xserver: Branch 'master' - 2 commits

[Michel Daenzer]

 exa/exa_accel.c |   78 ++++++++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 59 insertions(+), 19 deletions(-)

New commits:
commit a3afa6f2fb80489f7b6a88d12def09281d32ed94
Author: Michel Dänzer <michel at tungstengraphics.com>
Date:   Wed Jul 30 18:30:37 2008 +0200

    EXA: Optimize GXcopy tiled fills.

diff --git a/exa/exa_accel.c b/exa/exa_accel.c
index 8bcc2ce..cf15709 100644
--- a/exa/exa_accel.c
+++ b/exa/exa_accel.c
@@ -1211,6 +1211,7 @@ exaFillRegionTiled (DrawablePtr	pDrawable,
     int nbox = REGION_NUM_RECTS (pRegion);
     BoxPtr pBox = REGION_RECTS (pRegion);
     Bool ret = FALSE;
+    int i;
 
     tileWidth = pTile->drawable.width;
     tileHeight = pTile->drawable.height;
@@ -1252,20 +1253,26 @@ exaFillRegionTiled (DrawablePtr	pDrawable,
 
     if ((*pExaScr->info->PrepareCopy) (pTile, pPixmap, 1, 1, alu, planemask))
     {
-	while (nbox--)
+	for (i = 0; i < nbox; i++)
 	{
-	    int height = pBox->y2 - pBox->y1;
-	    int dstY = pBox->y1;
+	    int height = pBox[i].y2 - pBox[i].y1;
+	    int dstY = pBox[i].y1;
 	    int tileY;
 
+	    if (alu == GXcopy)
+		height = min(height, tileHeight);
+
 	    modulus(dstY - yoff - pDrawable->y - pPatOrg->y, tileHeight, tileY);
 
 	    while (height > 0) {
-		int width = pBox->x2 - pBox->x1;
-		int dstX = pBox->x1;
+		int width = pBox[i].x2 - pBox[i].x1;
+		int dstX = pBox[i].x1;
 		int tileX;
 		int h = tileHeight - tileY;
 
+		if (alu == GXcopy)
+		    width = min(width, tileWidth);
+
 		if (h > height)
 		    h = height;
 		height -= h;
@@ -1287,12 +1294,51 @@ exaFillRegionTiled (DrawablePtr	pDrawable,
 		dstY += h;
 		tileY = 0;
 	    }
-	    pBox++;
 	}
 	(*pExaScr->info->DoneCopy) (pPixmap);
+
 	exaMarkSync(pDrawable->pScreen);
 
-	ret = TRUE;
+	/* With GXcopy, we only need to do the basic algorithm up to the tile
+	 * size; then, we can just keep doubling the destination in each
+	 * direction until it fills the box. This way, the number of copy
+	 * operations is O(log(rx)) + O(log(ry)) instead of O(rx * ry), where
+	 * rx/ry is the ratio between box and tile width/height. This can make
+	 * a big difference if each driver copy incurs a significant constant
+	 * overhead.
+	 */
+	if (alu != GXcopy)
+	    ret = TRUE;
+	else if ((*pExaScr->info->PrepareCopy) (pPixmap, pPixmap, 1, 1, alu,
+						planemask)) {
+	    for (i = 0; i < nbox; i++)
+	    {
+		int width = min(pBox[i].x2 - pBox[i].x1, tileWidth);
+		int height = min(pBox[i].y2 - pBox[i].y1, tileHeight);
+		int dstX = pBox[i].x1 + width;
+		int dstY = pBox[i].y1 + height;
+
+		while (dstX < pBox[i].x2) {
+		    (*pExaScr->info->Copy) (pPixmap, pBox[i].x1, pBox[i].y1,
+					    dstX, pBox[i].y1, width, height);
+		    dstX += width;
+		    width = min(pBox[i].x2 - dstX, width * 2);
+		}
+
+		width = pBox[i].x2 - pBox[i].x1;
+
+		while (dstY < pBox[i].y2) {
+		    (*pExaScr->info->Copy) (pPixmap, pBox[i].x1, pBox[i].y1,
+					    pBox[i].x1, dstY, width, height);
+		    dstY += height;
+		    height = min(pBox[i].y2 - dstY, height * 2);
+		}
+	    }
+
+	    (*pExaScr->info->DoneCopy) (pPixmap);
+
+	    ret = TRUE;
+	}
     }
 
 out:
commit 37087bc10630ee7740df1369b3e56a44fd2ad2b0
Author: Michel Dänzer <michel at tungstengraphics.com>
Date:   Wed Jul 30 18:27:33 2008 +0200

    EXA: Replace open coded CopyArea logic with GC op call.

diff --git a/exa/exa_accel.c b/exa/exa_accel.c
index 1dbb269..8bcc2ce 100644
--- a/exa/exa_accel.c
+++ b/exa/exa_accel.c
@@ -261,22 +261,16 @@ exaDoShmPutImage(DrawablePtr pDrawable, GCPtr pGC, int depth,
 
     if (format == ZPixmap)
     {
-	PixmapPtr pPixmap;
-	ExaPixmapPriv(exaGetDrawablePixmap(pDrawable));
+	PixmapPtr pPixmap =
+	    GetScratchPixmapHeader(pDrawable->pScreen, w, h, depth,
+				   BitsPerPixel(depth), PixmapBytePad(w, depth),
+				   (pointer)data);
 
-	pPixmap = GetScratchPixmapHeader(pDrawable->pScreen, w, h, depth,
-		BitsPerPixel(depth), PixmapBytePad(w, depth), (pointer)data);
 	if (!pPixmap)
 	    return FALSE;
 
-	if (exaGCReadsDestination(pDrawable, pGC->planemask, pGC->fillStyle,
-				  pGC->alu))
-	    exaPrepareAccess (pDrawable, EXA_PREPARE_DEST);
-	else
-	    exaPrepareAccessReg (pDrawable, EXA_PREPARE_DEST,
-				 DamagePendingRegion(pExaPixmap->pDamage));
-	fbCopyArea((DrawablePtr)pPixmap, pDrawable, pGC, sx, sy, sw, sh, dx, dy);
-	exaFinishAccess(pDrawable, EXA_PREPARE_DEST);
+	pGC->ops->CopyArea(&pPixmap->drawable, pDrawable, pGC, sx, sy, sw, sh,
+			   dx, dy);
 
 	FreeScratchPixmapHeader(pPixmap);