From c7e2ba0c9b9b1fc1aed8f91f86471c4c8e650b78 Mon Sep 17 00:00:00 2001 From: Peter Hutterer Date: Tue, 24 Apr 2007 21:34:47 +0930 Subject: [PATCH 01/47] Use DevicePresence events to tell the client about enabled/disabled devices. Include the device id in the event sent to the client. --- dix/devices.c | 33 +++++++++++++++++++++++++++++---- hw/xfree86/common/xf86Xinput.c | 25 +++++++++++++++++++++++++ hw/xfree86/common/xf86Xinput.h | 3 +++ 3 files changed, 57 insertions(+), 4 deletions(-) diff --git a/dix/devices.c b/dix/devices.c index 7d8fd0398..e8ca5dfa7 100644 --- a/dix/devices.c +++ b/dix/devices.c @@ -75,6 +75,7 @@ SOFTWARE. #include "swaprep.h" #include "dixevents.h" +#include #include #include "exglobals.h" #include "exevents.h" @@ -157,6 +158,8 @@ EnableDevice(DeviceIntPtr dev) { DeviceIntPtr *prev; int ret; + DeviceIntRec dummyDev; + devicePresenceNotify ev; for (prev = &inputInfo.off_devices; *prev && (*prev != dev); @@ -175,6 +178,14 @@ EnableDevice(DeviceIntPtr dev) *prev = dev; dev->next = NULL; + ev.type = DevicePresenceNotify; + ev.time = currentTime.milliseconds; + ev.devchange = DeviceEnabled; + ev.deviceid = dev->id; + dummyDev.id = 0; + SendEventToAllWindows(&dummyDev, DevicePresenceNotifyMask, + (xEvent *) &ev, 1); + return TRUE; } @@ -182,6 +193,8 @@ Bool DisableDevice(DeviceIntPtr dev) { DeviceIntPtr *prev; + DeviceIntRec dummyDev; + devicePresenceNotify ev; for (prev = &inputInfo.devices; *prev && (*prev != dev); @@ -194,6 +207,15 @@ DisableDevice(DeviceIntPtr dev) *prev = dev->next; dev->next = inputInfo.off_devices; inputInfo.off_devices = dev; + + ev.type = DevicePresenceNotify; + ev.time = currentTime.milliseconds; + ev.devchange = DeviceDisabled; + ev.deviceid = dev->id; + dummyDev.id = 0; + SendEventToAllWindows(&dummyDev, DevicePresenceNotifyMask, + (xEvent *) &ev, 1); + return TRUE; } @@ -212,8 +234,8 @@ ActivateDevice(DeviceIntPtr dev) ev.type = DevicePresenceNotify; ev.time = currentTime.milliseconds; - ev.devchange = 0; - ev.deviceid = 0; + ev.devchange = DeviceAdded; + ev.deviceid = dev->id; dummyDev.id = 0; SendEventToAllWindows(&dummyDev, DevicePresenceNotifyMask, (xEvent *) &ev, 1); @@ -547,12 +569,15 @@ RemoveDevice(DeviceIntPtr dev) int ret = BadMatch; devicePresenceNotify ev; DeviceIntRec dummyDev; + int deviceid; DebugF("(dix) removing device %d\n", dev->id); if (!dev || dev == inputInfo.keyboard || dev == inputInfo.pointer) return BadImplementation; + deviceid = dev->id; + prev = NULL; for (tmp = inputInfo.devices; tmp; (prev = tmp), (tmp = next)) { next = tmp->next; @@ -587,8 +612,8 @@ RemoveDevice(DeviceIntPtr dev) inputInfo.numDevices--; ev.type = DevicePresenceNotify; ev.time = currentTime.milliseconds; - ev.devchange = 0; - ev.deviceid = 0; + ev.devchange = DeviceRemoved; + ev.deviceid = deviceid; dummyDev.id = 0; SendEventToAllWindows(&dummyDev, DevicePresenceNotifyMask, (xEvent *) &ev, 1); diff --git a/hw/xfree86/common/xf86Xinput.c b/hw/xfree86/common/xf86Xinput.c index 6ebb087cf..0af1c6a59 100644 --- a/hw/xfree86/common/xf86Xinput.c +++ b/hw/xfree86/common/xf86Xinput.c @@ -764,4 +764,29 @@ xf86InitValuatorDefaults(DeviceIntPtr dev, int axnum) } } + +/** + * Deactivate a device. Call this function from the driver if you receive a + * read error or something else that spoils your day. + * Device will be moved to the off_devices list, but it will still be there + * until you really clean up after it. + * Notifies the client about an inactive device. + */ +_X_EXPORT void +xf86DisableDevice(DeviceIntPtr dev) +{ + DisableDevice(dev); +} + +/** + * Reactivate a device. Call this function from the driver if you just found + * out that the read error wasn't quite that bad after all. + * Device will be re-activated, and an event sent to the client. + */ +_X_EXPORT void +xf86EnableDevice(DeviceIntPtr dev) +{ + EnableDevice(dev); +} + /* end of xf86Xinput.c */ diff --git a/hw/xfree86/common/xf86Xinput.h b/hw/xfree86/common/xf86Xinput.h index b2bc8dec1..f5beb6c94 100644 --- a/hw/xfree86/common/xf86Xinput.h +++ b/hw/xfree86/common/xf86Xinput.h @@ -187,6 +187,8 @@ void xf86InitValuatorAxisStruct(DeviceIntPtr dev, int axnum, int minval, void xf86InitValuatorDefaults(DeviceIntPtr dev, int axnum); void xf86AddEnabledDevice(InputInfoPtr pInfo); void xf86RemoveEnabledDevice(InputInfoPtr pInfo); +void xf86DisableDevice(DeviceIntPtr dev); +void xf86EnableDevice(DeviceIntPtr dev); /* xf86Helper.c */ void xf86AddInputDriver(InputDriverPtr driver, pointer module, int flags); @@ -204,6 +206,7 @@ int xf86GetMotionEvents(DeviceIntPtr dev, xTimecoord *buff, void xf86CollectInputOptions(InputInfoPtr pInfo, const char **defaultOpts, pointer extraOpts); + /* Legacy hatred */ #define SendCoreEvents 59 #define DontSendCoreEvents 60 From 09436fb7c38a9819bde770c4c21143591671c4d7 Mon Sep 17 00:00:00 2001 From: Peter Hutterer Date: Tue, 24 Apr 2007 22:52:33 +0930 Subject: [PATCH 02/47] Disable devices before removing, remove unrecoverable devices. --- dix/devices.c | 1 + hw/xfree86/common/xf86Xinput.c | 23 +++++++++++++++++++++-- hw/xfree86/common/xf86Xinput.h | 2 +- 3 files changed, 23 insertions(+), 3 deletions(-) diff --git a/dix/devices.c b/dix/devices.c index e8ca5dfa7..e227617ef 100644 --- a/dix/devices.c +++ b/dix/devices.c @@ -577,6 +577,7 @@ RemoveDevice(DeviceIntPtr dev) return BadImplementation; deviceid = dev->id; + DisableDevice(dev); prev = NULL; for (tmp = inputInfo.devices; tmp; (prev = tmp), (tmp = next)) { diff --git a/hw/xfree86/common/xf86Xinput.c b/hw/xfree86/common/xf86Xinput.c index 0af1c6a59..2d0a37674 100644 --- a/hw/xfree86/common/xf86Xinput.c +++ b/hw/xfree86/common/xf86Xinput.c @@ -771,11 +771,30 @@ xf86InitValuatorDefaults(DeviceIntPtr dev, int axnum) * Device will be moved to the off_devices list, but it will still be there * until you really clean up after it. * Notifies the client about an inactive device. + * + * @param panic True if device is unrecoverable and needs to be removed. */ _X_EXPORT void -xf86DisableDevice(DeviceIntPtr dev) +xf86DisableDevice(DeviceIntPtr dev, Bool panic) { - DisableDevice(dev); + devicePresenceNotify ev; + DeviceIntRec dummyDev; + + if(!panic) + { + DisableDevice(dev); + } else + { + ev.type = DevicePresenceNotify; + ev.time = currentTime.milliseconds; + ev.devchange = DeviceUnrecoverable; + ev.deviceid = dev->id; + dummyDev.id = 0; + SendEventToAllWindows(&dummyDev, DevicePresenceNotifyMask, + (xEvent *) &ev, 1); + + DeleteInputDeviceRequest(dev); + } } /** diff --git a/hw/xfree86/common/xf86Xinput.h b/hw/xfree86/common/xf86Xinput.h index f5beb6c94..7ef28ed73 100644 --- a/hw/xfree86/common/xf86Xinput.h +++ b/hw/xfree86/common/xf86Xinput.h @@ -187,7 +187,7 @@ void xf86InitValuatorAxisStruct(DeviceIntPtr dev, int axnum, int minval, void xf86InitValuatorDefaults(DeviceIntPtr dev, int axnum); void xf86AddEnabledDevice(InputInfoPtr pInfo); void xf86RemoveEnabledDevice(InputInfoPtr pInfo); -void xf86DisableDevice(DeviceIntPtr dev); +void xf86DisableDevice(DeviceIntPtr dev, Bool panic); void xf86EnableDevice(DeviceIntPtr dev); /* xf86Helper.c */ From 077a5d4555676d5775e990468a697b6890c6d609 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 12:57:55 -0400 Subject: [PATCH 03/47] Add functions fbCompositeSrcSrc_nxn() and fbCompositeTrans_0565xnx0565 from xserver via pixman. Add READ/WRITE and fbFinishAccess as appropriate. --- fb/fbpict.c | 181 +++++++++++++++++++++++++++++++++++++++++++++++++++- fb/fbpict.h | 8 +++ 2 files changed, 188 insertions(+), 1 deletion(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index cd6cac277..bc737d8ec 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 op, # 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 434526e32..76cab5bf8 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; \ } \ From 13e1d5ea55b0a3b7729316c8e37d3d8fca2075b5 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 12:59:18 -0400 Subject: [PATCH 04/47] Fix format vs formatCode in previous commit --- fb/fbpict.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index bc737d8ec..072f8d891 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; From fde4a5adf02d3067a064ebf6bdd666aa5784cfe9 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 13:30:43 -0400 Subject: [PATCH 05/47] 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. --- fb/fbpict.c | 632 ++++++++++++++++++++++++++++++++++++++++++---------- 1 file changed, 512 insertions(+), 120 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 072f8d891..660112c7d 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 op, 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 op, 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 op, 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 op, 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 op, 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; From 2d9a7a768747ca39a800475f12c424c298018dc6 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 14:46:59 -0400 Subject: [PATCH 06/47] 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. --- fb/fbpict.c | 704 +++++++++++++++++++++++++++------------------------- fb/fbpict.h | 4 +- 2 files changed, 373 insertions(+), 335 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 660112c7d..d027e3571 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)); \ +/* + * 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)); \ + } -#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; \ @@ -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 op, { 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 op, while (height--) { - // fixme: cleanup unused + /* fixme: cleanup unused */ unsigned int wt,wd; CARD32 workingiDest; CARD32 *widst; @@ -407,61 +405,152 @@ fbCompositeSolidMask_nx8x0565 (CARD8 op, 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; - - fbComposeGetStart (pDst, xDst, yDst, CARD16, dstStride, dstLine, 1); - fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); + CARD32 src, srca8, srca5; + CARD16 *dstLine, *dst; + CARD16 d; + CARD32 t; + CARD8 *maskLine, *mask, m; + FbStride dstStride, maskStride; + CARD16 w,src16; - while (height--) - { - dst = dstLine; - dstLine += dstStride; - mask = maskLine; - maskLine += maskStride; - w = width; + 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, 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++; - } - } + 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 op, 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 op, /* * 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 op, 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 op, 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 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; 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 - - // 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 op, 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 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); + 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; + + 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, - // 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); + 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 76cab5bf8..5246cd5ac 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))) From 7e16da7b78c422f96387502b9cc29eaa1741543f Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 18:15:34 -0400 Subject: [PATCH 07/47] Remove #if 0'ed blocks --- fb/fbpict.c | 41 ----------------------------------------- 1 file changed, 41 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index d027e3571..8ca733604 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, From 0a2fe443d25b1ca25349aba3f748df986952e20f Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Tue, 24 Apr 2007 19:02:44 -0400 Subject: [PATCH 08/47] Use READ/WRITE macros for new functions introduced in previous commits. --- fb/fbpict.c | 515 ++++++++++++++++++++++++++-------------------------- 1 file changed, 257 insertions(+), 258 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 8ca733604..44bee1bca 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 op, 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++; + rd = READ(edst++); + gd = READ(edst++); + bd = READ(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; - } + 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 op, CARD16 w,src16; fbComposeGetSolid(pSrc, src, pDst->format); - - if (src == 0) return; @@ -437,23 +435,23 @@ fbCompositeSolidMask_nx8x0565 (CARD8 op, 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 op, t = fbIn (src, m); t = fbOver (t, cvt0565to0888 (d)); } - *dst++ = cvt8888to0565 (t); + WRITE(dst++, cvt8888to0565 (t)); } } } @@ -517,34 +515,34 @@ fbCompositeSolidMask_nx8888x0565 (CARD8 op, 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 op, 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 op, 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 op, { 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 op, /* 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 op, 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) +{ + 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; - 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; + 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; - } - } - } - } - } - + 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 From 3ba1e8ab6d69566e1a3f8f0eb4605631aeffc8e5 Mon Sep 17 00:00:00 2001 From: Aaron Plattner Date: Tue, 24 Apr 2007 17:20:14 -0700 Subject: [PATCH 09/47] Include xf86Rename.h in xf86RandR12.h. --- hw/xfree86/modes/xf86RandR12.h | 3 +++ 1 file changed, 3 insertions(+) diff --git a/hw/xfree86/modes/xf86RandR12.h b/hw/xfree86/modes/xf86RandR12.h index 0d3346a77..4fd855cf5 100644 --- a/hw/xfree86/modes/xf86RandR12.h +++ b/hw/xfree86/modes/xf86RandR12.h @@ -24,6 +24,9 @@ #define _XF86_RANDR_H_ #include #include +#if XF86_MODES_RENAME +#include "xf86Rename.h" +#endif Bool xf86RandR12CreateScreenResources (ScreenPtr pScreen); Bool xf86RandR12Init(ScreenPtr pScreen); From d322608dc929d5f8cda07a53143a4f28423e0460 Mon Sep 17 00:00:00 2001 From: Adam Jackson Date: Wed, 25 Apr 2007 16:29:48 -0400 Subject: [PATCH 10/47] Fix a buffer overrun on machines with excessively large PCI busses. Formerly we sized an array with a compile time constant, then initialized its size to the same constant, but the Linux PCI init code would increase that "constant". So if you happened to have more than 128 PCI devices, you'd happily scribble into whatever variables happened to be in .bss after that array. Only really fixed for Linux atm. Other OSes will simply (still) fail to work on video devices above the 128th PCI device. --- hw/xfree86/os-support/bus/Pci.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/hw/xfree86/os-support/bus/Pci.c b/hw/xfree86/os-support/bus/Pci.c index b80371d83..bc5e11fb8 100644 --- a/hw/xfree86/os-support/bus/Pci.c +++ b/hw/xfree86/os-support/bus/Pci.c @@ -232,14 +232,14 @@ _X_EXPORT int pciNumBuses = 0; /* Actual number of PCI buses */ int pciMaxBusNum = MAX_PCI_BUSES; static Bool inProbe = FALSE; -static pciConfigPtr pci_devp[MAX_PCI_DEVICES + 1] = {NULL, }; +static pciConfigPtr *pci_devp = NULL; static int readPciBios( PCITAG Tag, CARD8* tmp, ADDRESS hostbase, unsigned char * buf, int len, PciBiosType BiosType ); static int (*pciOSHandleBIOS)(PCITAG Tag, int basereg, unsigned char *buf, int len); -int xf86MaxPciDevs = MAX_PCI_DEVICES; +int xf86MaxPciDevs = 0; /* * Platform specific PCI function pointers. @@ -272,6 +272,14 @@ pciInit() if (pciNumBuses <= 0) ARCH_PCI_OS_INIT(); #endif + if (xf86MaxPciDevs == 0) { + xf86Msg(X_WARNING, + "OS did not count PCI devices, guessing wildly\n"); + xf86MaxPciDevs = MAX_PCI_DEVICES; + } + if (pci_devp) + xfree(pci_devp); + pci_devp = xnfcalloc(xf86MaxPciDevs + 1, sizeof(pciConfigPtr)); } void pciSetOSBIOSPtr(int (*bios_fn)(PCITAG Tag, int basereg, unsigned char * buf, int len)) @@ -920,7 +928,7 @@ xf86scanpci(int flags) * result in an endless recursion if platform/OS specific PCI * bus probing code calls this function from with in it. */ - if (done || pci_devp[0]) + if (done || pci_devp) return pci_devp; done = TRUE; From 9c80eda826448822328bb678a7d284cc43fffb17 Mon Sep 17 00:00:00 2001 From: Adam Jackson Date: Wed, 25 Apr 2007 16:35:04 -0400 Subject: [PATCH 11/47] Disable RANDR's fake Xinerama protocol when there's more than one screen. ... in the protocol sense. Xinerama doesn't have any provision for more than one protocol screen each with its own geometry. Red Hat bug #231257. --- randr/rrxinerama.c | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/randr/rrxinerama.c b/randr/rrxinerama.c index 1db27f14c..2a57e4e32 100644 --- a/randr/rrxinerama.c +++ b/randr/rrxinerama.c @@ -428,6 +428,14 @@ RRXineramaExtensionInit(void) return; #endif + /* + * Xinerama isn't capable enough to have multiple protocol screens each + * with their own output geometry. So if there's more than one protocol + * screen, just don't even try. + */ + if (screenInfo.numScreens > 1) + return; + (void) AddExtension(PANORAMIX_PROTOCOL_NAME, 0,0, ProcRRXineramaDispatch, SProcRRXineramaDispatch, From c09e68ce30dabd6b7068b163b9d2382d85d0d0bc Mon Sep 17 00:00:00 2001 From: Adam Jackson Date: Wed, 25 Apr 2007 16:46:26 -0400 Subject: [PATCH 12/47] Paper over a crash at exit during GLX teardown. --- GL/glx/glxglcore.c | 12 +++++++----- 1 file changed, 7 insertions(+), 5 deletions(-) diff --git a/GL/glx/glxglcore.c b/GL/glx/glxglcore.c index ecfa4d773..b50740c3c 100644 --- a/GL/glx/glxglcore.c +++ b/GL/glx/glxglcore.c @@ -258,12 +258,14 @@ __glXMesaScreenDestroy(__GLXscreen *screen) __GLXMESAscreen *mesaScreen = (__GLXMESAscreen *) screen; int i; - for (i = 0; i < mesaScreen->num_vis; i++) { - if (mesaScreen->xm_vis[i]) - XMesaDestroyVisual(mesaScreen->xm_vis[i]); - } + if (mesaScreen->xm_vis) { + for (i = 0; i < mesaScreen->num_vis; i++) { + if (mesaScreen->xm_vis[i]) + XMesaDestroyVisual(mesaScreen->xm_vis[i]); + } - xfree(mesaScreen->xm_vis); + xfree(mesaScreen->xm_vis); + } __glXScreenDestroy(screen); From 66ba3d758a368bf83d75bab8b08bdb6b34925e40 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Wed, 25 Apr 2007 10:31:38 -0400 Subject: [PATCH 13/47] Various fixes from xserver via pixman (Billy Biggs) --- fb/fbpict.c | 16 ++++++++-------- 1 file changed, 8 insertions(+), 8 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 44bee1bca..db70872b4 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -123,7 +123,7 @@ fbIn (CARD32 x, CARD8 y) #if IMAGE_BYTE_ORDER == LSBFirst -#define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(int)where; \ +#define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(long)where; \ temp=count&3; \ where-=temp; \ workingWhere=(CARD32 *)where; \ @@ -136,7 +136,7 @@ fbIn (CARD32 x, CARD8 y) #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; \ + #define setupPackedReader(count,temp,where,workingWhere,workingVal) count=(long)where; \ temp=count&3; \ where-=temp; \ workingWhere=(CARD32 *)where; \ @@ -331,7 +331,7 @@ fbCompositeSolidMask_nx8x0888 (CARD8 op, while (height--) { /* fixme: cleanup unused */ - unsigned int wt, wd; + unsigned long wt, wd; CARD32 workingiDest; CARD32 *widst; @@ -1074,7 +1074,7 @@ fbCompositeTrans_0565xnx0565(CARD8 op, srcLine += srcStride; w = width; - if(((int)src&1)==1) + if(((long)src&1)==1) { s_16 = READ(src++); d_16 = READ(dst); @@ -1082,7 +1082,7 @@ fbCompositeTrans_0565xnx0565(CARD8 op, w--; } isrc=(CARD32 *)src; - if(((int)dst&1)==0) + if(((long)dst&1)==0) { idst=(CARD32 *)dst; while (w>1) @@ -1173,7 +1173,7 @@ fbCompositeTrans_0888xnx0888(CARD8 op, fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 3); { - unsigned int ws,wt; + unsigned long ws,wt; CARD32 workingSource; CARD32 *wsrc, *wdst, *widst; CARD32 rs, rd, nd; @@ -1196,7 +1196,7 @@ fbCompositeTrans_0888xnx0888(CARD8 op, setupPackedReader(ws,wt,isrc,wsrc,workingSource); /* get to word aligned */ - switch(!(int)src&3) + switch(~(long)dst&3) { case 1: readPackedSource(rs); @@ -1272,7 +1272,7 @@ fbCompositeTrans_0888xnx0888(CARD8 op, srcLine += srcStride; w = width*3; /* get to word aligned */ - switch(!(int)src&3) + switch(~(long)src&3) { case 1: rd=alphamaskCombine24(READ(src++), READ(dst))>>8; From 48c73dfc369fdf8f6023436ebe82bb604f76bb80 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Wed, 25 Apr 2007 12:09:22 -0400 Subject: [PATCH 14/47] Add function fbCompositeSrcAdd_8888x8x8(), and fix a bug where srcRepeat = FALSE would be set in the wrong place. --- fb/fbpict.c | 71 +++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 69 insertions(+), 2 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index db70872b4..b9c463aa5 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -894,6 +894,61 @@ fbCompositeSrcAdd_8888x8888 (CARD8 op, fbFinishAccess (pSrc->pDrawable); } +static void +fbCompositeSrcAdd_8888x8x8 (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 *maskLine, *mask; + FbStride dstStride, maskStride; + CARD16 w; + CARD32 src; + CARD8 sa; + + fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1); + fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); + fbComposeGetSolid (pSrc, src, pDst->format); + sa = (src >> 24); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + mask = maskLine; + maskLine += maskStride; + w = width; + + while (w--) + { + CARD16 tmp; + CARD16 a; + CARD32 m, d; + CARD32 r; + + a = READ(mask++); + d = READ(dst); + + m = FbInU (sa, 0, a, tmp); + r = FbAdd (m, d, 0, tmp); + + WRITE(dst++, r); + } + } + + fbFinishAccess(pDst->pDrawable); + fbFinishAccess(pMask->pDrawable); +} + void fbCompositeSrcAdd_1000x1000 (CARD8 op, PicturePtr pSrc, @@ -1587,6 +1642,8 @@ fbComposite (CARD8 op, default: break; } + if (func != fbCompositeGeneral) + srcRepeat = FALSE; } else if (! srcRepeat) /* has mask and non-repeating source */ { @@ -1669,8 +1726,6 @@ fbComposite (CARD8 op, } } } - if (func != fbCompositeGeneral) - srcRepeat = FALSE; } else if (maskRepeat && pMask->pDrawable->width == 1 && @@ -1887,6 +1942,18 @@ fbComposite (CARD8 op, break; } } + else + { + if ((pSrc->format == PICT_a8r8g8b8 || + pSrc->format == PICT_a8b8g8r8) && + fbCanGetSolid (pSrc) && + pMask->format == PICT_a8 && + pDst->format == PICT_a8) + { + srcRepeat = FALSE; + func = fbCompositeSrcAdd_8888x8x8; + } + } break; } From c19ece1d8c32dc81740a4036a642661f54064e75 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Wed, 25 Apr 2007 12:34:19 -0400 Subject: [PATCH 15/47] Integrate optimization from xserver from David Reveman where repeats get handled by fbFetchTransformed() rather than in the region walking code. --- fb/fbpict.c | 31 +++++++++++++++++++++++++++++-- 1 file changed, 29 insertions(+), 2 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index b9c463aa5..b4cddd767 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1463,6 +1463,8 @@ fbComposite (CARD8 op, CompositeFunc func = NULL; Bool srcRepeat = pSrc->pDrawable && pSrc->repeat; Bool maskRepeat = FALSE; + Bool srcTransform = pSrc->pDrawable && pSrc->transform; + Bool maskTransform = FALSE; Bool srcAlphaMap = pSrc->alphaMap != 0; Bool maskAlphaMap = FALSE; Bool dstAlphaMap = pDst->alphaMap != 0; @@ -1476,23 +1478,42 @@ fbComposite (CARD8 op, mmx_setup = TRUE; } #endif - + + if (pSrc->filter == PictFilterConvolution) + srcTransform = TRUE; + xDst += pDst->pDrawable->x; yDst += pDst->pDrawable->y; if (pSrc->pDrawable) { xSrc += pSrc->pDrawable->x; ySrc += pSrc->pDrawable->y; } + + if (srcRepeat && srcTransform && + pSrc->pDrawable->width == 1 && + pSrc->pDrawable->height == 1) + srcTransform = FALSE; + if (pMask && pMask->pDrawable) { xMask += pMask->pDrawable->x; yMask += pMask->pDrawable->y; maskRepeat = pMask->repeat == RepeatNormal; + + if (pMask->filter == PictFilterConvolution) + maskTransform = TRUE; + maskAlphaMap = pMask->alphaMap != 0; + + if (maskRepeat && maskTransform && + pMask->pDrawable->width == 1 && + pMask->pDrawable->height == 1) + maskTransform = FALSE; + } if (pSrc->pDrawable && (!pMask || pMask->pDrawable) - && !pSrc->transform && !(pMask && pMask->transform) + && !srcTransform && !maskTransform && !maskAlphaMap && !srcAlphaMap && !dstAlphaMap && (pSrc->filter != PictFilterConvolution) && (!pMask || pMask->filter != PictFilterConvolution)) @@ -1977,6 +1998,12 @@ fbComposite (CARD8 op, height)) return; + /* if we are transforming, we handle repeats in fbFetchTransformed */ + if (srcTransform) + srcRepeat = FALSE; + if (maskTransform) + maskRepeat = FALSE; + n = REGION_NUM_RECTS (®ion); pbox = REGION_RECTS (®ion); while (n--) From c056ce95d89ef1df57edf47149fc34cd3925496e Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Wed, 25 Apr 2007 13:21:47 -0400 Subject: [PATCH 16/47] Port MSVC++ CPU detection code from pixman. (Vladimir Vukicevic). --- fb/fbpict.c | 70 ++++++++++++++++++++++++++++++++++++++++++++++++----- 1 file changed, 64 insertions(+), 6 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index b4cddd767..f9f536863 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1509,7 +1509,6 @@ fbComposite (CARD8 op, pMask->pDrawable->width == 1 && pMask->pDrawable->height == 1) maskTransform = FALSE; - } if (pSrc->pDrawable && (!pMask || pMask->pDrawable) @@ -2116,7 +2115,7 @@ enum CPUFeatures { static unsigned int detectCPUFeatures(void) { unsigned int features = 0; - unsigned int result; + unsigned int result = 0; #ifdef HAVE_GETISAX if (getisax(&result, 1)) { @@ -2133,8 +2132,13 @@ static unsigned int detectCPUFeatures(void) { } #else char vendor[13]; +#ifdef _MSC_VER + int vendor0 = 0, vendor1, vendor2; +#endif vendor[0] = 0; vendor[12] = 0; + +#ifdef __GNUC__ /* 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 @@ -2153,7 +2157,7 @@ static unsigned int detectCPUFeatures(void) { "pop %%eax\n" "mov $0x0, %%edx\n" "xor %%ecx, %%eax\n" - "jz 1\n" + "jz 1f\n" "mov $0x00000000, %%eax\n" "push %%ebx\n" @@ -2177,6 +2181,45 @@ static unsigned int detectCPUFeatures(void) { : "%eax", "%ecx", "%edx" ); +#elif defined (_MSC_VER) + + _asm { + pushfd + pop eax + mov ecx, eax + xor eax, 00200000h + push eax + popfd + pushfd + pop eax + mov edx, 0 + xor eax, ecx + jz nocpuid + + mov eax, 0 + push ebx + cpuid + mov eax, ebx + pop ebx + mov vendor0, eax + mov vendor1, edx + mov vendor2, ecx + mov eax, 1 + push ebx + cpuid + pop ebx + nocpuid: + mov result, edx + } + memmove (vendor+0, &vendor0, 4); + memmove (vendor+4, &vendor1, 4); + memmove (vendor+8, &vendor2, 4); + +#else +# error unsupported compiler +#endif + + features = 0; if (result) { /* result now contains the standard feature bits */ if (result & (1 << 15)) @@ -2191,14 +2234,13 @@ static unsigned int detectCPUFeatures(void) { (strcmp(vendor, "AuthenticAMD") == 0 || strcmp(vendor, "Geode by NSC") == 0)) { /* check for AMD MMX extensions */ - - unsigned int result; +#ifdef __GNUC__ __asm__("push %%ebx\n" "mov $0x80000000, %%eax\n" "cpuid\n" "xor %%edx, %%edx\n" "cmp $0x1, %%eax\n" - "jge 2\n" + "jge 2f\n" "mov $0x80000001, %%eax\n" "cpuid\n" "2:\n" @@ -2208,11 +2250,27 @@ static unsigned int detectCPUFeatures(void) { : : "%eax", "%ecx", "%edx" ); +#elif defined _MSC_VER + _asm { + push ebx + mov eax, 80000000h + cpuid + xor edx, edx + cmp eax, 1 + jge notamd + mov eax, 80000001h + cpuid + notamd: + pop ebx + mov result, edx + } +#endif if (result & (1<<22)) features |= MMX_Extensions; } } #endif /* HAVE_GETISAX */ + return features; } From 67347739b0571b2978468e8088480b105f505ad2 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Wed, 25 Apr 2007 14:19:39 -0400 Subject: [PATCH 17/47] Don't treat convolution filters as transformations. Some rearrangement of code to get it closer to pixman. --- fb/fbpict.c | 37 +++++++++++++++++-------------------- 1 file changed, 17 insertions(+), 20 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index f9f536863..4fb949dcc 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1461,9 +1461,9 @@ fbComposite (CARD8 op, int n; BoxPtr pbox; CompositeFunc func = NULL; - Bool srcRepeat = pSrc->pDrawable && pSrc->repeat; + Bool srcRepeat = pSrc->pDrawable && pSrc->repeatType == RepeatNormal; Bool maskRepeat = FALSE; - Bool srcTransform = pSrc->pDrawable && pSrc->transform; + Bool srcTransform = pSrc->transform != 0; Bool maskTransform = FALSE; Bool srcAlphaMap = pSrc->alphaMap != 0; Bool maskAlphaMap = FALSE; @@ -1479,9 +1479,6 @@ fbComposite (CARD8 op, } #endif - if (pSrc->filter == PictFilterConvolution) - srcTransform = TRUE; - xDst += pDst->pDrawable->x; yDst += pDst->pDrawable->y; if (pSrc->pDrawable) { @@ -1517,21 +1514,6 @@ fbComposite (CARD8 op, && (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; - } - else -#endif - if (pMask == 0) - { - if (pSrc->format == pDst->format) - func = fbCompositeSrcSrc_nxn; - } - break; case PictOpOver: if (pMask) { @@ -1975,6 +1957,21 @@ fbComposite (CARD8 op, } } break; + case PictOpSrc: +#ifdef USE_MMX + if (!pMask && pSrc->format == pDst->format && + pSrc->format != PICT_a8 && pSrc->pDrawable != pDst->pDrawable) + { + func = fbCompositeCopyAreammx; + } + else +#endif + if (pMask == 0) + { + if (pSrc->format == pDst->format) + func = fbCompositeSrcSrc_nxn; + } + break; } if (!func) { From c0346e57e6d3857994f7af76060c502c2fdea294 Mon Sep 17 00:00:00 2001 From: Peter Hutterer Date: Thu, 26 Apr 2007 12:02:45 +0930 Subject: [PATCH 18/47] Require inputproto 1.4.2. Requirement was introduced with c7e2ba0c9b9b1fc1aed8f91f86471c4c8e650b78. --- configure.ac | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/configure.ac b/configure.ac index 872f376ee..9d4a50a01 100644 --- a/configure.ac +++ b/configure.ac @@ -628,7 +628,7 @@ XEXT_LIB='$(top_builddir)/Xext/libXext.la' XEXTXORG_LIB='$(top_builddir)/Xext/libXextbuiltin.la' dnl Core modules for most extensions, et al. -REQUIRED_MODULES="[randrproto >= 1.2] renderproto [fixesproto >= 4.0] [damageproto >= 1.1] xcmiscproto xextproto xproto xtrans [scrnsaverproto >= 1.1] bigreqsproto resourceproto fontsproto [inputproto >= 1.4] [kbproto >= 1.0.3]" +REQUIRED_MODULES="[randrproto >= 1.2] renderproto [fixesproto >= 4.0] [damageproto >= 1.1] xcmiscproto xextproto xproto xtrans [scrnsaverproto >= 1.1] bigreqsproto resourceproto fontsproto [inputproto >= 1.4.2] [kbproto >= 1.0.3]" REQUIRED_LIBS="xfont xau fontenc" if test "x$DBUS" = xauto; then From 3571b8e65b0857322d12e291305cfe29ea497c3c Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 09:45:11 -0400 Subject: [PATCH 19/47] More pixman merging - Move some code around to minimize diff noise - Add mmx ops from pixman that never made it into X - Merge Jeff Muizelaar's bugfixes to fbCompositeSrc_8888x8x8888mmx and fbCompositeSrc_x888x8x8888mmx. --- fb/fbmmx.c | 739 ++++++++++++++++++++++++++++++++++++++++++----------- fb/fbmmx.h | 62 +++++ 2 files changed, 651 insertions(+), 150 deletions(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index 5bbede10b..875fec01f 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -1140,23 +1140,22 @@ fbCompositeSrc_8888x8x8888mmx (CARD8 op, { CARD32 *dstLine, *dst; CARD32 *srcLine, *src; - CARD8 *maskLine; CARD32 mask; __m64 vmask; - FbStride dstStride, srcStride, maskStride; + FbStride dstStride, srcStride; CARD16 w; __m64 srca; - + CHECKPOINT(); - + fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1); fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1); - fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); - mask = *maskLine << 24 | *maskLine << 16 | *maskLine << 8 | *maskLine; + fbComposeGetSolid (pMask, mask, pDst->format); + mask = mask | mask >> 8 | mask >> 16 | mask >> 24; vmask = load8888 (mask); srca = MC(4x00ff); - + while (height--) { dst = dstLine; @@ -1169,9 +1168,93 @@ fbCompositeSrc_8888x8x8888mmx (CARD8 op, { __m64 s = load8888 (*src); __m64 d = load8888 (*dst); - + + *dst = store8888 (in_over (s, expand_alpha (s), vmask, d)); + + w--; + dst++; + src++; + } + + while (w >= 2) + { + __m64 vs = *(__m64 *)src; + __m64 vd = *(__m64 *)dst; + __m64 vsrc0 = expand8888 (vs, 0); + __m64 vsrc1 = expand8888 (vs, 1); + + *(__m64 *)dst = pack8888 ( + in_over (vsrc0, expand_alpha (vsrc0), vmask, expand8888 (vd, 0)), + in_over (vsrc1, expand_alpha (vsrc1), vmask, expand8888 (vd, 1))); + + w -= 2; + dst += 2; + src += 2; + } + + while (w) + { + __m64 s = load8888 (*src); + __m64 d = load8888 (*dst); + + *dst = store8888 (in_over (s, expand_alpha (s), vmask, d)); + + w--; + dst++; + src++; + } + } + + _mm_empty(); +} + +void +fbCompositeSrc_x888x8x8888mmx (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; + CARD32 mask; + __m64 vmask; + FbStride dstStride, srcStride; + CARD16 w; + __m64 srca; + + CHECKPOINT(); + + fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1); + fbComposeGetStart (pSrc, xSrc, ySrc, CARD32, srcStride, srcLine, 1); + fbComposeGetSolid (pMask, mask, pDst->format); + + mask = mask | mask >> 8 | mask >> 16 | mask >> 24; + vmask = load8888 (mask); + srca = MC(4x00ff); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + src = srcLine; + srcLine += srcStride; + w = width; + + while (w && (unsigned long)dst & 7) + { + __m64 s = load8888 (*src); + __m64 d = load8888 (*dst); + *dst = store8888 (in_over (s, srca, vmask, d)); - + w--; dst++; src++; @@ -1197,39 +1280,39 @@ fbCompositeSrc_8888x8x8888mmx (CARD8 op, __m64 vs6 = *(__m64 *)(src + 12); __m64 vs7 = *(__m64 *)(src + 14); - vd0 = (__m64)pack8888 ( + vd0 = pack8888 ( in_over (expand8888 (vs0, 0), srca, vmask, expand8888 (vd0, 0)), in_over (expand8888 (vs0, 1), srca, vmask, expand8888 (vd0, 1))); - - vd1 = (__m64)pack8888 ( + + vd1 = pack8888 ( in_over (expand8888 (vs1, 0), srca, vmask, expand8888 (vd1, 0)), in_over (expand8888 (vs1, 1), srca, vmask, expand8888 (vd1, 1))); - - vd2 = (__m64)pack8888 ( + + vd2 = pack8888 ( in_over (expand8888 (vs2, 0), srca, vmask, expand8888 (vd2, 0)), in_over (expand8888 (vs2, 1), srca, vmask, expand8888 (vd2, 1))); - - vd3 = (__m64)pack8888 ( + + vd3 = pack8888 ( in_over (expand8888 (vs3, 0), srca, vmask, expand8888 (vd3, 0)), in_over (expand8888 (vs3, 1), srca, vmask, expand8888 (vd3, 1))); - - vd4 = (__m64)pack8888 ( + + vd4 = pack8888 ( in_over (expand8888 (vs4, 0), srca, vmask, expand8888 (vd4, 0)), in_over (expand8888 (vs4, 1), srca, vmask, expand8888 (vd4, 1))); - - vd5 = (__m64)pack8888 ( + + vd5 = pack8888 ( in_over (expand8888 (vs5, 0), srca, vmask, expand8888 (vd5, 0)), in_over (expand8888 (vs5, 1), srca, vmask, expand8888 (vd5, 1))); - - vd6 = (__m64)pack8888 ( + + vd6 = pack8888 ( in_over (expand8888 (vs6, 0), srca, vmask, expand8888 (vd6, 0)), in_over (expand8888 (vs6, 1), srca, vmask, expand8888 (vd6, 1))); - - vd7 = (__m64)pack8888 ( + + vd7 = pack8888 ( in_over (expand8888 (vs7, 0), srca, vmask, expand8888 (vd7, 0)), in_over (expand8888 (vs7, 1), srca, vmask, expand8888 (vd7, 1))); - *(__m64 *)(dst + 0) = vd0; + *(__m64 *)(dst + 0) = vd0; *(__m64 *)(dst + 2) = vd1; *(__m64 *)(dst + 4) = vd2; *(__m64 *)(dst + 6) = vd3; @@ -1237,26 +1320,26 @@ fbCompositeSrc_8888x8x8888mmx (CARD8 op, *(__m64 *)(dst + 10) = vd5; *(__m64 *)(dst + 12) = vd6; *(__m64 *)(dst + 14) = vd7; - + w -= 16; dst += 16; src += 16; } - + while (w) { __m64 s = load8888 (*src); __m64 d = load8888 (*dst); - + *dst = store8888 (in_over (s, srca, vmask, d)); - + w--; dst++; src++; } } - _mm_empty(); + _mm_empty(); } void @@ -1552,6 +1635,256 @@ fbCompositeSolidMask_nx8x8888mmx (CARD8 op, } +Bool +fbSolidFillmmx (DrawablePtr pDraw, + int x, + int y, + int width, + int height, + FbBits xor) +{ + FbStride stride; + int bpp; + ullong fill; + __m64 vfill; + CARD32 byte_width; + CARD8 *byte_line; + FbBits *bits; + int xoff, yoff; + __m64 v1, v2, v3, v4, v5, v6, v7; + + CHECKPOINT(); + + fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff); + + if (bpp == 16 && (xor >> 16 != (xor & 0xffff))) + return FALSE; + + if (bpp != 16 && bpp != 32) + return FALSE; + + if (bpp == 16) + { + stride = stride * sizeof (FbBits) / 2; + byte_line = (CARD8 *)(((CARD16 *)bits) + stride * (y + yoff) + (x + xoff)); + byte_width = 2 * width; + stride *= 2; + } + else + { + stride = stride * sizeof (FbBits) / 4; + byte_line = (CARD8 *)(((CARD32 *)bits) + stride * (y + yoff) + (x + xoff)); + byte_width = 4 * width; + stride *= 4; + } + + fill = ((ullong)xor << 32) | xor; + vfill = (__m64)fill; + + __asm__ ( + "movq %7, %0\n" + "movq %7, %1\n" + "movq %7, %2\n" + "movq %7, %3\n" + "movq %7, %4\n" + "movq %7, %5\n" + "movq %7, %6\n" + : "=y" (v1), "=y" (v2), "=y" (v3), + "=y" (v4), "=y" (v5), "=y" (v6), "=y" (v7) + : "y" (vfill)); + + while (height--) + { + int w; + CARD8 *d = byte_line; + byte_line += stride; + w = byte_width; + + while (w >= 2 && ((unsigned long)d & 3)) + { + *(CARD16 *)d = xor; + w -= 2; + d += 2; + } + + while (w >= 4 && ((unsigned long)d & 7)) + { + *(CARD32 *)d = xor; + + w -= 4; + d += 4; + } + + while (w >= 64) + { + __asm__ ( + "movq %1, (%0)\n" + "movq %2, 8(%0)\n" + "movq %3, 16(%0)\n" + "movq %4, 24(%0)\n" + "movq %5, 32(%0)\n" + "movq %6, 40(%0)\n" + "movq %7, 48(%0)\n" + "movq %8, 56(%0)\n" + : + : "r" (d), + "y" (vfill), "y" (v1), "y" (v2), "y" (v3), + "y" (v4), "y" (v5), "y" (v6), "y" (v7) + : "memory"); + + w -= 64; + d += 64; + } + + while (w >= 4) + { + *(CARD32 *)d = xor; + + w -= 4; + d += 4; + } + if (w >= 2) + { + *(CARD16 *)d = xor; + w -= 2; + d += 2; + } + } + + _mm_empty(); + return TRUE; +} + +void +fbCompositeSolidMaskSrc_nx8x8888mmx (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; + CARD8 *maskLine, *mask; + FbStride dstStride, maskStride; + CARD16 w; + __m64 vsrc, vsrca; + ullong srcsrc; + + CHECKPOINT(); + + fbComposeGetSolid(pSrc, src, pDst->format); + + srca = src >> 24; + if (srca == 0) + { + fbSolidFillmmx (pDst->pDrawable, xDst, yDst, width, height, 0); + return; + } + + srcsrc = (ullong)src << 32 | src; + + fbComposeGetStart (pDst, xDst, yDst, CARD32, dstStride, dstLine, 1); + fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); + + vsrc = load8888 (src); + vsrca = expand_alpha (vsrc); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + mask = maskLine; + maskLine += maskStride; + w = width; + + CHECKPOINT(); + + while (w && (unsigned long)dst & 7) + { + ullong m = *mask; + + if (m) + { + __m64 vdest = in(vsrc, expand_alpha_rev ((__m64)m)); + *dst = store8888(vdest); + } + else + { + *dst = 0; + } + + w--; + mask++; + dst++; + } + + CHECKPOINT(); + + while (w >= 2) + { + ullong m0, m1; + m0 = *mask; + m1 = *(mask + 1); + + if (srca == 0xff && (m0 & m1) == 0xff) + { + *(ullong *)dst = srcsrc; + } + else if (m0 | m1) + { + __m64 vdest; + __m64 dest0, dest1; + + vdest = *(__m64 *)dst; + + dest0 = in(vsrc, expand_alpha_rev ((__m64)m0)); + dest1 = in(vsrc, expand_alpha_rev ((__m64)m1)); + + *(__m64 *)dst = pack8888(dest0, dest1); + } + else + { + *dst = 0; + } + + mask += 2; + dst += 2; + w -= 2; + } + + CHECKPOINT(); + + while (w) + { + ullong m = *mask; + + if (m) + { + __m64 vdest = load8888(*dst); + vdest = in(vsrc, expand_alpha_rev ((__m64)m)); + *dst = store8888(vdest); + } + else + { + *dst = 0; + } + + w--; + mask++; + dst++; + } + } + + _mm_empty(); +} + void fbCompositeSolidMask_nx8x0565mmx (CARD8 op, PicturePtr pSrc, @@ -2002,6 +2335,232 @@ fbCompositeSolidMask_nx8888x0565Cmmx (CARD8 op, _mm_empty (); } +void +fbCompositeIn_nx8x8mmx (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 *maskLine, *mask; + FbStride dstStride, maskStride; + CARD16 w; + CARD32 src; + CARD8 sa; + __m64 vsrc, vsrca; + + fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1); + fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); + + fbComposeGetSolid(pSrc, src, pDst->format); + + sa = src >> 24; + if (sa == 0) + return; + + vsrc = load8888(src); + vsrca = expand_alpha(vsrc); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + mask = maskLine; + maskLine += maskStride; + w = width; + + if ((((unsigned long)pDst & 3) == 0) && + (((unsigned long)pSrc & 3) == 0)) + { + while (w >= 4) + { + CARD32 m; + __m64 vmask; + __m64 vdest; + + m = 0; + + vmask = load8888 (*(CARD32 *)mask); + vdest = load8888 (*(CARD32 *)dst); + + *(CARD32 *)dst = store8888 (in (in (vsrca, vmask), vdest)); + + dst += 4; + mask += 4; + w -= 4; + } + } + + while (w--) + { + CARD16 tmp; + CARD8 a; + CARD32 m, d; + CARD32 r; + + a = *mask++; + d = *dst; + + m = FbInU (sa, 0, a, tmp); + r = FbInU (m, 0, d, tmp); + + *dst++ = r; + } + } + + _mm_empty(); +} + +void +fbCompositeIn_8x8mmx (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 srcStride, dstStride; + CARD16 w; + + fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1); + fbComposeGetStart (pSrc, xSrc, ySrc, CARD8, srcStride, srcLine, 1); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + src = srcLine; + srcLine += srcStride; + w = width; + + if ((((unsigned long)pDst & 3) == 0) && + (((unsigned long)pSrc & 3) == 0)) + { + while (w >= 4) + { + CARD32 *s = (CARD32 *)src; + CARD32 *d = (CARD32 *)dst; + + *d = store8888 (in (load8888 (*s), load8888 (*d))); + + w -= 4; + dst += 4; + src += 4; + } + } + + while (w--) + { + CARD8 s, d; + CARD16 tmp; + + s = *src; + d = *dst; + + *dst = FbInU (s, 0, d, tmp); + + src++; + dst++; + } + } + + _mm_empty (); +} + +void +fbCompositeSrcAdd_8888x8x8mmx (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 *maskLine, *mask; + FbStride dstStride, maskStride; + CARD16 w; + CARD32 src; + CARD8 sa; + __m64 vsrc, vsrca; + + fbComposeGetStart (pDst, xDst, yDst, CARD8, dstStride, dstLine, 1); + fbComposeGetStart (pMask, xMask, yMask, CARD8, maskStride, maskLine, 1); + + fbComposeGetSolid(pSrc, src, pDst->format); + + sa = src >> 24; + if (sa == 0) + return; + + vsrc = load8888(src); + vsrca = expand_alpha(vsrc); + + while (height--) + { + dst = dstLine; + dstLine += dstStride; + mask = maskLine; + maskLine += maskStride; + w = width; + + if ((((unsigned long)pMask & 3) == 0) && + (((unsigned long)pDst & 3) == 0)) + { + while (w >= 4) + { + __m64 vmask = load8888 (*(CARD32 *)mask); + __m64 vdest = load8888 (*(CARD32 *)dst); + + *(CARD32 *)dst = store8888 (_mm_adds_pu8 (in (vsrca, vmask), vdest)); + + w -= 4; + dst += 4; + mask += 4; + } + } + + while (w--) + { + CARD16 tmp; + CARD16 a; + CARD32 m, d; + CARD32 r; + + a = *mask++; + d = *dst; + + m = FbInU (sa, 0, a, tmp); + r = FbAdd (m, d, 0, tmp); + + *dst++ = r; + } + } + + _mm_empty(); +} + void fbCompositeSrcAdd_8000x8000mmx (CARD8 op, PicturePtr pSrc, @@ -2134,126 +2693,6 @@ fbCompositeSrcAdd_8888x8888mmx (CARD8 op, _mm_empty(); } -Bool -fbSolidFillmmx (DrawablePtr pDraw, - int x, - int y, - int width, - int height, - FbBits xor) -{ - FbStride stride; - int bpp; - ullong fill; - __m64 vfill; - CARD32 byte_width; - CARD8 *byte_line; - FbBits *bits; - int xoff, yoff; - __m64 v1, v2, v3, v4, v5, v6, v7; - - CHECKPOINT(); - - fbGetDrawable(pDraw, bits, stride, bpp, xoff, yoff); - - if (bpp == 16 && (xor >> 16 != (xor & 0xffff))) - return FALSE; - - if (bpp != 16 && bpp != 32) - return FALSE; - - if (bpp == 16) - { - stride = stride * sizeof (FbBits) / 2; - byte_line = (CARD8 *)(((CARD16 *)bits) + stride * (y + yoff) + (x + xoff)); - byte_width = 2 * width; - stride *= 2; - } - else - { - stride = stride * sizeof (FbBits) / 4; - byte_line = (CARD8 *)(((CARD32 *)bits) + stride * (y + yoff) + (x + xoff)); - byte_width = 4 * width; - stride *= 4; - } - - fill = ((ullong)xor << 32) | xor; - vfill = (__m64)fill; - - __asm__ ( - "movq %7, %0\n" - "movq %7, %1\n" - "movq %7, %2\n" - "movq %7, %3\n" - "movq %7, %4\n" - "movq %7, %5\n" - "movq %7, %6\n" - : "=y" (v1), "=y" (v2), "=y" (v3), - "=y" (v4), "=y" (v5), "=y" (v6), "=y" (v7) - : "y" (vfill)); - - while (height--) - { - int w; - CARD8 *d = byte_line; - byte_line += stride; - w = byte_width; - - while (w >= 2 && ((unsigned long)d & 3)) - { - *(CARD16 *)d = xor; - w -= 2; - d += 2; - } - - while (w >= 4 && ((unsigned long)d & 7)) - { - *(CARD32 *)d = xor; - - w -= 4; - d += 4; - } - - while (w >= 64) - { - __asm__ ( - "movq %1, (%0)\n" - "movq %2, 8(%0)\n" - "movq %3, 16(%0)\n" - "movq %4, 24(%0)\n" - "movq %5, 32(%0)\n" - "movq %6, 40(%0)\n" - "movq %7, 48(%0)\n" - "movq %8, 56(%0)\n" - : - : "r" (d), - "y" (vfill), "y" (v1), "y" (v2), "y" (v3), - "y" (v4), "y" (v5), "y" (v6), "y" (v7) - : "memory"); - - w -= 64; - d += 64; - } - - while (w >= 4) - { - *(CARD32 *)d = xor; - - w -= 4; - d += 4; - } - if (w >= 2) - { - *(CARD16 *)d = xor; - w -= 2; - d += 2; - } - } - - _mm_empty(); - return TRUE; -} - Bool fbCopyAreammx (DrawablePtr pSrc, DrawablePtr pDst, diff --git a/fb/fbmmx.h b/fb/fbmmx.h index b3e4d71ef..ca9c7e796 100644 --- a/fb/fbmmx.h +++ b/fb/fbmmx.h @@ -82,6 +82,32 @@ void fbCompositeSrc_8888x8888mmx (CARD8 op, INT16 yDst, CARD16 width, CARD16 height); +void +fbCompositeSolidMaskSrc_nx8x8888mmx (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 +fbCompositeSrc_x888x8x8888mmx (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 fbCompositeSolidMask_nx8888x8888Cmmx (CARD8 op, PicturePtr pSrc, PicturePtr pMask, @@ -106,6 +132,42 @@ void fbCompositeSolidMask_nx8x8888mmx (CARD8 op, INT16 yDst, CARD16 width, CARD16 height); +void fbCompositeIn_nx8x8mmx (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 fbCompositeIn_8x8mmx (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 fbCompositeSrcAdd_8888x8x8mmx (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 fbCompositeSrcAdd_8000x8000mmx (CARD8 op, PicturePtr pSrc, PicturePtr pMask, From a54ef54db19dcd36ed86b33cff2bc369f9690a15 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 10:24:25 -0400 Subject: [PATCH 20/47] Pixman merge Make sure fbCompositeSrc_x888x8x8888mmx and fbCompositeSrc_8888x8x8888mmx are used when possible. --- fb/fbpict.c | 77 ++++++++++++++++++++++++++++++----------------------- 1 file changed, 44 insertions(+), 33 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 4fb949dcc..1932b3e65 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1647,7 +1647,7 @@ fbComposite (CARD8 op, if (func != fbCompositeGeneral) srcRepeat = FALSE; } - else if (! srcRepeat) /* has mask and non-repeating source */ + else if (!srcRepeat) /* has mask and non-repeating source */ { if (pSrc->pDrawable == pMask->pDrawable && xSrc == xMask && ySrc == yMask && @@ -1712,44 +1712,55 @@ fbComposite (CARD8 op, } break; } - else + else if (maskRepeat && + pMask->pDrawable->width == 1 && + pMask->pDrawable->height == 1) { - /* non-repeating source, repeating mask => translucent window */ - if (fbCanGetSolid(pMask)) - { - if (pSrc->format == PICT_x8r8g8b8 && - pDst->format == PICT_x8r8g8b8 && - pMask->format == PICT_a8) - { + switch (pSrc->format) { + case PICT_r5g6b5: + case PICT_b5g6r5: + if (pDst->format == pSrc->format) + func = fbCompositeTrans_0565xnx0565; + break; + case PICT_r8g8b8: + case PICT_b8g8r8: + if (pDst->format == pSrc->format) + func = fbCompositeTrans_0888xnx0888; + break; #ifdef USE_MMX - if (fbHaveMMX()) - func = fbCompositeSrc_8888x8x8888mmx; + case PICT_x8r8g8b8: + if ((pDst->format == PICT_a8r8g8b8 || + pDst->format == PICT_x8r8g8b8) && + pMask->format == PICT_a8 && fbHaveMMX()) + func = fbCompositeSrc_x888x8x8888mmx; + break; + case PICT_x8b8g8r8: + if ((pDst->format == PICT_a8b8g8r8 || + pDst->format == PICT_x8b8g8r8) && + pMask->format == PICT_a8 && fbHaveMMX()) + func = fbCompositeSrc_x888x8x8888mmx; + break; + case PICT_a8r8g8b8: + if ((pDst->format == PICT_a8r8g8b8 || + pDst->format == PICT_x8r8g8b8) && + pMask->format == PICT_a8 && fbHaveMMX()) + func = fbCompositeSrc_8888x8x8888mmx; + break; + case PICT_a8b8g8r8: + if ((pDst->format == PICT_a8b8g8r8 || + pDst->format == PICT_x8b8g8r8) && + pMask->format == PICT_a8 && fbHaveMMX()) + func = fbCompositeSrc_8888x8x8888mmx; + break; #endif - } + default: + break; } + + if (func != fbCompositeGeneral) + maskRepeat = 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; - case PICT_r8g8b8: - case PICT_b8g8r8: - if (pDst->format == pSrc->format) - func = fbCompositeTrans_0888xnx0888; - break; - default: - break; - } - if (func != fbCompositeGeneral) - maskRepeat = FALSE; - } } else /* no mask */ { From 701ccb4a22cfd646ccb7f19b7b3a476aeb5ce2da Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 10:49:06 -0400 Subject: [PATCH 21/47] Pixman merge - Remove stray default label - Integrate new MMX ops SolidMaskSrc_nx8x8888mmx, In_8x8mmx, and In_nx8x8mmx - Formatting changes to reduce diff noise --- fb/fbpict.c | 88 +++++++++++++++++++++++++++++++++++++++-------------- 1 file changed, 65 insertions(+), 23 deletions(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 1932b3e65..ace19fd1f 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1641,8 +1641,6 @@ fbComposite (CARD8 op, default: break; } - default: - break; } if (func != fbCompositeGeneral) srcRepeat = FALSE; @@ -1969,19 +1967,73 @@ fbComposite (CARD8 op, } break; case PictOpSrc: -#ifdef USE_MMX - if (!pMask && pSrc->format == pDst->format && - pSrc->format != PICT_a8 && pSrc->pDrawable != pDst->pDrawable) + if (pMask) { - func = fbCompositeCopyAreammx; +#ifdef USE_MMX + if (fbCanGetSolid (pSrc)) + { + if (pMask->format == PICT_a8) + { + switch (pDst->format) + { + case PICT_a8r8g8b8: + case PICT_x8r8g8b8: + case PICT_a8b8g8r8: + case PICT_x8b8g8r8: + if (fbHaveMMX()) + { + srcRepeat = FALSE; + func = fbCompositeSolidMaskSrc_nx8x8888mmx; + } + break; + default: + break; + } + } + } +#endif } else -#endif - if (pMask == 0) + { + if (pSrc->format == pDst->format) { - if (pSrc->format == pDst->format) +#ifdef USE_MMX + if (pSrc->pDrawable != pDst->pDrawable && fbHaveMMX() && + (PICT_FORMAT_BPP (pSrc->format) == 16 || + PICT_FORMAT_BPP (pSrc->format) == 32)) + func = fbCompositeCopyAreammx; + else +#endif func = fbCompositeSrcSrc_nxn; } + } + break; + case PictOpIn: +#ifdef USE_MMX + if (pSrc->format == PICT_a8 && + pDst->format == PICT_a8 && + !pMask) + { + if (fbHaveMMX()) + func = fbCompositeIn_8x8mmx; + } + else if (srcRepeat && pMask && !pMask->componentAlpha && + (pSrc->format == PICT_a8r8g8b8 || + pSrc->format == PICT_a8b8g8r8) && + (pMask->format == PICT_a8) && + pDst->format == PICT_a8) + { + if (fbHaveMMX()) + { + srcRepeat = FALSE; + func = fbCompositeIn_nx8x8mmx; + } + } +#else + func = NULL; +#endif + break; + default: break; } @@ -1991,26 +2043,16 @@ fbComposite (CARD8 op, return; } - if (!miComputeCompositeRegion (®ion, - pSrc, - pMask, - pDst, - xSrc, - ySrc, - xMask, - yMask, - xDst, - yDst, - width, - height)) - return; - /* if we are transforming, we handle repeats in fbFetchTransformed */ if (srcTransform) srcRepeat = FALSE; if (maskTransform) maskRepeat = FALSE; + 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--) From 0ff7c94fcf6497ee8575f81cf97eeeb3a857739e Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 10:56:02 -0400 Subject: [PATCH 22/47] Pixman merge Make use of fbCompositeSrcAdd_8888x8x8mmx --- fb/fbpict.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index ace19fd1f..0a08affd7 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1962,7 +1962,12 @@ fbComposite (CARD8 op, pDst->format == PICT_a8) { srcRepeat = FALSE; - func = fbCompositeSrcAdd_8888x8x8; +#ifdef USE_MMX + if (fbHaveMMX()) + func = fbCompositeSrcAdd_8888x8x8mmx; + else +#endif + func = fbCompositeSrcAdd_8888x8x8; } } break; From 4fe918b38553133c27e5ae672e5c43984a9bbaea Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 12:41:01 -0400 Subject: [PATCH 23/47] Fix bug 8871 - scrolling corruption with a compositing manager Call miHandleExposures() in CopyArea/CopyPlane explicitly in cw to generate GraphicsExposes correctly. --- miext/cw/cw_ops.c | 37 ++++++++++++++++++++----------------- 1 file changed, 20 insertions(+), 17 deletions(-) diff --git a/miext/cw/cw_ops.c b/miext/cw/cw_ops.c index 80b72806c..82ec99999 100644 --- a/miext/cw/cw_ops.c +++ b/miext/cw/cw_ops.c @@ -30,6 +30,7 @@ #include "gcstruct.h" #include "pixmapstr.h" #include "cw.h" +#include "mi.h" #define SETUP_BACKING_DST(_pDst, _pGC) \ cwGCPtr pGCPrivate = getCwGC (_pGC); \ @@ -185,7 +186,7 @@ cwCopyArea(DrawablePtr pSrc, DrawablePtr pDst, GCPtr pGC, int srcx, int srcy, int w, int h, int dstx, int dsty) { int odstx, odsty; - RegionPtr exposed = NULL; + int osrcx, osrcy; SETUP_BACKING_DST(pDst, pGC); SETUP_BACKING_SRC(pSrc, pGC); @@ -193,19 +194,20 @@ cwCopyArea(DrawablePtr pSrc, DrawablePtr pDst, GCPtr pGC, int srcx, int srcy, odstx = dstx; odsty = dsty; + osrcx = srcx; + osrcy = srcy; CW_OFFSET_XY_DST(dstx, dsty); CW_OFFSET_XY_SRC(srcx, srcy); - exposed = (*pBackingGC->ops->CopyArea)(pBackingSrc, pBackingDst, - pBackingGC, srcx, srcy, w, h, - dstx, dsty); - - if (exposed != NULL) - REGION_TRANSLATE(pDst->pScreen, exposed, odstx - dstx, odsty - dsty); - + (*pBackingGC->ops->CopyArea)(pBackingSrc, pBackingDst, + pBackingGC, srcx, srcy, w, h, + dstx, dsty); + EPILOGUE(pGC); - return exposed; + return miHandleExposures(pSrc, pDst, pGC, + osrcx, osrcy, w, h, + odstx, odsty, 0); } static RegionPtr @@ -213,7 +215,7 @@ cwCopyPlane(DrawablePtr pSrc, DrawablePtr pDst, GCPtr pGC, int srcx, int srcy, int w, int h, int dstx, int dsty, unsigned long plane) { int odstx, odsty; - RegionPtr exposed = NULL; + int osrcx, osrcy; SETUP_BACKING_DST(pDst, pGC); SETUP_BACKING_SRC(pSrc, pGC); @@ -221,19 +223,20 @@ cwCopyPlane(DrawablePtr pSrc, DrawablePtr pDst, GCPtr pGC, int srcx, int srcy, odstx = dstx; odsty = dsty; + osrcx = srcx; + osrcy = srcy; CW_OFFSET_XY_DST(dstx, dsty); CW_OFFSET_XY_SRC(srcx, srcy); - exposed = (*pBackingGC->ops->CopyPlane)(pBackingSrc, pBackingDst, - pBackingGC, srcx, srcy, w, h, - dstx, dsty, plane); - - if (exposed != NULL) - REGION_TRANSLATE(pDst->pScreen, exposed, odstx - dstx, odsty - dsty); + (*pBackingGC->ops->CopyPlane)(pBackingSrc, pBackingDst, + pBackingGC, srcx, srcy, w, h, + dstx, dsty, plane); EPILOGUE(pGC); - return exposed; + return miHandleExposures(pSrc, pDst, pGC, + osrcx, osrcy, w, h, + odstx, odsty, plane); } static void From 0ebe48be59368b55c618f60d4656300bd7f52ed9 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 14:36:32 -0400 Subject: [PATCH 24/47] Pixman merge - Changes to support MS Visual C++ - use inline instead of __inline__ - Fix rounding errors (Billy Biggs, from xserver via pixman) --- fb/fbmmx.c | 141 ++++++++++++++++++++++++++++++++++++----------------- 1 file changed, 95 insertions(+), 46 deletions(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index 875fec01f..7f1a7b12b 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -85,30 +85,40 @@ typedef unsigned long long ullong; +#ifdef __GNUC__ +typedef ullong mmxdatafield; +#endif +#ifdef _MSC_VER +typedef unsigned __int64 ullong; +typedef __m64 mmxdatafield; +#endif + typedef struct { - ullong mmx_4x00ff; - ullong mmx_4x0080; - ullong mmx_565_rgb; - ullong mmx_565_unpack_multiplier; - ullong mmx_565_r; - ullong mmx_565_g; - ullong mmx_565_b; - ullong mmx_mask_0; - ullong mmx_mask_1; - ullong mmx_mask_2; - ullong mmx_mask_3; - ullong mmx_full_alpha; - ullong mmx_ffff0000ffff0000; - ullong mmx_0000ffff00000000; - ullong mmx_000000000000ffff; + mmxdatafield mmx_4x00ff; + mmxdatafield mmx_4x0080; + mmxdatafield mmx_565_rgb; + mmxdatafield mmx_565_unpack_multiplier; + mmxdatafield mmx_565_r; + mmxdatafield mmx_565_g; + mmxdatafield mmx_565_b; + mmxdatafield mmx_mask_0; + mmxdatafield mmx_mask_1; + mmxdatafield mmx_mask_2; + mmxdatafield mmx_mask_3; + mmxdatafield mmx_full_alpha; + mmxdatafield mmx_ffff0000ffff0000; + mmxdatafield mmx_0000ffff00000000; + mmxdatafield mmx_000000000000ffff; } MMXData; static const MMXData c = { +#ifdef __GNUC__ .mmx_4x00ff = 0x00ff00ff00ff00ffULL, .mmx_4x0080 = 0x0080008000800080ULL, .mmx_565_rgb = 0x000001f0003f001fULL, + .mmx_565_unpack_multiplier = 0x0000008404100840ULL, .mmx_565_r = 0x000000f800000000ULL, .mmx_565_g = 0x0000000000fc0000ULL, .mmx_565_b = 0x00000000000000f8ULL, @@ -117,15 +127,42 @@ static const MMXData c = .mmx_mask_2 = 0xffff0000ffffffffULL, .mmx_mask_3 = 0x0000ffffffffffffULL, .mmx_full_alpha = 0x00ff000000000000ULL, - .mmx_565_unpack_multiplier = 0x0000008404100840ULL, .mmx_ffff0000ffff0000 = 0xffff0000ffff0000ULL, .mmx_0000ffff00000000 = 0x0000ffff00000000ULL, .mmx_000000000000ffff = 0x000000000000ffffULL, +#endif +#ifdef _MSC_VER + { 0x00ff00ff00ff00ffUI64 }, + { 0x0080008000800080UI64 }, + { 0x000001f0003f001fUI64 }, + { 0x0000008404100840UI64 }, + { 0x000000f800000000UI64 }, + { 0x0000000000fc0000UI64 }, + { 0x00000000000000f8UI64 }, + { 0xffffffffffff0000UI64 }, + { 0xffffffff0000ffffUI64 }, + { 0xffff0000ffffffffUI64 }, + { 0x0000ffffffffffffUI64 }, + { 0x00ff000000000000UI64 }, + { 0xffff0000ffff0000UI64 }, + { 0x0000ffff00000000UI64 }, + { 0x000000000000ffffUI64 }, +#endif }; -#define MC(x) ((__m64) c.mmx_##x) +#ifdef _MSC_VER +#undef inline +#define inline __forceinline +#endif -static __inline__ __m64 +#ifdef __GNUC__ +#define MC(x) ((__m64) c.mmx_##x) +#endif +#ifdef _MSC_VER +#define MC(x) c.mmx_##x +#endif + +static inline __m64 shift (__m64 v, int s) { if (s > 0) @@ -136,13 +173,13 @@ shift (__m64 v, int s) return v; } -static __inline__ __m64 +static inline __m64 negate (__m64 mask) { return _mm_xor_si64 (mask, MC(4x00ff)); } -static __inline__ __m64 +static inline __m64 pix_multiply (__m64 a, __m64 b) { __m64 res; @@ -155,7 +192,7 @@ pix_multiply (__m64 a, __m64 b) return res; } -static __inline__ __m64 +static inline __m64 pix_add (__m64 a, __m64 b) { return _mm_adds_pu8 (a, b); @@ -163,19 +200,19 @@ pix_add (__m64 a, __m64 b) #ifdef USE_SSE -static __inline__ __m64 +static inline __m64 expand_alpha (__m64 pixel) { return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 3, 3, 3)); } -static __inline__ __m64 +static inline __m64 expand_alpha_rev (__m64 pixel) { return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(0, 0, 0, 0)); } -static __inline__ __m64 +static inline __m64 invert_colors (__m64 pixel) { return _mm_shuffle_pi16 (pixel, _MM_SHUFFLE(3, 0, 1, 2)); @@ -183,7 +220,7 @@ invert_colors (__m64 pixel) #else -static __inline__ __m64 +static inline __m64 expand_alpha (__m64 pixel) { __m64 t1, t2; @@ -197,7 +234,7 @@ expand_alpha (__m64 pixel) return t1; } -static __inline__ __m64 +static inline __m64 expand_alpha_rev (__m64 pixel) { __m64 t1, t2; @@ -214,7 +251,7 @@ expand_alpha_rev (__m64 pixel) return t1; } -static __inline__ __m64 +static inline __m64 invert_colors (__m64 pixel) { __m64 x, y, z; @@ -236,13 +273,13 @@ invert_colors (__m64 pixel) #endif -static __inline__ __m64 +static inline __m64 over (__m64 src, __m64 srca, __m64 dest) { return _mm_adds_pu8 (src, pix_multiply(dest, negate(srca))); } -static __inline__ __m64 +static inline __m64 over_rev_non_pre (__m64 src, __m64 dest) { __m64 srca = expand_alpha (src); @@ -251,14 +288,15 @@ over_rev_non_pre (__m64 src, __m64 dest) return over(pix_multiply(invert_colors(src), srcfaaa), srca, dest); } -static __inline__ __m64 +static inline __m64 in (__m64 src, __m64 mask) { return pix_multiply (src, mask); } -static __inline__ __m64 +#ifndef _MSC_VER +static inline __m64 in_over (__m64 src, __m64 srca, __m64 mask, @@ -266,20 +304,23 @@ in_over (__m64 src, { return over(in(src, mask), pix_multiply(srca, mask), dest); } +#else +#define in_over(src, srca, mask, dest) over(in(src, mask), pix_multiply(srca, mask), dest) +#endif -static __inline__ __m64 +static inline __m64 load8888 (CARD32 v) { return _mm_unpacklo_pi8 (_mm_cvtsi32_si64 (v), _mm_setzero_si64()); } -static __inline__ __m64 +static inline __m64 pack8888 (__m64 lo, __m64 hi) { return _mm_packs_pu16 (lo, hi); } -static __inline__ CARD32 +static inline CARD32 store8888 (__m64 v) { return _mm_cvtsi64_si32(pack8888(v, _mm_setzero_si64())); @@ -299,7 +340,7 @@ store8888 (__m64 v) * Note the trick here - the top word is shifted by another nibble to * avoid it bumping into the middle word */ -static __inline__ __m64 +static inline __m64 expand565 (__m64 pixel, int pos) { __m64 p = pixel; @@ -319,7 +360,7 @@ expand565 (__m64 pixel, int pos) return _mm_srli_pi16 (pixel, 8); } -static __inline__ __m64 +static inline __m64 expand8888 (__m64 in, int pos) { if (pos == 0) @@ -328,7 +369,7 @@ expand8888 (__m64 in, int pos) return _mm_unpackhi_pi8 (in, _mm_setzero_si64()); } -static __inline__ __m64 +static inline __m64 pack565 (__m64 pixel, __m64 target, int pos) { __m64 p = pixel; @@ -358,20 +399,28 @@ pack565 (__m64 pixel, __m64 target, int pos) return _mm_or_si64 (b, p); } -static __inline__ __m64 +#ifndef _MSC_VER +static inline __m64 pix_add_mul (__m64 x, __m64 a, __m64 y, __m64 b) { - x = _mm_mullo_pi16 (x, a); - y = _mm_mullo_pi16 (y, b); - x = _mm_srli_pi16(x, 1); - y = _mm_srli_pi16(y, 1); - x = _mm_adds_pu16 (x, y); - x = _mm_adds_pu16 (x, _mm_srli_pi16 (x, 8)); + x = _mm_mullo_pi16 (x, a); + y = _mm_mullo_pi16 (y, b); x = _mm_adds_pu16 (x, MC(4x0080)); - x = _mm_srli_pi16 (x, 7); + x = _mm_adds_pu16 (x, y); + x = _mm_adds_pu16 (x, _mm_srli_pi16 (x, 8)); + x = _mm_srli_pi16 (x, 8); return x; } +#else +#define pix_add_mul(x, a, y, b) \ +( x = _mm_mullo_pi16 (x, a), \ + y = _mm_mullo_pi16 (y, b), \ + x = _mm_adds_pu16 (x, MC(4x0080)), \ + x = _mm_adds_pu16 (x, y), \ + x = _mm_adds_pu16 (x, _mm_srli_pi16 (x, 8)), \ + _mm_srli_pi16 (x, 8) ) +#endif /* --------------- MMX code patch for fbcompose.c --------------------- */ @@ -590,7 +639,7 @@ mmxCombineSaturateU (CARD32 *dest, const CARD32 *src, int width) CARD32 da = ~d >> 24; if (sa > da) { - __m64 msa = load8888(FbIntDiv(da, sa)); + __m64 msa = load8888(FbIntDiv(da, sa))<<24; msa = expand_alpha_rev(msa); ms = pix_multiply(ms, msa); } From a300ef84cee26febfbe08c497d0d063588130bdd Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 14:37:53 -0400 Subject: [PATCH 25/47] Fix typo in previous commit --- fb/fbmmx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index 7f1a7b12b..164776aaa 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -639,7 +639,7 @@ mmxCombineSaturateU (CARD32 *dest, const CARD32 *src, int width) CARD32 da = ~d >> 24; if (sa > da) { - __m64 msa = load8888(FbIntDiv(da, sa))<<24; + __m64 msa = load8888(FbIntDiv(da, sa) << 24); msa = expand_alpha_rev(msa); ms = pix_multiply(ms, msa); } From 2208c6087d6bffcb24a30891a56430e28735874c Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Thu, 26 Apr 2007 14:40:30 -0400 Subject: [PATCH 26/47] Change expand_alpha_rev to expand_alpha in mmxSaturateU --- fb/fbmmx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index 164776aaa..0e9074648 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -640,7 +640,7 @@ mmxCombineSaturateU (CARD32 *dest, const CARD32 *src, int width) if (sa > da) { __m64 msa = load8888(FbIntDiv(da, sa) << 24); - msa = expand_alpha_rev(msa); + msa = expand_alpha(msa); ms = pix_multiply(ms, msa); } md = pix_add(md, ms); From 6c8152d6ee9eeb21a68a8bbfed1540939e5bcd1f Mon Sep 17 00:00:00 2001 From: Adam Jackson Date: Thu, 26 Apr 2007 14:59:04 -0400 Subject: [PATCH 27/47] Remove old edid_modes.c, it lives in hw/xfree86/modes/ now. --- hw/xfree86/ddc/edid_modes.c | 361 ------------------------------------ 1 file changed, 361 deletions(-) delete mode 100644 hw/xfree86/ddc/edid_modes.c diff --git a/hw/xfree86/ddc/edid_modes.c b/hw/xfree86/ddc/edid_modes.c deleted file mode 100644 index 926bc8921..000000000 --- a/hw/xfree86/ddc/edid_modes.c +++ /dev/null @@ -1,361 +0,0 @@ -/* - * Copyright 2006 Luc Verhaegen. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sub license, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice (including the - * next paragraph) shall be included in all copies or substantial portions - * of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL - * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER - * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING - * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER - * DEALINGS IN THE SOFTWARE. - */ - -#ifdef HAVE_XORG_CONFIG_H -#include -#endif - -#include "xf86.h" -#include "xf86DDC.h" -#include -#include "property.h" -#include "propertyst.h" -#include "xf86DDC.h" - -/* - * TODO: - * - for those with access to the VESA DMT standard; review please. - */ -#define MODEPREFIX(name) NULL, NULL, name, 0,M_T_DRIVER -#define MODESUFFIX 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,FALSE,FALSE,0,NULL,0,0.0,0.0 - -DisplayModeRec DDCEstablishedModes[17] = { - { MODEPREFIX("800x600"), 40000, 800, 840, 968, 1056, 0, 600, 601, 605, 628, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 800x600@60Hz */ - { MODEPREFIX("800x600"), 36000, 800, 824, 896, 1024, 0, 600, 601, 603, 625, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 800x600@56Hz */ - { MODEPREFIX("640x480"), 31500, 640, 656, 720, 840, 0, 480, 481, 484, 500, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 640x480@75Hz */ - { MODEPREFIX("640x480"), 31500, 640, 664, 704, 832, 0, 480, 489, 491, 520, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 640x480@72Hz */ - { MODEPREFIX("640x480"), 30240, 640, 704, 768, 864, 0, 480, 483, 486, 525, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 640x480@67Hz */ - { MODEPREFIX("640x480"), 25200, 640, 656, 752, 800, 0, 480, 490, 492, 525, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 640x480@60Hz */ - { MODEPREFIX("720x400"), 35500, 720, 738, 846, 900, 0, 400, 421, 423, 449, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 720x400@88Hz */ - { MODEPREFIX("720x400"), 28320, 720, 738, 846, 900, 0, 400, 412, 414, 449, 0, V_NHSYNC | V_PVSYNC, MODESUFFIX }, /* 720x400@70Hz */ - { MODEPREFIX("1280x1024"), 135000, 1280, 1296, 1440, 1688, 0, 1024, 1025, 1028, 1066, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 1280x1024@75Hz */ - { MODEPREFIX("1024x768"), 78800, 1024, 1040, 1136, 1312, 0, 768, 769, 772, 800, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 1024x768@75Hz */ - { MODEPREFIX("1024x768"), 75000, 1024, 1048, 1184, 1328, 0, 768, 771, 777, 806, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 1024x768@70Hz */ - { MODEPREFIX("1024x768"), 65000, 1024, 1048, 1184, 1344, 0, 768, 771, 777, 806, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 1024x768@60Hz */ - { MODEPREFIX("1024x768"), 44900, 1024, 1032, 1208, 1264, 0, 768, 768, 776, 817, 0, V_PHSYNC | V_PVSYNC | V_INTERLACE, MODESUFFIX }, /* 1024x768@43Hz */ - { MODEPREFIX("832x624"), 57284, 832, 864, 928, 1152, 0, 624, 625, 628, 667, 0, V_NHSYNC | V_NVSYNC, MODESUFFIX }, /* 832x624@75Hz */ - { MODEPREFIX("800x600"), 49500, 800, 816, 896, 1056, 0, 600, 601, 604, 625, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 800x600@75Hz */ - { MODEPREFIX("800x600"), 50000, 800, 856, 976, 1040, 0, 600, 637, 643, 666, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 800x600@72Hz */ - { MODEPREFIX("1152x864"), 108000, 1152, 1216, 1344, 1600, 0, 864, 865, 868, 900, 0, V_PHSYNC | V_PVSYNC, MODESUFFIX }, /* 1152x864@75Hz */ -}; - -static DisplayModePtr -DDCModesFromEstablished(int scrnIndex, struct established_timings *timing) -{ - DisplayModePtr Modes = NULL, Mode = NULL; - CARD32 bits = (timing->t1) | (timing->t2 << 8) | - ((timing->t_manu & 0x80) << 9); - int i; - - for (i = 0; i < 17; i++) { - if (bits & (0x01 << i)) { - Mode = xf86DuplicateMode(&DDCEstablishedModes[i]); - Modes = xf86ModesAdd(Modes, Mode); - } - } - - return Modes; -} - -/* - * - */ -static DisplayModePtr -DDCModesFromStandardTiming(int scrnIndex, struct std_timings *timing) -{ - DisplayModePtr Modes = NULL, Mode = NULL; - int i; - - for (i = 0; i < STD_TIMINGS; i++) { - if (timing[i].hsize && timing[i].vsize && timing[i].refresh) { - Mode = xf86CVTMode(timing[i].hsize, timing[i].vsize, - timing[i].refresh, FALSE, FALSE); - Mode->type = M_T_DRIVER; - Modes = xf86ModesAdd(Modes, Mode); - } - } - - return Modes; -} - -/* - * - */ -static DisplayModePtr -DDCModeFromDetailedTiming(int scrnIndex, struct detailed_timings *timing, - int preferred) -{ - DisplayModePtr Mode; - - /* - * Refuse to create modes that are insufficiently large. 64 is a random - * number, maybe the spec says something about what the minimum is. In - * particular I see this frequently with _old_ EDID, 1.0 or so, so maybe - * our parser is just being too aggresive there. - */ - if (timing->h_active < 64 || timing->v_active < 64) { - xf86DrvMsg(scrnIndex, X_INFO, - "%s: Ignoring tiny %dx%d mode\n", __func__, - timing->h_active, timing->v_active); - return NULL; - } - - /* We don't do stereo */ - if (timing->stereo) { - xf86DrvMsg(scrnIndex, X_INFO, - "%s: Ignoring: We don't handle stereo.\n", __func__); - return NULL; - } - - /* We only do seperate sync currently */ - if (timing->sync != 0x03) { - xf86DrvMsg(scrnIndex, X_INFO, - "%s: %dx%d Warning: We only handle seperate" - " sync.\n", __func__, timing->h_active, timing->v_active); - } - - Mode = xnfalloc(sizeof(DisplayModeRec)); - memset(Mode, 0, sizeof(DisplayModeRec)); - - Mode->type = M_T_DRIVER; - if (preferred) - Mode->type |= M_T_PREFERRED; - - Mode->Clock = timing->clock / 1000.0; - - Mode->HDisplay = timing->h_active; - Mode->HSyncStart = timing->h_active + timing->h_sync_off; - Mode->HSyncEnd = Mode->HSyncStart + timing->h_sync_width; - Mode->HTotal = timing->h_active + timing->h_blanking; - - Mode->VDisplay = timing->v_active; - Mode->VSyncStart = timing->v_active + timing->v_sync_off; - Mode->VSyncEnd = Mode->VSyncStart + timing->v_sync_width; - Mode->VTotal = timing->v_active + timing->v_blanking; - - xf86SetModeDefaultName(Mode); - - /* We ignore h/v_size and h/v_border for now. */ - - if (timing->interlaced) - Mode->Flags |= V_INTERLACE; - - if (timing->misc & 0x02) - Mode->Flags |= V_PHSYNC; - else - Mode->Flags |= V_NHSYNC; - - if (timing->misc & 0x01) - Mode->Flags |= V_PVSYNC; - else - Mode->Flags |= V_NVSYNC; - - return Mode; -} - -/* - * - */ -static void -DDCGuessRangesFromModes(int scrnIndex, MonPtr Monitor, DisplayModePtr Modes) -{ - DisplayModePtr Mode = Modes; - - if (!Monitor || !Modes) - return; - - /* set up the ranges for scanning through the modes */ - Monitor->nHsync = 1; - Monitor->hsync[0].lo = 1024.0; - Monitor->hsync[0].hi = 0.0; - - Monitor->nVrefresh = 1; - Monitor->vrefresh[0].lo = 1024.0; - Monitor->vrefresh[0].hi = 0.0; - - while (Mode) { - if (!Mode->HSync) - Mode->HSync = ((float) Mode->Clock ) / ((float) Mode->HTotal); - - if (!Mode->VRefresh) - Mode->VRefresh = (1000.0 * ((float) Mode->Clock)) / - ((float) (Mode->HTotal * Mode->VTotal)); - - if (Mode->HSync < Monitor->hsync[0].lo) - Monitor->hsync[0].lo = Mode->HSync; - - if (Mode->HSync > Monitor->hsync[0].hi) - Monitor->hsync[0].hi = Mode->HSync; - - if (Mode->VRefresh < Monitor->vrefresh[0].lo) - Monitor->vrefresh[0].lo = Mode->VRefresh; - - if (Mode->VRefresh > Monitor->vrefresh[0].hi) - Monitor->vrefresh[0].hi = Mode->VRefresh; - - Mode = Mode->next; - } -} - -DisplayModePtr -xf86DDCGetModes(int scrnIndex, xf86MonPtr DDC) -{ - int preferred, i; - DisplayModePtr Modes = NULL, Mode; - - preferred = PREFERRED_TIMING_MODE(DDC->features.msc); - - /* Add established timings */ - Mode = DDCModesFromEstablished(scrnIndex, &DDC->timings1); - Modes = xf86ModesAdd(Modes, Mode); - - /* Add standard timings */ - Mode = DDCModesFromStandardTiming(scrnIndex, DDC->timings2); - Modes = xf86ModesAdd(Modes, Mode); - - for (i = 0; i < DET_TIMINGS; i++) { - struct detailed_monitor_section *det_mon = &DDC->det_mon[i]; - - switch (det_mon->type) { - case DT: - Mode = DDCModeFromDetailedTiming(scrnIndex, - &det_mon->section.d_timings, - preferred); - preferred = 0; - Modes = xf86ModesAdd(Modes, Mode); - break; - case DS_STD_TIMINGS: - Mode = DDCModesFromStandardTiming(scrnIndex, - det_mon->section.std_t); - Modes = xf86ModesAdd(Modes, Mode); - break; - default: - break; - } - } - - return Modes; -} - -/* - * Fill out MonPtr with xf86MonPtr information. - */ -void -xf86DDCMonitorSet(int scrnIndex, MonPtr Monitor, xf86MonPtr DDC) -{ - DisplayModePtr Modes = NULL, Mode; - int i, clock; - Bool have_hsync = FALSE, have_vrefresh = FALSE; - - if (!Monitor || !DDC) - return; - - Monitor->DDC = DDC; - - Monitor->widthmm = 10 * DDC->features.hsize; - Monitor->heightmm = 10 * DDC->features.vsize; - - /* If this is a digital display, then we can use reduced blanking */ - if (DDC->features.input_type) - Monitor->reducedblanking = TRUE; - /* Allow the user to also enable this through config */ - - Modes = xf86DDCGetModes(scrnIndex, DDC); - - /* Skip EDID ranges if they were specified in the config file */ - have_hsync = (Monitor->nHsync != 0); - have_vrefresh = (Monitor->nVrefresh != 0); - - /* Go through the detailed monitor sections */ - for (i = 0; i < DET_TIMINGS; i++) { - switch (DDC->det_mon[i].type) { - case DS_RANGES: - if (!have_hsync) { - if (!Monitor->nHsync) - xf86DrvMsg(scrnIndex, X_INFO, - "Using EDID range info for horizontal sync\n"); - Monitor->hsync[Monitor->nHsync].lo = - DDC->det_mon[i].section.ranges.min_h; - Monitor->hsync[Monitor->nHsync].hi = - DDC->det_mon[i].section.ranges.max_h; - Monitor->nHsync++; - } else { - xf86DrvMsg(scrnIndex, X_INFO, - "Using hsync ranges from config file\n"); - } - - if (!have_vrefresh) { - if (!Monitor->nVrefresh) - xf86DrvMsg(scrnIndex, X_INFO, - "Using EDID range info for vertical refresh\n"); - Monitor->vrefresh[Monitor->nVrefresh].lo = - DDC->det_mon[i].section.ranges.min_v; - Monitor->vrefresh[Monitor->nVrefresh].hi = - DDC->det_mon[i].section.ranges.max_v; - Monitor->nVrefresh++; - } else { - xf86DrvMsg(scrnIndex, X_INFO, - "Using vrefresh ranges from config file\n"); - } - - clock = DDC->det_mon[i].section.ranges.max_clock * 1000; - if (clock > Monitor->maxPixClock) - Monitor->maxPixClock = clock; - - break; - default: - break; - } - } - - if (Modes) { - /* Print Modes */ - xf86DrvMsg(scrnIndex, X_INFO, "Printing DDC gathered Modelines:\n"); - - Mode = Modes; - while (Mode) { - xf86PrintModeline(scrnIndex, Mode); - Mode = Mode->next; - } - - /* Do we still need ranges to be filled in? */ - if (!Monitor->nHsync || !Monitor->nVrefresh) - DDCGuessRangesFromModes(scrnIndex, Monitor, Modes); - - /* look for last Mode */ - Mode = Modes; - - while (Mode->next) - Mode = Mode->next; - - /* add to MonPtr */ - if (Monitor->Modes) { - Monitor->Last->next = Modes; - Modes->prev = Monitor->Last; - Monitor->Last = Mode; - } else { - Monitor->Modes = Modes; - Monitor->Last = Mode; - } - } -} From ae04f2cb0a068cdc1e519627bf745de0c9e4a85a Mon Sep 17 00:00:00 2001 From: Adam Jackson Date: Thu, 26 Apr 2007 15:28:04 -0400 Subject: [PATCH 28/47] Fix the 'relink' target for kdrive servers. --- hw/kdrive/Makefile.am | 12 +++++++++--- hw/kdrive/ati/Makefile.am | 3 +++ hw/kdrive/chips/Makefile.am | 3 +++ hw/kdrive/ephyr/Makefile.am | 3 +++ hw/kdrive/epson/Makefile.am | 3 +++ hw/kdrive/fake/Makefile.am | 3 +++ hw/kdrive/fbdev/Makefile.am | 3 +++ hw/kdrive/i810/Makefile.am | 3 +++ hw/kdrive/mach64/Makefile.am | 3 +++ hw/kdrive/mga/Makefile.am | 3 +++ hw/kdrive/neomagic/Makefile.am | 3 +++ hw/kdrive/nvidia/Makefile.am | 3 +++ hw/kdrive/pm2/Makefile.am | 3 +++ hw/kdrive/r128/Makefile.am | 3 +++ hw/kdrive/sdl/Makefile.am | 3 +++ hw/kdrive/sis300/Makefile.am | 3 +++ hw/kdrive/smi/Makefile.am | 3 +++ hw/kdrive/vesa/Makefile.am | 3 +++ hw/kdrive/via/Makefile.am | 3 +++ 19 files changed, 63 insertions(+), 3 deletions(-) diff --git a/hw/kdrive/Makefile.am b/hw/kdrive/Makefile.am index e07804948..267d772e2 100644 --- a/hw/kdrive/Makefile.am +++ b/hw/kdrive/Makefile.am @@ -23,14 +23,20 @@ if KDRIVELINUX LINUX_SUBDIRS = linux endif -SUBDIRS = \ - src \ - $(LINUX_SUBDIRS) \ +SERVER_SUBDIRS = \ $(XSDL_SUBDIRS) \ $(FBDEV_SUBDIRS) \ $(VESA_SUBDIRS) \ $(XEPHYR_SUBDIRS) \ $(XFAKE_SUBDIRS) +SUBDIRS = \ + src \ + $(LINUX_SUBDIRS) \ + $(SERVER_SUBDIRS) + DIST_SUBDIRS = vesa ati chips epson i810 mach64 mga neomagic nvidia pm2 r128 \ smi via fbdev sdl ephyr src linux fake sis300 + +relink: + @for i in $(SERVER_SUBDIRS) ; do make -C $$i relink ; done diff --git a/hw/kdrive/ati/Makefile.am b/hw/kdrive/ati/Makefile.am index 76635fb52..61c1c844b 100644 --- a/hw/kdrive/ati/Makefile.am +++ b/hw/kdrive/ati/Makefile.am @@ -62,3 +62,6 @@ Xati_LDADD = \ $(ATI_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/chips/Makefile.am b/hw/kdrive/chips/Makefile.am index 2f8a88da0..a0a10d715 100644 --- a/hw/kdrive/chips/Makefile.am +++ b/hw/kdrive/chips/Makefile.am @@ -24,3 +24,6 @@ Xchips_LDADD = \ $(CHIPS_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/ephyr/Makefile.am b/hw/kdrive/ephyr/Makefile.am index c201fe9d6..1e820c097 100644 --- a/hw/kdrive/ephyr/Makefile.am +++ b/hw/kdrive/ephyr/Makefile.am @@ -29,3 +29,6 @@ Xephyr_LDADD = \ ../../../exa/libexa.la \ @KDRIVE_LIBS@ \ @XEPHYR_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/epson/Makefile.am b/hw/kdrive/epson/Makefile.am index a5bc70b02..d36230abe 100644 --- a/hw/kdrive/epson/Makefile.am +++ b/hw/kdrive/epson/Makefile.am @@ -24,3 +24,6 @@ Xepson_LDADD = \ $(EPSON_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/fake/Makefile.am b/hw/kdrive/fake/Makefile.am index d7ebfc243..69fdf59d5 100644 --- a/hw/kdrive/fake/Makefile.am +++ b/hw/kdrive/fake/Makefile.am @@ -20,3 +20,6 @@ Xfake_LDADD = \ libfake.a \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/fbdev/Makefile.am b/hw/kdrive/fbdev/Makefile.am index cb7180184..b7a863bf6 100644 --- a/hw/kdrive/fbdev/Makefile.am +++ b/hw/kdrive/fbdev/Makefile.am @@ -18,4 +18,7 @@ Xfbdev_LDADD = \ libfbdev.a \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) endif diff --git a/hw/kdrive/i810/Makefile.am b/hw/kdrive/i810/Makefile.am index 503958571..79093da60 100644 --- a/hw/kdrive/i810/Makefile.am +++ b/hw/kdrive/i810/Makefile.am @@ -27,3 +27,6 @@ Xi810_LDADD = \ $(I810_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/mach64/Makefile.am b/hw/kdrive/mach64/Makefile.am index 67712e262..b4d9a4eef 100644 --- a/hw/kdrive/mach64/Makefile.am +++ b/hw/kdrive/mach64/Makefile.am @@ -31,3 +31,6 @@ Xmach64_LDADD = \ $(MACH64_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/mga/Makefile.am b/hw/kdrive/mga/Makefile.am index ee0798915..db1a9563d 100644 --- a/hw/kdrive/mga/Makefile.am +++ b/hw/kdrive/mga/Makefile.am @@ -26,3 +26,6 @@ Xmga_LDADD = \ $(MGA_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/neomagic/Makefile.am b/hw/kdrive/neomagic/Makefile.am index 9a1af990c..33bc3a911 100644 --- a/hw/kdrive/neomagic/Makefile.am +++ b/hw/kdrive/neomagic/Makefile.am @@ -38,3 +38,6 @@ Xneomagic_LDADD = \ $(NEOMAGIC_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/nvidia/Makefile.am b/hw/kdrive/nvidia/Makefile.am index 67eff6961..79d2738a0 100644 --- a/hw/kdrive/nvidia/Makefile.am +++ b/hw/kdrive/nvidia/Makefile.am @@ -27,3 +27,6 @@ Xnvidia_LDADD = \ $(NVIDIA_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/pm2/Makefile.am b/hw/kdrive/pm2/Makefile.am index a7b0f0088..ec70276c5 100644 --- a/hw/kdrive/pm2/Makefile.am +++ b/hw/kdrive/pm2/Makefile.am @@ -25,3 +25,6 @@ Xpm2_LDADD = \ $(PM2_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/r128/Makefile.am b/hw/kdrive/r128/Makefile.am index eab80cce0..1ca1a605b 100644 --- a/hw/kdrive/r128/Makefile.am +++ b/hw/kdrive/r128/Makefile.am @@ -24,3 +24,6 @@ Xr128_LDADD = \ $(R128_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/sdl/Makefile.am b/hw/kdrive/sdl/Makefile.am index f5abb86e8..fa09640d2 100644 --- a/hw/kdrive/sdl/Makefile.am +++ b/hw/kdrive/sdl/Makefile.am @@ -11,3 +11,6 @@ Xsdl_LDADD = @KDRIVE_PURE_LIBS@ \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ \ @XSDL_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/sis300/Makefile.am b/hw/kdrive/sis300/Makefile.am index 98020745c..62f3266e1 100644 --- a/hw/kdrive/sis300/Makefile.am +++ b/hw/kdrive/sis300/Makefile.am @@ -38,3 +38,6 @@ Xsis_LDADD = \ $(SIS_LIBS) \ @KDRIVE_LIBS@ \ $(TSLIB_FLAG) + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/smi/Makefile.am b/hw/kdrive/smi/Makefile.am index 0fd9729fc..a6ac474b0 100644 --- a/hw/kdrive/smi/Makefile.am +++ b/hw/kdrive/smi/Makefile.am @@ -29,3 +29,6 @@ Xsmi_LDADD = \ $(SMI_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/vesa/Makefile.am b/hw/kdrive/vesa/Makefile.am index 54a6f47ee..e062fe777 100644 --- a/hw/kdrive/vesa/Makefile.am +++ b/hw/kdrive/vesa/Makefile.am @@ -23,3 +23,6 @@ Xvesa_LDADD = \ libvesa.a \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) diff --git a/hw/kdrive/via/Makefile.am b/hw/kdrive/via/Makefile.am index 0ea88816a..1c5796963 100644 --- a/hw/kdrive/via/Makefile.am +++ b/hw/kdrive/via/Makefile.am @@ -25,3 +25,6 @@ Xvia_LDADD = \ $(VIA_LIBS) \ @KDRIVE_LIBS@ \ @XSERVER_LIBS@ + +relink: + rm -f $(bin_PROGRAMS) && make $(bin_PROGRAMS) From 6c4f1826bf2c5f30f5fe6e489a02b6375478b380 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Fri, 27 Apr 2007 08:13:08 -0400 Subject: [PATCH 29/47] Bug fix in fbCompositeIn_nx8x8888 Make sure both halves of the dst word is set to zero when the masks are both 0. --- fb/fbmmx.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index 0e9074648..cea8ad9d5 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -1900,7 +1900,7 @@ fbCompositeSolidMaskSrc_nx8x8888mmx (CARD8 op, } else { - *dst = 0; + *(ullong *)dst = 0; } mask += 2; From 78a20455356ccc310f73cfc65ad65a7677eee7e5 Mon Sep 17 00:00:00 2001 From: Soren Sandmann Pedersen Date: Fri, 27 Apr 2007 15:20:24 -0400 Subject: [PATCH 30/47] Pixman merging More msvc++ porting --- fb/fbmmx.c | 35 ++++++++++++++++++++++++++++++++++- 1 file changed, 34 insertions(+), 1 deletion(-) diff --git a/fb/fbmmx.c b/fb/fbmmx.c index cea8ad9d5..a322bec2c 100644 --- a/fb/fbmmx.c +++ b/fb/fbmmx.c @@ -1700,7 +1700,9 @@ fbSolidFillmmx (DrawablePtr pDraw, CARD8 *byte_line; FbBits *bits; int xoff, yoff; +#ifdef __GNUC__ __m64 v1, v2, v3, v4, v5, v6, v7; +#endif CHECKPOINT(); @@ -1730,6 +1732,7 @@ fbSolidFillmmx (DrawablePtr pDraw, fill = ((ullong)xor << 32) | xor; vfill = (__m64)fill; +#ifdef __GNUC__ __asm__ ( "movq %7, %0\n" "movq %7, %1\n" @@ -1741,6 +1744,7 @@ fbSolidFillmmx (DrawablePtr pDraw, : "=y" (v1), "=y" (v2), "=y" (v3), "=y" (v4), "=y" (v5), "=y" (v6), "=y" (v7) : "y" (vfill)); +#endif while (height--) { @@ -1766,6 +1770,7 @@ fbSolidFillmmx (DrawablePtr pDraw, while (w >= 64) { +#ifdef __GNUC__ __asm__ ( "movq %1, (%0)\n" "movq %2, 8(%0)\n" @@ -1780,7 +1785,16 @@ fbSolidFillmmx (DrawablePtr pDraw, "y" (vfill), "y" (v1), "y" (v2), "y" (v3), "y" (v4), "y" (v5), "y" (v6), "y" (v7) : "memory"); - +#else + *(__m64*) (d + 0) = vfill; + *(__m64*) (d + 8) = vfill; + *(__m64*) (d + 16) = vfill; + *(__m64*) (d + 24) = vfill; + *(__m64*) (d + 32) = vfill; + *(__m64*) (d + 40) = vfill; + *(__m64*) (d + 48) = vfill; + *(__m64*) (d + 56) = vfill; +#endif w -= 64; d += 64; } @@ -2823,6 +2837,7 @@ fbCopyAreammx (DrawablePtr pSrc, while (w >= 64) { +#ifdef __GNUC__ __asm__ ( "movq (%1), %%mm0\n" "movq 8(%1), %%mm1\n" @@ -2846,6 +2861,24 @@ fbCopyAreammx (DrawablePtr pSrc, : "memory", "%mm0", "%mm1", "%mm2", "%mm3", "%mm4", "%mm5", "%mm6", "%mm7"); +#else + __m64 v0 = *(__m64 *)(s + 0); + __m64 v1 = *(__m64 *)(s + 8); + __m64 v2 = *(__m64 *)(s + 16); + __m64 v3 = *(__m64 *)(s + 24); + __m64 v4 = *(__m64 *)(s + 32); + __m64 v5 = *(__m64 *)(s + 40); + __m64 v6 = *(__m64 *)(s + 48); + __m64 v7 = *(__m64 *)(s + 56); + *(__m64 *)(d + 0) = v0; + *(__m64 *)(d + 8) = v1; + *(__m64 *)(d + 16) = v2; + *(__m64 *)(d + 24) = v3; + *(__m64 *)(d + 32) = v4; + *(__m64 *)(d + 40) = v5; + *(__m64 *)(d + 48) = v6; + *(__m64 *)(d + 56) = v7; +#endif w -= 64; s += 64; From 2866e0bac9b8dd3892c5e68abcfc6c97cebaf88a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:38:13 +0200 Subject: [PATCH 31/47] Fix a couple of picture repeat fields incorrectly compared to RepeatNormal. --- exa/exa_render.c | 4 ++-- fb/fbpict.c | 2 +- hw/xgl/xglcompose.c | 2 +- 3 files changed, 4 insertions(+), 4 deletions(-) diff --git a/exa/exa_render.c b/exa/exa_render.c index b78d7282c..d48a1425f 100644 --- a/exa/exa_render.c +++ b/exa/exa_render.c @@ -627,8 +627,8 @@ exaComposite(CARD8 op, pMask->repeat = 0; if (pExaScr->info->PrepareComposite && - (!pSrc->repeat || pSrc->repeat == RepeatNormal) && - (!pMask || !pMask->repeat || pMask->repeat == RepeatNormal) && + (!pSrc->repeat || pSrc->repeatType == RepeatNormal) && + (!pMask || !pMask->repeat || pMask->repeatType == RepeatNormal) && !pSrc->alphaMap && (!pMask || !pMask->alphaMap) && !pDst->alphaMap) { Bool isSrcSolid; diff --git a/fb/fbpict.c b/fb/fbpict.c index 0a08affd7..2c1039834 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -1495,7 +1495,7 @@ fbComposite (CARD8 op, { xMask += pMask->pDrawable->x; yMask += pMask->pDrawable->y; - maskRepeat = pMask->repeat == RepeatNormal; + maskRepeat = pMask->repeatType == RepeatNormal; if (pMask->filter == PictFilterConvolution) maskTransform = TRUE; diff --git a/hw/xgl/xglcompose.c b/hw/xgl/xglcompose.c index d2aead05b..34f7a0c43 100644 --- a/hw/xgl/xglcompose.c +++ b/hw/xgl/xglcompose.c @@ -177,7 +177,7 @@ xglCompositeGeneral (CARD8 op, { if (!pSrc->transform && pSrc->filter != PictFilterConvolution) { - if (pSrc->pDrawable && pSrc->repeat == RepeatNormal) + if (pSrc->pDrawable && pSrc->repeatType == RepeatNormal) { XGL_PIXMAP_PRIV ((PixmapPtr) pSrc->pDrawable); From 5e4b3232dafe3b0dec65bf639bebaba4774210b7 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:38:22 +0200 Subject: [PATCH 32/47] Fix fbCompositeTrans_0888xnx0888 build for wfb on big endian. --- fb/fbpict.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/fb/fbpict.c b/fb/fbpict.c index 2c1039834..a7359673d 100644 --- a/fb/fbpict.c +++ b/fb/fbpict.c @@ -140,7 +140,7 @@ fbIn (CARD32 x, CARD8 y) temp=count&3; \ where-=temp; \ workingWhere=(CARD32 *)where; \ - workingVal=READ(workingWhere)++; \ + workingVal=READ(workingWhere++); \ count=4-temp; \ workingVal<<=(8*temp) #define readPacked(where,x,y,z) {if(!(x)) { (x)=4; y = READ(z++); } where=(y)>>24; (y)<<=8; (x)--;} From d3f8667341bfe6dc7d0258c4ad69377f37d88d95 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:44:27 +0200 Subject: [PATCH 33/47] EXA: Fix exaEnableDisableFBAccess for nested disables and enables. --- exa/exa_offscreen.c | 6 ++++-- exa/exa_priv.h | 1 + 2 files changed, 5 insertions(+), 2 deletions(-) diff --git a/exa/exa_offscreen.c b/exa/exa_offscreen.c index 7708dd739..38ad58f60 100644 --- a/exa/exa_offscreen.c +++ b/exa/exa_offscreen.c @@ -341,13 +341,15 @@ exaEnableDisableFBAccess (int index, Bool enable) ScreenPtr pScreen = screenInfo.screens[index]; ExaScreenPriv (pScreen); - if (!enable) { + if (!enable && pExaScr->disableFbCount++ == 0) { if (pExaScr->info->exa_minor < 1) ExaOffscreenSwapOut (pScreen); else ExaOffscreenEjectPixmaps (pScreen); pExaScr->swappedOut = TRUE; - } else { + } + + if (enable && --pExaScr->disableFbCount == 0) { if (pExaScr->info->exa_minor < 1) ExaOffscreenSwapIn (pScreen); pExaScr->swappedOut = FALSE; diff --git a/exa/exa_priv.h b/exa/exa_priv.h index 984cb669b..382f0591e 100644 --- a/exa/exa_priv.h +++ b/exa/exa_priv.h @@ -113,6 +113,7 @@ typedef struct { enum ExaMigrationHeuristic migration; Bool hideOffscreenPixmapData; Bool checkDirtyCorrectness; + unsigned disableFbCount; } ExaScreenPrivRec, *ExaScreenPrivPtr; /* From e869573b52fac69fb88cea120daaeec59c7a3461 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:45:48 +0200 Subject: [PATCH 34/47] EXA: exaAssertNotDirty improvements. * Return early if the valid region is empty or the pixmap is pinned. * Fix loop for several cliprects. --- exa/exa_migration.c | 21 ++++++++------------- 1 file changed, 8 insertions(+), 13 deletions(-) diff --git a/exa/exa_migration.c b/exa/exa_migration.c index eedc5fd03..70d8e1235 100644 --- a/exa/exa_migration.c +++ b/exa/exa_migration.c @@ -464,12 +464,10 @@ exaAssertNotDirty (PixmapPtr pPixmap) BoxPtr pBox = REGION_RECTS(pValidReg); Bool ret = TRUE; - if (pExaPixmap == NULL || pExaPixmap->fb_ptr == NULL) + if (!nbox || exaPixmapIsPinned(pPixmap) || pExaPixmap->fb_ptr == NULL) return ret; - dst = pExaPixmap->sys_ptr; dst_pitch = pExaPixmap->sys_pitch; - src = pExaPixmap->fb_ptr; src_pitch = pExaPixmap->fb_pitch; cpp = pPixmap->drawable.bitsPerPixel / 8; @@ -486,21 +484,18 @@ exaAssertNotDirty (PixmapPtr pPixmap) continue; rowbytes = (pBox->x2 - pBox->x1) * cpp; - src += pBox->y1 * src_pitch + pBox->x1 * cpp; - dst += pBox->y1 * dst_pitch + pBox->x1 * cpp; + src = pExaPixmap->fb_ptr + pBox->y1 * src_pitch + pBox->x1 * cpp; + dst = pExaPixmap->sys_ptr + pBox->y1 * dst_pitch + pBox->x1 * cpp; - for (y = pBox->y2 - pBox->y1; y; y--) { - if (memcmp(dst + pBox->y1 * dst_pitch + pBox->x1 * cpp, - src + pBox->y1 * src_pitch + pBox->x1 * cpp, - (pBox->x2 - pBox->x1) * cpp) != 0) { + for (y = pBox->y1; y < pBox->y2; + y++, src += src_pitch, dst += dst_pitch) { + if (memcmp(dst, src, rowbytes) != 0) { ret = FALSE; + exaPixmapDirty(pPixmap, pBox->x1, pBox->y1, pBox->x2, + pBox->y2); break; } - src += src_pitch; - dst += dst_pitch; } - src -= pBox->y1 * src_pitch + pBox->x1 * cpp; - dst -= pBox->y1 * dst_pitch + pBox->x1 * cpp; } exaFinishAccess(&pPixmap->drawable, EXA_PREPARE_SRC); From 567f18a09bfb05f448be40c7ebe0f210f955601c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:46:49 +0200 Subject: [PATCH 35/47] EXA: FillRegion{Solid,Tiled} improvements. * Support planemasks, different ALUs and arbitrary tile origin. * Leave damage tracking and non-trivial fallbacks to callers. * Always migrate for fallbacks. This is in preparation for using these from more other functions. --- exa/exa.c | 2 +- exa/exa_accel.c | 102 ++++++++++++++++++++++++++++++------------------ exa/exa_priv.h | 8 ++++ 3 files changed, 72 insertions(+), 40 deletions(-) diff --git a/exa/exa.c b/exa/exa.c index 23fe55516..3d77800ff 100644 --- a/exa/exa.c +++ b/exa/exa.c @@ -126,7 +126,7 @@ exaGetDrawablePixmap(DrawablePtr pDrawable) * the backing drawable. These coordinates are nonzero only for redirected * windows. */ -static void +void exaGetDrawableDeltas (DrawablePtr pDrawable, PixmapPtr pPixmap, int *xp, int *yp) { diff --git a/exa/exa_accel.c b/exa/exa_accel.c index e633d80a5..95084fc37 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -1043,10 +1043,12 @@ exaCopyWindow(WindowPtr pWin, DDXPointRec ptOldOrg, RegionPtr prgnSrc) REGION_UNINIT(pWin->drawable.pScreen, &rgnDst); } -static void +static Bool exaFillRegionSolid (DrawablePtr pDrawable, RegionPtr pRegion, - Pixel pixel) + Pixel pixel, + CARD32 planemask, + CARD32 alu) { ExaScreenPriv(pDrawable->pScreen); PixmapPtr pPixmap; @@ -1062,22 +1064,19 @@ exaFillRegionSolid (DrawablePtr pDrawable, if (pPixmap->drawable.width > pExaScr->info->maxX || pPixmap->drawable.height > pExaScr->info->maxY) { - exaDoMigration (pixmaps, 1, FALSE); goto fallback; } else { exaDoMigration (pixmaps, 1, TRUE); } if ((pPixmap = exaGetOffscreenPixmap (pDrawable, &xoff, &yoff)) && - (*pExaScr->info->PrepareSolid) (pPixmap, GXcopy, FB_ALLONES, pixel)) + (*pExaScr->info->PrepareSolid) (pPixmap, alu, planemask, pixel)) { while (nbox--) { (*pExaScr->info->Solid) (pPixmap, pBox->x1 + xoff, pBox->y1 + yoff, pBox->x2 + xoff, pBox->y2 + yoff); - exaPixmapDirty (pPixmap, pBox->x1 + xoff, pBox->y1 + yoff, - pBox->x2 + xoff, pBox->y2 + yoff); pBox++; } (*pExaScr->info->DoneSolid) (pPixmap); @@ -1086,27 +1085,30 @@ exaFillRegionSolid (DrawablePtr pDrawable, else { fallback: + if (alu != GXcopy || planemask != FB_ALLONES) + return FALSE; EXA_FALLBACK(("to %p (%c)\n", pDrawable, exaDrawableLocation(pDrawable))); + exaDoMigration (pixmaps, 1, FALSE); exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); fbFillRegionSolid (pDrawable, pRegion, 0, fbReplicatePixel (pixel, pDrawable->bitsPerPixel)); exaFinishAccess (pDrawable, EXA_PREPARE_DEST); - while (nbox--) - { - exaDrawableDirty (pDrawable, pBox->x1, pBox->y1, pBox->x2, pBox->y2); - pBox++; - } } + + return TRUE; } /* Try to do an accelerated tile of the pTile into pRegion of pDrawable. * Based on fbFillRegionTiled(), fbTile(). */ -static void +Bool exaFillRegionTiled (DrawablePtr pDrawable, RegionPtr pRegion, - PixmapPtr pTile) + PixmapPtr pTile, + DDXPointPtr pPatOrg, + CARD32 planemask, + CARD32 alu) { ExaScreenPriv(pDrawable->pScreen); PixmapPtr pPixmap; @@ -1122,10 +1124,10 @@ exaFillRegionTiled (DrawablePtr pDrawable, /* If we're filling with a solid color, grab it out and go to * FillRegionSolid, saving numerous copies. */ - if (tileWidth == 1 && tileHeight == 1) { - exaFillRegionSolid(pDrawable, pRegion, exaGetPixmapFirstPixel (pTile)); - return; - } + if (tileWidth == 1 && tileHeight == 1) + return exaFillRegionSolid(pDrawable, pRegion, + exaGetPixmapFirstPixel (pTile), planemask, + alu); pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = FALSE; @@ -1139,7 +1141,6 @@ exaFillRegionTiled (DrawablePtr pDrawable, tileWidth > pExaScr->info->maxX || tileHeight > pExaScr->info->maxY) { - exaDoMigration (pixmaps, 2, FALSE); goto fallback; } else { exaDoMigration (pixmaps, 2, TRUE); @@ -1153,8 +1154,9 @@ exaFillRegionTiled (DrawablePtr pDrawable, if (!exaPixmapIsOffscreen(pTile)) goto fallback; - if ((*pExaScr->info->PrepareCopy) (exaGetOffscreenPixmap((DrawablePtr)pTile, &tileXoff, &tileYoff), pPixmap, 0, 0, GXcopy, - FB_ALLONES)) + if ((*pExaScr->info->PrepareCopy) (exaGetOffscreenPixmap((DrawablePtr)pTile, + &tileXoff, &tileYoff), + pPixmap, 0, 0, alu, planemask)) { while (nbox--) { @@ -1162,7 +1164,7 @@ exaFillRegionTiled (DrawablePtr pDrawable, int dstY = pBox->y1; int tileY; - tileY = (dstY - pDrawable->y) % tileHeight; + tileY = (dstY - pDrawable->y - pPatOrg->y) % tileHeight; while (height > 0) { int width = pBox->x2 - pBox->x1; int dstX = pBox->x1; @@ -1173,7 +1175,7 @@ exaFillRegionTiled (DrawablePtr pDrawable, h = height; height -= h; - tileX = (dstX - pDrawable->x) % tileWidth; + tileX = (dstX - pDrawable->x - pPatOrg->x) % tileWidth; while (width > 0) { int w = tileWidth - tileX; if (w > width) @@ -1190,38 +1192,44 @@ exaFillRegionTiled (DrawablePtr pDrawable, dstY += h; tileY = 0; } - exaPixmapDirty (pPixmap, pBox->x1 + xoff, pBox->y1 + yoff, - pBox->x2 + xoff, pBox->y2 + yoff); pBox++; } (*pExaScr->info->DoneCopy) (pPixmap); exaMarkSync(pDrawable->pScreen); - return; + return TRUE; } fallback: + if (alu != GXcopy || planemask != FB_ALLONES) + return FALSE; EXA_FALLBACK(("from %p to %p (%c,%c)\n", pTile, pDrawable, exaDrawableLocation(&pTile->drawable), exaDrawableLocation(pDrawable))); + exaDoMigration (pixmaps, 2, FALSE); exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); exaPrepareAccess ((DrawablePtr)pTile, EXA_PREPARE_SRC); fbFillRegionTiled (pDrawable, pRegion, pTile); exaFinishAccess ((DrawablePtr)pTile, EXA_PREPARE_SRC); exaFinishAccess (pDrawable, EXA_PREPARE_DEST); - while (nbox--) - { - exaDrawableDirty (pDrawable, pBox->x1, pBox->y1, pBox->x2, pBox->y2); - pBox++; - } + + return TRUE; } void exaPaintWindow(WindowPtr pWin, RegionPtr pRegion, int what) { ExaScreenPriv (pWin->drawable.pScreen); - if (!REGION_NUM_RECTS(pRegion)) + PixmapPtr pPixmap = exaGetDrawablePixmap((DrawablePtr)pWin); + int xoff, yoff; + BoxPtr pBox; + int nbox = REGION_NUM_RECTS(pRegion); + + if (!nbox) return; + if (!pExaScr->swappedOut) { + DDXPointRec zeros = { 0, 0 }; + switch (what) { case PW_BACKGROUND: switch (pWin->backgroundState) { @@ -1235,25 +1243,41 @@ exaPaintWindow(WindowPtr pWin, RegionPtr pRegion, int what) what); return; case BackgroundPixel: - exaFillRegionSolid((DrawablePtr)pWin, pRegion, pWin->background.pixel); - return; + exaFillRegionSolid((DrawablePtr)pWin, pRegion, pWin->background.pixel, + FB_ALLONES, GXcopy); + goto damage; case BackgroundPixmap: - exaFillRegionTiled((DrawablePtr)pWin, pRegion, pWin->background.pixmap); - return; + exaFillRegionTiled((DrawablePtr)pWin, pRegion, pWin->background.pixmap, + &zeros, FB_ALLONES, GXcopy); + goto damage; } break; case PW_BORDER: if (pWin->borderIsPixel) { - exaFillRegionSolid((DrawablePtr)pWin, pRegion, pWin->border.pixel); - return; + exaFillRegionSolid((DrawablePtr)pWin, pRegion, pWin->border.pixel, + FB_ALLONES, GXcopy); + goto damage; } else { - exaFillRegionTiled((DrawablePtr)pWin, pRegion, pWin->border.pixmap); - return; + exaFillRegionTiled((DrawablePtr)pWin, pRegion, pWin->border.pixmap, + &zeros, FB_ALLONES, GXcopy); + goto damage; } break; } } ExaCheckPaintWindow (pWin, pRegion, what); + +damage: + exaGetDrawableDeltas((DrawablePtr)pWin, pPixmap, &xoff, &yoff); + + pBox = REGION_RECTS(pRegion); + + while (nbox--) + { + exaPixmapDirty (pPixmap, pBox->x1 + xoff, pBox->y1 + yoff, + pBox->x2 + xoff, pBox->y2 + yoff); + pBox++; + } } /** diff --git a/exa/exa_priv.h b/exa/exa_priv.h index 382f0591e..b97608214 100644 --- a/exa/exa_priv.h +++ b/exa/exa_priv.h @@ -288,6 +288,10 @@ exaGetPixmapFirstPixel (PixmapPtr pPixmap); void exaCopyWindow(WindowPtr pWin, DDXPointRec ptOldOrg, RegionPtr prgnSrc); +Bool +exaFillRegionTiled (DrawablePtr pDrawable, RegionPtr pRegion, PixmapPtr pTile, + DDXPointPtr pPatOrg, CARD32 planemask, CARD32 alu); + void exaPaintWindow(WindowPtr pWin, RegionPtr pRegion, int what); @@ -343,6 +347,10 @@ exaFinishAccess(DrawablePtr pDrawable, int index); void exaPixmapDirty(PixmapPtr pPix, int x1, int y1, int x2, int y2); +void +exaGetDrawableDeltas (DrawablePtr pDrawable, PixmapPtr pPixmap, + int *xp, int *yp); + void exaDrawableDirty(DrawablePtr pDrawable, int x1, int y1, int x2, int y2); From 0c8905ebc91cf654facef84ee52231a358deec5c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:47:08 +0200 Subject: [PATCH 36/47] EXA: PolyFillRect improvements. * Convert rects to region and use it for damage tracking. * When possible, defer to exaFillRegion{Solid,Tiled} using converted region. * Always migrate for fallbacks. * Move damage tracking out of ExaCheckPolyFillRect. --- exa/exa_accel.c | 83 ++++++++++++++++++++++++++++++++++++----------- exa/exa_unaccel.c | 31 +++--------------- 2 files changed, 69 insertions(+), 45 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index 95084fc37..721882b4c 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -618,6 +618,9 @@ exaPolySegment (DrawablePtr pDrawable, GCPtr pGC, int nseg, DEALLOCATE_LOCAL(prect); } +static Bool exaFillRegionSolid (DrawablePtr pDrawable, RegionPtr pRegion, + Pixel pixel, CARD32 planemask, CARD32 alu); + static void exaPolyFillRect(DrawablePtr pDrawable, GCPtr pGC, @@ -626,7 +629,7 @@ exaPolyFillRect(DrawablePtr pDrawable, { ExaScreenPriv (pDrawable->pScreen); RegionPtr pClip = fbGetCompositeClip(pGC); - PixmapPtr pPixmap; + PixmapPtr pPixmap = exaGetDrawablePixmap(pDrawable); register BoxPtr pbox; BoxPtr pextent; int extentX1, extentX2, extentY1, extentY2; @@ -635,39 +638,80 @@ exaPolyFillRect(DrawablePtr pDrawable, int xoff, yoff; int xorg, yorg; int n; - ExaMigrationRec pixmaps[1]; + ExaMigrationRec pixmaps[2]; + RegionPtr pReg = RECTS_TO_REGION(pScreen, nrect, prect, CT_UNSORTED); + RegionPtr pDamageReg = DamageRegion(ExaGetPixmapPriv(pPixmap)->pDamage); + + /* Compute intersection of rects and clip region */ + REGION_TRANSLATE(pScreen, pReg, pDrawable->x, pDrawable->y); + REGION_INTERSECT(pScreen, pReg, pClip, pReg); + + if (!REGION_NUM_RECTS(pReg)) { + REGION_DESTROY(pScreen, pReg); + return; + } pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = FALSE; - pixmaps[0].pPix = pPixmap = exaGetDrawablePixmap (pDrawable); + pixmaps[0].pPix = pPixmap; + exaGetDrawableDeltas(pDrawable, pPixmap, &xoff, &yoff); + if (pExaScr->swappedOut || - pGC->fillStyle != FillSolid || pPixmap->drawable.width > pExaScr->info->maxX || pPixmap->drawable.height > pExaScr->info->maxY) { - exaDoMigration (pixmaps, 1, FALSE); - ExaCheckPolyFillRect (pDrawable, pGC, nrect, prect); - while (nrect-- >= 0) { - exaDrawableDirty(pDrawable, - pDrawable->x + prect->x, - pDrawable->y + prect->y, - pDrawable->x + prect->x + prect->width, - pDrawable->y + prect->y + prect->height); - prect++; - } - return; - } else { - exaDoMigration (pixmaps, 1, TRUE); + goto fallback; } - if (!(pPixmap = exaGetOffscreenPixmap (pDrawable, &xoff, &yoff)) || + /* For ROPs where overlaps don't matter, convert rectangles to region and + * call exaFillRegion{Solid,Tiled}. + */ + if ((pGC->fillStyle == FillSolid || pGC->fillStyle == FillTiled) && + (pGC->alu == GXcopy || pGC->alu == GXclear || pGC->alu == GXnoop || + pGC->alu == GXcopyInverted || pGC->alu == GXset)) { + if (((pGC->fillStyle == FillSolid || pGC->tileIsPixel) && + exaFillRegionSolid(pDrawable, pReg, pGC->fillStyle == FillSolid ? + pGC->fgPixel : pGC->tile.pixel, pGC->planemask, + pGC->alu)) || + (pGC->fillStyle == FillTiled && !pGC->tileIsPixel && + exaFillRegionTiled(pDrawable, pReg, pGC->tile.pixmap, &pGC->patOrg, + pGC->planemask, pGC->alu))) { + goto damage; + } + } + + if (pGC->fillStyle != FillSolid && + !(pGC->tileIsPixel && pGC->fillStyle == FillTiled)) + { + goto fallback; + } + + exaDoMigration (pixmaps, 1, TRUE); + + if (!exaPixmapIsOffscreen (pPixmap) || !(*pExaScr->info->PrepareSolid) (pPixmap, pGC->alu, pGC->planemask, pGC->fgPixel)) { +fallback: + if (pGC->fillStyle == FillTiled && !pGC->tileIsPixel) { + pixmaps[1].as_dst = FALSE; + pixmaps[1].as_src = TRUE; + pixmaps[1].pPix = pGC->tile.pixmap; + exaDoMigration (pixmaps, 2, FALSE); + } else { + exaDoMigration (pixmaps, 1, FALSE); + } + ExaCheckPolyFillRect (pDrawable, pGC, nrect, prect); + +damage: + REGION_TRANSLATE(pScreen, pReg, xoff, yoff); + REGION_UNION(pScreen, pDamageReg, pReg, pDamageReg); + REGION_DESTROY(pScreen, pReg); + return; } @@ -715,7 +759,8 @@ exaPolyFillRect(DrawablePtr pDrawable, pbox = REGION_RECTS(pClip); /* * clip the rectangle to each box in the clip region - * this is logically equivalent to calling Intersect() + * this is logically equivalent to calling Intersect(), + * but rectangles may overlap each other here. */ while(n--) { diff --git a/exa/exa_unaccel.c b/exa/exa_unaccel.c index 7713a08c9..a7f939845 100644 --- a/exa/exa_unaccel.c +++ b/exa/exa_unaccel.c @@ -201,32 +201,11 @@ ExaCheckPolyFillRect (DrawablePtr pDrawable, GCPtr pGC, { EXA_FALLBACK(("to %p (%c)\n", pDrawable, exaDrawableLocation(pDrawable))); - if (nrect) { - int x1 = max(prect->x, 0), y1 = max(prect->y, 0); - int x2 = min(prect->x + prect->width, pDrawable->width); - int y2 = min(prect->y + prect->height, pDrawable->height); - - exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); - exaPrepareAccessGC (pGC); - fbPolyFillRect (pDrawable, pGC, nrect, prect); - exaFinishAccessGC (pGC); - exaFinishAccess (pDrawable, EXA_PREPARE_DEST); - - /* Only track bounding box of damage, as this path can degenerate to - * zillions of damage boxes - */ - while (--nrect) - { - prect++; - x1 = min(x1, prect->x); - x2 = max(x2, prect->x + prect->width); - y1 = min(y1, prect->y); - y2 = max(y2, prect->y + prect->height); - } - - exaDrawableDirty (pDrawable, pDrawable->x + x1, pDrawable->y + y1, - pDrawable->x + x2, pDrawable->y + y2); - } + exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); + exaPrepareAccessGC (pGC); + fbPolyFillRect (pDrawable, pGC, nrect, prect); + exaFinishAccessGC (pGC); + exaFinishAccess (pDrawable, EXA_PREPARE_DEST); } void From ce317a5b76c053f449122c46e1372bf8e067cb4c Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:47:16 +0200 Subject: [PATCH 37/47] EXA: Glyphs improvements. * Don't waste effort on invisible glyphs. * Add damage tracking where necessary. * Always migrate for fallbacks. --- exa/exa_render.c | 42 ++++++++++++++++++++++++++++++++---------- 1 file changed, 32 insertions(+), 10 deletions(-) diff --git a/exa/exa_render.c b/exa/exa_render.c index d48a1425f..2234cdbda 100644 --- a/exa/exa_render.c +++ b/exa/exa_render.c @@ -845,10 +845,11 @@ exaGlyphs (CARD8 op, PixmapPtr pPixmap = NULL; PicturePtr pPicture; PixmapPtr pMaskPixmap = NULL; + PixmapPtr pDstPixmap = exaGetDrawablePixmap(pDst->pDrawable); PicturePtr pMask; ScreenPtr pScreen = pDst->pDrawable->pScreen; int width = 0, height = 0; - int x, y; + int x, y, x1, y1, xoff, yoff; int xDst = list->xOff, yDst = list->yOff; int n; int error; @@ -892,7 +893,12 @@ exaGlyphs (CARD8 op, xRectangle rect; miGlyphExtents (nlist, list, glyphs, &extents); - + + extents.x1 = max(extents.x1, 0); + extents.y1 = max(extents.y1, 0); + extents.x2 = min(extents.x2, pDst->pDrawable->width); + extents.y2 = min(extents.y2, pDst->pDrawable->height); + if (extents.x2 <= extents.x1 || extents.y2 <= extents.y1) return; width = extents.x2 - extents.x1; @@ -918,6 +924,7 @@ exaGlyphs (CARD8 op, rect.width = width; rect.height = height; (*pGC->ops->PolyFillRect) (&pMaskPixmap->drawable, pGC, 1, &rect); + exaPixmapDirty(pMaskPixmap, 0, 0, width, height); FreeScratchGC (pGC); x = -extents.x1; y = -extents.y1; @@ -929,6 +936,8 @@ exaGlyphs (CARD8 op, y = 0; } + exaGetDrawableDeltas(pDst->pDrawable, pDstPixmap, &xoff, &yoff); + while (nlist--) { GCPtr pGC = NULL; @@ -983,13 +992,21 @@ exaGlyphs (CARD8 op, pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = TRUE; pixmaps[0].pPix = pPixmap; - exaDoMigration (pixmaps, 1, TRUE); + exaDoMigration (pixmaps, 1, pExaScr->info->PrepareComposite != NULL); while (n--) { GlyphPtr glyph = *glyphs++; pointer glyphdata = (pointer) (glyph + 1); - + DrawablePtr pCmpDrw = (maskFormat ? pMask : pDst)->pDrawable; + + x1 = x - glyph->info.x; + y1 = y - glyph->info.y; + + if (x1 >= pCmpDrw->width || y1 >= pCmpDrw->height || + (x1 + glyph->info.width) <= 0 || (y1 + glyph->info.height) <= 0) + goto nextglyph; + (*pScreen->ModifyPixmapHeader) (pScratchPixmap, glyph->info.width, glyph->info.height, @@ -1048,17 +1065,22 @@ exaGlyphs (CARD8 op, if (maskFormat) { exaComposite (PictOpAdd, pPicture, NULL, pMask, 0, 0, 0, 0, - x - glyph->info.x, y - glyph->info.y, - glyph->info.width, glyph->info.height); + x1, y1, glyph->info.width, glyph->info.height); + exaPixmapDirty(pMaskPixmap, x1, y1, x1 + glyph->info.width, + y1 + glyph->info.height); } else { exaComposite (op, pSrc, pPicture, pDst, - xSrc + (x - glyph->info.x) - xDst, - ySrc + (y - glyph->info.y) - yDst, - 0, 0, x - glyph->info.x, y - glyph->info.y, - glyph->info.width, glyph->info.height); + xSrc + x1 - xDst, ySrc + y1 - yDst, + 0, 0, x1, y1, glyph->info.width, + glyph->info.height); + x1 += pDst->pDrawable->x + xoff; + y1 += pDst->pDrawable->y + yoff; + exaPixmapDirty(pDstPixmap, x1, y1, x1 + glyph->info.width, + y1 + glyph->info.height); } +nextglyph: x += glyph->info.xOff; y += glyph->info.yOff; } From 81b055605c34b5823f6c5f63cc0f92f43c6b7252 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:47:43 +0200 Subject: [PATCH 38/47] EXA: Composite improvements. * Defer to simpler hooks in more cases (inspired by XAA behaviour). * Move damage tracking from lower to higher level functions. * Always migrate for fallbacks. --- exa/exa_render.c | 79 +++++++++++++++++++++++++++--------------------- 1 file changed, 44 insertions(+), 35 deletions(-) diff --git a/exa/exa_render.c b/exa/exa_render.c index 2234cdbda..63a412c5f 100644 --- a/exa/exa_render.c +++ b/exa/exa_render.c @@ -297,15 +297,15 @@ exaTryDriverSolidFill(PicturePtr pSrc, nbox = REGION_NUM_RECTS(®ion); pbox = REGION_RECTS(®ion); + while (nbox--) { (*pExaScr->info->Solid) (pDstPix, pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, pbox->x2 + dst_off_x, pbox->y2 + dst_off_y); - exaPixmapDirty (pDstPix, pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, - pbox->x2 + dst_off_x, pbox->y2 + dst_off_y); pbox++; } + (*pExaScr->info->DoneSolid) (pDstPix); exaMarkSync(pDst->pDrawable->pScreen); @@ -446,8 +446,6 @@ exaTryDriverComposite(CARD8 op, pbox->y1 + dst_off_y, pbox->x2 - pbox->x1, pbox->y2 - pbox->y1); - exaPixmapDirty (pDstPix, pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, - pbox->x2 + dst_off_x, pbox->y2 + dst_off_y); pbox++; } (*pExaScr->info->DoneComposite) (pDstPix); @@ -521,6 +519,9 @@ exaTryMagicTwoPassCompositeHelper(CARD8 op, CARD16 height) { ExaScreenPriv (pDst->pDrawable->pScreen); + DrawablePtr pDstDraw = pDst->pDrawable; + PixmapPtr pDstPixmap = exaGetDrawablePixmap(pDstDraw); + int xoff, yoff; assert(op == PictOpOver); @@ -539,6 +540,12 @@ exaTryMagicTwoPassCompositeHelper(CARD8 op, exaComposite(PictOpOutReverse, pSrc, pMask, pDst, xSrc, ySrc, xMask, yMask, xDst, yDst, width, height); + exaGetDrawableDeltas(pDstDraw, pDstPixmap, &xoff, &yoff); + xoff += pDstDraw->x; + yoff += pDstDraw->y; + exaPixmapDirty(pDstPixmap, xDst + xoff, yDst + yoff, xDst + xoff + width, + yDst + yoff + height); + /* Then, add in the source value times the destination alpha factors (1.0). */ exaComposite(PictOpAdd, pSrc, pMask, pDst, xSrc, ySrc, xMask, yMask, @@ -565,6 +572,28 @@ exaComposite(CARD8 op, int ret = -1; Bool saveSrcRepeat = pSrc->repeat; Bool saveMaskRepeat = pMask ? pMask->repeat : 0; + ExaMigrationRec pixmaps[3]; + int npixmaps = 1; + PixmapPtr pSrcPixmap = NULL; + + pixmaps[0].as_dst = TRUE; + pixmaps[0].as_src = exaOpReadsDestination(op); + pixmaps[0].pPix = exaGetDrawablePixmap (pDst->pDrawable); + + if (pSrc->pDrawable) { + pSrcPixmap = exaGetDrawablePixmap (pSrc->pDrawable); + pixmaps[npixmaps].as_dst = FALSE; + pixmaps[npixmaps].as_src = TRUE; + pixmaps[npixmaps].pPix = pSrcPixmap; + npixmaps++; + } + + if (pMask && pMask->pDrawable) { + pixmaps[npixmaps].as_dst = FALSE; + pixmaps[npixmaps].as_src = TRUE; + pixmaps[npixmaps].pPix = exaGetDrawablePixmap (pMask->pDrawable); + npixmaps++; + } /* We currently don't support acceleration of gradients, or other pictures * with a NULL pDrawable. @@ -583,19 +612,24 @@ exaComposite(CARD8 op, if (!pMask) { - if (op == PictOpSrc) + if ((op == PictOpSrc && + ((pSrc->format == pDst->format) || + (pSrc->format==PICT_a8r8g8b8 && pDst->format==PICT_x8r8g8b8) || + (pSrc->format==PICT_a8b8g8r8 && pDst->format==PICT_x8b8g8r8))) || + (op == PictOpOver && !pSrc->alphaMap && !pDst->alphaMap && + pSrc->format == pDst->format && + (pSrc->format==PICT_x8r8g8b8 || pSrc->format==PICT_x8b8g8r8))) { if (pSrc->pDrawable->width == 1 && - pSrc->pDrawable->height == 1 && pSrc->repeat && - pSrc->repeatType == RepeatNormal) + pSrc->pDrawable->height == 1 && + pSrc->repeat) { ret = exaTryDriverSolidFill(pSrc, pDst, xSrc, ySrc, xDst, yDst, width, height); if (ret == 1) goto done; } - else if (!pSrc->repeat && !pSrc->transform && - pSrc->format == pDst->format) + else if (pSrcPixmap && !pSrc->repeat && !pSrc->transform) { RegionRec region; @@ -660,39 +694,14 @@ exaComposite(CARD8 op, } } - if (ret != 0) { - ExaMigrationRec pixmaps[3]; - /* failure to accelerate was not due to pixmaps being in the wrong - * locations. - */ - pixmaps[0].as_dst = TRUE; - pixmaps[0].as_src = exaOpReadsDestination(op); - pixmaps[0].pPix = exaGetDrawablePixmap (pDst->pDrawable); - pixmaps[1].as_dst = FALSE; - pixmaps[1].as_src = TRUE; - pixmaps[1].pPix = exaGetDrawablePixmap (pSrc->pDrawable); - if (pMask) { - pixmaps[2].as_dst = FALSE; - pixmaps[2].as_src = TRUE; - pixmaps[2].pPix = exaGetDrawablePixmap (pMask->pDrawable); - exaDoMigration(pixmaps, 3, FALSE); - } else { - exaDoMigration(pixmaps, 2, FALSE); - } - } - fallback: #if DEBUG_TRACE_FALL exaPrintCompositeFallback (op, pSrc, pMask, pDst); #endif + exaDoMigration(pixmaps, npixmaps, FALSE); ExaCheckComposite (op, pSrc, pMask, pDst, xSrc, ySrc, xMask, yMask, xDst, yDst, width, height); - exaDrawableDirty(pDst->pDrawable, - pDst->pDrawable->x + xDst, - pDst->pDrawable->y + yDst, - pDst->pDrawable->x + xDst + width, - pDst->pDrawable->y + yDst + height); done: pSrc->repeat = saveSrcRepeat; From a8d6ebdf9338dc2f6ff9a532e6fec460a70d3b1e Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:47:53 +0200 Subject: [PATCH 39/47] EXA: Defer to FillRegionTiled in Composite when possible. Committed separately as this case is hard to hit and has only been tested lightly. --- exa/exa_render.c | 39 +++++++++++++++++++++++++++++++++++++++ 1 file changed, 39 insertions(+) diff --git a/exa/exa_render.c b/exa/exa_render.c index 63a412c5f..06881e39f 100644 --- a/exa/exa_render.c +++ b/exa/exa_render.c @@ -651,6 +651,45 @@ exaComposite(CARD8 op, REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); goto done; } + else if (pSrcPixmap && !pSrc->transform && + pSrc->repeatType == RepeatNormal) + { + RegionRec region; + DDXPointRec srcOrg; + + /* Let's see if the driver can do the repeat in one go */ + if (pExaScr->info->PrepareComposite && !pSrc->alphaMap && + !pDst->alphaMap) + { + ret = exaTryDriverComposite(op, pSrc, pMask, pDst, xSrc, + ySrc, xMask, yMask, xDst, yDst, + width, height); + if (ret == 1) + goto done; + } + + /* Now see if we can use exaFillRegionTiled() */ + xDst += pDst->pDrawable->x; + yDst += pDst->pDrawable->y; + xSrc += pSrc->pDrawable->x; + ySrc += pSrc->pDrawable->y; + + if (!miComputeCompositeRegion (®ion, pSrc, pMask, pDst, xSrc, + ySrc, xMask, yMask, xDst, yDst, + width, height)) + goto done; + + srcOrg.x = (xSrc - xDst) % pSrcPixmap->drawable.width; + srcOrg.y = (ySrc - yDst) % pSrcPixmap->drawable.height; + + ret = exaFillRegionTiled(pDst->pDrawable, ®ion, pSrcPixmap, + &srcOrg, FB_ALLONES, GXcopy); + + REGION_UNINIT(pDst->pDrawable->pScreen, ®ion); + + if (ret) + goto done; + } } } From 7fca16901187ade48e83e6a2684ef464b1912357 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:48:11 +0200 Subject: [PATCH 40/47] EXA: ImageGlyphBlt improvements. * Don't waste effort on invisible glyphs. * Only track damage for bounding box instead of each glyph separately. * Always migrate for fallbacks. --- exa/exa_accel.c | 129 +++++++++++++++++++++--------------------------- 1 file changed, 57 insertions(+), 72 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index 721882b4c..fc1f3d7e6 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -915,12 +915,36 @@ exaImageGlyphBlt (DrawablePtr pDrawable, int dstBpp; int dstXoff, dstYoff; FbBits depthMask; + PixmapPtr pPixmap = exaGetDrawablePixmap(pDrawable); + ExaMigrationRec pixmaps[1]; + int xBack, widthBack, yBack, heightBack; + + for (ppci = ppciInit, n = nglyph, widthBack = 0; n; n--) + widthBack += (*ppci++)->metrics.characterWidth; + + xBack = x; + if (widthBack < 0) + { + xBack += widthBack; + widthBack = -widthBack; + } + yBack = y - FONTASCENT(pGC->font); + heightBack = FONTASCENT(pGC->font) + FONTDESCENT(pGC->font); + + if (xBack >= pDrawable->width || yBack >= pDrawable->height || + (xBack + widthBack) <= 0 || (yBack + heightBack) <= 0) + return; + + pixmaps[0].as_dst = TRUE; + pixmaps[0].as_src = TRUE; + pixmaps[0].pPix = pPixmap; depthMask = FbFullMask(pDrawable->depth); if ((pGC->planemask & depthMask) != depthMask) { + exaDoMigration(pixmaps, 1, FALSE); ExaCheckImageGlyphBlt(pDrawable, pGC, x, y, nglyph, ppciInit, pglyphBase); - return; + goto damage; } glyph = NULL; switch (pDrawable->bitsPerPixel) { @@ -932,6 +956,8 @@ exaImageGlyphBlt (DrawablePtr pDrawable, x += pDrawable->x; y += pDrawable->y; + xBack += pDrawable->x; + yBack += pDrawable->y; if (TERMINALFONT (pGC->font) && !glyph) { @@ -939,23 +965,6 @@ exaImageGlyphBlt (DrawablePtr pDrawable, } else { - int xBack, widthBack; - int yBack, heightBack; - - ppci = ppciInit; - n = nglyph; - widthBack = 0; - while (n--) - widthBack += (*ppci++)->metrics.characterWidth; - - xBack = x; - if (widthBack < 0) - { - xBack += widthBack; - widthBack = -widthBack; - } - yBack = y - FONTASCENT(pGC->font); - heightBack = FONTASCENT(pGC->font) + FONTDESCENT(pGC->font); exaSolidBoxClipped (pDrawable, fbGetCompositeClip(pGC), pGC->planemask, @@ -968,74 +977,50 @@ exaImageGlyphBlt (DrawablePtr pDrawable, } EXA_FALLBACK(("to %p (%c)\n", pDrawable, exaDrawableLocation(pDrawable))); + exaDoMigration(pixmaps, 1, FALSE); exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); exaPrepareAccessGC (pGC); fbGetDrawable (pDrawable, dst, dstStride, dstBpp, dstXoff, dstYoff); - ppci = ppciInit; - while (nglyph--) + for (ppci = ppciInit; nglyph; nglyph--, x += pci->metrics.characterWidth) { pci = *ppci++; - pglyph = FONTGLYPHBITS(pglyphBase, pci); gWidth = GLYPHWIDTHPIXELS(pci); gHeight = GLYPHHEIGHTPIXELS(pci); - if (gWidth && gHeight) + gx = x + pci->metrics.leftSideBearing; + gy = y - pci->metrics.ascent; + + if (!gWidth || !gHeight || (gx + gWidth) <= xBack || + (gy + gHeight) <= yBack || gx >= (xBack + widthBack) || + gy >= (yBack + heightBack)) + continue; + + pglyph = FONTGLYPHBITS(pglyphBase, pci); + + if (glyph && gWidth <= sizeof (FbStip) * 8 && + fbGlyphIn (fbGetCompositeClip(pGC), gx, gy, gWidth, gHeight)) { - gx = x + pci->metrics.leftSideBearing; - gy = y - pci->metrics.ascent; - if (glyph && gWidth <= sizeof (FbStip) * 8 && - fbGlyphIn (fbGetCompositeClip(pGC), gx, gy, gWidth, gHeight)) - { - (*glyph) (dst + (gy + dstYoff) * dstStride, - dstStride, - dstBpp, - (FbStip *) pglyph, - pPriv->fg, - gx + dstXoff, - gHeight); - exaDrawableDirty (pDrawable, gx, gy, gx + gWidth, gy + gHeight); - } - else - { - RegionPtr pClip = fbGetCompositeClip(pGC); - int nbox; - BoxPtr pbox; - - gStride = GLYPHWIDTHBYTESPADDED(pci) / sizeof (FbStip); - fbPutXYImage (pDrawable, - pClip, - pPriv->fg, - pPriv->bg, - pPriv->pm, - GXcopy, - opaque, - - gx, - gy, - gWidth, gHeight, - - (FbStip *) pglyph, - gStride, - 0); - - for (nbox = REGION_NUM_RECTS(pClip), pbox = REGION_RECTS(pClip); - nbox--; pbox++) { - int x1 = max(gx, pbox->x1), x2 = min(gx + gWidth, pbox->x2); - int y1 = max(gy, pbox->y1), y2 = min(gy + gHeight, pbox->y2); - - if (x1 >= x2 || y1 >= y2) - continue; - - exaDrawableDirty (pDrawable, gx, gy, gx + gWidth, - gy + gHeight); - } - } + (*glyph) (dst + (gy + dstYoff) * dstStride, dstStride, dstBpp, + (FbStip *) pglyph, pPriv->fg, gx + dstXoff, gHeight); + } + else + { + RegionPtr pClip = fbGetCompositeClip(pGC); + + gStride = GLYPHWIDTHBYTESPADDED(pci) / sizeof (FbStip); + fbPutXYImage (pDrawable, pClip, pPriv->fg, pPriv->bg, pPriv->pm, + GXcopy, opaque, gx, gy, gWidth, gHeight, + (FbStip *) pglyph, gStride, 0); } - x += pci->metrics.characterWidth; } exaFinishAccessGC (pGC); exaFinishAccess (pDrawable, EXA_PREPARE_DEST); + +damage: + exaGetDrawableDeltas(pDrawable, pPixmap, &dstXoff, &dstYoff); + exaPixmapDirty(pPixmap, xBack + dstXoff, yBack + dstYoff, + xBack + dstXoff + widthBack, yBack + dstYoff + heightBack); } const GCOps exaOps = { From 0880aaac9c83019fec2e3d32871f74c7a407f8b3 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:48:19 +0200 Subject: [PATCH 41/47] EXA: PutImage improvements. * Migrate for fallbacks when appropriate. * Add damage tracking in ExaCheckPutImage. --- exa/exa_accel.c | 29 ++++++++++++++++------------- exa/exa_unaccel.c | 5 +++++ 2 files changed, 21 insertions(+), 13 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index fc1f3d7e6..dd02fef63 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -154,8 +154,9 @@ exaPutImage (DrawablePtr pDrawable, GCPtr pGC, int depth, int x, int y, int xoff, yoff; int src_stride, bpp = pDrawable->bitsPerPixel; - if (pExaScr->swappedOut || pExaScr->info->UploadToScreen == NULL) - goto migrate_and_fallback; + pixmaps[0].as_dst = TRUE; + pixmaps[0].as_src = FALSE; + pixmaps[0].pPix = exaGetDrawablePixmap (pDrawable); /* Don't bother with under 8bpp, XYPixmaps. */ if (format != ZPixmap || bpp < 8) @@ -165,10 +166,14 @@ exaPutImage (DrawablePtr pDrawable, GCPtr pGC, int depth, int x, int y, if (!EXA_PM_IS_SOLID(pDrawable, pGC->planemask) || pGC->alu != GXcopy) goto migrate_and_fallback; - pixmaps[0].as_dst = TRUE; - pixmaps[0].as_src = FALSE; - pixmaps[0].pPix = exaGetDrawablePixmap (pDrawable); + if (pExaScr->swappedOut) + goto fallback; + exaDoMigration (pixmaps, 1, TRUE); + + if (pExaScr->info->UploadToScreen == NULL) + goto fallback; + pPix = exaGetOffscreenPixmap (pDrawable, &xoff, &yoff); if (pPix == NULL) @@ -221,25 +226,23 @@ exaPutImage (DrawablePtr pDrawable, GCPtr pGC, int depth, int x, int y, fbBltStip((FbStip *)bits + (y1 - y) * (src_stride / sizeof(FbStip)), src_stride / sizeof(FbStip), - (x1 - x) * bpp, - dst + (y1 + yoff) * dst_stride, + (x1 - x) * dstBpp, + dst + (y1 + dstYoff) * dst_stride, dst_stride, - (x1 + xoff) * bpp, - (x2 - x1) * bpp, + (x1 + dstXoff) * dstBpp, + (x2 - x1) * dstBpp, y2 - y1, - GXcopy, FB_ALLONES, bpp); + GXcopy, FB_ALLONES, dstBpp); exaFinishAccess(pDrawable, EXA_PREPARE_DEST); } + exaPixmapDirty(pPix, x1 + xoff, y1 + yoff, x2 + xoff, y2 + yoff); } return; migrate_and_fallback: - pixmaps[0].as_dst = TRUE; - pixmaps[0].as_src = FALSE; - pixmaps[0].pPix = exaGetDrawablePixmap (pDrawable); exaDoMigration (pixmaps, 1, FALSE); fallback: diff --git a/exa/exa_unaccel.c b/exa/exa_unaccel.c index a7f939845..708d1eac6 100644 --- a/exa/exa_unaccel.c +++ b/exa/exa_unaccel.c @@ -88,10 +88,15 @@ ExaCheckPutImage (DrawablePtr pDrawable, GCPtr pGC, int depth, int x, int y, int w, int h, int leftPad, int format, char *bits) { + PixmapPtr pPixmap = exaGetDrawablePixmap(pDrawable); + int xoff, yoff; + EXA_FALLBACK(("to %p (%c)\n", pDrawable, exaDrawableLocation(pDrawable))); exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); fbPutImage (pDrawable, pGC, depth, x, y, w, h, leftPad, format, bits); exaFinishAccess (pDrawable, EXA_PREPARE_DEST); + exaGetDrawableDeltas(pDrawable, pPixmap, &xoff, &yoff); + exaPixmapDirty(pPixmap, x + xoff, y + yoff, x + xoff + w, y + yoff + h); } RegionPtr From d2245386eed200e77a8c84bdda36ab29e39fd593 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:48:31 +0200 Subject: [PATCH 42/47] EXA: GetImage improvements. Only migrate when appropriate. In particular, don't migrate to offscreen in the no-fallback case as copying from system memory should usually be as fast if not faster than DownloadFromScreen, in particular if the bits need to be uploaded to offscreen first. --- exa/exa_accel.c | 21 ++++++++------------- 1 file changed, 8 insertions(+), 13 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index dd02fef63..6a0fbb70f 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -1330,27 +1330,22 @@ exaGetImage (DrawablePtr pDrawable, int x, int y, int w, int h, int xoff, yoff; Bool ok; - if (pExaScr->swappedOut || pExaScr->info->DownloadFromScreen == NULL) - goto fallback; + if (pExaScr->info->DownloadFromScreen == NULL) + goto migrate_and_fallback; /* Only cover the ZPixmap, solid copy case. */ if (format != ZPixmap || !EXA_PM_IS_SOLID(pDrawable, planeMask)) - goto fallback; + goto migrate_and_fallback; /* Only try to handle the 8bpp and up cases, since we don't want to think * about <8bpp. */ if (pDrawable->bitsPerPixel < 8) + goto migrate_and_fallback; + + if (pExaScr->swappedOut) goto fallback; - /* Migrate, but assume that we could accelerate the download. It is up to - * the migration scheme to ensure that this case doesn't result in bad - * moving of pixmaps. - */ - pixmaps[0].as_dst = FALSE; - pixmaps[0].as_src = TRUE; - pixmaps[0].pPix = exaGetDrawablePixmap (pDrawable); - exaDoMigration (pixmaps, 1, TRUE); pPix = exaGetOffscreenPixmap (pDrawable, &xoff, &yoff); if (pPix == NULL) goto fallback; @@ -1365,12 +1360,12 @@ exaGetImage (DrawablePtr pDrawable, int x, int y, int w, int h, return; } -fallback: +migrate_and_fallback: pixmaps[0].as_dst = FALSE; pixmaps[0].as_src = TRUE; pixmaps[0].pPix = exaGetDrawablePixmap (pDrawable); exaDoMigration (pixmaps, 1, FALSE); - +fallback: ExaCheckGetImage (pDrawable, x, y, w, h, format, planeMask, d); } From 982d7c2c0b948ba04c8eefa475d660981e6ed9f9 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:48:59 +0200 Subject: [PATCH 43/47] EXA: CopyNtoN improvements. * Centralize handling of fallbacks and damage tracking. * Always migrate for fallbacks. --- exa/exa_accel.c | 83 +++++++++++++++++++++++++------------------------ 1 file changed, 43 insertions(+), 40 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index 6a0fbb70f..800c4f0fb 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -390,6 +390,7 @@ exaCopyNtoN (DrawablePtr pSrcDrawable, int src_off_x, src_off_y; int dst_off_x, dst_off_y; ExaMigrationRec pixmaps[2]; + Bool fallback = FALSE; pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = FALSE; @@ -407,62 +408,64 @@ exaCopyNtoN (DrawablePtr pSrcDrawable, pDstPixmap->drawable.width > pExaScr->info->maxX || pDstPixmap->drawable.height > pExaScr->info->maxY) { - exaDoMigration (pixmaps, 2, FALSE); - goto fallback; + fallback = TRUE; } else { exaDoMigration (pixmaps, 2, TRUE); } /* Mixed directions must be handled specially if the card is lame */ - if (pExaScr->info->flags & EXA_TWO_BITBLT_DIRECTIONS && + if (!fallback && (pExaScr->info->flags & EXA_TWO_BITBLT_DIRECTIONS) && reverse != upsidedown) { - if (!exaCopyNtoNTwoDir(pSrcDrawable, pDstDrawable, pGC, pbox, nbox, + if (exaCopyNtoNTwoDir(pSrcDrawable, pDstDrawable, pGC, pbox, nbox, dx, dy)) - goto fallback; - return; + return; + fallback = TRUE; } - if ((pSrcPixmap = exaGetOffscreenPixmap (pSrcDrawable, &src_off_x, &src_off_y)) && - (pDstPixmap = exaGetOffscreenPixmap (pDstDrawable, &dst_off_x, &dst_off_y)) && - (*pExaScr->info->PrepareCopy) (pSrcPixmap, pDstPixmap, - reverse ? -1 : 1, upsidedown ? -1 : 1, - pGC ? pGC->alu : GXcopy, - pGC ? pGC->planemask : FB_ALLONES)) + pSrcPixmap = exaGetDrawablePixmap (pSrcDrawable); + pDstPixmap = exaGetDrawablePixmap (pDstDrawable); + + exaGetDrawableDeltas (pSrcDrawable, pSrcPixmap, &src_off_x, &src_off_y); + exaGetDrawableDeltas (pDstDrawable, pDstPixmap, &dst_off_x, &dst_off_y); + + if (fallback || !exaPixmapIsOffscreen(pSrcPixmap) || + !exaPixmapIsOffscreen(pDstPixmap) || + !(*pExaScr->info->PrepareCopy) (pSrcPixmap, pDstPixmap, reverse ? -1 : 1, + upsidedown ? -1 : 1, + pGC ? pGC->alu : GXcopy, + pGC ? pGC->planemask : FB_ALLONES)) { + fallback = TRUE; + EXA_FALLBACK(("from %p to %p (%c,%c)\n", pSrcDrawable, pDstDrawable, + exaDrawableLocation(pSrcDrawable), + exaDrawableLocation(pDstDrawable))); + exaDoMigration (pixmaps, 2, FALSE); + exaPrepareAccess (pDstDrawable, EXA_PREPARE_DEST); + exaPrepareAccess (pSrcDrawable, EXA_PREPARE_SRC); + fbCopyNtoN (pSrcDrawable, pDstDrawable, pGC, + pbox, nbox, dx, dy, reverse, upsidedown, + bitplane, closure); + exaFinishAccess (pSrcDrawable, EXA_PREPARE_SRC); + exaFinishAccess (pDstDrawable, EXA_PREPARE_DEST); + } + + while (nbox--) { - while (nbox--) - { + if (!fallback) (*pExaScr->info->Copy) (pDstPixmap, pbox->x1 + dx + src_off_x, pbox->y1 + dy + src_off_y, pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, - pbox->x2 - pbox->x1, - pbox->y2 - pbox->y1); - exaPixmapDirty (pDstPixmap, - pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, - pbox->x2 + dst_off_x, pbox->y2 + dst_off_y); - pbox++; - } - (*pExaScr->info->DoneCopy) (pDstPixmap); - exaMarkSync(pDstDrawable->pScreen); - return; - } - -fallback: - EXA_FALLBACK(("from %p to %p (%c,%c)\n", pSrcDrawable, pDstDrawable, - exaDrawableLocation(pSrcDrawable), - exaDrawableLocation(pDstDrawable))); - exaPrepareAccess (pDstDrawable, EXA_PREPARE_DEST); - exaPrepareAccess (pSrcDrawable, EXA_PREPARE_SRC); - fbCopyNtoN (pSrcDrawable, pDstDrawable, pGC, - pbox, nbox, dx, dy, reverse, upsidedown, - bitplane, closure); - exaFinishAccess (pSrcDrawable, EXA_PREPARE_SRC); - exaFinishAccess (pDstDrawable, EXA_PREPARE_DEST); - while (nbox--) - { - exaDrawableDirty (pDstDrawable, pbox->x1, pbox->y1, pbox->x2, pbox->y2); + pbox->x2 - pbox->x1, pbox->y2 - pbox->y1); + exaPixmapDirty (pDstPixmap, pbox->x1 + dst_off_x, pbox->y1 + dst_off_y, + pbox->x2 + dst_off_x, pbox->y2 + dst_off_y); pbox++; } + + if (fallback) + return; + + (*pExaScr->info->DoneCopy) (pDstPixmap); + exaMarkSync (pDstDrawable->pScreen); } RegionPtr From 584697a2231ac782f362a925e1489c15483a8791 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:49:09 +0200 Subject: [PATCH 44/47] EXA: SolidBoxClipped improvements. * Centralize handling of fallbacks and damage tracking. * Always migrate for fallbacks. --- exa/exa_accel.c | 17 ++++++++--------- 1 file changed, 8 insertions(+), 9 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index 800c4f0fb..4aa2d63d8 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -826,20 +826,19 @@ exaSolidBoxClipped (DrawablePtr pDrawable, pPixmap->drawable.width > pExaScr->info->maxX || pPixmap->drawable.height > pExaScr->info->maxY) { - exaDoMigration (pixmaps, 1, FALSE); - goto fallback; + fallback = TRUE; } else { exaDoMigration (pixmaps, 1, TRUE); } - pPixmap = exaGetOffscreenPixmap (pDrawable, &xoff, &yoff); + exaGetDrawableDeltas (pDrawable, pPixmap, &xoff, &yoff); - if (!pPixmap || + if (fallback || !exaPixmapIsOffscreen(pPixmap) || !(*pExaScr->info->PrepareSolid) (pPixmap, GXcopy, pm, fg)) { -fallback: EXA_FALLBACK(("to %p (%c)\n", pDrawable, exaDrawableLocation(pDrawable))); + exaDoMigration (pixmaps, 1, FALSE); fallback = TRUE; exaPrepareAccess (pDrawable, EXA_PREPARE_DEST); fg = fbReplicatePixel (fg, pDrawable->bitsPerPixel); @@ -878,10 +877,10 @@ fallback: (*pExaScr->info->Solid) (pPixmap, partX1 + xoff, partY1 + yoff, partX2 + xoff, partY2 + yoff); - exaPixmapDirty (pPixmap, partX1 + xoff, partY1 + yoff, - partX2 + xoff, partY2 + yoff); - } else - exaDrawableDirty (pDrawable, partX1, partY1, partX2, partY2); + } + + exaPixmapDirty (pPixmap, partX1 + xoff, partY1 + yoff, partX2 + xoff, + partY2 + yoff); } if (fallback) From b1b6674a919943a8ac37e54d02e8d0d23a642b1d Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:49:28 +0200 Subject: [PATCH 45/47] EXA: FillSpans improvements. * Don't need to track damage. * Always migrate for fallbacks. --- exa/exa_accel.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-) diff --git a/exa/exa_accel.c b/exa/exa_accel.c index 4aa2d63d8..bf63f2c4f 100644 --- a/exa/exa_accel.c +++ b/exa/exa_accel.c @@ -74,6 +74,7 @@ exaFillSpans(DrawablePtr pDrawable, GCPtr pGC, int n, pGC->planemask, pGC->fgPixel)) { + exaDoMigration (pixmaps, 1, FALSE); ExaCheckFillSpans (pDrawable, pGC, n, ppt, pwidth, fSorted); return; } @@ -109,8 +110,6 @@ exaFillSpans(DrawablePtr pDrawable, GCPtr pGC, int n, (*pExaScr->info->Solid) (pPixmap, fullX1 + off_x, fullY1 + off_y, fullX2 + off_x, fullY1 + 1 + off_y); - exaPixmapDirty (pPixmap, fullX1 + off_x, fullY1 + off_y, - fullX2 + off_x, fullY1 + 1 + off_y); } else { @@ -129,8 +128,6 @@ exaFillSpans(DrawablePtr pDrawable, GCPtr pGC, int n, (*pExaScr->info->Solid) (pPixmap, partX1 + off_x, fullY1 + off_y, partX2 + off_x, fullY1 + 1 + off_y); - exaPixmapDirty (pPixmap, partX1 + off_x, fullY1 + off_y, - partX2 + off_x, fullY1 + 1 + off_y); } } pbox++; From a261e1325057974d58440812b93c00c0caa4423a Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:49:35 +0200 Subject: [PATCH 46/47] EXA: Remove DrawableDirty. Convert the remaining callers to PixmapDirty. --- exa/exa.c | 23 ----------------------- exa/exa_priv.h | 3 --- exa/exa_render.c | 14 ++++++++++---- 3 files changed, 10 insertions(+), 30 deletions(-) diff --git a/exa/exa.c b/exa/exa.c index 3d77800ff..c5b5a666a 100644 --- a/exa/exa.c +++ b/exa/exa.c @@ -172,29 +172,6 @@ exaPixmapDirty (PixmapPtr pPix, int x1, int y1, int x2, int y2) REGION_UNINIT(pScreen, ®ion); } -/** - * exaDrawableDirty() marks a pixmap backing a drawable as dirty, allowing for - * optimizations in pixmap migration when no changes have occurred. - */ -void -exaDrawableDirty (DrawablePtr pDrawable, int x1, int y1, int x2, int y2) -{ - PixmapPtr pPix = exaGetDrawablePixmap(pDrawable); - int xoff, yoff; - - x1 = max(x1, pDrawable->x); - y1 = max(y1, pDrawable->y); - x2 = min(x2, pDrawable->x + pDrawable->width); - y2 = min(y2, pDrawable->y + pDrawable->height); - - if (x1 >= x2 || y1 >= y2) - return; - - exaGetDrawableDeltas(pDrawable, pPix, &xoff, &yoff); - - exaPixmapDirty(pPix, x1 + xoff, y1 + yoff, x2 + xoff, y2 + yoff); -} - static Bool exaDestroyPixmap (PixmapPtr pPixmap) { diff --git a/exa/exa_priv.h b/exa/exa_priv.h index b97608214..ece589851 100644 --- a/exa/exa_priv.h +++ b/exa/exa_priv.h @@ -351,9 +351,6 @@ void exaGetDrawableDeltas (DrawablePtr pDrawable, PixmapPtr pPixmap, int *xp, int *yp); -void -exaDrawableDirty(DrawablePtr pDrawable, int x1, int y1, int x2, int y2); - Bool exaDrawableIsOffscreen (DrawablePtr pDrawable); diff --git a/exa/exa_render.c b/exa/exa_render.c index 06881e39f..5e7c67feb 100644 --- a/exa/exa_render.c +++ b/exa/exa_render.c @@ -764,6 +764,7 @@ exaRasterizeTrapezoid (PicturePtr pPicture, xTrapezoid *trap, { DrawablePtr pDraw = pPicture->pDrawable; ExaMigrationRec pixmaps[1]; + int xoff, yoff; pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = TRUE; @@ -772,8 +773,10 @@ exaRasterizeTrapezoid (PicturePtr pPicture, xTrapezoid *trap, exaPrepareAccess(pDraw, EXA_PREPARE_DEST); fbRasterizeTrapezoid(pPicture, trap, x_off, y_off); - exaDrawableDirty(pDraw, pDraw->x, pDraw->y, - pDraw->x + pDraw->width, pDraw->y + pDraw->height); + exaGetDrawableDeltas(pDraw, pixmaps[0].pPix, &xoff, &yoff); + exaPixmapDirty(pixmaps[0].pPix, pDraw->x + xoff, pDraw->y + yoff, + pDraw->x + xoff + pDraw->width, + pDraw->y + yoff + pDraw->height); exaFinishAccess(pDraw, EXA_PREPARE_DEST); } @@ -787,6 +790,7 @@ exaAddTriangles (PicturePtr pPicture, INT16 x_off, INT16 y_off, int ntri, { DrawablePtr pDraw = pPicture->pDrawable; ExaMigrationRec pixmaps[1]; + int xoff, yoff; pixmaps[0].as_dst = TRUE; pixmaps[0].as_src = TRUE; @@ -795,8 +799,10 @@ exaAddTriangles (PicturePtr pPicture, INT16 x_off, INT16 y_off, int ntri, exaPrepareAccess(pDraw, EXA_PREPARE_DEST); fbAddTriangles(pPicture, x_off, y_off, ntri, tris); - exaDrawableDirty(pDraw, pDraw->x, pDraw->y, - pDraw->x + pDraw->width, pDraw->y + pDraw->height); + exaGetDrawableDeltas(pDraw, pixmaps[0].pPix, &xoff, &yoff); + exaPixmapDirty(pixmaps[0].pPix, pDraw->x + xoff, pDraw->y + yoff, + pDraw->x + xoff + pDraw->width, + pDraw->y + yoff + pDraw->height); exaFinishAccess(pDraw, EXA_PREPARE_DEST); } From 3c91a993e8c752002adf85c317216e1487c20780 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?Michel=20D=C3=A4nzer?= Date: Sun, 29 Apr 2007 23:49:41 +0200 Subject: [PATCH 47/47] EXA: Fix OffscreenValidate build with DEBUG_OFFSCREEN enabled. --- exa/exa_offscreen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/exa/exa_offscreen.c b/exa/exa_offscreen.c index 38ad58f60..6fe646a88 100644 --- a/exa/exa_offscreen.c +++ b/exa/exa_offscreen.c @@ -54,7 +54,7 @@ ExaOffscreenValidate (ScreenPtr pScreen) assert (area->offset >= area->base_offset && area->offset < (area->base_offset + area->size)); if (prev) - assert (prev->base_offset + prev->area.size == area->base_offset); + assert (prev->base_offset + prev->size == area->base_offset); prev = area; } assert (prev->base_offset + prev->size == pExaScr->info->memorySize);