--- Mesa-10.0.2/src/mesa/main/readpix.c	2014-02-03 12:42:39.000000000 -0500
+++ Mesa-10.0.2/src/mesa/main/readpix.c	2014-02-06 09:14:09.886460677 -0500
@@ -422,7 +422,19 @@ read_rgba_pixels_swizzle(struct gl_conte
    GLubyte *dst, *map;
    int dstStride, stride, j;
    GLboolean swizzle_rb = GL_FALSE, copy_xrgb = GL_FALSE;
+   GLboolean klconvert_888_to_argb_v1 = GL_FALSE;
+   GLboolean klconvert_888_to_argb_v2 = GL_FALSE;
+   GLboolean klconvert_x888_to_argb = GL_FALSE;

+   //printf("[glReadPixels]%s(%d, %d, %s[%s - 0x%x], %s, %p)\n", __func__, width, height, _mesa_lookup_enum_by_nr(format), _mesa_lookup_enum_by_nr(rb->Format), rb->Format, _mesa_lookup_enum_by_nr(type), pixels);
+
+   if (rb->Format == MESA_FORMAT_XRGB8888 && format == GL_BGRA && type == GL_UNSIGNED_BYTE) {
+       klconvert_x888_to_argb = GL_TRUE;
+   } else
+   if (rb->Format == MESA_FORMAT_RGB888 && format == GL_BGRA && type == GL_UNSIGNED_BYTE) {
+      klconvert_888_to_argb_v2 = GL_TRUE;
+      //klconvert_888_to_argb_v1 = GL_TRUE;
+   } else
    /* XXX we could check for other swizzle/special cases here as needed */
    if (rb->Format == MESA_FORMAT_RGBA8888_REV &&
        format == GL_BGRA &&
@@ -465,6 +477,63 @@ read_rgba_pixels_swizzle(struct gl_conte
          dst += dstStride;
          map += stride;
       }
+   } else if (klconvert_888_to_argb_v1) {
+        /* KL: convert rgb -> argb, faster - but not fast enough */
+        unsigned char *psrc = map;
+        unsigned char *pdst = dst;
+        for (j = 0; j < height; j++) {
+                for (int k = 0; k < width; k++) {
+                        *(pdst++) /* B */ = *(psrc++);
+                        *(pdst++) /* G */ = *(psrc++);
+                        *(pdst++) /* R */ = *(psrc++);
+                        *(pdst++) /* A */ = 0x00;
+                }
+        }
+   } else if (klconvert_888_to_argb_v2) {
+        /* KL: convert rgb -> argb, faster again */
+        /* KL: Vectored SIMD move would improve */
+        unsigned char *psrc = map;
+        unsigned int *ddst = (unsigned int *)dst;
+        register unsigned int p[12];
+        register unsigned int a, b, c, d;
+        for (j = 0; j < (height * width); j += 4) {
+
+                /* read 4 pixels as BGRBGRBGRBGR */
+                p[0] = *(psrc++);
+                p[1] = *(psrc++);
+                p[2] = *(psrc++);
+                p[3] = *(psrc++);
+                p[4] = *(psrc++);
+                p[5] = *(psrc++);
+                p[6] = *(psrc++);
+                p[7] = *(psrc++);
+                p[8] = *(psrc++);
+                p[9] = *(psrc++);
+                p[10] = *(psrc++);
+                p[11] = *(psrc++);
+
+               /* Alpha channel defaults to 0 */
+                a = p[0] | p[1] << 8 | p[2] << 16;
+                b = p[3] | p[4] << 8 | p[5] << 16;
+                c = p[6] | p[7] << 8 | p[8] << 16;
+                d = p[9] | p[10] << 8 | p[11] << 16;
+
+                *(ddst++) = a;
+                *(ddst++) = b;
+                *(ddst++) = c;
+                *(ddst++) = d;
+        }
+   } else if (klconvert_x888_to_argb) {
+       if (dstStride == stride) {
+           /* Special case, convert identical strides into a single memcpy. */
+           width = width * height;
+           height = 1;
+       }
+       for (j = 0; j < height; j++) {
+          memcpy(dst, map, width * 4);
+          dst += dstStride;
+          map += stride;
+       }
    } else if (copy_xrgb) {
       /* convert xrgb -> argb */
       for (j = 0; j < height; j++) {