gallium: add a generic vertex (or other) buffer translation module
authorKeith Whitwell <keith@tungstengraphics.com>
Tue, 15 Apr 2008 13:35:29 +0000 (14:35 +0100)
committerKeith Whitwell <keith@tungstengraphics.com>
Tue, 15 Apr 2008 13:35:29 +0000 (14:35 +0100)
src/gallium/auxiliary/translate/Makefile [new file with mode: 0644]
src/gallium/auxiliary/translate/SConscript [new file with mode: 0644]
src/gallium/auxiliary/translate/translate.h [new file with mode: 0644]
src/gallium/auxiliary/translate/translate_generic.c [new file with mode: 0644]

diff --git a/src/gallium/auxiliary/translate/Makefile b/src/gallium/auxiliary/translate/Makefile
new file mode 100644 (file)
index 0000000..051987b
--- /dev/null
@@ -0,0 +1,12 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = translate
+
+C_SOURCES = \
+       translate_generic.c
+
+include ../../Makefile.template
+
+symlinks:
+
diff --git a/src/gallium/auxiliary/translate/SConscript b/src/gallium/auxiliary/translate/SConscript
new file mode 100644 (file)
index 0000000..947b943
--- /dev/null
@@ -0,0 +1,9 @@
+Import('*')
+
+cso_cache = env.ConvenienceLibrary(
+       target = 'translate',
+       source = [
+               'translate_generic.c',
+       ])
+
+auxiliaries.insert(0, translate)
diff --git a/src/gallium/auxiliary/translate/translate.h b/src/gallium/auxiliary/translate/translate.h
new file mode 100644 (file)
index 0000000..ac6bc70
--- /dev/null
@@ -0,0 +1,82 @@
+/*
+ * Copyright 2008 Tungsten Graphics, inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * on the rights to use, copy, modify, merge, publish, distribute, sub
+ * license, and/or sell copies of the Software, and to permit persons to whom
+ * the Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.  IN NO EVENT SHALL
+ * TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ * DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR
+ * OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE
+ * USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+
+/**
+ * Vertex fetch/store/convert code.  This functionality is used in two places:
+ * 1. Vertex fetch/convert - to grab vertex data from incoming vertex
+ *    arrays and convert to format needed by vertex shaders.
+ * 2. Vertex store/emit - to convert simple float[][4] vertex attributes
+ *    (which is the organization used throughout the draw/prim pipeline) to
+ *    hardware-specific formats and emit into hardware vertex buffers.
+ *
+ *
+ * Authors:
+ *    Keith Whitwell <keithw@tungstengraphics.com>
+ */
+
+#ifndef _TRANSLATE_H
+#define _TRANSLATE_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_format.h"
+
+struct translate_element 
+{
+   enum pipe_format input_format;
+   unsigned input_buffer;
+   unsigned input_offset;
+
+   enum pipe_format output_format;
+   unsigned output_offset;
+};
+
+
+struct translate {
+   void (*destroy)( struct translate * );
+
+   void (*set_buffer)( struct translate *,
+                      unsigned i,
+                      const void *ptr,
+                      unsigned stride );
+
+   void (*run_elts)( struct translate *,
+                    const unsigned *elts,
+                    unsigned count,
+                    void *output_buffer);
+};
+
+
+
+struct translate *translate_sse2_create( unsigned output_stride,
+                                        const struct translate_element *elements,
+                                        unsigned nr_elements );
+
+struct translate *translate_generic_create( unsigned output_stride,
+                                           const struct translate_element *elements,
+                                           unsigned nr_elements );
+
+
+#endif
diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c
new file mode 100644 (file)
index 0000000..643f643
--- /dev/null
@@ -0,0 +1,630 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  */
+
+#include "pipe/p_util.h"
+#include "pipe/p_state.h"
+#include "translate.h"
+
+
+#define DRAW_DBG 0
+
+typedef void (*fetch_func)(const void *ptr, float *attrib);
+typedef void (*emit_func)(const float *attrib, void *ptr);
+
+
+
+struct translate_generic {
+   struct translate translate;
+
+   struct {
+      fetch_func fetch;
+      unsigned buffer;
+      unsigned input_offset;
+
+      emit_func emit;
+      unsigned output_offset;
+      
+      char *input_ptr;
+      unsigned input_stride;
+
+   } attrib[PIPE_MAX_ATTRIBS];
+
+   unsigned nr_attrib;
+   unsigned output_stride;
+};
+
+
+static struct translate_generic *translate_generic( struct translate *translate )
+{
+   return (struct translate_generic *)translate;
+}
+
+/**
+ * Fetch a float[4] vertex attribute from memory, doing format/type
+ * conversion as needed.
+ *
+ * This is probably needed/dupliocated elsewhere, eg format
+ * conversion, texture sampling etc.
+ */
+#define ATTRIB( NAME, SZ, TYPE, FROM, TO )             \
+static void                                            \
+fetch_##NAME(const void *ptr, float *attrib)           \
+{                                                      \
+   const float defaults[4] = { 0,0,0,1 };      \
+   int i;                                              \
+                                                       \
+   for (i = 0; i < SZ; i++) {                          \
+      attrib[i] = FROM(i);                             \
+   }                                                   \
+                                                       \
+   for (; i < 4; i++) {                                        \
+      attrib[i] = defaults[i];                         \
+   }                                                   \
+}                                                      \
+                                                       \
+static void                                            \
+emit_##NAME(const float *attrib, void *ptr)            \
+{  \
+   unsigned i;                                         \
+   TYPE *out = (TYPE *)ptr;                            \
+                                                       \
+   for (i = 0; i < SZ; i++) {                          \
+      out[i] = TO(attrib[i]);                          \
+   }                                                   \
+}
+
+
+#define FROM_64_FLOAT(i)   ((float) ((double *) ptr)[i])
+#define FROM_32_FLOAT(i)   (((float *) ptr)[i])
+
+#define FROM_8_USCALED(i)  ((float) ((unsigned char *) ptr)[i])
+#define FROM_16_USCALED(i) ((float) ((unsigned short *) ptr)[i])
+#define FROM_32_USCALED(i) ((float) ((unsigned int *) ptr)[i])
+
+#define FROM_8_SSCALED(i)  ((float) ((char *) ptr)[i])
+#define FROM_16_SSCALED(i) ((float) ((short *) ptr)[i])
+#define FROM_32_SSCALED(i) ((float) ((int *) ptr)[i])
+
+#define FROM_8_UNORM(i)    ((float) ((unsigned char *) ptr)[i] / 255.0f)
+#define FROM_16_UNORM(i)   ((float) ((unsigned short *) ptr)[i] / 65535.0f)
+#define FROM_32_UNORM(i)   ((float) ((unsigned int *) ptr)[i] / 4294967295.0f)
+
+#define FROM_8_SNORM(i)    ((float) ((char *) ptr)[i] / 127.0f)
+#define FROM_16_SNORM(i)   ((float) ((short *) ptr)[i] / 32767.0f)
+#define FROM_32_SNORM(i)   ((float) ((int *) ptr)[i] / 2147483647.0f)
+
+#define TO_64_FLOAT(f)   ((double) f)
+#define TO_32_FLOAT(f)   (f)
+
+#define TO_8_USCALED(f)  ((unsigned char) f)
+#define TO_16_USCALED(f) ((unsigned short) f)
+#define TO_32_USCALED(f) ((unsigned int) f)
+
+#define TO_8_SSCALED(f)  ((char) f)
+#define TO_16_SSCALED(f) ((short) f)
+#define TO_32_SSCALED(f) ((int) f)
+
+#define TO_8_UNORM(f)    ((unsigned char) (f * 255.0f))
+#define TO_16_UNORM(f)   ((unsigned short) (f * 65535.0f))
+#define TO_32_UNORM(f)   ((unsigned int) (f * 4294967295.0f))
+
+#define TO_8_SNORM(f)    ((char) (f * 127.0f))
+#define TO_16_SNORM(f)   ((short) (f * 32767.0f))
+#define TO_32_SNORM(f)   ((int) (f * 2147483647.0f))
+
+
+
+ATTRIB( R64G64B64A64_FLOAT,   4, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64G64B64_FLOAT,      3, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64G64_FLOAT,         2, double, FROM_64_FLOAT, TO_64_FLOAT )
+ATTRIB( R64_FLOAT,            1, double, FROM_64_FLOAT, TO_64_FLOAT )
+
+ATTRIB( R32G32B32A32_FLOAT,   4, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32G32B32_FLOAT,      3, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32G32_FLOAT,         2, float, FROM_32_FLOAT, TO_32_FLOAT )
+ATTRIB( R32_FLOAT,            1, float, FROM_32_FLOAT, TO_32_FLOAT )
+
+ATTRIB( R32G32B32A32_USCALED, 4, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32G32B32_USCALED,    3, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32G32_USCALED,       2, unsigned, FROM_32_USCALED, TO_32_USCALED )
+ATTRIB( R32_USCALED,          1, unsigned, FROM_32_USCALED, TO_32_USCALED )
+
+ATTRIB( R32G32B32A32_SSCALED, 4, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32G32B32_SSCALED,    3, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32G32_SSCALED,       2, int, FROM_32_SSCALED, TO_32_SSCALED )
+ATTRIB( R32_SSCALED,          1, int, FROM_32_SSCALED, TO_32_SSCALED )
+
+ATTRIB( R32G32B32A32_UNORM, 4, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32G32B32_UNORM,    3, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32G32_UNORM,       2, unsigned, FROM_32_UNORM, TO_32_UNORM )
+ATTRIB( R32_UNORM,          1, unsigned, FROM_32_UNORM, TO_32_UNORM )
+
+ATTRIB( R32G32B32A32_SNORM, 4, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32G32B32_SNORM,    3, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32G32_SNORM,       2, int, FROM_32_SNORM, TO_32_SNORM )
+ATTRIB( R32_SNORM,          1, int, FROM_32_SNORM, TO_32_SNORM )
+
+ATTRIB( R16G16B16A16_USCALED, 4, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16G16B16_USCALED,    3, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16G16_USCALED,       2, ushort, FROM_16_USCALED, TO_16_USCALED )
+ATTRIB( R16_USCALED,          1, ushort, FROM_16_USCALED, TO_16_USCALED )
+
+ATTRIB( R16G16B16A16_SSCALED, 4, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16G16B16_SSCALED,    3, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16G16_SSCALED,       2, short, FROM_16_SSCALED, TO_16_SSCALED )
+ATTRIB( R16_SSCALED,          1, short, FROM_16_SSCALED, TO_16_SSCALED )
+
+ATTRIB( R16G16B16A16_UNORM, 4, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16G16B16_UNORM,    3, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16G16_UNORM,       2, ushort, FROM_16_UNORM, TO_16_UNORM )
+ATTRIB( R16_UNORM,          1, ushort, FROM_16_UNORM, TO_16_UNORM )
+
+ATTRIB( R16G16B16A16_SNORM, 4, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16G16B16_SNORM,    3, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16G16_SNORM,       2, short, FROM_16_SNORM, TO_16_SNORM )
+ATTRIB( R16_SNORM,          1, short, FROM_16_SNORM, TO_16_SNORM )
+
+ATTRIB( R8G8B8A8_USCALED,   4, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8G8B8_USCALED,     3, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8G8_USCALED,       2, ubyte, FROM_8_USCALED, TO_8_USCALED )
+ATTRIB( R8_USCALED,         1, ubyte, FROM_8_USCALED, TO_8_USCALED )
+
+ATTRIB( R8G8B8A8_SSCALED,  4, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8G8B8_SSCALED,    3, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8G8_SSCALED,      2, char, FROM_8_SSCALED, TO_8_SSCALED )
+ATTRIB( R8_SSCALED,        1, char, FROM_8_SSCALED, TO_8_SSCALED )
+
+ATTRIB( R8G8B8A8_UNORM,  4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8G8B8_UNORM,    3, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8G8_UNORM,      2, ubyte, FROM_8_UNORM, TO_8_UNORM )
+ATTRIB( R8_UNORM,        1, ubyte, FROM_8_UNORM, TO_8_UNORM )
+
+ATTRIB( R8G8B8A8_SNORM,  4, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8G8B8_SNORM,    3, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8G8_SNORM,      2, char, FROM_8_SNORM, TO_8_SNORM )
+ATTRIB( R8_SNORM,        1, char, FROM_8_SNORM, TO_8_SNORM )
+
+ATTRIB( A8R8G8B8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+//ATTRIB( R8G8B8A8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )
+
+
+
+static void
+fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib)
+{
+   attrib[2] = FROM_8_UNORM(0);
+   attrib[1] = FROM_8_UNORM(1);
+   attrib[0] = FROM_8_UNORM(2);
+   attrib[3] = FROM_8_UNORM(3);
+}
+
+static void
+emit_B8G8R8A8_UNORM( const float *attrib, void *ptr)
+{
+   ubyte *out = (ubyte *)ptr;
+   out[2] = TO_8_UNORM(out[0]);
+   out[1] = TO_8_UNORM(out[1]);
+   out[0] = TO_8_UNORM(out[2]);
+   out[3] = TO_8_UNORM(out[3]);
+}
+
+static void 
+fetch_NULL( const void *ptr, float *attrib )
+{
+   attrib[0] = 0;
+   attrib[1] = 0;
+   attrib[2] = 0;
+   attrib[3] = 1;
+}
+
+static void 
+emit_NULL( const float *attrib, void *ptr )
+{
+   /* do nothing is the only sensible option */
+}
+
+static fetch_func get_fetch_func( enum pipe_format format )
+{
+   switch (format) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return fetch_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return fetch_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return fetch_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return fetch_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return fetch_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return fetch_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return fetch_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return fetch_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return fetch_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return fetch_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return fetch_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return fetch_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return fetch_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return fetch_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return fetch_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return fetch_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return fetch_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return fetch_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return fetch_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return fetch_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return fetch_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return fetch_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return fetch_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return fetch_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return fetch_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return fetch_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return fetch_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return fetch_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return fetch_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return fetch_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return fetch_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return fetch_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return fetch_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return fetch_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return fetch_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return fetch_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return fetch_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return fetch_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return fetch_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return fetch_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return fetch_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return fetch_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return fetch_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return fetch_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return fetch_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return fetch_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return fetch_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return fetch_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return fetch_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return fetch_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return fetch_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return fetch_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return fetch_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return fetch_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return fetch_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return fetch_R8G8B8A8_SSCALED;
+
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return fetch_A8R8G8B8_UNORM;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return fetch_B8G8R8A8_UNORM;
+
+   default:
+      assert(0); 
+      return fetch_NULL;
+   }
+}
+
+
+
+
+static emit_func get_emit_func( enum pipe_format format )
+{
+   switch (format) {
+   case PIPE_FORMAT_R64_FLOAT:
+      return emit_R64_FLOAT;
+   case PIPE_FORMAT_R64G64_FLOAT:
+      return emit_R64G64_FLOAT;
+   case PIPE_FORMAT_R64G64B64_FLOAT:
+      return emit_R64G64B64_FLOAT;
+   case PIPE_FORMAT_R64G64B64A64_FLOAT:
+      return emit_R64G64B64A64_FLOAT;
+
+   case PIPE_FORMAT_R32_FLOAT:
+      return emit_R32_FLOAT;
+   case PIPE_FORMAT_R32G32_FLOAT:
+      return emit_R32G32_FLOAT;
+   case PIPE_FORMAT_R32G32B32_FLOAT:
+      return emit_R32G32B32_FLOAT;
+   case PIPE_FORMAT_R32G32B32A32_FLOAT:
+      return emit_R32G32B32A32_FLOAT;
+
+   case PIPE_FORMAT_R32_UNORM:
+      return emit_R32_UNORM;
+   case PIPE_FORMAT_R32G32_UNORM:
+      return emit_R32G32_UNORM;
+   case PIPE_FORMAT_R32G32B32_UNORM:
+      return emit_R32G32B32_UNORM;
+   case PIPE_FORMAT_R32G32B32A32_UNORM:
+      return emit_R32G32B32A32_UNORM;
+
+   case PIPE_FORMAT_R32_USCALED:
+      return emit_R32_USCALED;
+   case PIPE_FORMAT_R32G32_USCALED:
+      return emit_R32G32_USCALED;
+   case PIPE_FORMAT_R32G32B32_USCALED:
+      return emit_R32G32B32_USCALED;
+   case PIPE_FORMAT_R32G32B32A32_USCALED:
+      return emit_R32G32B32A32_USCALED;
+
+   case PIPE_FORMAT_R32_SNORM:
+      return emit_R32_SNORM;
+   case PIPE_FORMAT_R32G32_SNORM:
+      return emit_R32G32_SNORM;
+   case PIPE_FORMAT_R32G32B32_SNORM:
+      return emit_R32G32B32_SNORM;
+   case PIPE_FORMAT_R32G32B32A32_SNORM:
+      return emit_R32G32B32A32_SNORM;
+
+   case PIPE_FORMAT_R32_SSCALED:
+      return emit_R32_SSCALED;
+   case PIPE_FORMAT_R32G32_SSCALED:
+      return emit_R32G32_SSCALED;
+   case PIPE_FORMAT_R32G32B32_SSCALED:
+      return emit_R32G32B32_SSCALED;
+   case PIPE_FORMAT_R32G32B32A32_SSCALED:
+      return emit_R32G32B32A32_SSCALED;
+
+   case PIPE_FORMAT_R16_UNORM:
+      return emit_R16_UNORM;
+   case PIPE_FORMAT_R16G16_UNORM:
+      return emit_R16G16_UNORM;
+   case PIPE_FORMAT_R16G16B16_UNORM:
+      return emit_R16G16B16_UNORM;
+   case PIPE_FORMAT_R16G16B16A16_UNORM:
+      return emit_R16G16B16A16_UNORM;
+
+   case PIPE_FORMAT_R16_USCALED:
+      return emit_R16_USCALED;
+   case PIPE_FORMAT_R16G16_USCALED:
+      return emit_R16G16_USCALED;
+   case PIPE_FORMAT_R16G16B16_USCALED:
+      return emit_R16G16B16_USCALED;
+   case PIPE_FORMAT_R16G16B16A16_USCALED:
+      return emit_R16G16B16A16_USCALED;
+
+   case PIPE_FORMAT_R16_SNORM:
+      return emit_R16_SNORM;
+   case PIPE_FORMAT_R16G16_SNORM:
+      return emit_R16G16_SNORM;
+   case PIPE_FORMAT_R16G16B16_SNORM:
+      return emit_R16G16B16_SNORM;
+   case PIPE_FORMAT_R16G16B16A16_SNORM:
+      return emit_R16G16B16A16_SNORM;
+
+   case PIPE_FORMAT_R16_SSCALED:
+      return emit_R16_SSCALED;
+   case PIPE_FORMAT_R16G16_SSCALED:
+      return emit_R16G16_SSCALED;
+   case PIPE_FORMAT_R16G16B16_SSCALED:
+      return emit_R16G16B16_SSCALED;
+   case PIPE_FORMAT_R16G16B16A16_SSCALED:
+      return emit_R16G16B16A16_SSCALED;
+
+   case PIPE_FORMAT_R8_UNORM:
+      return emit_R8_UNORM;
+   case PIPE_FORMAT_R8G8_UNORM:
+      return emit_R8G8_UNORM;
+   case PIPE_FORMAT_R8G8B8_UNORM:
+      return emit_R8G8B8_UNORM;
+   case PIPE_FORMAT_R8G8B8A8_UNORM:
+      return emit_R8G8B8A8_UNORM;
+
+   case PIPE_FORMAT_R8_USCALED:
+      return emit_R8_USCALED;
+   case PIPE_FORMAT_R8G8_USCALED:
+      return emit_R8G8_USCALED;
+   case PIPE_FORMAT_R8G8B8_USCALED:
+      return emit_R8G8B8_USCALED;
+   case PIPE_FORMAT_R8G8B8A8_USCALED:
+      return emit_R8G8B8A8_USCALED;
+
+   case PIPE_FORMAT_R8_SNORM:
+      return emit_R8_SNORM;
+   case PIPE_FORMAT_R8G8_SNORM:
+      return emit_R8G8_SNORM;
+   case PIPE_FORMAT_R8G8B8_SNORM:
+      return emit_R8G8B8_SNORM;
+   case PIPE_FORMAT_R8G8B8A8_SNORM:
+      return emit_R8G8B8A8_SNORM;
+
+   case PIPE_FORMAT_R8_SSCALED:
+      return emit_R8_SSCALED;
+   case PIPE_FORMAT_R8G8_SSCALED:
+      return emit_R8G8_SSCALED;
+   case PIPE_FORMAT_R8G8B8_SSCALED:
+      return emit_R8G8B8_SSCALED;
+   case PIPE_FORMAT_R8G8B8A8_SSCALED:
+      return emit_R8G8B8A8_SSCALED;
+
+   case PIPE_FORMAT_A8R8G8B8_UNORM:
+      return emit_A8R8G8B8_UNORM;
+
+   case PIPE_FORMAT_B8G8R8A8_UNORM:
+      return emit_B8G8R8A8_UNORM;
+
+   default:
+      assert(0); 
+      return emit_NULL;
+   }
+}
+
+
+
+/**
+ * Fetch vertex attributes for 'count' vertices.
+ */
+static void generic_run_elts( struct translate *translate,
+                             const unsigned *elts,
+                             unsigned count,
+                             void *output_buffer )
+{
+   struct translate_generic *tg = translate_generic(translate);
+   char *vert = output_buffer;
+   unsigned nr_attrs = tg->nr_attrib;
+   unsigned attr;
+   unsigned i;
+
+   /* loop over vertex attributes (vertex shader inputs)
+    */
+   for (i = 0; i < count; i++) {
+      unsigned elt = *elts++;
+
+      for (attr = 0; attr < nr_attrs; attr++) {
+        float data[4];
+
+        const char *src = (tg->attrib[attr].input_ptr + 
+                           tg->attrib[attr].input_stride * elt);
+
+        char *dst = (vert + 
+                     tg->attrib[attr].output_offset);
+
+        tg->attrib[attr].fetch( src, data );
+        tg->attrib[attr].emit( data, dst );
+      }
+      
+      vert += tg->output_stride;
+   }
+}
+
+
+                              
+static void generic_set_buffer( struct translate *translate,
+                               unsigned buf,
+                               const void *ptr,
+                               unsigned stride )
+{
+   struct translate_generic *tg = translate_generic(translate);
+   unsigned i;
+
+   for (i = 0; i < tg->nr_attrib; i++) {
+      if (tg->attrib[i].buffer == buf) {
+        tg->attrib[i].input_ptr = ((char *)ptr +
+                                   tg->attrib[i].input_offset);
+        tg->attrib[i].input_stride = stride;
+      }
+   }
+}
+
+
+static void generic_destroy( struct translate *translate )
+{
+   FREE(translate);
+}
+
+struct translate *translate_generic_create( unsigned output_stride,
+                                           const struct translate_element *elements,
+                                           unsigned nr_elements )
+{
+   struct translate_generic *tg = CALLOC_STRUCT(translate_generic);
+   unsigned i;
+
+   if (tg == NULL)
+      return NULL;
+
+   tg->translate.destroy = generic_destroy;
+   tg->translate.set_buffer = generic_set_buffer;
+   tg->translate.run_elts = generic_run_elts;
+
+   for (i = 0; i < nr_elements; i++) {
+      tg->attrib[i].fetch = get_fetch_func(elements[i].input_format);
+      tg->attrib[i].buffer = elements[i].input_buffer;
+      tg->attrib[i].input_offset = elements[i].input_offset;
+
+      tg->attrib[i].emit = get_emit_func(elements[i].output_format);
+      tg->attrib[i].output_offset = elements[i].output_offset;
+   }
+
+   tg->nr_attrib = nr_elements;
+
+
+   return &tg->translate;
+}