3b7729316c8e37d3d8fca2075b5 (from 077a5d4555676d5775e990468a697b6890c6d609)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 12:59:18 2007 -0400
Fix format vs formatCode in previous commit
diff --git a/fb/fbpict.c b/fb/fbpict.c
index bc737d8..072f8d8 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -1085,7 +1085,7 @@ fbComposite (CARD8 op,
#endif
if (pMask == 0)
{
- if (pSrc->format_code == pDst->format_code)
+ if (pSrc->format == pDst->format)
func = fbCompositeSrcSrc_nxn;
}
break;
diff-tree 077a5d4555676d5775e990468a697b6890c6d609 (from 09436fb7c38a9819bde770c4c21143591671c4d7)
Author: Soren Sandmann Pedersen <ssp at dhcp83-218.boston.redhat.com>
Date: Tue Apr 24 12:57:55 2007 -0400
Add functions fbCompositeSrcSrc_nxn() and fbCompositeTrans_0565xnx0565
from xserver via pixman. Add READ/WRITE and fbFinishAccess as
appropriate.
diff --git a/fb/fbpict.c b/fb/fbpict.c
index cd6cac2..bc737d8 100644
--- a/fb/fbpict.c
+++ b/fb/fbpict.c
@@ -64,6 +64,20 @@ fbOver (CARD32 x, CARD32 y)
return m|n|o|p;
}
+static CARD32
+fbIn24 (CARD32 x, CARD8 y)
+{
+ CARD16 a = y;
+ CARD16 t;
+ CARD32 m,n,o,p;
+
+ m = FbInU(x,0,a,t);
+ n = FbInU(x,8,a,t);
+ o = FbInU(x,16,a,t);
+ p = (y << 24);
+ return m|n|o|p;
+}
+
CARD32
fbOver24 (CARD32 x, CARD32 y)
{
@@ -864,6 +878,148 @@ fbCompositeSolidMask_nx1xn (CARD8 o
# define mod(a,b) ((b) == 1 ? 0 : (a) >= 0 ? (a) % (b) : (b) - (-a) % (b))
+/*
+ * Apply a constant alpha value in an over computation
+ */
+
+static void
+fbCompositeTrans_0565xnx0565(CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height)
+{
+ CARD16 *dstLine, *dst;
+ CARD16 *srcLine, *src;
+ FbStride dstStride, srcStride;
+ CARD16 w;
+ FbBits mask;
+ CARD8 maskAlpha;
+ CARD16 s_16, d_16, r_16;
+ CARD32 s_32, d_32, i_32, r_32;
+
+ fbComposeGetSolid (pMask, mask, pDst->format);
+ maskAlpha = mask >> 24;
+
+ if (!maskAlpha)
+ return;
+ if (maskAlpha == 0xff)
+ {
+ fbCompositeSrc_0565x0565 (op, pSrc, pMask, pDst,
+ xSrc, ySrc, xMask, yMask, xDst, yDst,
+ width, height);
+ return;
+ }
+
+ fbComposeGetStart (pSrc, xSrc, ySrc, CARD16, srcStride, srcLine, 1);
+ fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1);
+
+ while (height--)
+ {
+ dst = dstLine;
+ dstLine += dstStride;
+ src = srcLine;
+ srcLine += srcStride;
+ w = width;
+
+ while (w--)
+ {
+ s_16 = READ(src++);
+ s_32 = cvt0565to8888(s_16);
+ d_16 = READ(dst);
+ d_32 = cvt0565to8888(d_16);
+
+ i_32 = fbIn24 (s_32, maskAlpha);
+ r_32 = fbOver24 (i_32, d_32);
+ r_16 = cvt8888to0565(r_32);
+ WRITE(dst++, r_16);
+ }
+ }
+
+ fbFinishAccess (pSrc->pDrawable);
+ fbFinishAccess (pDst->pDrawable);
+}
+
+/*
+ * Simple bitblt
+ */
+
+static void
+fbCompositeSrcSrc_nxn (CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height)
+{
+ FbBits *dst;
+ FbBits *src;
+ FbStride dstStride, srcStride;
+ int srcXoff, srcYoff;
+ int dstXoff, dstYoff;
+ int srcBpp;
+ int dstBpp;
+ Bool reverse = FALSE;
+ Bool upsidedown = FALSE;
+
+ fbGetDrawable(pSrc->pDrawable,src,srcStride,srcBpp,srcXoff,srcYoff);
+ fbGetDrawable(pDst->pDrawable,dst,dstStride,dstBpp,dstXoff,dstYoff);
+
+ fbBlt (src + (ySrc + srcYoff) * srcStride,
+ srcStride,
+ (xSrc + srcXoff) * srcBpp,
+
+ dst + (yDst + dstYoff) * dstStride,
+ dstStride,
+ (xDst + dstXoff) * dstBpp,
+
+ (width) * dstBpp,
+ (height),
+
+ GXcopy,
+ FB_ALLONES,
+ dstBpp,
+
+ reverse,
+ upsidedown);
+
+ fbFinishAccess(pSrc->pDrawable);
+ fbFinishAccess(pDst->pDrawable);
+}
+
+/*
+ * Solid fill
+void
+fbCompositeSolidSrc_nxn (CARD8 op,
+ PicturePtr pSrc,
+ PicturePtr pMask,
+ PicturePtr pDst,
+ INT16 xSrc,
+ INT16 ySrc,
+ INT16 xMask,
+ INT16 yMask,
+ INT16 xDst,
+ INT16 yDst,
+ CARD16 width,
+ CARD16 height)
+{
+
+}
+ */
+
void
fbComposite (CARD8 op,
PicturePtr pSrc,
@@ -925,7 +1081,13 @@ fbComposite (CARD8 op,
{
func = fbCompositeCopyAreammx;
}
+ else
#endif
+ if (pMask == 0)
+ {
+ if (pSrc->format_code == pDst->format_code)
+ func = fbCompositeSrcSrc_nxn;
+ }
break;
case PictOpOver:
if (pMask)
@@ -933,7 +1095,6 @@ fbComposite (CARD8 op,
if (fbCanGetSolid(pSrc) &&
!maskRepeat)
{
- srcRepeat = FALSE;
if (PICT_FORMAT_COLOR(pSrc->format)) {
switch (pMask->format) {
case PICT_a8:
@@ -1120,6 +1281,24 @@ fbComposite (CARD8 op,
}
}
}
+ if (func != fbCompositeGeneral)
+ srcRepeat = FALSE;
+ }
+ else if (maskRepeat &&
+ pMask->pDrawable->width == 1 &&
+ pMask->pDrawable->height == 1)
+ {
+ switch (pSrc->format) {
+ case PICT_r5g6b5:
+ case PICT_b5g6r5:
+ if (pDst->format == pSrc->format)
+ func = fbCompositeTrans_0565xnx0565;
+ break;
+ default:
+ break;
+ }
+ if (func != fbCompositeGeneral)
+ maskRepeat = FALSE;
}
}
else /* no mask */
diff --git a/fb/fbpict.h b/fb/fbpict.h
index 434526e..76cab5b 100644
--- a/fb/fbpict.h
+++ b/fb/fbpict.h
@@ -123,6 +123,14 @@ fbCanGetSolid(PicturePtr pict)
(bits) = READ((CARD16 *) __bits__); \
(bits) = cvt0565to8888(bits); \
break; \
+ case 8: \
+ (bits) = READ((CARD8 *) __bits__); \
+ (bits) = (bits) << 24; \
+ break; \
+ case 1: \
+ (bits) = READ((CARD32 *) __bits__); \
+ (bits) = FbLeftStipBits((bits),1) ? 0xff000000 : 0x00000000;\
+ break; \
default: \
return; \
} \
More information about the xorg-commit
mailing list