13 files changed, 258 insertions, 336 deletions

diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.c b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
index 7dbc646370..c9b88b0ae1 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.c
@@ -30,6 +30,7 @@
 #include "intel_decode.h"
 #include "intel_reg.h"
 #include "intel_bufmgr.h"
+#include "intel_buffers.h"
 
 /* Relocations in kernel space:
  *    - pass dma buffer seperately
@@ -133,6 +134,9 @@ do_flush_locked(struct intel_batchbuffer *batch,
 {
    struct intel_context *intel = batch->intel;
    int ret = 0;
+   unsigned int num_cliprects = 0;
+   struct drm_clip_rect *cliprects = NULL;
+   int x_off = 0, y_off = 0;
 
    if (batch->buffer)
       dri_bo_subdata (batch->buf, 0, used, batch->buffer);
@@ -142,23 +146,21 @@ do_flush_locked(struct intel_batchbuffer *batch,
    batch->map = NULL;
    batch->ptr = NULL;
 
-   /* Throw away non-effective packets.  Won't work once we have
-    * hardware contexts which would preserve statechanges beyond a
-    * single buffer.
-    */
 
-   if (!(intel->numClipRects == 0 &&
-	 batch->cliprect_mode == LOOP_CLIPRECTS) || intel->no_hw) {
-      dri_bo_exec(batch->buf, used,
-		  intel->pClipRects,
-		  batch->cliprect_mode != LOOP_CLIPRECTS ?
-		  0 : intel->numClipRects,
-		  (((GLuint) intel->drawX) & 0xffff) |
-		  (((GLuint) intel->drawY) << 16));
+   if (batch->cliprect_mode == LOOP_CLIPRECTS) {
+      intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
+   }
+   /* Dispatch the batchbuffer, if it has some effect (nonzero cliprects).
+    * Can't short-circuit like this once we have hardware contexts, but we
+    * should always be in DRI2 mode by then anyway.
+    */
+   if ((batch->cliprect_mode != LOOP_CLIPRECTS ||
+	num_cliprects != 0) && !intel->no_hw) {
+      dri_bo_exec(batch->buf, used, cliprects, num_cliprects,
+		  (x_off & 0xffff) | (y_off << 16));
    }
 
-   if (intel->numClipRects == 0 &&
-       batch->cliprect_mode == LOOP_CLIPRECTS) {
+   if (batch->cliprect_mode == LOOP_CLIPRECTS && num_cliprects == 0) {
       if (allow_unlock) {
 	 /* If we are not doing any actual user-visible rendering,
 	  * do a sched_yield to keep the app from pegging the cpu while
diff --git a/src/mesa/drivers/dri/intel/intel_batchbuffer.h b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
index 1f8096b32e..8129996979 100644
--- a/src/mesa/drivers/dri/intel/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/intel/intel_batchbuffer.h
@@ -19,6 +19,9 @@ enum cliprect_mode {
    /**
     * Batchbuffer contents require looping over per cliprect at batch submit
     * time.
+    *
+    * This will be upgraded to NO_LOOP_CLIPRECTS when there's a single
+    * constant cliprect, as in DRI2 or FBO rendering.
     */
    LOOP_CLIPRECTS,
    /**
@@ -29,8 +32,10 @@ enum cliprect_mode {
    /**
     * Batchbuffer contents contain drawing that already handles cliprects, such
     * as 2D drawing to front/back/depth that doesn't respect DRAWING_RECTANGLE.
+    *
     * Equivalent behavior to NO_LOOP_CLIPRECTS, but may not persist in batch
-    * outside of LOCK/UNLOCK.
+    * outside of LOCK/UNLOCK.  This is upgraded to just NO_LOOP_CLIPRECTS when
+    * there's a constant cliprect, as in DRI2 or FBO rendering.
     */
    REFERENCES_CLIPRECTS
 };
@@ -115,6 +120,11 @@ intel_batchbuffer_require_space(struct intel_batchbuffer *batch,
    if (intel_batchbuffer_space(batch) < sz)
       intel_batchbuffer_flush(batch);
 
+   if ((cliprect_mode == LOOP_CLIPRECTS ||
+	cliprect_mode == REFERENCES_CLIPRECTS) &&
+       batch->intel->constant_cliprect)
+      cliprect_mode = NO_LOOP_CLIPRECTS;
+
    if (cliprect_mode != IGNORE_CLIPRECTS) {
       if (batch->cliprect_mode == IGNORE_CLIPRECTS) {
 	 batch->cliprect_mode = cliprect_mode;
diff --git a/src/mesa/drivers/dri/intel/intel_blit.c b/src/mesa/drivers/dri/intel/intel_blit.c
index 2917401e02..e1046f4a5d 100644
--- a/src/mesa/drivers/dri/intel/intel_blit.c
+++ b/src/mesa/drivers/dri/intel/intel_blit.c
@@ -272,24 +272,53 @@ intelEmitCopyBlit(struct intel_context *intel,
 		  GLshort w, GLshort h,
 		  GLenum logic_op)
 {
-   GLuint CMD, BR13;
+   GLuint CMD, BR13, pass = 0;
    int dst_y2 = dst_y + h;
    int dst_x2 = dst_x + w;
    dri_bo *aper_array[3];
    BATCH_LOCALS;
 
    /* do space/cliprects check before going any further */
-   intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
- again:
-   aper_array[0] = intel->batch->buf;
-   aper_array[1] = dst_buffer;
-   aper_array[2] = src_buffer;
-
-   if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
-      intel_batchbuffer_flush(intel->batch);
-      goto again;
+   do {
+       aper_array[0] = intel->batch->buf;
+       aper_array[1] = dst_buffer;
+       aper_array[2] = src_buffer;
+
+       if (dri_bufmgr_check_aperture_space(aper_array, 3) != 0) {
+           intel_batchbuffer_flush(intel->batch);
+           pass++;
+       } else
+           break;
+   } while (pass < 2);
+
+   if (pass >= 2) {
+       GLboolean locked = GL_FALSE;       
+       if (!intel->locked) {
+           LOCK_HARDWARE(intel);
+           locked = GL_TRUE;
+       }
+
+       dri_bo_map(dst_buffer, GL_TRUE);
+       dri_bo_map(src_buffer, GL_FALSE);
+       _mesa_copy_rect((GLubyte *)dst_buffer->virtual + dst_offset,
+                       cpp,
+                       dst_pitch,
+                       dst_x, dst_y, 
+                       w, h, 
+                       (GLubyte *)src_buffer->virtual + src_offset, 
+                       src_pitch,
+                       src_x, src_y);
+       
+       dri_bo_unmap(src_buffer);
+       dri_bo_unmap(dst_buffer);
+       
+       if (locked)
+           UNLOCK_HARDWARE(intel);
+
+       return;
    }
 
+   intel_batchbuffer_require_space(intel->batch, 8 * 4, NO_LOOP_CLIPRECTS);
    DBG("%s src:buf(%p)/%d+%d %d,%d dst:buf(%p)/%d+%d %d,%d sz:%dx%d\n",
        __FUNCTION__,
        src_buffer, src_pitch, src_offset, src_x, src_y,
@@ -393,6 +422,9 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
    struct gl_framebuffer *fb = ctx->DrawBuffer;
    GLuint clear_depth;
    GLbitfield skipBuffers = 0;
+   unsigned int num_cliprects;
+   struct drm_clip_rect *cliprects;
+   int x_off, y_off;
    BATCH_LOCALS;
 
    /*
@@ -417,7 +449,8 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
    intelFlush(&intel->ctx);
    LOCK_HARDWARE(intel);
 
-   if (intel->numClipRects) {
+   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
+   if (num_cliprects) {
       GLint cx, cy, cw, ch;
       drm_clip_rect_t clear;
       int i;
@@ -432,15 +465,15 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
          /* clearing a window */
 
          /* flip top to bottom */
-         clear.x1 = cx + intel->drawX;
+         clear.x1 = cx + x_off;
          clear.y1 = intel->driDrawable->y + intel->driDrawable->h - cy - ch;
          clear.x2 = clear.x1 + cw;
          clear.y2 = clear.y1 + ch;
       }
       else {
          /* clearing FBO */
-         assert(intel->numClipRects == 1);
-         assert(intel->pClipRects == &intel->fboRect);
+         assert(num_cliprects == 1);
+         assert(cliprects == &intel->fboRect);
          clear.x1 = cx;
          clear.y1 = cy;
          clear.x2 = clear.x1 + cw;
@@ -448,8 +481,8 @@ intelClearWithBlit(GLcontext *ctx, GLbitfield mask)
          /* no change to mask */
       }
 
-      for (i = 0; i < intel->numClipRects; i++) {
-         const drm_clip_rect_t *box = &intel->pClipRects[i];
+      for (i = 0; i < num_cliprects; i++) {
+         const drm_clip_rect_t *box = &cliprects[i];
          drm_clip_rect_t b;
          GLuint buf;
          GLuint clearMask = mask;      /* use copy, since we modify it below */
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index f5eaf765f3..f8f009c6a3 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -123,99 +123,40 @@ intel_readbuf_region(struct intel_context *intel)
       return NULL;
 }
 
-
-
-/**
- * Update the following fields for rendering to a user-created FBO:
- *   intel->numClipRects
- *   intel->pClipRects
- *   intel->drawX
- *   intel->drawY
- */
-static void
-intelSetRenderbufferClipRects(struct intel_context *intel)
-{
-   /* If the batch contents require looping over cliprects, flush them before
-    * we go changing which cliprects get referenced when that happens.
-    */
-   if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
-       (intel->fboRect.x2 != intel->ctx.DrawBuffer->Width ||
-	intel->fboRect.x2 != intel->ctx.DrawBuffer->Height))
-      intel_batchbuffer_flush(intel->batch);
-
-   assert(intel->ctx.DrawBuffer->Width > 0);
-   assert(intel->ctx.DrawBuffer->Height > 0);
-   intel->fboRect.x1 = 0;
-   intel->fboRect.y1 = 0;
-   intel->fboRect.x2 = intel->ctx.DrawBuffer->Width;
-   intel->fboRect.y2 = intel->ctx.DrawBuffer->Height;
-   intel->numClipRects = 1;
-   intel->pClipRects = &intel->fboRect;
-   intel->drawX = 0;
-   intel->drawY = 0;
-}
-
-
-/**
- * As above, but for rendering to front buffer of a window.
- * \sa intelSetRenderbufferClipRects
- */
-static void
-intelSetFrontClipRects(struct intel_context *intel)
-{
-   __DRIdrawablePrivate *dPriv = intel->driDrawable;
-
-   if (!dPriv)
-      return;
-
-   /* If the batch contents require looping over cliprects, flush them before
-    * we go changing which cliprects get referenced when that happens.
-    */
-   if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
-       intel->pClipRects != dPriv->pClipRects)
-      intel_batchbuffer_flush(intel->batch);
-   intel->numClipRects = dPriv->numClipRects;
-   intel->pClipRects = dPriv->pClipRects;
-   intel->drawX = dPriv->x;
-   intel->drawY = dPriv->y;
-}
-
-
-/**
- * As above, but for rendering to back buffer of a window.
- */
-static void
-intelSetBackClipRects(struct intel_context *intel)
+void
+intel_get_cliprects(struct intel_context *intel,
+		    struct drm_clip_rect **cliprects,
+		    unsigned int *num_cliprects,
+		    int *x_off, int *y_off)
 {
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
-   struct intel_framebuffer *intel_fb;
-
-   if (!dPriv)
-      return;
-
-   intel_fb = dPriv->driverPrivate;
+   struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
 
-   if (intel_fb->pf_active || dPriv->numBackClipRects == 0) {
+   if (intel->constant_cliprect) {
+      /* FBO or DRI2 rendering, which can just use the fb's size. */
+      intel->fboRect.x1 = 0;
+      intel->fboRect.y1 = 0;
+      intel->fboRect.x2 = intel->ctx.DrawBuffer->Width;
+      intel->fboRect.y2 = intel->ctx.DrawBuffer->Height;
+
+      *cliprects = &intel->fboRect;
+      *num_cliprects = 1;
+      *x_off = 0;
+      *y_off = 0;
+   } else if (intel->front_cliprects ||
+       intel_fb->pf_active || dPriv->numBackClipRects == 0) {
       /* use the front clip rects */
-      if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
-	  intel->pClipRects != dPriv->pClipRects)
-	 intel_batchbuffer_flush(intel->batch);
-
-      intel->numClipRects = dPriv->numClipRects;
-      intel->pClipRects = dPriv->pClipRects;
-      intel->drawX = dPriv->x;
-      intel->drawY = dPriv->y;
+      *cliprects = dPriv->pClipRects;
+      *num_cliprects = dPriv->numClipRects;
+      *x_off = dPriv->x;
+      *y_off = dPriv->y;
    }
    else {
       /* use the back clip rects */
-      if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
-	  intel->pClipRects != dPriv->pBackClipRects)
-	 intel_batchbuffer_flush(intel->batch);
-
-      intel->numClipRects = dPriv->numBackClipRects;
-      intel->pClipRects = dPriv->pBackClipRects;
-      intel->drawX = dPriv->backX;
-      intel->drawY = dPriv->backY;
+      *num_cliprects = dPriv->numBackClipRects;
+      *cliprects = dPriv->pBackClipRects;
+      *x_off = dPriv->backX;
+      *y_off = dPriv->backY;
    }
 }
 
@@ -300,29 +241,6 @@ intelWindowMoved(struct intel_context *intel)
    __DRIdrawablePrivate *dPriv = intel->driDrawable;
    struct intel_framebuffer *intel_fb = dPriv->driverPrivate;
 
-   if (!intel->ctx.DrawBuffer) {
-      /* when would this happen? -BP */
-      intelSetFrontClipRects(intel);
-   }
-   else if (intel->ctx.DrawBuffer->Name != 0) {
-      /* drawing to user-created FBO - do nothing */
-      /* Cliprects would be set from intelDrawBuffer() */
-   }
-   else {
-      /* drawing to a window */
-      switch (intel_fb->Base._ColorDrawBufferIndexes[0]) {
-      case BUFFER_FRONT_LEFT:
-         intelSetFrontClipRects(intel);
-         break;
-      case BUFFER_BACK_LEFT:
-         intelSetBackClipRects(intel);
-         break;
-      default:
-         intelSetFrontClipRects(intel);
-      }
-	
-   }
-
    if (!intel->intelScreen->driScrnPriv->dri2.enabled &&
        intel->intelScreen->driScrnPriv->ddx_version.minor >= 7) {
       volatile struct drm_i915_sarea *sarea = intel->sarea;
@@ -894,7 +812,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    struct intel_context *intel = intel_context(ctx);
    struct intel_region *colorRegions[MAX_DRAW_BUFFERS], *depthRegion = NULL;
    struct intel_renderbuffer *irbDepth = NULL, *irbStencil = NULL;
-   int front = 0;               /* drawing to front color buffer? */
 
    if (!fb) {
       /* this can happen during the initial context initialization */
@@ -927,52 +844,44 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
     */
    if (fb->_NumColorDrawBuffers == 0) {
       /* writing to 0  */
-      FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_TRUE);
       colorRegions[0] = NULL;
-
-      if (fb->Name != 0)
-	 intelSetRenderbufferClipRects(intel);
+      intel->constant_cliprect = GL_TRUE;
    } else if (fb->_NumColorDrawBuffers > 1) {
        int i;
        struct intel_renderbuffer *irb;
-       FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE);
 
-       if (fb->Name != 0)
-           intelSetRenderbufferClipRects(intel);
        for (i = 0; i < fb->_NumColorDrawBuffers; i++) {
            irb = intel_renderbuffer(fb->_ColorDrawBuffers[i]);
-           colorRegions[i] = (irb && irb->region) ? irb->region : NULL;
+           colorRegions[i] = irb ? irb->region : NULL;
        }
+       intel->constant_cliprect = GL_TRUE;
    }
    else {
-      /* draw to exactly one color buffer */
-      /*_mesa_debug(ctx, "Hardware rendering\n");*/
-      FALLBACK(intel, INTEL_FALLBACK_DRAW_BUFFER, GL_FALSE);
-      if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
-         front = 1;
-      }
-
-      /*
-       * Get the intel_renderbuffer for the colorbuffer we're drawing into.
-       * And set up cliprects.
+      /* Get the intel_renderbuffer for the single colorbuffer we're drawing
+       * into, and set up cliprects if it's .
        */
       if (fb->Name == 0) {
+	 intel->constant_cliprect = intel->driScreen->dri2.enabled;
 	 /* drawing to window system buffer */
-	 if (front) {
-	    intelSetFrontClipRects(intel);
+	 if (fb->_ColorDrawBufferIndexes[0] == BUFFER_FRONT_LEFT) {
+	    if (!intel->constant_cliprect && !intel->front_cliprects)
+	       intel_batchbuffer_flush(intel->batch);
+	    intel->front_cliprects = GL_TRUE;
 	    colorRegions[0] = intel_get_rb_region(fb, BUFFER_FRONT_LEFT);
 	 }
 	 else {
-	    intelSetBackClipRects(intel);
+	    if (!intel->constant_cliprect && intel->front_cliprects)
+	       intel_batchbuffer_flush(intel->batch);
+	    intel->front_cliprects = GL_FALSE;
 	    colorRegions[0]= intel_get_rb_region(fb, BUFFER_BACK_LEFT);
 	 }
       }
       else {
 	 /* drawing to user-created FBO */
 	 struct intel_renderbuffer *irb;
-	 intelSetRenderbufferClipRects(intel);
 	 irb = intel_renderbuffer(fb->_ColorDrawBuffers[0]);
 	 colorRegions[0] = (irb && irb->region) ? irb->region : NULL;
+	 intel->constant_cliprect = GL_TRUE;
       }
    }
 
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.h b/src/mesa/drivers/dri/intel/intel_buffers.h
index a669a85431..e5afb37dd1 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.h
+++ b/src/mesa/drivers/dri/intel/intel_buffers.h
@@ -29,6 +29,8 @@
 #ifndef INTEL_BUFFERS_H
 #define INTEL_BUFFERS_H
 
+#include "dri_util.h"
+#include "drm.h"
 
 struct intel_context;
 struct intel_framebuffer;
@@ -53,4 +55,9 @@ extern void intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb);
 
 extern void intelInitBufferFuncs(struct dd_function_table *functions);
 
+void intel_get_cliprects(struct intel_context *intel,
+			 struct drm_clip_rect **cliprects,
+			 unsigned int *num_cliprects,
+			 int *x_off, int *y_off);
+
 #endif /* INTEL_BUFFERS_H */
diff --git a/src/mesa/drivers/dri/intel/intel_chipset.h b/src/mesa/drivers/dri/intel/intel_chipset.h
index 170efd060a..d1b4941601 100644
--- a/src/mesa/drivers/dri/intel/intel_chipset.h
+++ b/src/mesa/drivers/dri/intel/intel_chipset.h
@@ -68,11 +68,12 @@
 				 devid == PCI_CHIP_I965_GME || \
 				 devid == PCI_CHIP_GM45_GM)
 
-#define IS_GM45_GM(devid)       (devid == PCI_CHIP_GM45_GM)
-#define IS_G4X(devid)           (devid == PCI_CHIP_IGD_E_G || \
+#define IS_G45(devid)           (devid == PCI_CHIP_IGD_E_G || \
                                  devid == PCI_CHIP_Q45_G || \
                                  devid == PCI_CHIP_G45_G || \
                                  devid == PCI_CHIP_G41_G)
+#define IS_GM45(devid)          (devid == PCI_CHIP_GM45_GM)
+#define IS_G4X(devid)		(IS_G45(devid) || IS_GM45(devid))
 
 #define IS_915(devid)		(devid == PCI_CHIP_I915_G || \
 				 devid == PCI_CHIP_E7221_G || \
@@ -91,7 +92,6 @@
 				 devid == PCI_CHIP_I965_GM || \
 				 devid == PCI_CHIP_I965_GME || \
 				 devid == PCI_CHIP_I946_GZ || \
-				 IS_GM45_GM(devid) || \
 				 IS_G4X(devid))
 
 #define IS_9XX(devid)		(IS_915(devid) || \
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 2b3a9b9d37..9ac18e6960 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -588,9 +588,6 @@ intelInitContext(struct intel_context *intel,
    intel->driFd = sPriv->fd;
    intel->driHwLock = sPriv->lock;
 
-   intel->width = intelScreen->width;
-   intel->height = intelScreen->height;
-
    driParseConfigFiles(&intel->optionCache, &intelScreen->optionCache,
                        intel->driScreen->myNum,
 		       IS_965(intelScreen->deviceID) ? "i965" : "i915");
@@ -932,38 +929,6 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
 		 sarea->ctxOwner, intel->hHWContext);
    }
 
-   if (sarea->width != intel->width || sarea->height != intel->height) {
-       int numClipRects = intel->numClipRects;
-
-       /*
-	* FIXME: Really only need to do this when drawing to a
-	* common back- or front buffer.
-	*/
-
-       /*
-	* This will essentially drop the outstanding batchbuffer on
-	* the floor.
-	*/
-       intel->numClipRects = 0;
-
-       if (intel->Fallback)
-	   _swrast_flush(&intel->ctx);
-
-       if (!IS_965(intel->intelScreen->deviceID))
-	   INTEL_FIREVERTICES(intel);
-
-       if (intel->batch->map != intel->batch->ptr)
-	   intel_batchbuffer_flush(intel->batch);
-
-       intel->numClipRects = numClipRects;
-
-       /* force window update */
-       intel->lastStamp = 0;
-
-       intel->width = sarea->width;
-       intel->height = sarea->height;
-   }
-
    /* Drawable changed?
     */
    if (dPriv && intel->lastStamp != dPriv->lastStamp) {
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 554159ac44..3938af4c72 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -235,10 +235,18 @@ struct intel_context
 
    /* These refer to the current drawing buffer:
     */
-   int drawX, drawY;            /**< origin of drawing area within region */
-   GLuint numClipRects;         /**< cliprects for drawing */
-   drm_clip_rect_t *pClipRects;
    struct gl_texture_object *frame_buffer_texobj;
+   /**
+    * Set to true if a single constant cliprect should be used in the
+    * batchbuffer.  Otherwise, cliprects must be calculated at batchbuffer
+    * flush time while the lock is held.
+    */
+   GLboolean constant_cliprect;
+   /**
+    * In !constant_cliprect mode, set to true if the front cliprects should be
+    * used instead of back.
+    */
+   GLboolean front_cliprects;
    drm_clip_rect_t fboRect;     /**< cliprect for FBO rendering */
 
    int perf_boxes;
@@ -271,10 +279,6 @@ struct intel_context
     */
    driOptionCache optionCache;
 
-   /* Last seen width/height of the screen */
-   int width;
-   int height;
-
    int64_t swap_ust;
    int64_t swap_missed_ust;
 
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
index 8963519893..0565197ea0 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_bitmap.c
@@ -191,11 +191,7 @@ do_blit_bitmap( GLcontext *ctx,
    color8888 = INTEL_PACKCOLOR8888(ubcolor[0], ubcolor[1], ubcolor[2], ubcolor[3]);
    color565 = INTEL_PACKCOLOR565(ubcolor[0], ubcolor[1], ubcolor[2]);
 
-   /* Does zoom apply to bitmaps?
-    */
-   if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F) ||
-       ctx->Pixel.ZoomX != 1.0F || 
-       ctx->Pixel.ZoomY != 1.0F)
+   if (!intel_check_blit_fragment_ops(ctx, tmpColor[3] == 1.0F))
       return GL_FALSE;
 
    LOCK_HARDWARE(intel);
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index 50518a6879..8ebbc95a1d 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -75,19 +75,21 @@ intel_texture_drawpixels(GLcontext * ctx,
    /* We're going to mess with texturing with no regard to existing texture
     * state, so if there is some set up we have to bail.
     */
-   if (ctx->Texture._EnabledUnits != 0)
+   if (ctx->Texture._EnabledUnits != 0) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glDrawPixels() fallback: texturing enabled\n");
       return GL_FALSE;
+   }
 
    /* Can't do textured DrawPixels with a fragment program, unless we were
     * to generate a new program that sampled our texture and put the results
     * in the fragment color before the user's program started.
     */
-   if (ctx->FragmentProgram.Enabled)
-      return GL_FALSE;
-
-   /* Don't even want to think about it */
-   if (format == GL_COLOR_INDEX)
+   if (ctx->FragmentProgram.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glDrawPixels() fallback: fragment program enabled\n");
       return GL_FALSE;
+   }
 
    /* We don't have a way to generate fragments with stencil values which *
     * will set the resulting stencil value.
@@ -108,8 +110,12 @@ intel_texture_drawpixels(GLcontext * ctx,
     * the color buffer, and sample the texture values into the fragment depth
     * in a program.
     */
-   if (format == GL_DEPTH_COMPONENT)
+   if (format == GL_DEPTH_COMPONENT) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr,
+		 "glDrawPixels() fallback: format == GL_DEPTH_COMPONENT\n");
       return GL_FALSE;
+   }
 
    _mesa_PushAttrib(GL_ENABLE_BIT | GL_TRANSFORM_BIT | GL_TEXTURE_BIT |
 		    GL_CURRENT_BIT);
@@ -141,22 +147,27 @@ intel_texture_drawpixels(GLcontext * ctx,
    _mesa_PushMatrix();
    _mesa_LoadIdentity();
 
+   /* Create the vertex buffer based on the current raster pos.  The x and y
+    * we're handed are ctx->Current.RasterPos[0,1] rounded to integers.
+    * We also apply the depth.  However, the W component is already multiplied
+    * into ctx->Current.RasterPos[0,1,2] and we can ignore it at this point.
+    */
    vertices[0][0] = x;
    vertices[0][1] = y;
    vertices[0][2] = ctx->Current.RasterPos[2];
-   vertices[0][3] = ctx->Current.RasterPos[3];
+   vertices[0][3] = 1.0;
    vertices[1][0] = x + width * ctx->Pixel.ZoomX;
    vertices[1][1] = y;
    vertices[1][2] = ctx->Current.RasterPos[2];
-   vertices[1][3] = ctx->Current.RasterPos[3];
+   vertices[1][3] = 1.0;
    vertices[2][0] = x + width * ctx->Pixel.ZoomX;
    vertices[2][1] = y + height * ctx->Pixel.ZoomY;
    vertices[2][2] = ctx->Current.RasterPos[2];
-   vertices[2][3] = ctx->Current.RasterPos[3];
+   vertices[2][3] = 1.0;
    vertices[3][0] = x;
    vertices[3][1] = y + height * ctx->Pixel.ZoomY;
    vertices[3][2] = ctx->Current.RasterPos[2];
-   vertices[3][3] = ctx->Current.RasterPos[3];
+   vertices[3][3] = 1.0;
 
    texcoords[0][0] = 0.0;
    texcoords[0][1] = 0.0;
@@ -212,8 +223,12 @@ intel_stencil_drawpixels(GLcontext * ctx,
       return GL_TRUE;
 
    /* Can't do a per-bit writemask while treating stencil as rgba data. */
-   if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff)
+   if ((ctx->Stencil.WriteMask[0] & 0xff) != 0xff) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glDrawPixels(STENCIL_INDEX) fallback: "
+		 "stencil mask enabled\n");
       return GL_FALSE;
+   }
 
    /* We use FBOs for our wrapping of the depthbuffer into a color
     * destination.
@@ -224,21 +239,29 @@ intel_stencil_drawpixels(GLcontext * ctx,
    /* We're going to mess with texturing with no regard to existing texture
     * state, so if there is some set up we have to bail.
     */
-   if (ctx->Texture._EnabledUnits != 0)
+   if (ctx->Texture._EnabledUnits != 0) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glDrawPixels(STENCIL_INDEX) fallback: "
+		 "texturing enabled\n");
       return GL_FALSE;
+   }
 
    /* Can't do textured DrawPixels with a fragment program, unless we were
     * to generate a new program that sampled our texture and put the results
     * in the fragment color before the user's program started.
     */
-   if (ctx->FragmentProgram.Enabled)
+   if (ctx->FragmentProgram.Enabled) {
+      if (INTEL_DEBUG & DEBUG_FALLBACKS)
+	 fprintf(stderr, "glDrawPixels(STENCIL_INDEX) fallback: "
+		 "fragment program enabled\n");
       return GL_FALSE;
+   }
 
    /* Check that we can load in a texture this big. */
    if (width > (1 << (ctx->Const.MaxTextureLevels - 1)) ||
        height > (1 << (ctx->Const.MaxTextureLevels - 1))) {
       if (INTEL_DEBUG & DEBUG_FALLBACKS)
-	 fprintf(stderr, "glDrawPixels(STENCIL_IDNEX) fallback: "
+	 fprintf(stderr, "glDrawPixels(STENCIL_INDEX) fallback: "
 		 "bitmap too large (%dx%d)\n",
 		 width, height);
       return GL_FALSE;
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index c21f408093..81a7386e42 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -29,6 +29,8 @@
 #define CMD_2D				(0x2 << 29)
 #define CMD_3D				(0x3 << 29)
 
+#define MI_NOOP				(CMD_MI | 0)
+
 #define MI_BATCH_BUFFER_END		(CMD_MI | 0xA << 23)
 
 #define MI_FLUSH			(CMD_MI | (4 << 23))
@@ -44,6 +46,9 @@
 #define _3DSTATE_LOAD_STATE_IMMEDIATE_1   (CMD_3D | (0x1d<<24) | (0x04<<16))
 #define I1_LOAD_S(n)                      (1<<(4+n))
 
+#define _3DSTATE_DRAWRECT_INFO		(CMD_3D | (0x1d<<24) | (0x80<<16) | 0x3)
+#define _3DSTATE_DRAWRECT_INFO_I965	(CMD_3D | (3 << 27) | (1 << 24) | 0x2)
+
 /** @{
  *
  * PIPE_CONTROL operation, a combination MI_FLUSH and register write with
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index 45faf64c71..8dbcc3050e 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -79,30 +79,6 @@ intel_region_unmap(struct intel_context *intel, struct intel_region *region)
    }
 }
 
-static int
-intel_set_region_tiling_gem(struct intel_context *intel,
-			    struct intel_region *region,
-			    uint32_t bo_handle)
-{
-   struct drm_i915_gem_get_tiling get_tiling;
-   int ret;
-
-   memset(&get_tiling, 0, sizeof(get_tiling));
-
-   get_tiling.handle = bo_handle;
-   ret = ioctl(intel->driFd, DRM_IOCTL_I915_GEM_GET_TILING, &get_tiling);
-   if (ret != 0) {
-      fprintf(stderr, "Failed to get tiling state for region: %s\n",
-	      strerror(errno));
-      return ret;
-   }
-
-   region->tiling = get_tiling.tiling_mode;
-   region->bit_6_swizzle = get_tiling.swizzle_mode;
-
-   return 0;
-}
-
 static struct intel_region *
 intel_region_alloc_internal(struct intel_context *intel,
 			    GLuint cpp,
@@ -151,6 +127,7 @@ intel_region_alloc_for_handle(struct intel_context *intel,
 {
    struct intel_region *region;
    dri_bo *buffer;
+   int ret;
 
    buffer = intel_bo_gem_create_from_name(intel->bufmgr, name, handle);
 
@@ -159,7 +136,14 @@ intel_region_alloc_for_handle(struct intel_context *intel,
    if (region == NULL)
       return region;
 
-   intel_set_region_tiling_gem(intel, region, handle);
+   ret = dri_bo_get_tiling(region->buffer, &region->tiling,
+			   &region->bit_6_swizzle);
+   if (ret != 0) {
+      fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n",
+	      handle, name, strerror(-ret));
+      intel_region_release(&region);
+      return NULL;
+   }
 
    return region;
 }
@@ -489,7 +473,14 @@ intel_recreate_static(struct intel_context *intel,
 						     name,
 						     region_desc->bo_handle);
 
-      intel_set_region_tiling_gem(intel, region, region_desc->bo_handle);
+      ret = dri_bo_get_tiling(region->buffer, &region->tiling,
+			      &region->bit_6_swizzle);
+      if (ret != 0) {
+	 fprintf(stderr, "Couldn't get tiling of buffer %d (%s): %s\n",
+		 region_desc->bo_handle, name, strerror(-ret));
+	 intel_region_release(&region);
+	 return NULL;
+      }
    } else {
       if (region->classic_map != NULL) {
 	 drmUnmap(region->classic_map,
diff --git a/src/mesa/drivers/dri/intel/intel_span.c b/src/mesa/drivers/dri/intel/intel_span.c
index 8e2b4456f8..8f4e681ffe 100644
--- a/src/mesa/drivers/dri/intel/intel_span.c
+++ b/src/mesa/drivers/dri/intel/intel_span.c
@@ -30,6 +30,7 @@
 #include "main/mtypes.h"
 #include "main/colormac.h"
 
+#include "intel_buffers.h"
 #include "intel_fbo.h"
 #include "intel_screen.h"
 #include "intel_span.h"
@@ -131,12 +132,8 @@ pwrite_8(struct intel_renderbuffer *irb, uint32_t offset, uint8_t val)
 }
 
 static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb,
-				struct intel_context *intel,
 				int x, int y)
 {
-	x += intel->drawX;
-	y += intel->drawY;
-
 	return (y * irb->region->pitch + x) * irb->region->cpp;
 }
 
@@ -145,7 +142,6 @@ static uint32_t no_tile_swizzle(struct intel_renderbuffer *irb,
  */
 
 static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
-			       struct intel_context *intel,
 			       int x, int y)
 {
 	int	tile_stride;
@@ -155,9 +151,6 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 	int	tile_off, tile_base;
 	
 	tile_stride = (irb->pfPitch * irb->region->cpp) << 3;
-	
-	x += intel->drawX;
-	y += intel->drawY;
 
 	xbyte = x * irb->region->cpp;
 
@@ -204,7 +197,6 @@ static uint32_t x_tile_swizzle(struct intel_renderbuffer *irb,
 }
 
 static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
-			       struct intel_context *intel,
 			       int x, int y)
 {
 	int	tile_stride;
@@ -214,9 +206,6 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 	int	tile_off, tile_base;
 	
 	tile_stride = (irb->pfPitch * irb->region->cpp) << 5;
-	
-	x += intel->drawX;
-	y += intel->drawY;
 
 	xbyte = x * irb->region->cpp;
 
@@ -268,8 +257,12 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
    const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
    const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1;	\
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
    GLuint p;								\
-   (void) p;
+   (void) p;								\
+   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
 
 /* XXX FBO: this is identical to the macro in spantmp2.h except we get
  * the cliprect info from the context, not the driDrawable.
@@ -277,12 +270,12 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
  */
 #define HW_CLIPLOOP()							\
    do {									\
-      int _nc = intel->numClipRects;					\
+      int _nc = num_cliprects;						\
       while ( _nc-- ) {							\
-	 int minx = intel->pClipRects[_nc].x1 - intel->drawX;		\
-	 int miny = intel->pClipRects[_nc].y1 - intel->drawY;		\
-	 int maxx = intel->pClipRects[_nc].x2 - intel->drawX;		\
-	 int maxy = intel->pClipRects[_nc].y2 - intel->drawY;
+	 int minx = cliprects[_nc].x1 - x_off;				\
+	 int miny = cliprects[_nc].y1 - y_off;				\
+	 int maxx = cliprects[_nc].x2 - x_off;				\
+	 int maxy = cliprects[_nc].y2 - y_off;
 	
 #if 0
       }}
@@ -295,6 +288,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define HW_UNLOCK()
 
+/* Convenience macros to avoid typing the swizzle argument over and over */
+#define NO_TILE(_X, _Y) no_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
+#define X_TILE(_X, _Y) x_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
+#define Y_TILE(_X, _Y) y_tile_swizzle(irb, (_X) + x_off, (_Y) + y_off)
+
 /* 16 bit, RGB565 color spanline and pixel functions
  */
 #define SPANTMP_PIXEL_FMT GL_RGB
@@ -302,8 +300,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel##x##_RGB565
 #define TAG2(x,y) intel##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, no_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, no_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_16(irb, NO_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_16(irb, NO_TILE(X, Y), V)
 #include "spantmp2.h"
 
 /* 32 bit, ARGB8888 color spanline and pixel functions
@@ -313,8 +311,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel##x##_ARGB8888
 #define TAG2(x,y) intel##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, no_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, no_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_32(irb, NO_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_32(irb, NO_TILE(X, Y), V)
 #include "spantmp2.h"
 
 /* 32 bit, xRGB8888 color spanline and pixel functions
@@ -324,8 +322,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel##x##_xRGB8888
 #define TAG2(x,y) intel##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, no_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, no_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_xrgb8888(irb, NO_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, NO_TILE(X, Y), V)
 #include "spantmp2.h"
 
 /* 16 bit RGB565 color tile spanline and pixel functions
@@ -336,8 +334,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_XTile_##x##_RGB565
 #define TAG2(x,y) intel_XTile_##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, x_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, x_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_16(irb, X_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_16(irb, X_TILE(X, Y), V)
 #include "spantmp2.h"
 
 #define SPANTMP_PIXEL_FMT GL_RGB
@@ -345,8 +343,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_YTile_##x##_RGB565
 #define TAG2(x,y) intel_YTile_##x##_RGB565##y
-#define GET_VALUE(X, Y) pread_16(irb, y_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_16(irb, y_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_16(irb, Y_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_16(irb, Y_TILE(X, Y), V)
 #include "spantmp2.h"
 
 /* 32 bit ARGB888 color tile spanline and pixel functions
@@ -357,8 +355,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_XTile_##x##_ARGB8888
 #define TAG2(x,y) intel_XTile_##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, x_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, x_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_32(irb, X_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_32(irb, X_TILE(X, Y), V)
 #include "spantmp2.h"
 
 #define SPANTMP_PIXEL_FMT GL_BGRA
@@ -366,8 +364,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_YTile_##x##_ARGB8888
 #define TAG2(x,y) intel_YTile_##x##_ARGB8888##y
-#define GET_VALUE(X, Y) pread_32(irb, y_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_32(irb, y_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_32(irb, Y_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_32(irb, Y_TILE(X, Y), V)
 #include "spantmp2.h"
 
 /* 32 bit xRGB888 color tile spanline and pixel functions
@@ -378,8 +376,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_XTile_##x##_xRGB8888
 #define TAG2(x,y) intel_XTile_##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, x_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, x_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_xrgb8888(irb, X_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, X_TILE(X, Y), V)
 #include "spantmp2.h"
 
 #define SPANTMP_PIXEL_FMT GL_BGRA
@@ -387,15 +385,19 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 #define TAG(x)    intel_YTile_##x##_xRGB8888
 #define TAG2(x,y) intel_YTile_##x##_xRGB8888##y
-#define GET_VALUE(X, Y) pread_xrgb8888(irb, y_tile_swizzle(irb, intel, X, Y))
-#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, y_tile_swizzle(irb, intel, X, Y), V)
+#define GET_VALUE(X, Y) pread_xrgb8888(irb, Y_TILE(X, Y))
+#define PUT_VALUE(X, Y, V) pwrite_xrgb8888(irb, Y_TILE(X, Y), V)
 #include "spantmp2.h"
 
 #define LOCAL_DEPTH_VARS						\
    struct intel_context *intel = intel_context(ctx);			\
    struct intel_renderbuffer *irb = intel_renderbuffer(rb);		\
    const GLint yScale = irb->RenderToTexture ? 1 : -1;			\
-   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1;
+   const GLint yBias = irb->RenderToTexture ? 0 : irb->Base.Height - 1; \
+   unsigned int num_cliprects;						\
+   struct drm_clip_rect *cliprects;					\
+   int x_off, y_off;							\
+   intel_get_cliprects(intel, &cliprects, &num_cliprects, &x_off, &y_off);
 
 
 #define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS
@@ -404,10 +406,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
  ** 16-bit depthbuffer functions.
  **/
 #define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) \
-   pwrite_16(irb, no_tile_swizzle(irb, intel, _x, _y), d)
-#define READ_DEPTH(d, _x, _y) \
-   d = pread_16(irb, no_tile_swizzle(irb, intel, _x, _y))
+#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, NO_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = pread_16(irb, NO_TILE(_x, _y))
 #define TAG(x) intel##x##_z16
 #include "depthtmp.h"
 
@@ -416,10 +416,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
  ** 16-bit x tile depthbuffer functions.
  **/
 #define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) \
-   pwrite_16(irb, x_tile_swizzle(irb, intel, _x, _y), d)
-#define READ_DEPTH(d, _x, _y) \
-   d = pread_16(irb, x_tile_swizzle(irb, intel, _x, _y))
+#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, X_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = pread_16(irb, X_TILE(_x, _y))
 #define TAG(x) intel_XTile_##x##_z16
 #include "depthtmp.h"
 
@@ -427,10 +425,8 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
  ** 16-bit y tile depthbuffer functions.
  **/
 #define VALUE_TYPE GLushort
-#define WRITE_DEPTH(_x, _y, d) \
-   pwrite_16(irb, y_tile_swizzle(irb, intel, _x, _y), d)
-#define READ_DEPTH(d, _x, _y) \
-   d = pread_16(irb, y_tile_swizzle(irb, intel, _x, _y))
+#define WRITE_DEPTH(_x, _y, d) pwrite_16(irb, Y_TILE(_x, _y), d)
+#define READ_DEPTH(d, _x, _y) d = pread_16(irb, Y_TILE(_x, _y))
 #define TAG(x) intel_YTile_##x##_z16
 #include "depthtmp.h"
 
@@ -445,12 +441,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 /* Change ZZZS -> SZZZ */
 #define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, no_tile_swizzle(irb, intel, _x, _y),		\
-	     ((d) >> 8) | ((d) << 24))
+   pwrite_32(irb, NO_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
 
 /* Change SZZZ -> ZZZS */
 #define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, no_tile_swizzle(irb, intel, _x, _y));	\
+   GLuint tmp = pread_32(irb, NO_TILE(_x, _y));			\
    d = (tmp << 8) | (tmp >> 24);				\
 }
 
@@ -468,12 +463,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 /* Change ZZZS -> SZZZ */
 #define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, x_tile_swizzle(irb, intel, _x, _y),		\
-	     ((d) >> 8) | ((d) << 24))				\
+   pwrite_32(irb, X_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
 
 /* Change SZZZ -> ZZZS */
 #define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, x_tile_swizzle(irb, intel, _x, _y));	\
+   GLuint tmp = pread_32(irb, X_TILE(_x, _y));		\
    d = (tmp << 8) | (tmp >> 24);				\
 }
 
@@ -490,12 +484,11 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 
 /* Change ZZZS -> SZZZ */
 #define WRITE_DEPTH(_x, _y, d)					\
-   pwrite_32(irb, y_tile_swizzle(irb, intel, _x, _y),		\
-	     ((d) >> 8) | ((d) << 24))
+   pwrite_32(irb, Y_TILE(_x, _y), ((d) >> 8) | ((d) << 24))
 
 /* Change SZZZ -> ZZZS */
 #define READ_DEPTH( d, _x, _y ) {				\
-   GLuint tmp = pread_32(irb, y_tile_swizzle(irb, intel, _x, _y));	\
+   GLuint tmp = pread_32(irb, Y_TILE(_x, _y));			\
    d = (tmp << 8) | (tmp >> 24);				\
 }
 
@@ -506,36 +499,24 @@ static uint32_t y_tile_swizzle(struct intel_renderbuffer *irb,
 /**
  ** 8-bit stencil function (XXX FBO: This is obsolete)
  **/
-#define WRITE_STENCIL(_x, _y, d)				\
-   pwrite_8(irb, no_tile_swizzle(irb, intel, _x, _y) + 3, d)
-
-#define READ_STENCIL(d, _x, _y)					\
-   d = pread_8(irb, no_tile_swizzle(irb, intel, _x, _y) + 3);
-
+#define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, NO_TILE(_x, _y) + 3, d)
+#define READ_STENCIL(d, _x, _y) d = pread_8(irb, NO_TILE(_x, _y) + 3);
 #define TAG(x) intel##x##_z24_s8
 #include "stenciltmp.h"
 
 /**
  ** 8-bit x-tile stencil function (XXX FBO: This is obsolete)
  **/
-#define WRITE_STENCIL(_x, _y, d)				\
-   pwrite_8(irb, x_tile_swizzle(irb, intel, _x, _y) + 3, d)
-
-#define READ_STENCIL(d, _x, _y)					\
-   d = pread_8(irb, x_tile_swizzle(irb, intel, _x, _y) + 3);
-
+#define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, X_TILE(_x, _y) + 3, d)
+#define READ_STENCIL(d, _x, _y) d = pread_8(irb, X_TILE(_x, _y) + 3);
 #define TAG(x) intel_XTile_##x##_z24_s8
 #include "stenciltmp.h"
 
 /**
  ** 8-bit y-tile stencil function (XXX FBO: This is obsolete)
  **/
-#define WRITE_STENCIL(_x, _y, d)				\
-   pwrite_8(irb, y_tile_swizzle(irb, intel, _x, _y) + 3, d)
-
-#define READ_STENCIL(d, _x, _y)					\
-   d = pread_8(irb, y_tile_swizzle(irb, intel, _x, _y) + 3)
-
+#define WRITE_STENCIL(_x, _y, d) pwrite_8(irb, Y_TILE(_x, _y) + 3, d)
+#define READ_STENCIL(d, _x, _y) d = pread_8(irb, Y_TILE(_x, _y) + 3)
 #define TAG(x) intel_YTile_##x##_z24_s8
 #include "stenciltmp.h"
 
@@ -602,14 +583,10 @@ intel_map_unmap_buffers(struct intel_context *intel, GLboolean map)
       if (tex) {
          /* render to texture */
          ASSERT(att->Renderbuffer);
-         if (map) {
-            struct gl_texture_image *texImg;
-            texImg = tex->Image[att->CubeMapFace][att->TextureLevel];
+         if (map)
             intel_tex_map_images(intel, intel_texture_object(tex));
-         }
-         else {
+         else
             intel_tex_unmap_images(intel, intel_texture_object(tex));
-         }
       }
    }