r698 : Improve performance of non-SIMD color conversion routines and use global constants to define colorspace extension parameters

24 files changed, 1227 insertions, 557 deletions

diff --git a/configure.ac b/configure.ac
index 44a3ac1..91ca8ff 100644
--- a/configure.ac
+++ b/configure.ac
@@ -173,6 +173,20 @@ AC_MSG_RESULT($VERSION_SCRIPT)
 AM_CONDITIONAL(VERSION_SCRIPT, test "x$VERSION_SCRIPT" = "xyes")
 AC_SUBST(VERSION_SCRIPT_FLAG)
 
+# Check for non-broken inline under various spellings
+AC_MSG_CHECKING(for inline)
+ljt_cv_inline=""
+AC_TRY_COMPILE(, [} __attribute__((always_inline)) int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__attribute__((always_inline))",
+AC_TRY_COMPILE(, [} __inline__ int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline__",
+AC_TRY_COMPILE(, [} __inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="__inline",
+AC_TRY_COMPILE(, [} inline int foo() { return 0; }
+int bar() { return foo();], ljt_cv_inline="inline"))))
+AC_MSG_RESULT($ljt_cv_inline)
+AC_DEFINE_UNQUOTED([INLINE],[$ljt_cv_inline],[How to obtain function inlining.])
+
 AC_MSG_CHECKING([whether to include arithmetic encoding support])
 AC_ARG_WITH([arith-enc],
     AC_HELP_STRING([--without-arith-enc], [Omit arithmetic encoding support]))
diff --git a/jccolext.c b/jccolext.c
new file mode 100644
index 0000000..acbfa23
--- /dev/null
+++ b/jccolext.c
@@ -0,0 +1,114 @@
+/*
+ * jccolext.c
+ *
+ * Copyright (C) 1991-1996, Thomas G. Lane.
+ * Copyright (C) 2009-2011, D. R. Commander.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains input colorspace conversion routines.
+ */
+
+
+/* This file is included by jccolor.c */
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ *
+ * Note that we change from the application's interleaved-pixel format
+ * to our internal noninterleaved, one-plane-per-component format.
+ * The input buffer is therefore three times as wide as the output buffer.
+ *
+ * A starting row offset is provided only for the output buffer.  The caller
+ * can easily adjust the passed input_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+rgb_ycc_convert_internal (j_compress_ptr cinfo,
+                          JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                          JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr0, outptr1, outptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr0 = output_buf[0][output_row];
+    outptr1 = output_buf[1][output_row];
+    outptr2 = output_buf[2][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
+       * must be too; we do not need an explicit range-limiting operation.
+       * Hence the value being shifted is never negative, and we don't
+       * need the general RIGHT_SHIFT macro.
+       */
+      /* Y */
+      outptr0[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+      /* Cb */
+      outptr1[col] = (JSAMPLE)
+		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
+		 >> SCALEBITS);
+      /* Cr */
+      outptr2[col] = (JSAMPLE)
+		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
+
+
+/**************** Cases other than RGB -> YCbCr **************/
+
+
+/*
+ * Convert some rows of samples to the JPEG colorspace.
+ * This version handles RGB->grayscale conversion, which is the same
+ * as the RGB->Y portion of RGB->YCbCr.
+ * We assume rgb_ycc_start has been called (we only use the Y tables).
+ */
+
+INLINE
+LOCAL(void)
+rgb_gray_convert_internal (j_compress_ptr cinfo,
+                           JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
+                           JDIMENSION output_row, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int r, g, b;
+  register INT32 * ctab = cconvert->rgb_ycc_tab;
+  register JSAMPROW inptr;
+  register JSAMPROW outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->image_width;
+
+  while (--num_rows >= 0) {
+    inptr = *input_buf++;
+    outptr = output_buf[0][output_row];
+    output_row++;
+    for (col = 0; col < num_cols; col++) {
+      r = GETJSAMPLE(inptr[RGB_RED]);
+      g = GETJSAMPLE(inptr[RGB_GREEN]);
+      b = GETJSAMPLE(inptr[RGB_BLUE]);
+      inptr += RGB_PIXELSIZE;
+      /* Y */
+      outptr[col] = (JSAMPLE)
+		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
+		 >> SCALEBITS);
+    }
+  }
+}
diff --git a/jccolor.c b/jccolor.c
index 9559aa5..12804f3 100644
--- a/jccolor.c
+++ b/jccolor.c
@@ -3,7 +3,7 @@
  *
  * Copyright (C) 1991-1996, Thomas G. Lane.
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright 2009-2011 D. R. Commander
+ * Copyright (C) 2009-2011, D. R. Commander.
  * This file is part of the Independent JPEG Group's software.
  * For conditions of distribution and use, see the accompanying README file.
  *
@@ -81,6 +81,99 @@ typedef my_color_converter * my_cconvert_ptr;
 #define TABLE_SIZE	(8*(MAXJSAMPLE+1))
 
 
+/* Include inline routines for colorspace extensions */
+
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED EXT_RGB_RED
+#define RGB_GREEN EXT_RGB_GREEN
+#define RGB_BLUE EXT_RGB_BLUE
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define rgb_ycc_convert_internal extrgb_ycc_convert_internal
+#define rgb_gray_convert_internal extrgb_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+#define RGB_RED EXT_RGBX_RED
+#define RGB_GREEN EXT_RGBX_GREEN
+#define RGB_BLUE EXT_RGBX_BLUE
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define rgb_ycc_convert_internal extrgbx_ycc_convert_internal
+#define rgb_gray_convert_internal extrgbx_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+#define RGB_RED EXT_BGR_RED
+#define RGB_GREEN EXT_BGR_GREEN
+#define RGB_BLUE EXT_BGR_BLUE
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define rgb_ycc_convert_internal extbgr_ycc_convert_internal
+#define rgb_gray_convert_internal extbgr_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+#define RGB_RED EXT_BGRX_RED
+#define RGB_GREEN EXT_BGRX_GREEN
+#define RGB_BLUE EXT_BGRX_BLUE
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define rgb_ycc_convert_internal extbgrx_ycc_convert_internal
+#define rgb_gray_convert_internal extbgrx_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+#define RGB_RED EXT_XBGR_RED
+#define RGB_GREEN EXT_XBGR_GREEN
+#define RGB_BLUE EXT_XBGR_BLUE
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define rgb_ycc_convert_internal extxbgr_ycc_convert_internal
+#define rgb_gray_convert_internal extxbgr_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+#define RGB_RED EXT_XRGB_RED
+#define RGB_GREEN EXT_XRGB_GREEN
+#define RGB_BLUE EXT_XRGB_BLUE
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define rgb_ycc_convert_internal extxrgb_ycc_convert_internal
+#define rgb_gray_convert_internal extxrgb_gray_convert_internal
+#include "jccolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef rgb_ycc_convert_internal
+#undef rgb_gray_convert_internal
+
+
 /*
  * Initialize for RGB->YCC colorspace conversion.
  */
@@ -119,14 +212,6 @@ rgb_ycc_start (j_compress_ptr cinfo)
 
 /*
  * Convert some rows of samples to the JPEG colorspace.
- *
- * Note that we change from the application's interleaved-pixel format
- * to our internal noninterleaved, one-plane-per-component format.
- * The input buffer is therefore three times as wide as the output buffer.
- *
- * A starting row offset is provided only for the output buffer.  The caller
- * can easily adjust the passed input_buf value to accommodate any row
- * offset required on that side.
  */
 
 METHODDEF(void)
@@ -134,47 +219,35 @@ rgb_ycc_convert (j_compress_ptr cinfo,
 		 JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
 		 JDIMENSION output_row, int num_rows)
 {
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr0, outptr1, outptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-  int rindex = rgb_red[cinfo->in_color_space];
-  int gindex = rgb_green[cinfo->in_color_space];
-  int bindex = rgb_blue[cinfo->in_color_space];
-  int rgbstride = rgb_pixelsize[cinfo->in_color_space];
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr0 = output_buf[0][output_row];
-    outptr1 = output_buf[1][output_row];
-    outptr2 = output_buf[2][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[rindex]);
-      g = GETJSAMPLE(inptr[gindex]);
-      b = GETJSAMPLE(inptr[bindex]);
-      inptr += rgbstride;
-      /* If the inputs are 0..MAXJSAMPLE, the outputs of these equations
-       * must be too; we do not need an explicit range-limiting operation.
-       * Hence the value being shifted is never negative, and we don't
-       * need the general RIGHT_SHIFT macro.
-       */
-      /* Y */
-      outptr0[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-      /* Cb */
-      outptr1[col] = (JSAMPLE)
-		((ctab[r+R_CB_OFF] + ctab[g+G_CB_OFF] + ctab[b+B_CB_OFF])
-		 >> SCALEBITS);
-      /* Cr */
-      outptr2[col] = (JSAMPLE)
-		((ctab[r+R_CR_OFF] + ctab[g+G_CR_OFF] + ctab[b+B_CR_OFF])
-		 >> SCALEBITS);
-    }
+  switch (cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+      extrgbx_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+      extbgrx_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+      extxbgr_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+      extxrgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    default:
+      rgb_ycc_convert_internal(cinfo, input_buf, output_buf, output_row,
+                               num_rows);
+      break;
   }
 }
 
@@ -184,9 +257,6 @@ rgb_ycc_convert (j_compress_ptr cinfo,
 
 /*
  * Convert some rows of samples to the JPEG colorspace.
- * This version handles RGB->grayscale conversion, which is the same
- * as the RGB->Y portion of RGB->YCbCr.
- * We assume rgb_ycc_start has been called (we only use the Y tables).
  */
 
 METHODDEF(void)
@@ -194,32 +264,35 @@ rgb_gray_convert (j_compress_ptr cinfo,
 		  JSAMPARRAY input_buf, JSAMPIMAGE output_buf,
 		  JDIMENSION output_row, int num_rows)
 {
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int r, g, b;
-  register INT32 * ctab = cconvert->rgb_ycc_tab;
-  register JSAMPROW inptr;
-  register JSAMPROW outptr;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->image_width;
-  int rindex = rgb_red[cinfo->in_color_space];
-  int gindex = rgb_green[cinfo->in_color_space];
-  int bindex = rgb_blue[cinfo->in_color_space];
-  int rgbstride = rgb_pixelsize[cinfo->in_color_space];
-
-  while (--num_rows >= 0) {
-    inptr = *input_buf++;
-    outptr = output_buf[0][output_row];
-    output_row++;
-    for (col = 0; col < num_cols; col++) {
-      r = GETJSAMPLE(inptr[rindex]);
-      g = GETJSAMPLE(inptr[gindex]);
-      b = GETJSAMPLE(inptr[bindex]);
-      inptr += rgbstride;
-      /* Y */
-      outptr[col] = (JSAMPLE)
-		((ctab[r+R_Y_OFF] + ctab[g+G_Y_OFF] + ctab[b+B_Y_OFF])
-		 >> SCALEBITS);
-    }
+  switch (cinfo->in_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_RGBX:
+      extrgbx_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                    num_rows);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGRX:
+      extbgrx_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                    num_rows);
+      break;
+    case JCS_EXT_XBGR:
+      extxbgr_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                    num_rows);
+      break;
+    case JCS_EXT_XRGB:
+      extxrgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                    num_rows);
+      break;
+    default:
+      rgb_gray_convert_internal(cinfo, input_buf, output_buf, output_row,
+                                num_rows);
+      break;
   }
 }
 
diff --git a/jconfig.h.in b/jconfig.h.in
index fbddf7d..4aef763 100644
--- a/jconfig.h.in
+++ b/jconfig.h.in
@@ -27,6 +27,9 @@
 /* Compiler does not support pointers to unspecified structures. */
 #undef INCOMPLETE_TYPES_BROKEN
 
+/* How to obtain function inlining. */
+#undef INLINE
+
 /* Compiler has <strings.h> rather than standard <string.h>. */
 #undef NEED_BSD_STRINGS
 
@@ -50,11 +53,5 @@
 /* Define to empty if `const' does not conform to ANSI C. */
 #undef const
 
-/* Define to `__inline__' or `__inline' if that's what the C compiler
-   calls it, or to nothing if 'inline' is not supported under any name.  */
-#ifndef __cplusplus
-#undef inline
-#endif
-
 /* Define to `unsigned int' if <sys/types.h> does not define. */
 #undef size_t
diff --git a/jdcolext.c b/jdcolext.c
new file mode 100644
index 0000000..d0ccca3
--- /dev/null
+++ b/jdcolext.c
@@ -0,0 +1,94 @@
+/*
+ * jdcolext.c
+ *
+ * Copyright (C) 1991-1997, Thomas G. Lane.
+ * Copyright (C) 2009, 2011, D. R. Commander.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains output colorspace conversion routines.
+ */
+
+
+/* This file is included by jdcolor.c */
+
+
+/*
+ * Convert some rows of samples to the output colorspace.
+ *
+ * Note that we change from noninterleaved, one-plane-per-component format
+ * to interleaved-pixel format.  The output buffer is therefore three times
+ * as wide as the input buffer.
+ * A starting row offset is provided only for the input buffer.  The caller
+ * can easily adjust the passed output_buf value to accommodate any row
+ * offset required on that side.
+ */
+
+INLINE
+LOCAL(void)
+ycc_rgb_convert_internal (j_decompress_ptr cinfo,
+                          JSAMPIMAGE input_buf, JDIMENSION input_row,
+                          JSAMPARRAY output_buf, int num_rows)
+{
+  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
+  register int y, cb, cr;
+  register JSAMPROW outptr;
+  register JSAMPROW inptr0, inptr1, inptr2;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  register int * Crrtab = cconvert->Cr_r_tab;
+  register int * Cbbtab = cconvert->Cb_b_tab;
+  register INT32 * Crgtab = cconvert->Cr_g_tab;
+  register INT32 * Cbgtab = cconvert->Cb_g_tab;
+  SHIFT_TEMPS
+
+  while (--num_rows >= 0) {
+    inptr0 = input_buf[0][input_row];
+    inptr1 = input_buf[1][input_row];
+    inptr2 = input_buf[2][input_row];
+    input_row++;
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      y  = GETJSAMPLE(inptr0[col]);
+      cb = GETJSAMPLE(inptr1[col]);
+      cr = GETJSAMPLE(inptr2[col]);
+      /* Range-limiting is essential due to noise introduced by DCT losses. */
+      outptr[RGB_RED] =   range_limit[y + Crrtab[cr]];
+      outptr[RGB_GREEN] = range_limit[y +
+			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
+						 SCALEBITS))];
+      outptr[RGB_BLUE] =  range_limit[y + Cbbtab[cb]];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
+
+
+/*
+ * Convert grayscale to RGB: just duplicate the graylevel three times.
+ * This is provided to support applications that don't want to cope
+ * with grayscale as a separate case.
+ */
+
+INLINE
+LOCAL(void)
+gray_rgb_convert_internal (j_decompress_ptr cinfo,
+                           JSAMPIMAGE input_buf, JDIMENSION input_row,
+                           JSAMPARRAY output_buf, int num_rows)
+{
+  register JSAMPROW inptr, outptr;
+  register JDIMENSION col;
+  JDIMENSION num_cols = cinfo->output_width;
+
+  while (--num_rows >= 0) {
+    inptr = input_buf[0][input_row++];
+    outptr = *output_buf++;
+    for (col = 0; col < num_cols; col++) {
+      /* We can dispense with GETJSAMPLE() here */
+      outptr[RGB_RED] = outptr[RGB_GREEN] = outptr[RGB_BLUE] = inptr[col];
+      outptr += RGB_PIXELSIZE;
+    }
+  }
+}
diff --git a/jdcolor.c b/jdcolor.c
index 0af024c..d6e5a10 100644
--- a/jdcolor.c
+++ b/jdcolor.c
@@ -3,7 +3,7 @@
  *
  * Copyright (C) 1991-1997, Thomas G. Lane.
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, D. R. Commander.
+ * Copyright (C) 2009, 2011, D. R. Commander.
  * This file is part of the Independent JPEG Group's software.
  * For conditions of distribution and use, see the accompanying README file.
  *
@@ -65,6 +65,99 @@ typedef my_color_deconverter * my_cconvert_ptr;
 #define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
 
 
+/* Include inline routines for colorspace extensions */
+
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED EXT_RGB_RED
+#define RGB_GREEN EXT_RGB_GREEN
+#define RGB_BLUE EXT_RGB_BLUE
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extrgb_convert_internal
+#define gray_rgb_convert_internal gray_extrgb_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+#define RGB_RED EXT_RGBX_RED
+#define RGB_GREEN EXT_RGBX_GREEN
+#define RGB_BLUE EXT_RGBX_BLUE
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extrgbx_convert_internal
+#define gray_rgb_convert_internal gray_extrgbx_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+#define RGB_RED EXT_BGR_RED
+#define RGB_GREEN EXT_BGR_GREEN
+#define RGB_BLUE EXT_BGR_BLUE
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extbgr_convert_internal
+#define gray_rgb_convert_internal gray_extbgr_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+#define RGB_RED EXT_BGRX_RED
+#define RGB_GREEN EXT_BGRX_GREEN
+#define RGB_BLUE EXT_BGRX_BLUE
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extbgrx_convert_internal
+#define gray_rgb_convert_internal gray_extbgrx_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+#define RGB_RED EXT_XBGR_RED
+#define RGB_GREEN EXT_XBGR_GREEN
+#define RGB_BLUE EXT_XBGR_BLUE
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extxbgr_convert_internal
+#define gray_rgb_convert_internal gray_extxbgr_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+#define RGB_RED EXT_XRGB_RED
+#define RGB_GREEN EXT_XRGB_GREEN
+#define RGB_BLUE EXT_XRGB_BLUE
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define ycc_rgb_convert_internal ycc_extxrgb_convert_internal
+#define gray_rgb_convert_internal gray_extxrgb_convert_internal
+#include "jdcolext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef ycc_rgb_convert_internal
+#undef gray_rgb_convert_internal
+
+
 /*
  * Initialize tables for YCC->RGB colorspace conversion.
  */
@@ -110,13 +203,6 @@ build_ycc_rgb_table (j_decompress_ptr cinfo)
 
 /*
  * Convert some rows of samples to the output colorspace.
- *
- * Note that we change from noninterleaved, one-plane-per-component format
- * to interleaved-pixel format.  The output buffer is therefore three times
- * as wide as the input buffer.
- * A starting row offset is provided only for the input buffer.  The caller
- * can easily adjust the passed output_buf value to accommodate any row
- * offset required on that side.
  */
 
 METHODDEF(void)
@@ -124,42 +210,35 @@ ycc_rgb_convert (j_decompress_ptr cinfo,
 		 JSAMPIMAGE input_buf, JDIMENSION input_row,
 		 JSAMPARRAY output_buf, int num_rows)
 {
-  my_cconvert_ptr cconvert = (my_cconvert_ptr) cinfo->cconvert;
-  register int y, cb, cr;
-  register JSAMPROW outptr;
-  register JSAMPROW inptr0, inptr1, inptr2;
-  register JDIMENSION col;
-  JDIMENSION num_cols = cinfo->output_width;
-  int rindex = rgb_red[cinfo->out_color_space];
-  int gindex = rgb_green[cinfo->out_color_space];
-  int bindex = rgb_blue[cinfo->out_color_space];
-  int rgbstride = rgb_pixelsize[cinfo->out_color_space];
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  register int * Crrtab = cconvert->Cr_r_tab;
-  register int * Cbbtab = cconvert->Cb_b_tab;
-  register INT32 * Crgtab = cconvert->Cr_g_tab;
-  register INT32 * Cbgtab = cconvert->Cb_g_tab;
-  SHIFT_TEMPS
-
-  while (--num_rows >= 0) {
-    inptr0 = input_buf[0][input_row];
-    inptr1 = input_buf[1][input_row];
-    inptr2 = input_buf[2][input_row];
-    input_row++;
-    outptr = *output_buf++;
-    for (col = 0; col < num_cols; col++) {
-      y  = GETJSAMPLE(inptr0[col]);
-      cb = GETJSAMPLE(inptr1[col]);
-      cr = GETJSAMPLE(inptr2[col]);
-      /* Range-limiting is essential due to noise introduced by DCT losses. */
-      outptr[rindex] =   range_limit[y + Crrtab[cr]];
-      outptr[gindex] = range_limit[y +
-			      ((int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr],
-						 SCALEBITS))];
-      outptr[bindex] =  range_limit[y + Cbbtab[cb]];
-      outptr += rgbstride;
-    }
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      ycc_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_RGBX:
+      ycc_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGR:
+      ycc_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                  num_rows);
+      break;
+    case JCS_EXT_BGRX:
+      ycc_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XBGR:
+      ycc_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_XRGB:
+      ycc_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    default:
+      ycc_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                               num_rows);
+      break;
   }
 }
 
@@ -215,9 +294,7 @@ grayscale_convert (j_decompress_ptr cinfo,
 
 
 /*
- * Convert grayscale to RGB: just duplicate the graylevel three times.
- * This is provided to support applications that don't want to cope
- * with grayscale as a separate case.
+ * Convert grayscale to RGB
  */
 
 METHODDEF(void)
@@ -225,22 +302,35 @@ gray_rgb_convert (j_decompress_ptr cinfo,
 		  JSAMPIMAGE input_buf, JDIMENSION input_row,
 		  JSAMPARRAY output_buf, int num_rows)
 {
-  register JSAMPROW inptr, outptr;
-  JSAMPLE *maxinptr;
-  JDIMENSION num_cols = cinfo->output_width;
-  int rindex = rgb_red[cinfo->out_color_space];
-  int gindex = rgb_green[cinfo->out_color_space];
-  int bindex = rgb_blue[cinfo->out_color_space];
-  int rgbstride = rgb_pixelsize[cinfo->out_color_space];
-
-  while (--num_rows >= 0) {
-    inptr = input_buf[0][input_row++];
-    maxinptr = &inptr[num_cols];
-    outptr = *output_buf++;
-    for (; inptr < maxinptr; inptr++, outptr += rgbstride) {
-      /* We can dispense with GETJSAMPLE() here */
-      outptr[rindex] = outptr[gindex] = outptr[bindex] = *inptr;
-    }
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      gray_extrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_RGBX:
+      gray_extrgbx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                    num_rows);
+      break;
+    case JCS_EXT_BGR:
+      gray_extbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                   num_rows);
+      break;
+    case JCS_EXT_BGRX:
+      gray_extbgrx_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                    num_rows);
+      break;
+    case JCS_EXT_XBGR:
+      gray_extxbgr_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                    num_rows);
+      break;
+    case JCS_EXT_XRGB:
+      gray_extxrgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                    num_rows);
+      break;
+    default:
+      gray_rgb_convert_internal(cinfo, input_buf, input_row, output_buf,
+                                num_rows);
+      break;
   }
 }
 
diff --git a/jdmerge.c b/jdmerge.c
index edf061a..c75497a 100644
--- a/jdmerge.c
+++ b/jdmerge.c
@@ -3,7 +3,7 @@
  *
  * Copyright (C) 1994-1996, Thomas G. Lane.
  * Copyright 2009 Pierre Ossman <ossman@cendio.se> for Cendio AB
- * Copyright (C) 2009, D. R. Commander.
+ * Copyright (C) 2009, 2011, D. R. Commander.
  * This file is part of the Independent JPEG Group's software.
  * For conditions of distribution and use, see the accompanying README file.
  *
@@ -77,6 +77,99 @@ typedef my_upsampler * my_upsample_ptr;
 #define FIX(x)		((INT32) ((x) * (1L<<SCALEBITS) + 0.5))
 
 
+/* Include inline routines for colorspace extensions */
+
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+
+#define RGB_RED EXT_RGB_RED
+#define RGB_GREEN EXT_RGB_GREEN
+#define RGB_BLUE EXT_RGB_BLUE
+#define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
+#define h2v1_merged_upsample_internal extrgb_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extrgb_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED EXT_RGBX_RED
+#define RGB_GREEN EXT_RGBX_GREEN
+#define RGB_BLUE EXT_RGBX_BLUE
+#define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
+#define h2v1_merged_upsample_internal extrgbx_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extrgbx_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED EXT_BGR_RED
+#define RGB_GREEN EXT_BGR_GREEN
+#define RGB_BLUE EXT_BGR_BLUE
+#define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
+#define h2v1_merged_upsample_internal extbgr_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extbgr_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED EXT_BGRX_RED
+#define RGB_GREEN EXT_BGRX_GREEN
+#define RGB_BLUE EXT_BGRX_BLUE
+#define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
+#define h2v1_merged_upsample_internal extbgrx_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extbgrx_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED EXT_XBGR_RED
+#define RGB_GREEN EXT_XBGR_GREEN
+#define RGB_BLUE EXT_XBGR_BLUE
+#define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
+#define h2v1_merged_upsample_internal extxbgr_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extxbgr_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+#define RGB_RED EXT_XRGB_RED
+#define RGB_GREEN EXT_XRGB_GREEN
+#define RGB_BLUE EXT_XRGB_BLUE
+#define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
+#define h2v1_merged_upsample_internal extxrgb_h2v1_merged_upsample_internal
+#define h2v2_merged_upsample_internal extxrgb_h2v2_merged_upsample_internal
+#include "jdmrgext.c"
+#undef RGB_RED
+#undef RGB_GREEN
+#undef RGB_BLUE
+#undef RGB_PIXELSIZE
+#undef h2v1_merged_upsample_internal
+#undef h2v2_merged_upsample_internal
+
+
 /*
  * Initialize tables for YCC->RGB colorspace conversion.
  * This is taken directly from jdcolor.c; see that file for more info.
@@ -230,55 +323,35 @@ h2v1_merged_upsample (j_decompress_ptr cinfo,
 		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
 		      JSAMPARRAY output_buf)
 {
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr;
-  JSAMPROW inptr0, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr0 = input_buf[0][in_row_group_ctr];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr = output_buf[0];
-  /* Loop for each pair of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 2 Y values and emit 2 pixels */
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr += rgb_pixelsize[cinfo->out_color_space];
-    y  = GETJSAMPLE(*inptr0++);
-    outptr[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr += rgb_pixelsize[cinfo->out_color_space];
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr0);
-    outptr[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                           output_buf);
+      break;
+    case JCS_EXT_RGBX:
+      extrgbx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                           output_buf);
+      break;
+    case JCS_EXT_BGRX:
+      extbgrx_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_XBGR:
+      extxbgr_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_XRGB:
+      extxrgb_h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    default:
+      h2v1_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                    output_buf);
+      break;
   }
 }
 
@@ -292,71 +365,35 @@ h2v2_merged_upsample (j_decompress_ptr cinfo,
 		      JSAMPIMAGE input_buf, JDIMENSION in_row_group_ctr,
 		      JSAMPARRAY output_buf)
 {
-  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
-  register int y, cred, cgreen, cblue;
-  int cb, cr;
-  register JSAMPROW outptr0, outptr1;
-  JSAMPROW inptr00, inptr01, inptr1, inptr2;
-  JDIMENSION col;
-  /* copy these pointers into registers if possible */
-  register JSAMPLE * range_limit = cinfo->sample_range_limit;
-  int * Crrtab = upsample->Cr_r_tab;
-  int * Cbbtab = upsample->Cb_b_tab;
-  INT32 * Crgtab = upsample->Cr_g_tab;
-  INT32 * Cbgtab = upsample->Cb_g_tab;
-  SHIFT_TEMPS
-
-  inptr00 = input_buf[0][in_row_group_ctr*2];
-  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
-  inptr1 = input_buf[1][in_row_group_ctr];
-  inptr2 = input_buf[2][in_row_group_ctr];
-  outptr0 = output_buf[0];
-  outptr1 = output_buf[1];
-  /* Loop for each group of output pixels */
-  for (col = cinfo->output_width >> 1; col > 0; col--) {
-    /* Do the chroma part of the calculation */
-    cb = GETJSAMPLE(*inptr1++);
-    cr = GETJSAMPLE(*inptr2++);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    /* Fetch 4 Y values and emit 4 pixels */
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr0[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr0[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr00++);
-    outptr0[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr0[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr0[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr0 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr1[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr1[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-    y  = GETJSAMPLE(*inptr01++);
-    outptr1[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr1[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr1[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    outptr1 += RGB_PIXELSIZE;
-  }
-  /* If image width is odd, do the last output column separately */
-  if (cinfo->output_width & 1) {
-    cb = GETJSAMPLE(*inptr1);
-    cr = GETJSAMPLE(*inptr2);
-    cred = Crrtab[cr];
-    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
-    cblue = Cbbtab[cb];
-    y  = GETJSAMPLE(*inptr00);
-    outptr0[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr0[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr0[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
-    y  = GETJSAMPLE(*inptr01);
-    outptr1[rgb_red[cinfo->out_color_space]] =   range_limit[y + cred];
-    outptr1[rgb_green[cinfo->out_color_space]] = range_limit[y + cgreen];
-    outptr1[rgb_blue[cinfo->out_color_space]] =  range_limit[y + cblue];
+  switch (cinfo->out_color_space) {
+    case JCS_EXT_RGB:
+      extrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                           output_buf);
+      break;
+    case JCS_EXT_RGBX:
+      extrgbx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_BGR:
+      extbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                           output_buf);
+      break;
+    case JCS_EXT_BGRX:
+      extbgrx_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_XBGR:
+      extxbgr_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    case JCS_EXT_XRGB:
+      extxrgb_h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                            output_buf);
+      break;
+    default:
+      h2v2_merged_upsample_internal(cinfo, input_buf, in_row_group_ctr,
+                                    output_buf);
+      break;
   }
 }
 
diff --git a/jdmrgext.c b/jdmrgext.c
new file mode 100644
index 0000000..95ddd55
--- /dev/null
+++ b/jdmrgext.c
@@ -0,0 +1,156 @@
+/*
+ * jdmrgext.c
+ *
+ * Copyright (C) 1994-1996, Thomas G. Lane.
+ * This file is part of the Independent JPEG Group's software.
+ * For conditions of distribution and use, see the accompanying README file.
+ *
+ * This file contains code for merged upsampling/color conversion.
+ */
+
+
+/* This file is included by jdmerge.c */
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 1:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v1_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr;
+  JSAMPROW inptr0, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr0 = input_buf[0][in_row_group_ctr];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr = output_buf[0];
+  /* Loop for each pair of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 2 Y values and emit 2 pixels */
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+    outptr += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr0++);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+    outptr += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr0);
+    outptr[RGB_RED] =   range_limit[y + cred];
+    outptr[RGB_GREEN] = range_limit[y + cgreen];
+    outptr[RGB_BLUE] =  range_limit[y + cblue];
+  }
+}
+
+
+/*
+ * Upsample and color convert for the case of 2:1 horizontal and 2:1 vertical.
+ */
+
+INLINE
+LOCAL(void)
+h2v2_merged_upsample_internal (j_decompress_ptr cinfo,
+                               JSAMPIMAGE input_buf,
+                               JDIMENSION in_row_group_ctr,
+                               JSAMPARRAY output_buf)
+{
+  my_upsample_ptr upsample = (my_upsample_ptr) cinfo->upsample;
+  register int y, cred, cgreen, cblue;
+  int cb, cr;
+  register JSAMPROW outptr0, outptr1;
+  JSAMPROW inptr00, inptr01, inptr1, inptr2;
+  JDIMENSION col;
+  /* copy these pointers into registers if possible */
+  register JSAMPLE * range_limit = cinfo->sample_range_limit;
+  int * Crrtab = upsample->Cr_r_tab;
+  int * Cbbtab = upsample->Cb_b_tab;
+  INT32 * Crgtab = upsample->Cr_g_tab;
+  INT32 * Cbgtab = upsample->Cb_g_tab;
+  SHIFT_TEMPS
+
+  inptr00 = input_buf[0][in_row_group_ctr*2];
+  inptr01 = input_buf[0][in_row_group_ctr*2 + 1];
+  inptr1 = input_buf[1][in_row_group_ctr];
+  inptr2 = input_buf[2][in_row_group_ctr];
+  outptr0 = output_buf[0];
+  outptr1 = output_buf[1];
+  /* Loop for each group of output pixels */
+  for (col = cinfo->output_width >> 1; col > 0; col--) {
+    /* Do the chroma part of the calculation */
+    cb = GETJSAMPLE(*inptr1++);
+    cr = GETJSAMPLE(*inptr2++);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    /* Fetch 4 Y values and emit 4 pixels */
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr00++);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    outptr0 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+    outptr1 += RGB_PIXELSIZE;
+    y  = GETJSAMPLE(*inptr01++);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+    outptr1 += RGB_PIXELSIZE;
+  }
+  /* If image width is odd, do the last output column separately */
+  if (cinfo->output_width & 1) {
+    cb = GETJSAMPLE(*inptr1);
+    cr = GETJSAMPLE(*inptr2);
+    cred = Crrtab[cr];
+    cgreen = (int) RIGHT_SHIFT(Cbgtab[cb] + Crgtab[cr], SCALEBITS);
+    cblue = Cbbtab[cb];
+    y  = GETJSAMPLE(*inptr00);
+    outptr0[RGB_RED] =   range_limit[y + cred];
+    outptr0[RGB_GREEN] = range_limit[y + cgreen];
+    outptr0[RGB_BLUE] =  range_limit[y + cblue];
+    y  = GETJSAMPLE(*inptr01);
+    outptr1[RGB_RED] =   range_limit[y + cred];
+    outptr1[RGB_GREEN] = range_limit[y + cgreen];
+    outptr1[RGB_BLUE] =  range_limit[y + cblue];
+  }
+}
diff --git a/jmorecfg.h b/jmorecfg.h
index c14a5f3..1448779 100644
--- a/jmorecfg.h
+++ b/jmorecfg.h
@@ -315,20 +315,54 @@ typedef int boolean;
 
 #define JPEG_NUMCS 12
 
+#define EXT_RGB_RED        0
+#define EXT_RGB_GREEN      1
+#define EXT_RGB_BLUE       2
+#define EXT_RGB_PIXELSIZE  3
+
+#define EXT_RGBX_RED       0
+#define EXT_RGBX_GREEN     1
+#define EXT_RGBX_BLUE      2
+#define EXT_RGBX_PIXELSIZE 4
+
+#define EXT_BGR_RED        2
+#define EXT_BGR_GREEN      1
+#define EXT_BGR_BLUE       0
+#define EXT_BGR_PIXELSIZE  3
+
+#define EXT_BGRX_RED       2
+#define EXT_BGRX_GREEN     1
+#define EXT_BGRX_BLUE      0
+#define EXT_BGRX_PIXELSIZE 4
+
+#define EXT_XBGR_RED       3
+#define EXT_XBGR_GREEN     2
+#define EXT_XBGR_BLUE      1
+#define EXT_XBGR_PIXELSIZE 4
+
+#define EXT_XRGB_RED       1
+#define EXT_XRGB_GREEN     2
+#define EXT_XRGB_BLUE      3
+#define EXT_XRGB_PIXELSIZE 4
+
 static const int rgb_red[JPEG_NUMCS] = {
-	-1, -1, RGB_RED, -1, -1, -1, 0, 0, 2, 2, 3, 1
+  -1, -1, RGB_RED, -1, -1, -1, EXT_RGB_RED, EXT_RGBX_RED,
+  EXT_BGR_RED, EXT_BGRX_RED, EXT_XBGR_RED, EXT_XRGB_RED
 };
 
 static const int rgb_green[JPEG_NUMCS] = {
-	-1, -1, RGB_GREEN, -1, -1, -1, 1, 1, 1, 1, 2, 2
+  -1, -1, RGB_GREEN, -1, -1, -1, EXT_RGB_GREEN, EXT_RGBX_GREEN,
+  EXT_BGR_GREEN, EXT_BGRX_GREEN, EXT_XBGR_GREEN, EXT_XRGB_GREEN
 };
 
 static const int rgb_blue[JPEG_NUMCS] = {
-	-1, -1, RGB_BLUE, -1, -1, -1, 2, 2, 0, 0, 1, 3
+  -1, -1, RGB_BLUE, -1, -1, -1, EXT_RGB_BLUE, EXT_RGBX_BLUE,
+  EXT_BGR_BLUE, EXT_BGRX_BLUE, EXT_XBGR_BLUE, EXT_XRGB_BLUE
 };
 
 static const int rgb_pixelsize[JPEG_NUMCS] = {
-	-1, -1, RGB_PIXELSIZE, -1, -1, -1, 3, 4, 3, 4, 4, 4
+  -1, -1, RGB_PIXELSIZE, -1, -1, -1, EXT_RGB_PIXELSIZE, EXT_RGBX_PIXELSIZE,
+  EXT_BGR_PIXELSIZE, EXT_BGRX_PIXELSIZE, EXT_XBGR_PIXELSIZE, EXT_XRGB_PIXELSIZE
 };
 
 /* Definitions for speed-related optimizations. */
diff --git a/simd/jccolmmx.asm b/simd/jccolmmx.asm
index d0f0d79..9650e47 100644
--- a/simd/jccolmmx.asm
+++ b/simd/jccolmmx.asm
@@ -60,10 +60,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extrgb_ycc_convert_mmx
 %include "jcclrmmx.asm"
 
@@ -71,10 +71,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extrgbx_ycc_convert_mmx
 %include "jcclrmmx.asm"
 
@@ -82,10 +82,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extbgr_ycc_convert_mmx
 %include "jcclrmmx.asm"
 
@@ -93,10 +93,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extbgrx_ycc_convert_mmx
 %include "jcclrmmx.asm"
 
@@ -104,10 +104,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extxbgr_ycc_convert_mmx
 %include "jcclrmmx.asm"
 
@@ -115,9 +115,9 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_mmx jsimd_extxrgb_ycc_convert_mmx
 %include "jcclrmmx.asm"
diff --git a/simd/jccolss2-64.asm b/simd/jccolss2-64.asm
index 18de456..ae60148 100644
--- a/simd/jccolss2-64.asm
+++ b/simd/jccolss2-64.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
 %include "jcclrss2-64.asm"
 
@@ -68,10 +68,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
 %include "jcclrss2-64.asm"
 
@@ -79,10 +79,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
 %include "jcclrss2-64.asm"
 
@@ -90,10 +90,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
 %include "jcclrss2-64.asm"
 
@@ -101,10 +101,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
 %include "jcclrss2-64.asm"
 
@@ -112,9 +112,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
 %include "jcclrss2-64.asm"
diff --git a/simd/jccolss2.asm b/simd/jccolss2.asm
index 7acb59c..ac001d1 100644
--- a/simd/jccolss2.asm
+++ b/simd/jccolss2.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extrgb_ycc_convert_sse2
 %include "jcclrss2.asm"
 
@@ -68,10 +68,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extrgbx_ycc_convert_sse2
 %include "jcclrss2.asm"
 
@@ -79,10 +79,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extbgr_ycc_convert_sse2
 %include "jcclrss2.asm"
 
@@ -90,10 +90,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extbgrx_ycc_convert_sse2
 %include "jcclrss2.asm"
 
@@ -101,10 +101,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extxbgr_ycc_convert_sse2
 %include "jcclrss2.asm"
 
@@ -112,9 +112,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_ycc_convert_sse2 jsimd_extxrgb_ycc_convert_sse2
 %include "jcclrss2.asm"
diff --git a/simd/jcgrammx.asm b/simd/jcgrammx.asm
index 8c96824..b8b8dd3 100644
--- a/simd/jcgrammx.asm
+++ b/simd/jcgrammx.asm
@@ -53,10 +53,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extrgb_gray_convert_mmx
 %include "jcgrymmx.asm"
 
@@ -64,10 +64,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extrgbx_gray_convert_mmx
 %include "jcgrymmx.asm"
 
@@ -75,10 +75,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extbgr_gray_convert_mmx
 %include "jcgrymmx.asm"
 
@@ -86,10 +86,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extbgrx_gray_convert_mmx
 %include "jcgrymmx.asm"
 
@@ -97,10 +97,10 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extxbgr_gray_convert_mmx
 %include "jcgrymmx.asm"
 
@@ -108,9 +108,9 @@ PD_ONEHALF	times 2 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_mmx jsimd_extxrgb_gray_convert_mmx
 %include "jcgrymmx.asm"
diff --git a/simd/jcgrass2-64.asm b/simd/jcgrass2-64.asm
index 232f329..ba28cc3 100644
--- a/simd/jcgrass2-64.asm
+++ b/simd/jcgrass2-64.asm
@@ -50,10 +50,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
 %include "jcgryss2-64.asm"
 
@@ -61,10 +61,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
 %include "jcgryss2-64.asm"
 
@@ -72,10 +72,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
 %include "jcgryss2-64.asm"
 
@@ -83,10 +83,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
 %include "jcgryss2-64.asm"
 
@@ -94,10 +94,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
 %include "jcgryss2-64.asm"
 
@@ -105,9 +105,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
 %include "jcgryss2-64.asm"
diff --git a/simd/jcgrass2.asm b/simd/jcgrass2.asm
index 45ea33a..998968e 100644
--- a/simd/jcgrass2.asm
+++ b/simd/jcgrass2.asm
@@ -50,10 +50,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extrgb_gray_convert_sse2
 %include "jcgryss2.asm"
 
@@ -61,10 +61,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extrgbx_gray_convert_sse2
 %include "jcgryss2.asm"
 
@@ -72,10 +72,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extbgr_gray_convert_sse2
 %include "jcgryss2.asm"
 
@@ -83,10 +83,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extbgrx_gray_convert_sse2
 %include "jcgryss2.asm"
 
@@ -94,10 +94,10 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extxbgr_gray_convert_sse2
 %include "jcgryss2.asm"
 
@@ -105,9 +105,9 @@ PD_ONEHALF	times 4 dd  (1 << (SCALEBITS-1))
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_rgb_gray_convert_sse2 jsimd_extxrgb_gray_convert_sse2
 %include "jcgryss2.asm"
diff --git a/simd/jdcolmmx.asm b/simd/jdcolmmx.asm
index 0834bab..5e4e47d 100644
--- a/simd/jdcolmmx.asm
+++ b/simd/jdcolmmx.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgb_convert_mmx
 %include "jdclrmmx.asm"
 
@@ -68,10 +68,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extrgbx_convert_mmx
 %include "jdclrmmx.asm"
 
@@ -79,10 +79,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgr_convert_mmx
 %include "jdclrmmx.asm"
 
@@ -90,10 +90,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extbgrx_convert_mmx
 %include "jdclrmmx.asm"
 
@@ -101,10 +101,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxbgr_convert_mmx
 %include "jdclrmmx.asm"
 
@@ -112,9 +112,9 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_mmx jsimd_ycc_extxrgb_convert_mmx
 %include "jdclrmmx.asm"
diff --git a/simd/jdcolss2-64.asm b/simd/jdcolss2-64.asm
index d14a28a..01b3dce 100644
--- a/simd/jdcolss2-64.asm
+++ b/simd/jdcolss2-64.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
 %include "jdclrss2-64.asm"
 
@@ -68,10 +68,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
 %include "jdclrss2-64.asm"
 
@@ -79,10 +79,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
 %include "jdclrss2-64.asm"
 
@@ -90,10 +90,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
 %include "jdclrss2-64.asm"
 
@@ -101,10 +101,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
 %include "jdclrss2-64.asm"
 
@@ -112,9 +112,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
 %include "jdclrss2-64.asm"
diff --git a/simd/jdcolss2.asm b/simd/jdcolss2.asm
index cab4dd0..1912d92 100644
--- a/simd/jdcolss2.asm
+++ b/simd/jdcolss2.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgb_convert_sse2
 %include "jdclrss2.asm"
 
@@ -68,10 +68,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extrgbx_convert_sse2
 %include "jdclrss2.asm"
 
@@ -79,10 +79,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgr_convert_sse2
 %include "jdclrss2.asm"
 
@@ -90,10 +90,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extbgrx_convert_sse2
 %include "jdclrss2.asm"
 
@@ -101,10 +101,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxbgr_convert_sse2
 %include "jdclrss2.asm"
 
@@ -112,9 +112,9 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_ycc_rgb_convert_sse2 jsimd_ycc_extxrgb_convert_sse2
 %include "jdclrss2.asm"
diff --git a/simd/jdmermmx.asm b/simd/jdmermmx.asm
index 75baaa5..7b86c74 100644
--- a/simd/jdmermmx.asm
+++ b/simd/jdmermmx.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgb_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgb_merged_upsample_mmx
 %include "jdmrgmmx.asm"
@@ -69,10 +69,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extrgbx_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extrgbx_merged_upsample_mmx
 %include "jdmrgmmx.asm"
@@ -81,10 +81,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgr_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgr_merged_upsample_mmx
 %include "jdmrgmmx.asm"
@@ -93,10 +93,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extbgrx_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extbgrx_merged_upsample_mmx
 %include "jdmrgmmx.asm"
@@ -105,10 +105,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxbgr_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxbgr_merged_upsample_mmx
 %include "jdmrgmmx.asm"
@@ -117,10 +117,10 @@ PD_ONEHALF	times 2 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_mmx jsimd_h2v1_extxrgb_merged_upsample_mmx
 %define jsimd_h2v2_merged_upsample_mmx jsimd_h2v2_extxrgb_merged_upsample_mmx
 %include "jdmrgmmx.asm"
diff --git a/simd/jdmerss2-64.asm b/simd/jdmerss2-64.asm
index a1fe963..a184ea6 100644
--- a/simd/jdmerss2-64.asm
+++ b/simd/jdmerss2-64.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
@@ -69,10 +69,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
@@ -81,10 +81,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
@@ -93,10 +93,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
@@ -105,10 +105,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
@@ -117,10 +117,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
 %include "jdmrgss2-64.asm"
diff --git a/simd/jdmerss2.asm b/simd/jdmerss2.asm
index 21881b4..e536c80 100644
--- a/simd/jdmerss2.asm
+++ b/simd/jdmerss2.asm
@@ -57,10 +57,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_RGB_RED
+%define RGB_GREEN EXT_RGB_GREEN
+%define RGB_BLUE EXT_RGB_BLUE
+%define RGB_PIXELSIZE EXT_RGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgb_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgb_merged_upsample_sse2
 %include "jdmrgss2.asm"
@@ -69,10 +69,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 0
-%define RGB_GREEN 1
-%define RGB_BLUE 2
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_RGBX_RED
+%define RGB_GREEN EXT_RGBX_GREEN
+%define RGB_BLUE EXT_RGBX_BLUE
+%define RGB_PIXELSIZE EXT_RGBX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extrgbx_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extrgbx_merged_upsample_sse2
 %include "jdmrgss2.asm"
@@ -81,10 +81,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 3
+%define RGB_RED EXT_BGR_RED
+%define RGB_GREEN EXT_BGR_GREEN
+%define RGB_BLUE EXT_BGR_BLUE
+%define RGB_PIXELSIZE EXT_BGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgr_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgr_merged_upsample_sse2
 %include "jdmrgss2.asm"
@@ -93,10 +93,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 2
-%define RGB_GREEN 1
-%define RGB_BLUE 0
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_BGRX_RED
+%define RGB_GREEN EXT_BGRX_GREEN
+%define RGB_BLUE EXT_BGRX_BLUE
+%define RGB_PIXELSIZE EXT_BGRX_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extbgrx_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extbgrx_merged_upsample_sse2
 %include "jdmrgss2.asm"
@@ -105,10 +105,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 3
-%define RGB_GREEN 2
-%define RGB_BLUE 1
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XBGR_RED
+%define RGB_GREEN EXT_XBGR_GREEN
+%define RGB_BLUE EXT_XBGR_BLUE
+%define RGB_PIXELSIZE EXT_XBGR_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxbgr_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxbgr_merged_upsample_sse2
 %include "jdmrgss2.asm"
@@ -117,10 +117,10 @@ PD_ONEHALF	times 4 dd  1 << (SCALEBITS-1)
 %undef RGB_GREEN
 %undef RGB_BLUE
 %undef RGB_PIXELSIZE
-%define RGB_RED 1
-%define RGB_GREEN 2
-%define RGB_BLUE 3
-%define RGB_PIXELSIZE 4
+%define RGB_RED EXT_XRGB_RED
+%define RGB_GREEN EXT_XRGB_GREEN
+%define RGB_BLUE EXT_XRGB_BLUE
+%define RGB_PIXELSIZE EXT_XRGB_PIXELSIZE
 %define jsimd_h2v1_merged_upsample_sse2 jsimd_h2v1_extxrgb_merged_upsample_sse2
 %define jsimd_h2v2_merged_upsample_sse2 jsimd_h2v2_extxrgb_merged_upsample_sse2
 %include "jdmrgss2.asm"
diff --git a/simd/jsimdcfg.inc.h b/simd/jsimdcfg.inc.h
index 4876038..13cff03 100644
--- a/simd/jsimdcfg.inc.h
+++ b/simd/jsimdcfg.inc.h
@@ -32,9 +32,38 @@ definev(DCTSIZE2)
 definev(RGB_RED)
 definev(RGB_GREEN)
 definev(RGB_BLUE)
-
 definev(RGB_PIXELSIZE)
 
+definev(EXT_RGB_RED)
+definev(EXT_RGB_GREEN)
+definev(EXT_RGB_BLUE)
+definev(EXT_RGB_PIXELSIZE)
+
+definev(EXT_RGBX_RED)
+definev(EXT_RGBX_GREEN)
+definev(EXT_RGBX_BLUE)
+definev(EXT_RGBX_PIXELSIZE)
+
+definev(EXT_BGR_RED)
+definev(EXT_BGR_GREEN)
+definev(EXT_BGR_BLUE)
+definev(EXT_BGR_PIXELSIZE)
+
+definev(EXT_BGRX_RED)
+definev(EXT_BGRX_GREEN)
+definev(EXT_BGRX_BLUE)
+definev(EXT_BGRX_PIXELSIZE)
+
+definev(EXT_XBGR_RED)
+definev(EXT_XBGR_GREEN)
+definev(EXT_XBGR_BLUE)
+definev(EXT_XBGR_PIXELSIZE)
+
+definev(EXT_XRGB_RED)
+definev(EXT_XRGB_GREEN)
+definev(EXT_XRGB_BLUE)
+definev(EXT_XRGB_PIXELSIZE)
+
 ; Representation of a single sample (pixel element value).
 ; On this SIMD implementation, this must be 'unsigned char'.
 ;
diff --git a/win/jconfig.h.in b/win/jconfig.h.in
index 3b835ea..373976d 100644
--- a/win/jconfig.h.in
+++ b/win/jconfig.h.in
@@ -32,7 +32,15 @@ typedef signed int INT32;
 #endif
 #define XMD_H                   /* prevent jmorecfg.h from redefining it */
 
-#define inline __inline
+#ifndef INLINE
+#if defined(__GNUC__)
+#define INLINE __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#define INLINE __forceinline
+#else
+#define INLINE
+#endif
+#endif
 
 #ifdef JPEG_INTERNALS
 
diff --git a/win/jsimdcfg.inc b/win/jsimdcfg.inc
index 68e22e8..a852849 100755
--- a/win/jsimdcfg.inc
+++ b/win/jsimdcfg.inc
@@ -13,6 +13,30 @@
 %define RGB_GREEN 1
 %define RGB_BLUE 2
 %define RGB_PIXELSIZE 3
+%define EXT_RGB_RED 0
+%define EXT_RGB_GREEN 1
+%define EXT_RGB_BLUE 2
+%define EXT_RGB_PIXELSIZE 3
+%define EXT_RGBX_RED 0
+%define EXT_RGBX_GREEN 1
+%define EXT_RGBX_BLUE 2
+%define EXT_RGBX_PIXELSIZE 4
+%define EXT_BGR_RED 2
+%define EXT_BGR_GREEN 1
+%define EXT_BGR_BLUE 0
+%define EXT_BGR_PIXELSIZE 3
+%define EXT_BGRX_RED 2
+%define EXT_BGRX_GREEN 1
+%define EXT_BGRX_BLUE 0
+%define EXT_BGRX_PIXELSIZE 4
+%define EXT_XBGR_RED 3
+%define EXT_XBGR_GREEN 2
+%define EXT_XBGR_BLUE 1
+%define EXT_XBGR_PIXELSIZE 4
+%define EXT_XRGB_RED 1
+%define EXT_XRGB_GREEN 2
+%define EXT_XRGB_BLUE 3
+%define EXT_XRGB_PIXELSIZE 4
 ; Representation of a single sample (pixel element value).
 ; On this SIMD implementation, this must be 'unsigned char'.
 ;