[Intel-gfx] [PATCH 1/2] Use ALIGN macro instead of open coding it.

mattst88 at gmail.com mattst88 at gmail.com
Sat Aug 21 23:32:24 CEST 2010


From: Matt Turner <mattst88 at gmail.com>

Signed-off-by: Matt Turner <mattst88 at gmail.com>
---
 src/intel_memory.c            |    2 +-
 src/intel_video.c             |   38 ++++++++++++++++----------------------
 src/legacy/i810/i810_accel.c  |    2 +-
 src/legacy/i810/i810_common.h |    2 ++
 src/legacy/i810/i810_driver.c |    3 +--
 src/legacy/i810/i810_video.c  |   18 +++++++++---------
 src/xvmc/i915_xvmc.c          |    4 +++-
 7 files changed, 33 insertions(+), 36 deletions(-)

diff --git a/src/intel_memory.c b/src/intel_memory.c
index 47444eb..b42e6d7 100644
--- a/src/intel_memory.c
+++ b/src/intel_memory.c
@@ -167,7 +167,7 @@ intel_check_display_stride(ScrnInfoPtr scrn, int stride, Bool tiling)
  */
 static inline int intel_pad_drawable_width(int width)
 {
-	return (width + 63) & ~63;
+	return ALIGN(width, 64);
 }
 
 /**
diff --git a/src/intel_video.c b/src/intel_video.c
index e0a5c53..cde596c 100644
--- a/src/intel_video.c
+++ b/src/intel_video.c
@@ -1277,10 +1277,10 @@ intel_clip_video_helper(ScrnInfoPtr scrn,
 
 	*top = y1 >> 16;
 	*left = (x1 >> 16) & ~1;
-	*npixels = ((((x2 + 0xffff) >> 16) + 1) & ~1) - *left;
+	*npixels = ALIGN(((x2 + 0xffff) >> 16), 2) - *left;
 	if (is_planar_fourcc(id)) {
 		*top &= ~1;
-		*nlines = ((((y2 + 0xffff) >> 16) + 1) & ~1) - *top;
+		*nlines = ALIGN(((y2 + 0xffff) >> 16), 2) - *top;
 	} else
 		*nlines = ((y2 + 0xffff) >> 16) - *top;
 
@@ -1385,32 +1385,32 @@ intel_setup_dst_params(ScrnInfoPtr scrn, intel_adaptor_private *adaptor_priv, sh
 		       int id)
 {
 	intel_screen_private *intel = intel_get_screen_private(scrn);
-	int pitchAlignMask;
+	int pitchAlign;
 
 	/* Only needs to be DWORD-aligned for textured on i915, but overlay has
 	 * stricter requirements.
 	 */
 	if (adaptor_priv->textured) {
-		pitchAlignMask = 3;
+		pitchAlign = 4;
 	} else {
 		if (IS_I965G(intel))
 			/* Actually the alignment is 64 bytes, too. But the
 			 * stride must be at least 512 bytes. Take the easy fix
 			 * and align on 512 bytes unconditionally. */
-			pitchAlignMask = 511;
+			pitchAlign = 512;
 		else if (IS_I830(intel) || IS_845G(intel))
 			/* Harsh, errata on these chipsets limit the stride to be
 			 * a multiple of 256 bytes.
 			 */
-			pitchAlignMask = 255;
+			pitchAlign = 256;
 		else
-			pitchAlignMask = 63;
+			pitchAlign = 64;
 	}
 
 #if INTEL_XVMC
 	/* for i915 xvmc, hw requires 1kb aligned surfaces */
 	if ((id == FOURCC_XVMC) && IS_I915(intel))
-		pitchAlignMask = 0x3ff;
+		pitchAlign = 1024;
 #endif
 
 	/* Determine the desired destination pitch (representing the chroma's pitch,
@@ -1418,26 +1418,20 @@ intel_setup_dst_params(ScrnInfoPtr scrn, intel_adaptor_private *adaptor_priv, sh
 	 */
 	if (is_planar_fourcc(id)) {
 		if (adaptor_priv->rotation & (RR_Rotate_90 | RR_Rotate_270)) {
-			*dstPitch =
-			    ((height / 2) + pitchAlignMask) & ~pitchAlignMask;
-			*dstPitch2 =
-			    (height + pitchAlignMask) & ~pitchAlignMask;
+			*dstPitch = ALIGN((height / 2), pitchAlign);
+			*dstPitch2 = ALIGN(height, pitchAlign);
 			*size = *dstPitch * width * 3;
 		} else {
-			*dstPitch =
-			    ((width / 2) + pitchAlignMask) & ~pitchAlignMask;
-			*dstPitch2 =
-			    (width + pitchAlignMask) & ~pitchAlignMask;
+			*dstPitch = ALIGN((width / 2), pitchAlign);
+			*dstPitch2 = ALIGN(width, pitchAlign);
 			*size = *dstPitch * height * 3;
 		}
 	} else {
 		if (adaptor_priv->rotation & (RR_Rotate_90 | RR_Rotate_270)) {
-			*dstPitch =
-			    ((height << 1) + pitchAlignMask) & ~pitchAlignMask;
+			*dstPitch = ALIGN((height << 1), pitchAlign);
 			*size = *dstPitch * width;
 		} else {
-			*dstPitch =
-			    ((width << 1) + pitchAlignMask) & ~pitchAlignMask;
+			*dstPitch = ALIGN((width << 1), pitchAlign);
 			*size = *dstPitch * height;
 		}
 		*dstPitch2 = 0;
@@ -1472,8 +1466,8 @@ intel_copy_video_data(ScrnInfoPtr scrn, intel_adaptor_private *adaptor_priv,
 	int size;
 
 	if (is_planar_fourcc(id)) {
-		srcPitch = (width + 0x3) & ~0x3;
-		srcPitch2 = ((width >> 1) + 0x3) & ~0x3;
+		srcPitch = ALIGN(width, 0x4);
+		srcPitch2 = ALIGN((width >> 1), 0x4);
 	} else {
 		srcPitch = width << 1;
 	}
diff --git a/src/legacy/i810/i810_accel.c b/src/legacy/i810/i810_accel.c
index ae4a654..9aa3e42 100644
--- a/src/legacy/i810/i810_accel.c
+++ b/src/legacy/i810/i810_accel.c
@@ -129,7 +129,7 @@ I810AccelInit(ScreenPtr pScreen)
     */
    if (pI810->Scratch.Size != 0) {
       int i;
-      int width = ((pScrn->displayWidth + 31) & ~31) / 8;
+      int width = ALIGN(pScrn->displayWidth, 32) / 8;
       int nr_buffers = pI810->Scratch.Size / width;
       unsigned char *ptr = pI810->FbBase + pI810->Scratch.Start;
 
diff --git a/src/legacy/i810/i810_common.h b/src/legacy/i810/i810_common.h
index ea28f4a..a526f73 100644
--- a/src/legacy/i810/i810_common.h
+++ b/src/legacy/i810/i810_common.h
@@ -50,6 +50,8 @@
 #define KB(x) ((x) * 1024)
 #define MB(x) ((x) * KB(1024))
 
+#define ALIGN(i,m) (((i) + (m) - 1) & ~((m) - 1))
+
 /* Using usleep() makes things noticably slow. */
 #if 0
 #define DELAY(x) usleep(x)
diff --git a/src/legacy/i810/i810_driver.c b/src/legacy/i810/i810_driver.c
index 3fc9d04..3637e25 100644
--- a/src/legacy/i810/i810_driver.c
+++ b/src/legacy/i810/i810_driver.c
@@ -1545,8 +1545,7 @@ I810AllocateFront(ScrnInfoPtr pScrn)
 
    if (!I810AllocLow(&(pI810->FrontBuffer),
 		     &(pI810->SysMem),
-		     ((pI810->FbMemBox.x2 *
-		       pI810->FbMemBox.y2 * pI810->cpp) + 4095) & ~4095)) {
+		     ALIGN((pI810->FbMemBox.x2 * pI810->FbMemBox.y2 * pI810->cpp), 4096))) {
       xf86DrvMsg(pScrn->scrnIndex,
 		 X_WARNING, "Framebuffer allocation failed\n");
       return FALSE;
diff --git a/src/legacy/i810/i810_video.c b/src/legacy/i810/i810_video.c
index 7e3db8c..68dc471 100644
--- a/src/legacy/i810/i810_video.c
+++ b/src/legacy/i810/i810_video.c
@@ -750,14 +750,14 @@ I810DisplayVideo(
     switch(id) {
     case FOURCC_YV12:
     case FOURCC_I420:
-	swidth = (width + 7) & ~7;
+	swidth = ALIGN(width, 8);
 	overlay->SWID = (swidth << 15) | swidth;
 	overlay->SWIDQW = (swidth << 12) | (swidth >> 3);
 	break;
     case FOURCC_UYVY:
     case FOURCC_YUY2:
     default:
-	swidth = ((width + 3) & ~3) << 1;
+	swidth = ALIGN(width, 4) << 1;
 	overlay->SWID = swidth;
 	overlay->SWIDQW = swidth >> 3;
 	break;
@@ -1013,15 +1013,15 @@ I810PutImage(
     switch(id) {
     case FOURCC_YV12:
     case FOURCC_I420:
-	 srcPitch = (width + 3) & ~3;
-	 dstPitch = ((width >> 1) + 7) & ~7;  /* of chroma */
+	 srcPitch = ALIGN(width, 4);
+	 dstPitch = ALIGN((width >> 1), 8);  /* of chroma */
 	 size =  dstPitch * height * 3;	
 	 break;
     case FOURCC_UYVY:
     case FOURCC_YUY2:
     default:
 	 srcPitch = (width << 1);
-	 dstPitch = (srcPitch + 7) & ~7;
+	 dstPitch = ALIGN(srcPitch, 8);
 	 size = dstPitch * height;
 	 break;
     }  
@@ -1062,13 +1062,13 @@ I810PutImage(
     /* copy data */
     top = y1 >> 16;
     left = (x1 >> 16) & ~1;
-    npixels = ((((x2 + 0xffff) >> 16) + 1) & ~1) - left;
+    npixels = ALIGN(((x2 + 0xffff) >> 16), 2) - left;
 
     switch(id) {
     case FOURCC_YV12:
     case FOURCC_I420:
 	top &= ~1;
-	nlines = ((((y2 + 0xffff) >> 16) + 1) & ~1) - top;
+	nlines = ALIGN(((y2 + 0xffff) >> 16), 2) - top;
 	I810CopyPlanarData(pScrn, buf, srcPitch, dstPitch,  height, top, left, 
                            nlines, npixels, id);
 	break;
@@ -1213,8 +1213,8 @@ I810AllocateSurface(
     if((w > 1024) || (h > 1024))
 	return BadAlloc;
 
-    w = (w + 1) & ~1;
-    pitch = ((w << 1) + 15) & ~15;
+    w = ALIGN(w, 2);
+    pitch = ALIGN((w << 1), 16);
     bpp = pScrn->bitsPerPixel >> 3;
     size = ((pitch * h) + bpp - 1) / bpp;
 
diff --git a/src/xvmc/i915_xvmc.c b/src/xvmc/i915_xvmc.c
index 9ad8d01..21a1149 100644
--- a/src/xvmc/i915_xvmc.c
+++ b/src/xvmc/i915_xvmc.c
@@ -31,7 +31,9 @@
 #include "i915_structs.h"
 #include "i915_program.h"
 
-#define STRIDE(w)               (((w) + 0x3ff) & ~0x3ff)
+#define ALIGN(i,m)		(((i) + (m) - 1) & ~((m) - 1))
+
+#define STRIDE(w)               (ALIGN((w), 1024))
 #define SIZE_Y420(w, h)         (h * STRIDE(w))
 #define SIZE_UV420(w, h)        ((h >> 1) * STRIDE(w >> 1))
 #define SIZE_YUV420(w, h)       (SIZE_Y420(w,h) + SIZE_UV420(w,h) * 2)
-- 
1.7.1




More information about the Intel-gfx mailing list