--- old/src/share/native/sun/java2d/opengl/OGLBlitLoops.c	2014-04-24 00:01:28.810066400 +0400
+++ new/src/share/native/sun/java2d/opengl/OGLBlitLoops.c	2014-04-24 00:01:28.665459000 +0400
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2003, 2013, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2003, 2014, Oracle and/or its affiliates. All rights reserved.
  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
  *
  * This code is free software; you can redistribute it and/or modify it
@@ -34,6 +34,9 @@
 #include "OGLSurfaceData.h"
 #include "GraphicsPrimitiveMgr.h"
 
+#include <stdlib.h> // malloc
+#include <string.h> // memcpy
+
 extern OGLPixelFormat PixelFormats[];
 
 /**
@@ -750,6 +753,8 @@
         dstOps->GetRasInfo(env, dstOps, &dstInfo);
         if (dstInfo.rasBase) {
             void *pDst = dstInfo.rasBase;
+            void* tempRow;
+            jint clippedStride;
 
             srcx = srcInfo.bounds.x1;
             srcy = srcInfo.bounds.y1;
@@ -758,7 +763,7 @@
             width = srcInfo.bounds.x2 - srcInfo.bounds.x1;
             height = srcInfo.bounds.y2 - srcInfo.bounds.y1;
 
-            j2d_glPixelStorei(GL_PACK_SKIP_PIXELS, dstx);
+            pDst = PtrAddBytes(pDst, dstx * dstInfo.pixelStride);   
             j2d_glPixelStorei(GL_PACK_ROW_LENGTH,
                               dstInfo.scanStride / dstInfo.pixelStride);
             j2d_glPixelStorei(GL_PACK_ALIGNMENT, pf.alignment);
@@ -780,16 +785,37 @@
             // this accounts for lower-left origin of the source region
             srcx = srcOps->xOffset + srcx;
             srcy = srcOps->yOffset + srcOps->height - (srcy + 1);
-
-            // we must read one scanline at a time because there is no way
-            // to read starting at the top-left corner of the source region
-            while (height > 0) {
-                j2d_glPixelStorei(GL_PACK_SKIP_ROWS, dsty);
-                j2d_glReadPixels(srcx, srcy, width, 1,
+            
+            // Note that glReadPixels() is extremely slow!
+            // So we try to call it only once and flip the image using memcpy.
+            clippedStride = dstInfo.pixelStride * width;
+            if (tempRow = malloc(clippedStride)) {
+                // fast path
+                jint i;
+                srcy = srcy - height + 1;
+                pDst = PtrAddBytes(pDst, dsty * dstInfo.scanStride);
+                j2d_glReadPixels(srcx, srcy, width, height,
                                  pf.format, pf.type, pDst);
-                srcy--;
-                dsty++;
-                height--;
+                for (i = 0; i < height / 2; ++i) {
+                    void * row1 = PtrAddBytes(pDst, (i * dstInfo.scanStride));
+                    void * row2 = PtrAddBytes(pDst,(height - i - 1) * dstInfo.scanStride);
+                    memcpy(tempRow, row1, clippedStride);
+                    memcpy(row1, row2, clippedStride);
+                    memcpy(row2, tempRow, clippedStride);
+                }
+                free(tempRow);
+            } else {
+                // slow path
+                // we must read one scanline at a time because there is no way
+                // to read starting at the top-left corner of the source region
+                while (height > 0) {
+                    j2d_glPixelStorei(GL_PACK_SKIP_ROWS, dsty);
+                    j2d_glReadPixels(srcx, srcy, width, 1,
+                                     pf.format, pf.type, pDst);
+                    srcy--;
+                    dsty++;
+                    height--;
+                }
             }
 
 #ifdef MACOSX