--- old/src/share/native/sun/java2d/opengl/OGLBlitLoops.c	2014-04-21 16:50:34.000000000 +0400
+++ new/src/share/native/sun/java2d/opengl/OGLBlitLoops.c	2014-04-21 16:50:34.000000000 +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 +753,7 @@
         dstOps->GetRasInfo(env, dstOps, &dstInfo);
         if (dstInfo.rasBase) {
             void *pDst = dstInfo.rasBase;
+            void* tempRow;
 
             srcx = srcInfo.bounds.x1;
             srcy = srcInfo.bounds.y1;
@@ -777,21 +781,40 @@
             J2dTraceLn2(J2D_TRACE_VERBOSE, "  dx=%d dy=%d",
                         dstx, dsty);
 
-            // 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.
+            if (tempRow = malloc(dstInfo.scanStride)) {
+                // fast path
+                jint i;
+                const jint stride = dstInfo.scanStride;
+                srcy = srcOps->yOffset + srcy;
+                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 * stride));
+                    void * row2 = PtrAddBytes(pDst,(height - i - 1) * stride);
+                    memcpy(tempRow, row1, stride);
+                    memcpy(row1, row2, stride);
+                    memcpy(row2, tempRow, stride);
+                }
+                free(tempRow);
+            } else {
+                // slow path
+                // this accounts for lower-left origin of the source region
+                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,
+                                     pf.format, pf.type, pDst);
+                    srcy--;
+                    dsty++;
+                    height--;
+                }
             }
-
 #ifdef MACOSX
             if (srcOps->isOpaque) {
                 j2d_glPixelTransferf(GL_ALPHA_BIAS, 0.0);