Re: Question about the performance of gtk+ in neo1973 mobile phone
- From: Koen Kooi <koen dominion kabel utwente nl>
- To: Daniel Amelang <daniel amelang gmail com>
- Cc: performance-list gnome org, mobile-devel-list gnome org
- Subject: Re: Question about the performance of gtk+ in neo1973 mobile phone
- Date: Wed, 23 May 2007 10:46:20 +0200
-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Daniel Amelang schreef:
> On 5/23/07, Koen Kooi <koen dominion kabel utwente nl> wrote:
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Daniel Amelang schreef:
>>
>>> If slow pixbuf rendering were a problem, you may be interested in a
>>> patch I submitted recently to xorg that speeded up pixbuf rendering on
>>> R5G6B5 framebuffers by 3.4x :
>>>
>>> http://lists.freedesktop.org/archives/xorg/2007-April/024286.html
>>>
>>> That speedup was measured on a N800, but the neo1973 should see
>>> similar improvement.
>> I gave the testapp and patch a spin:
>>
>> ipaq hx4700:
>>
>> unpatched Xw100: 6.04 FPS
>> patched Xw100: 6.04 FPS
>> unpatched Xfbdev: 6.04 FPS
>> patched Xfbdev: 6.04 FPS
>>
>> ipaq h2200:
>>
>> unpatched Xfbdev: 4.18 FPS
>> patched Xfbdev: 4.17 FPS
>>
>> Any clues why the patch doesn't seem to have an effect?
>
> Assuming you didn't make any errors in applying the patch and getting
> the patched server compiled and up and running (I do have faith in you
> :), it's likely that for your combination of Gdk+Xserver+hardware, the
> test program doesn't hit the exact same composite case
> (fbCompositeSrc_8888RevNPx0565) as it does on my N800.
gtk 2.10.12 + xserver-xorg 1.2.0, built as kdrive.
> Easiest way to tell that would to add a couple ErrorF calls in
> fbpict.c to see what functions are called. ErrorF messages get spat
> out to stderr, if I remember correctly.
>
> You didn't happen to apply that patch by hand, did you? I can think of
> a couple tricky places that might have tripped you up if you did.
> Especially the part about adding the new function to the fbComposite
> "switch case of doom". Care to send me your patched fbpict.c?
attached.
regards,
Koen
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.5 (Darwin)
iD8DBQFGU/9cMkyGM64RGpERAst6AJwKBtrF10aXL64PINzp64UeS2MG5ACeKOc3
plSyJuOetHvaqxZyw8OfCTM=
=N20f
-----END PGP SIGNATURE-----
/*
*
* Copyright © 2000 SuSE, Inc.
*
* Permission to use, copy, modify, distribute, and sell this software and its
* documentation for any purpose is hereby granted without fee, provided that
* the above copyright notice appear in all copies and that both that
* copyright notice and this permission notice appear in supporting
* documentation, and that the name of SuSE not be used in advertising or
* publicity pertaining to distribution of the software without specific,
* written prior permission. SuSE makes no representations about the
* suitability of this software for any purpose. It is provided "as is"
* without express or implied warranty.
*
* SuSE DISCLAIMS ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SuSE
* BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
*
* Author: Keith Packard, SuSE, Inc.
*/
#ifdef HAVE_DIX_CONFIG_H
#include <dix-config.h>
#endif
#include <string.h>
#include "fb.h"
#ifdef RENDER
#include "picturestr.h"
#include "mipict.h"
#include "fbpict.h"
#include "fbmmx.h"
typedef void (*CompositeFunc) (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
fbOver (CARD32 x, CARD32 y)
{
CARD16 a = ~x >> 24;
CARD16 t;
CARD32 m,n,o,p;
m = FbOverU(x,y,0,a,t);
n = FbOverU(x,y,8,a,t);
o = FbOverU(x,y,16,a,t);
p = FbOverU(x,y,24,a,t);
return m|n|o|p;
}
CARD32
fbOver24 (CARD32 x, CARD32 y)
{
CARD16 a = ~x >> 24;
CARD16 t;
CARD32 m,n,o;
m = FbOverU(x,y,0,a,t);
n = FbOverU(x,y,8,a,t);
o = FbOverU(x,y,16,a,t);
return m|n|o;
}
CARD32
fbIn (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 = FbInU(x,24,a,t);
return m|n|o|p;
}
/*
* Naming convention:
*
* opSRCxMASKxDST
*/
void
fbCompositeSolidMask_nx8x8888 (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, srca;
CARD32 *dstLine, *dst, d, dstMask;
CARD8 *maskLine, *mask, m;
FbStride dstStride, maskStride;
CARD16 w;
fbComposeGetSolid(pSrc, src, pDst->format);
dstMask = FbFullMask (pDst->pDrawable->depth);
srca = src >> 24;
if (src == 0)
return;
fbComposeGetStart (pDst, xDst, yDst, CARD32, 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 = src & dstMask;
else
*dst = fbOver (src, *dst) & dstMask;
}
else if (m)
{
d = fbIn (src, m);
*dst = fbOver (d, *dst) & dstMask;
}
dst++;
}
}
}
void
fbCompositeSolidMask_nx8888x8888C (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, srca;
CARD32 *dstLine, *dst, d, dstMask;
CARD32 *maskLine, *mask, ma;
FbStride dstStride, maskStride;
CARD16 w;
CARD32 m, n, o, p;
fbComposeGetSolid(pSrc, src, pDst->format);
dstMask = FbFullMask (pDst->pDrawable->depth);
srca = src >> 24;
if (src == 0)
return;
fbComposeGetStart (pDst, xDst, yDst, CARD32, 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--)
{
ma = *mask++;
if (ma == 0xffffffff)
{
if (srca == 0xff)
*dst = src & dstMask;
else
*dst = fbOver (src, *dst) & dstMask;
}
else if (ma)
{
d = *dst;
#define FbInOverC(src,srca,msk,dst,i,result) { \
CARD16 __a = FbGet8(msk,i); \
CARD32 __t, __ta; \
CARD32 __i; \
__t = FbIntMult (FbGet8(src,i), __a,__i); \
__ta = (CARD8) ~FbIntMult (srca, __a,__i); \
__t = __t + FbIntMult(FbGet8(dst,i),__ta,__i); \
__t = (CARD32) (CARD8) (__t | (-(__t >> 8))); \
result = __t << (i); \
}
FbInOverC (src, srca, ma, d, 0, m);
FbInOverC (src, srca, ma, d, 8, n);
FbInOverC (src, srca, ma, d, 16, o);
FbInOverC (src, srca, ma, d, 24, p);
*dst = m|n|o|p;
}
dst++;
}
}
}
void
fbCompositeSolidMask_nx8x0888 (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, srca;
CARD8 *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, 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 = *mask++;
if (m == 0xff)
{
if (srca == 0xff)
d = src;
else
{
d = Fetch24(dst);
d = fbOver24 (src, d);
}
Store24(dst,d);
}
else if (m)
{
d = fbOver24 (fbIn(src,m), Fetch24(dst));
Store24(dst,d);
}
dst += 3;
}
}
}
void
fbCompositeSolidMask_nx8x0565 (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, 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 = *mask++;
if (m == 0xff)
{
if (srca == 0xff)
d = src;
else
{
d = *dst;
d = fbOver24 (src, cvt0565to8888(d));
}
*dst = cvt8888to0565(d);
}
else if (m)
{
d = *dst;
d = fbOver24 (fbIn(src,m), cvt0565to8888(d));
*dst = cvt8888to0565(d);
}
dst++;
}
}
}
void
fbCompositeSolidMask_nx8888x0565C (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, srca;
CARD16 src16;
CARD16 *dstLine, *dst;
CARD32 d;
CARD32 *maskLine, *mask, ma;
FbStride dstStride, maskStride;
CARD16 w;
CARD32 m, n, o;
fbComposeGetSolid(pSrc, src, pDst->format);
srca = src >> 24;
if (src == 0)
return;
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--)
{
ma = *mask++;
if (ma == 0xffffffff)
{
if (srca == 0xff)
{
*dst = src16;
}
else
{
d = *dst;
d = fbOver24 (src, cvt0565to8888(d));
*dst = cvt8888to0565(d);
}
}
else if (ma)
{
d = *dst;
d = cvt0565to8888(d);
FbInOverC (src, srca, ma, d, 0, m);
FbInOverC (src, srca, ma, d, 8, n);
FbInOverC (src, srca, ma, d, 16, o);
d = m|n|o;
*dst = cvt8888to0565(d);
}
dst++;
}
}
}
void
fbCompositeSrc_8888x8888 (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 *dstLine, *dst, dstMask;
CARD32 *srcLine, *src, s;
FbStride dstStride, srcStride;
CARD8 a;
CARD16 w;
fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
dstMask = FbFullMask (pDst->pDrawable->depth);
while (height--)
{
dst = dstLine;
dstLine += dstStride;
src = srcLine;
srcLine += srcStride;
w = width;
while (w--)
{
s = *src++;
a = s >> 24;
if (a == 0xff)
*dst = s & dstMask;
else if (a)
*dst = fbOver (s, *dst) & dstMask;
dst++;
}
}
}
void
fbCompositeSrc_8888x0888 (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;
CARD32 d;
CARD32 *srcLine, *src, s;
CARD8 a;
FbStride dstStride, srcStride;
CARD16 w;
fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3);
fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
while (height--)
{
dst = dstLine;
dstLine += dstStride;
src = srcLine;
srcLine += srcStride;
w = width;
while (w--)
{
s = *src++;
a = s >> 24;
if (a)
{
if (a == 0xff)
d = s;
else
d = fbOver24 (s, Fetch24(dst));
Store24(dst,d);
}
dst += 3;
}
}
}
void
fbCompositeSrc_8888x0565 (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;
CARD32 d;
CARD32 *srcLine, *src, s;
CARD8 a;
FbStride dstStride, srcStride;
CARD16 w;
fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, 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 = *src++;
a = s >> 24;
if (a)
{
if (a == 0xff)
d = s;
else
{
d = *dst;
d = fbOver24 (s, cvt0565to8888(d));
}
*dst = cvt8888to0565(d);
}
dst++;
}
}
}
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--)
*dst++ = *src++;
}
}
#define cvt8888Revto0565(s) ((((s) >> 19) & 0x001f) | \
(((s) >> 5) & 0x07e0) | \
(((s) << 8) & 0xf800))
#define FbOverU_8888RevNPx565(s, d) \
\
/* Extract alpha */ \
s_a = (s) >> 24; \
\
/* Extract r8g8b8 color channels */ \
s_r = ( (s) & 0xff); \
s_g = (((s) >> 8) & 0xff); \
s_b = (((s) >> 16) & 0xff); \
\
/* Extract r5g6b5 color channels */ \
d_r = ((d) >> 8) & 0xf8; \
d_g = ((d) >> 3) & 0xfc; \
d_b = ((d) << 3) & 0xf8; \
\
/* Use the higher bits of the channel to fill out the bottom */ \
d_r |= (d_r >> 5); \
d_g |= (d_g >> 6); \
d_b |= (d_b >> 5); \
\
/* Blend */ \
d_r = (s_r - d_r) * s_a + (d_r << 8); \
d_g = (s_g - d_g) * s_a + (d_g << 8); \
d_b = (s_b - d_b) * s_a + (d_b << 8); \
\
/* Pack result as r5g6b5 */ \
(d) = (d_r & 0xf800) | ((d_g & 0xfc00) >> 5) | (d_b >> 11)
void
fbCompositeSrc_8888RevNPx0565 (FbComposeData *params)
{
CARD16 *dstLine, *dst;
CARD32 *srcLine, *src;
FbStride dstStride, srcStride;
int w, h;
fbComposeGetStart (params->src, params->xSrc, params->ySrc, CARD32,
srcStride, srcLine, 1);
fbComposeGetStart (params->dest, params->xDest, params->yDest, CARD16,
dstStride, dstLine, 1);
h = params->height;
while (h--)
{
CARD32 s1, s2, s3, s4;
int d_r, d_g, d_b, s_r, s_g, s_b, s_a;
CARD32 *dst_2px_wide;
src = srcLine;
srcLine += srcStride;
dst_2px_wide = (CARD32 *) dstLine;
dstLine += dstStride;
w = params->width - 4;
while (w >= 0)
{
s1 = *src;
s2 = *(src + 1);
s3 = *(src + 2);
s4 = *(src + 3);
w -= 4;
src += 4;
/* Check if the next 4 pixels are opaque */
if ((s1 & s2 & s3 & s4) > 0xfeffffff)
{
/* In this case, we just perform a SOURCE for all 4 pixels */
#if X_BYTE_ORDER == X_BIG_ENDIAN
*dst_2px_wide++ = (cvt8888Revto0565 (s1) << 16) |
cvt8888Revto0565 (s2);
*dst_2px_wide++ = (cvt8888Revto0565 (s3) << 16) |
cvt8888Revto0565 (s4);
#else
*dst_2px_wide++ = cvt8888Revto0565 (s1) |
(cvt8888Revto0565 (s2) << 16);
*dst_2px_wide++ = cvt8888Revto0565 (s3) |
(cvt8888Revto0565 (s4) << 16);
#endif
}
/* Next, check if the next 4 pixels have any alpha in them at all */
else if ((s1 | s2 | s3 | s4) > 0x00ffffff)
{
/* In which case, we perform OVER on each one of them */
CARD32 d1, d2, d3, d4;
#if X_BYTE_ORDER == X_BIG_ENDIAN
d1 = (*dst_2px_wide >> 16);
d2 = (*dst_2px_wide & 0xffff);
FbOverU_8888RevNPx565 (s1, d1);
FbOverU_8888RevNPx565 (s2, d2);
*dst_2px_wide++ = (d1 << 16) | d2;
#else
d2 = (*dst_2px_wide >> 16);
d1 = (*dst_2px_wide & 0xffff);
FbOverU_8888RevNPx565 (s1, d1);
FbOverU_8888RevNPx565 (s2, d2);
*dst_2px_wide++ = d1 | (d2 << 16);
#endif
#if X_BYTE_ORDER == X_BIG_ENDIAN
d3 = (*dst_2px_wide >> 16);
d4 = (*dst_2px_wide & 0xffff);
FbOverU_8888RevNPx565 (s3, d3);
FbOverU_8888RevNPx565 (s4, d4);
*dst_2px_wide++ = (d3 << 16) | d4;
#else
d4 = (*dst_2px_wide >> 16);
d3 = (*dst_2px_wide & 0xffff);
FbOverU_8888RevNPx565 (s3, d3);
FbOverU_8888RevNPx565 (s4, d4);
*dst_2px_wide++ = d3 | (d4 << 16);
#endif
}
else
{
/* Do nothing, since the source pixels are all transparent */
dst_2px_wide += 2;
}
}
/* Deal with left over pixels */
for (dst = (CARD16 *) dst_2px_wide; w > -4; w--)
{
CARD32 d = *dst;
CARD32 s = *src++;
FbOverU_8888RevNPx565 (s, d);
*dst++ = d;
}
}
}
void
fbCompositeSrcAdd_8000x8000 (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;
CARD8 *srcLine, *src;
FbStride dstStride, srcStride;
CARD16 w;
CARD8 s, d;
CARD16 t;
fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1);
fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1);
while (height--)
{
dst = dstLine;
dstLine += dstStride;
src = srcLine;
srcLine += srcStride;
w = width;
while (w--)
{
s = *src++;
if (s)
{
if (s != 0xff)
{
d = *dst;
t = d + s;
s = t | (0 - (t >> 8));
}
*dst = s;
}
dst++;
}
}
}
void
fbCompositeSrcAdd_8888x8888 (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 *dstLine, *dst;
CARD32 *srcLine, *src;
FbStride dstStride, srcStride;
CARD16 w;
CARD32 s, d;
CARD16 t;
CARD32 m,n,o,p;
fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1);
fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1);
while (height--)
{
dst = dstLine;
dstLine += dstStride;
src = srcLine;
srcLine += srcStride;
w = width;
while (w--)
{
s = *src++;
if (s)
{
if (s != 0xffffffff)
{
d = *dst;
if (d)
{
m = FbAdd(s,d,0,t);
n = FbAdd(s,d,8,t);
o = FbAdd(s,d,16,t);
p = FbAdd(s,d,24,t);
s = m|n|o|p;
}
}
*dst = s;
}
dst++;
}
}
}
void
fbCompositeSrcAdd_1000x1000 (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 *dstBits, *srcBits;
FbStride dstStride, srcStride;
int dstBpp, srcBpp;
int dstXoff, dstYoff;
int srcXoff, srcYoff;
fbGetDrawable(pSrc->pDrawable, srcBits, srcStride, srcBpp, srcXoff, srcYoff);
fbGetDrawable(pDst->pDrawable, dstBits, dstStride, dstBpp, dstXoff, dstYoff);
fbBlt (srcBits + srcStride * (ySrc + srcYoff),
srcStride,
xSrc + srcXoff,
dstBits + dstStride * (yDst + dstYoff),
dstStride,
xDst + dstXoff,
width,
height,
GXor,
FB_ALLONES,
srcBpp,
FALSE,
FALSE);
}
void
fbCompositeSolidMask_nx1xn (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 *dstBits;
FbStip *maskBits;
FbStride dstStride, maskStride;
int dstBpp, maskBpp;
int dstXoff, dstYoff;
int maskXoff, maskYoff;
FbBits src;
fbComposeGetSolid(pSrc, src, pDst->format);
if ((src & 0xff000000) != 0xff000000)
{
fbCompositeGeneral (op, pSrc, pMask, pDst,
xSrc, ySrc, xMask, yMask, xDst, yDst,
width, height);
return;
}
fbGetStipDrawable (pMask->pDrawable, maskBits, maskStride, maskBpp, maskXoff, maskYoff);
fbGetDrawable (pDst->pDrawable, dstBits, dstStride, dstBpp, dstXoff, dstYoff);
switch (dstBpp) {
case 32:
break;
case 24:
break;
case 16:
src = cvt8888to0565(src);
break;
}
src = fbReplicatePixel (src, dstBpp);
fbBltOne (maskBits + maskStride * (yMask + maskYoff),
maskStride,
xMask + maskXoff,
dstBits + dstStride * (yDst + dstYoff),
dstStride,
(xDst + dstXoff) * dstBpp,
dstBpp,
width * dstBpp,
height,
0x0,
src,
FB_ALLONES,
0x0);
}
# define mod(a,b) ((b) == 1 ? 0 : (a) >= 0 ? (a) % (b) : (b) - (-a) % (b))
void
fbComposite (CARD8 op,
PicturePtr pSrc,
PicturePtr pMask,
PicturePtr pDst,
INT16 xSrc,
INT16 ySrc,
INT16 xMask,
INT16 yMask,
INT16 xDst,
INT16 yDst,
CARD16 width,
CARD16 height)
{
RegionRec region;
int n;
BoxPtr pbox;
CompositeFunc func = NULL;
Bool srcRepeat = pSrc->pDrawable && pSrc->repeat;
Bool maskRepeat = FALSE;
Bool srcAlphaMap = pSrc->alphaMap != 0;
Bool maskAlphaMap = FALSE;
Bool dstAlphaMap = pDst->alphaMap != 0;
int x_msk, y_msk, x_src, y_src, x_dst, y_dst;
int w, h, w_this, h_this;
#ifdef USE_MMX
static Bool mmx_setup = FALSE;
if (!mmx_setup) {
fbComposeSetupMMX();
mmx_setup = TRUE;
}
#endif
xDst += pDst->pDrawable->x;
yDst += pDst->pDrawable->y;
if (pSrc->pDrawable) {
xSrc += pSrc->pDrawable->x;
ySrc += pSrc->pDrawable->y;
}
if (pMask && pMask->pDrawable)
{
xMask += pMask->pDrawable->x;
yMask += pMask->pDrawable->y;
maskRepeat = pMask->repeat == RepeatNormal;
maskAlphaMap = pMask->alphaMap != 0;
}
if (pSrc->pDrawable && (!pMask || pMask->pDrawable)
&& !pSrc->transform && !(pMask && pMask->transform)
&& !maskAlphaMap && !srcAlphaMap && !dstAlphaMap
&& (pSrc->filter != PictFilterConvolution)
&& (!pMask || pMask->filter != PictFilterConvolution))
switch (op) {
case PictOpSrc:
#ifdef USE_MMX
if (!pMask && pSrc->format == pDst->format &&
pSrc->format != PICT_a8 && pSrc->pDrawable != pDst->pDrawable)
{
func = fbCompositeCopyAreammx;
}
#endif
break;
case PictOpOver:
if (pMask)
{
if (fbCanGetSolid(pSrc) &&
!maskRepeat)
{
srcRepeat = FALSE;
if (PICT_FORMAT_COLOR(pSrc->format)) {
switch (pMask->format) {
case PICT_a8:
switch (pDst->format) {
case PICT_r5g6b5:
case PICT_b5g6r5:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8x0565mmx;
else
#endif
func = fbCompositeSolidMask_nx8x0565;
break;
case PICT_r8g8b8:
case PICT_b8g8r8:
func = fbCompositeSolidMask_nx8x0888;
break;
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8x8888mmx;
else
#endif
func = fbCompositeSolidMask_nx8x8888;
break;
default:
break;
}
break;
case PICT_a8r8g8b8:
if (pMask->componentAlpha) {
switch (pDst->format) {
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8888x8888Cmmx;
else
#endif
func = fbCompositeSolidMask_nx8888x8888C;
break;
case PICT_r5g6b5:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8888x0565Cmmx;
else
#endif
func = fbCompositeSolidMask_nx8888x0565C;
break;
default:
break;
}
}
break;
case PICT_a8b8g8r8:
if (pMask->componentAlpha) {
switch (pDst->format) {
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8888x8888Cmmx;
else
#endif
func = fbCompositeSolidMask_nx8888x8888C;
break;
case PICT_b5g6r5:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSolidMask_nx8888x0565Cmmx;
else
#endif
func = fbCompositeSolidMask_nx8888x0565C;
break;
default:
break;
}
}
break;
case PICT_a1:
switch (pDst->format) {
case PICT_r5g6b5:
case PICT_b5g6r5:
case PICT_r8g8b8:
case PICT_b8g8r8:
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
func = fbCompositeSolidMask_nx1xn;
break;
default:
break;
}
break;
default:
break;
}
default:
break;
}
}
else if (! srcRepeat) /* has mask and non-repeating source */
{
if (pSrc->pDrawable == pMask->pDrawable &&
xSrc == xMask && ySrc == yMask &&
!pMask->componentAlpha && !maskRepeat)
{
/* source == mask: non-premultiplied data */
switch (pSrc->format) {
case PICT_x8b8g8r8:
switch (pMask->format) {
case PICT_a8r8g8b8:
case PICT_a8b8g8r8:
switch (pDst->format) {
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888RevNPx8888mmx;
#endif
break;
case PICT_r5g6b5:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888RevNPx0565mmx;
#endif
break;
default:
break;
}
break;
default:
break;
}
break;
case PICT_x8r8g8b8:
switch (pMask->format) {
case PICT_a8r8g8b8:
case PICT_a8b8g8r8:
switch (pDst->format) {
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888RevNPx8888mmx;
#endif
break;
case PICT_r5g6b5:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888RevNPx0565mmx;
else
#endif
func = fbCompositeSrc_8888RevNPx0565;
break;
default:
break;
}
break;
default:
break;
}
break;
default:
break;
}
break;
}
else
{
/* non-repeating source, repeating mask => translucent window */
if (fbCanGetSolid(pMask))
{
if (pSrc->format == PICT_x8r8g8b8 &&
pDst->format == PICT_x8r8g8b8 &&
pMask->format == PICT_a8)
{
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888x8x8888mmx;
#endif
}
}
}
}
}
else /* no mask */
{
if (fbCanGetSolid(pSrc))
{
/* no mask and repeating source */
switch (pSrc->format) {
case PICT_a8r8g8b8:
switch (pDst->format) {
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
{
srcRepeat = FALSE;
func = fbCompositeSolid_nx8888mmx;
}
#endif
break;
case PICT_r5g6b5:
#ifdef USE_MMX
if (fbHaveMMX())
{
srcRepeat = FALSE;
func = fbCompositeSolid_nx0565mmx;
}
#endif
break;
default:
break;
}
break;
default:
break;
}
}
else if (! srcRepeat)
{
switch (pSrc->format) {
case PICT_a8r8g8b8:
switch (pDst->format) {
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888x8888mmx;
else
#endif
func = fbCompositeSrc_8888x8888;
break;
case PICT_r8g8b8:
func = fbCompositeSrc_8888x0888;
break;
case PICT_r5g6b5:
func = fbCompositeSrc_8888x0565;
break;
default:
break;
}
break;
case PICT_x8r8g8b8:
switch (pDst->format) {
case PICT_a8r8g8b8:
case PICT_x8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeCopyAreammx;
#endif
break;
default:
break;
}
case PICT_x8b8g8r8:
switch (pDst->format) {
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeCopyAreammx;
#endif
break;
default:
break;
}
break;
case PICT_a8b8g8r8:
switch (pDst->format) {
case PICT_a8b8g8r8:
case PICT_x8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrc_8888x8888mmx;
else
#endif
func = fbCompositeSrc_8888x8888;
break;
case PICT_b8g8r8:
func = fbCompositeSrc_8888x0888;
break;
case PICT_b5g6r5:
func = fbCompositeSrc_8888x0565;
break;
default:
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;
}
}
}
break;
case PictOpAdd:
if (pMask == 0)
{
switch (pSrc->format) {
case PICT_a8r8g8b8:
switch (pDst->format) {
case PICT_a8r8g8b8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrcAdd_8888x8888mmx;
else
#endif
func = fbCompositeSrcAdd_8888x8888;
break;
default:
break;
}
break;
case PICT_a8b8g8r8:
switch (pDst->format) {
case PICT_a8b8g8r8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrcAdd_8888x8888mmx;
else
#endif
func = fbCompositeSrcAdd_8888x8888;
break;
default:
break;
}
break;
case PICT_a8:
switch (pDst->format) {
case PICT_a8:
#ifdef USE_MMX
if (fbHaveMMX())
func = fbCompositeSrcAdd_8000x8000mmx;
else
#endif
func = fbCompositeSrcAdd_8000x8000;
break;
default:
break;
}
break;
case PICT_a1:
switch (pDst->format) {
case PICT_a1:
func = fbCompositeSrcAdd_1000x1000;
break;
default:
break;
}
break;
default:
break;
}
}
break;
}
if (!func) {
/* no fast path, use the general code */
fbCompositeGeneral(op, pSrc, pMask, pDst, xSrc, ySrc, xMask, yMask, xDst, yDst, width, height);
return;
}
if (!miComputeCompositeRegion (®ion,
pSrc,
pMask,
pDst,
xSrc,
ySrc,
xMask,
yMask,
xDst,
yDst,
width,
height))
return;
n = REGION_NUM_RECTS (®ion);
pbox = REGION_RECTS (®ion);
while (n--)
{
h = pbox->y2 - pbox->y1;
y_src = pbox->y1 - yDst + ySrc;
y_msk = pbox->y1 - yDst + yMask;
y_dst = pbox->y1;
while (h)
{
h_this = h;
w = pbox->x2 - pbox->x1;
x_src = pbox->x1 - xDst + xSrc;
x_msk = pbox->x1 - xDst + xMask;
x_dst = pbox->x1;
if (maskRepeat)
{
y_msk = mod (y_msk - pMask->pDrawable->y, pMask->pDrawable->height);
if (h_this > pMask->pDrawable->height - y_msk)
h_this = pMask->pDrawable->height - y_msk;
y_msk += pMask->pDrawable->y;
}
if (srcRepeat)
{
y_src = mod (y_src - pSrc->pDrawable->y, pSrc->pDrawable->height);
if (h_this > pSrc->pDrawable->height - y_src)
h_this = pSrc->pDrawable->height - y_src;
y_src += pSrc->pDrawable->y;
}
while (w)
{
w_this = w;
if (maskRepeat)
{
x_msk = mod (x_msk - pMask->pDrawable->x, pMask->pDrawable->width);
if (w_this > pMask->pDrawable->width - x_msk)
w_this = pMask->pDrawable->width - x_msk;
x_msk += pMask->pDrawable->x;
}
if (srcRepeat)
{
x_src = mod (x_src - pSrc->pDrawable->x, pSrc->pDrawable->width);
if (w_this > pSrc->pDrawable->width - x_src)
w_this = pSrc->pDrawable->width - x_src;
x_src += pSrc->pDrawable->x;
}
(*func) (op, pSrc, pMask, pDst,
x_src, y_src, x_msk, y_msk, x_dst, y_dst,
w_this, h_this);
w -= w_this;
x_src += w_this;
x_msk += w_this;
x_dst += w_this;
}
h -= h_this;
y_src += h_this;
y_msk += h_this;
y_dst += h_this;
}
pbox++;
}
REGION_UNINIT (pDst->pDrawable->pScreen, ®ion);
}
#endif /* RENDER */
Bool
fbPictureInit (ScreenPtr pScreen, PictFormatPtr formats, int nformats)
{
#ifdef RENDER
PictureScreenPtr ps;
if (!miPictureInit (pScreen, formats, nformats))
return FALSE;
ps = GetPictureScreen(pScreen);
ps->Composite = fbComposite;
ps->Glyphs = miGlyphs;
ps->CompositeRects = miCompositeRects;
ps->RasterizeTrapezoid = fbRasterizeTrapezoid;
ps->AddTraps = fbAddTraps;
ps->AddTriangles = fbAddTriangles;
#endif /* RENDER */
return TRUE;
}
#ifdef USE_MMX
/* The CPU detection code needs to be in a file not compiled with
* "-mmmx -msse", as gcc would generate CMOV instructions otherwise
* that would lead to SIGILL instructions on old CPUs that don't have
* it.
*/
#if !defined(__amd64__) && !defined(__x86_64__)
#ifdef HAVE_GETISAX
#include <sys/auxv.h>
#endif
enum CPUFeatures {
NoFeatures = 0,
MMX = 0x1,
MMX_Extensions = 0x2,
SSE = 0x6,
SSE2 = 0x8,
CMOV = 0x10
};
static unsigned int detectCPUFeatures(void) {
unsigned int features = 0;
unsigned int result;
#ifdef HAVE_GETISAX
if (getisax(&result, 1)) {
if (result & AV_386_CMOV)
features |= CMOV;
if (result & AV_386_MMX)
features |= MMX;
if (result & AV_386_AMD_MMX)
features |= MMX_Extensions;
if (result & AV_386_SSE)
features |= SSE;
if (result & AV_386_SSE2)
features |= SSE2;
}
#else
char vendor[13];
vendor[0] = 0;
vendor[12] = 0;
/* see p. 118 of amd64 instruction set manual Vol3 */
/* We need to be careful about the handling of %ebx and
* %esp here. We can't declare either one as clobbered
* since they are special registers (%ebx is the "PIC
* register" holding an offset to global data, %esp the
* stack pointer), so we need to make sure they have their
* original values when we access the output operands.
*/
__asm__ ("pushf\n"
"pop %%eax\n"
"mov %%eax, %%ecx\n"
"xor $0x00200000, %%eax\n"
"push %%eax\n"
"popf\n"
"pushf\n"
"pop %%eax\n"
"mov $0x0, %%edx\n"
"xor %%ecx, %%eax\n"
"jz 1\n"
"mov $0x00000000, %%eax\n"
"push %%ebx\n"
"cpuid\n"
"mov %%ebx, %%eax\n"
"pop %%ebx\n"
"mov %%eax, %1\n"
"mov %%edx, %2\n"
"mov %%ecx, %3\n"
"mov $0x00000001, %%eax\n"
"push %%ebx\n"
"cpuid\n"
"pop %%ebx\n"
"1:\n"
"mov %%edx, %0\n"
: "=r" (result),
"=m" (vendor[0]),
"=m" (vendor[4]),
"=m" (vendor[8])
:
: "%eax", "%ecx", "%edx"
);
if (result) {
/* result now contains the standard feature bits */
if (result & (1 << 15))
features |= CMOV;
if (result & (1 << 23))
features |= MMX;
if (result & (1 << 25))
features |= SSE;
if (result & (1 << 26))
features |= SSE2;
if ((result & MMX) && !(result & SSE) && (strcmp(vendor, "AuthenticAMD") == 0)) {
/* check for AMD MMX extensions */
unsigned int result;
__asm__("push %%ebx\n"
"mov $0x80000000, %%eax\n"
"cpuid\n"
"xor %%edx, %%edx\n"
"cmp $0x1, %%eax\n"
"jge 2\n"
"mov $0x80000001, %%eax\n"
"cpuid\n"
"2:\n"
"pop %%ebx\n"
"mov %%edx, %0\n"
: "=r" (result)
:
: "%eax", "%ecx", "%edx"
);
if (result & (1<<22))
features |= MMX_Extensions;
}
}
#endif /* HAVE_GETISAX */
return features;
}
Bool
fbHaveMMX (void)
{
static Bool initialized = FALSE;
static Bool mmx_present;
if (!initialized)
{
unsigned int features = detectCPUFeatures();
mmx_present = (features & (MMX|MMX_Extensions)) == (MMX|MMX_Extensions);
initialized = TRUE;
}
return mmx_present;
}
#endif /* __amd64__ */
#endif
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]
