Import unichrome dri driver from mesa 7.0.3.7.0.4 7.0.3

author: Luc Verhaegen <libv@skynet.be> 2010-03-11 14:18:41 +0100
committer: Luc Verhaegen <libv@skynet.be> 2010-03-11 14:18:41 +0100
commit: bf6a158b95d810d06344b4ca9395e4ebbca8bda9 (patch)
tree: 36a709a98723329d45cf43154e0fb93278a07cbd
parent: 5dee9b7b19c1aa3a13618b08bc24f00677b5364b (diff)
29 files changed, 11738 insertions, 7 deletions
diff --git a/configure.ac b/configure.ac
index 2a78cfd..4ce1052 100644
--- a/configure.ac
+++ b/configure.ac
@@ -1,7 +1,7 @@
 # Process this file with autoconf to produce a configure script
 
 AC_PREREQ(2.57)
-AC_INIT([mesa-dri-xxx], 7.0.3, [], mesa-dri-xxx)
+AC_INIT([mesa-dri-unichrome], 7.0.3, [], mesa-dri-unichrome)
 
 AM_INIT_AUTOMAKE([dist-bzip2])
 
@@ -16,7 +16,8 @@ AC_PROG_CC
 AC_HEADER_STDC
 
 PKG_CHECK_MODULES([DRM], [libdrm >= 2.3.0])
-PKG_CHECK_MODULES([DRI], [libmesadri = 7.0.3 libmesadricommon = 7.0.3])
+PKG_CHECK_MODULES([DRI], [libmesadri >= 7.0.3 libmesadri < 7.1.0
+			  libmesadricommon >= 7.0.3 libmesadricommon < 7.1.0])
 
 AC_OUTPUT([
 	Makefile
diff --git a/src/Makefile.am b/src/Makefile.am
index aa854c5..4984403 100644
--- a/src/Makefile.am
+++ b/src/Makefile.am
@@ -1,7 +1,19 @@
 AM_CFLAGS = -DIN_DRI_DRIVER -DGLX_DIRECT_RENDERING -DGLX_INDIRECT_RENDERING
 
-xxx_dri_la_LTLIBRARIES = xxx_dri.la
-xxx_dri_la_CFLAGS = $(AM_CFLAGS) $(DRM_CFLAGS) $(DRI_CFLAGS) -Iserver
-xxx_dri_la_LDFLAGS = -module -noprefix -lm -ldl $(DRM_LIBS) $(DRI_LIBS)
-xxx_dri_ladir = @libdir@/dri
-xxx_dri_la_SOURCES = \
+unichrome_dri_la_LTLIBRARIES = unichrome_dri.la
+unichrome_dri_la_CFLAGS = $(AM_CFLAGS) $(DRM_CFLAGS) $(DRI_CFLAGS) -Iserver
+unichrome_dri_la_LDFLAGS = -module -noprefix -avoid-version -lm -ldl \
+	$(DRM_LIBS) $(DRI_LIBS)
+unichrome_dri_ladir = @libdir@/dri
+unichrome_dri_la_SOURCES = \
+	via_context.c \
+	via_fb.c \
+	via_tex.c \
+	via_ioctl.c \
+	via_memcpy.c \
+	via_render.c \
+	via_screen.c \
+	via_span.c \
+	via_state.c \
+	via_texcombine.c \
+	via_tris.c 
diff --git a/src/server/via.h b/src/server/via.h
new file mode 100644
index 0000000..2cfe663
--- /dev/null
+++ b/src/server/via.h
@@ -0,0 +1,11 @@
+#ifndef __VIA_H__
+#define __VIA_H__
+
+typedef struct VIAInfo
+{
+    size_t registerSize;
+    void * registerHandle;
+    void * data;
+} * VIAInfoPtr;
+
+#endif /* __VIA_H__ */
diff --git a/src/server/via_dri.c b/src/server/via_dri.c
new file mode 100644
index 0000000..6944bd6
--- /dev/null
+++ b/src/server/via_dri.c
@@ -0,0 +1,1252 @@
+/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_dri.c,v 1.4 2003/09/24 02:43:30 dawes Exp $ */
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+ 
+#include "driver.h"
+#include "drm.h"
+#include "imports.h"
+
+#include "dri_util.h"
+
+#include "via_context.h"
+#include "via_dri.h"
+#include "via_driver.h"
+#include "xf86drm.h"
+
+static void VIAEnableMMIO(DRIDriverContext * ctx);
+static void VIADisableMMIO(DRIDriverContext * ctx);
+static void VIADisableExtendedFIFO(DRIDriverContext *ctx);
+static void VIAEnableExtendedFIFO(DRIDriverContext *ctx);
+static void VIAInitialize2DEngine(DRIDriverContext *ctx);
+static void VIAInitialize3DEngine(DRIDriverContext *ctx);
+
+static int VIADRIScreenInit(DRIDriverContext * ctx);
+static void VIADRICloseScreen(DRIDriverContext * ctx);
+static int VIADRIFinishScreenInit(DRIDriverContext * ctx);
+
+/* _SOLO : missing macros normally defined by X code */
+#define xf86DrvMsg(a, b, ...) fprintf(stderr, __VA_ARGS__)
+#define MMIO_IN8(base, addr) ((*(((volatile u_int8_t*)base)+(addr)))+0)
+#define MMIO_OUT8(base, addr, val) ((*(((volatile u_int8_t*)base)+(addr)))=((u_int8_t)val))
+#define MMIO_OUT16(base, addr, val) ((*(volatile u_int16_t*)(((u_int8_t*)base)+(addr)))=((u_int16_t)val))
+
+#define VIDEO	0 
+#define AGP		1
+#define AGP_PAGE_SIZE 4096
+#define AGP_PAGES 8192
+#define AGP_SIZE (AGP_PAGE_SIZE * AGP_PAGES)
+#define AGP_CMDBUF_PAGES 512
+#define AGP_CMDBUF_SIZE (AGP_PAGE_SIZE * AGP_CMDBUF_PAGES)
+
+static char VIAKernelDriverName[] = "via";
+static char VIAClientDriverName[] = "unichrome";
+
+static int VIADRIAgpInit(const DRIDriverContext *ctx, VIAPtr pVia);
+static int VIADRIPciInit(DRIDriverContext * ctx, VIAPtr pVia);
+static int VIADRIFBInit(DRIDriverContext * ctx, VIAPtr pVia);
+static int VIADRIKernelInit(DRIDriverContext * ctx, VIAPtr pVia);
+static int VIADRIMapInit(DRIDriverContext * ctx, VIAPtr pVia);
+
+static void VIADRIIrqInit( DRIDriverContext *ctx )
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI = pVia->devPrivate;
+
+    pVIADRI->irqEnabled = drmGetInterruptFromBusID(pVia->drmFD,
+					   ctx->pciBus,
+					   ctx->pciDevice,
+					   ctx->pciFunc);
+
+    if ((drmCtlInstHandler(pVia->drmFD, pVIADRI->irqEnabled))) {
+	xf86DrvMsg(pScreen->myNum, X_WARNING,
+		   "[drm] Failure adding irq handler. "
+		   "Falling back to irq-free operation.\n");
+	pVIADRI->irqEnabled = 0;
+    }
+
+    if (pVIADRI->irqEnabled)
+	xf86DrvMsg(pScreen->myNum, X_INFO,
+		   "[drm] Irq handler installed, using IRQ %d.\n",
+		   pVIADRI->irqEnabled);
+}
+
+static void VIADRIIrqExit( DRIDriverContext *ctx ) {
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI = pVia->devPrivate;
+
+    if (pVIADRI->irqEnabled) {
+	if (drmCtlUninstHandler(pVia->drmFD)) {
+	    xf86DrvMsg(pScreen-myNum, X_INFO,"[drm] Irq handler uninstalled.\n");
+	} else {
+	    xf86DrvMsg(pScreen->myNum, X_ERROR,
+		       "[drm] Could not uninstall irq handler.\n");
+	}
+    }
+}
+	    
+static void VIADRIRingBufferCleanup(DRIDriverContext *ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI = pVia->devPrivate;
+    drm_via_dma_init_t ringBufInit;
+
+    if (pVIADRI->ringBufActive) {
+	xf86DrvMsg(pScreen->myNum, X_INFO, 
+		   "[drm] Cleaning up DMA ring-buffer.\n");
+	ringBufInit.func = VIA_CLEANUP_DMA;
+	if (drmCommandWrite(pVia->drmFD, DRM_VIA_DMA_INIT, &ringBufInit,
+			    sizeof(ringBufInit))) {
+	    xf86DrvMsg(pScreen->myNum, X_WARNING, 
+		       "[drm] Failed to clean up DMA ring-buffer: %d\n", errno);
+	}
+	pVIADRI->ringBufActive = 0;
+    }
+}
+
+static int VIADRIRingBufferInit(DRIDriverContext *ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI = pVia->devPrivate;
+    drm_via_dma_init_t ringBufInit;
+    drmVersionPtr drmVer;
+
+    pVIADRI->ringBufActive = 0;
+
+    if (NULL == (drmVer = drmGetVersion(pVia->drmFD))) {
+	return GL_FALSE;
+    }
+
+    if (((drmVer->version_major <= 1) && (drmVer->version_minor <= 3))) {
+	return GL_FALSE;
+    } 
+
+    /*
+     * Info frome code-snippet on DRI-DEVEL list; Erdi Chen.
+     */
+
+    switch (pVia->ChipId) {
+    case PCI_CHIP_VT3259:
+    	ringBufInit.reg_pause_addr = 0x40c;
+	break;
+    default:
+    	ringBufInit.reg_pause_addr = 0x418;
+	break;
+    }
+   
+    ringBufInit.offset = pVia->agpSize;
+    ringBufInit.size = AGP_CMDBUF_SIZE;
+    ringBufInit.func = VIA_INIT_DMA;
+    if (drmCommandWrite(pVia->drmFD, DRM_VIA_DMA_INIT, &ringBufInit,
+			sizeof(ringBufInit))) {
+	xf86DrvMsg(pScreen->myNum, X_ERROR, 
+		   "[drm] Failed to initialize DMA ring-buffer: %d\n", errno);
+	return GL_FALSE;
+    }
+    xf86DrvMsg(pScreen->myNum, X_INFO, 
+	       "[drm] Initialized AGP ring-buffer, size 0x%lx at AGP offset 0x%lx.\n",
+	       ringBufInit.size, ringBufInit.offset);
+   
+    pVIADRI->ringBufActive = 1;
+    return GL_TRUE;
+}	    
+
+static int VIADRIAgpInit(const DRIDriverContext *ctx, VIAPtr pVia)
+{
+    unsigned long  agp_phys;
+    drmAddress agpaddr;
+    VIADRIPtr pVIADRI;
+    pVIADRI = pVia->devPrivate;
+    pVia->agpSize = 0;
+
+    if (drmAgpAcquire(pVia->drmFD) < 0) {
+        xf86DrvMsg(pScreen->myNum, X_ERROR, "[drm] drmAgpAcquire failed %d\n", errno);
+        return GL_FALSE;
+    }
+
+    if (drmAgpEnable(pVia->drmFD, drmAgpGetMode(pVia->drmFD)&~0x0) < 0) {
+         xf86DrvMsg(pScreen->myNum, X_ERROR, "[drm] drmAgpEnable failed\n");
+        return GL_FALSE;
+    }
+    
+    xf86DrvMsg(pScreen->myNum, X_INFO, "[drm] drmAgpEnabled succeeded\n");
+
+    if (drmAgpAlloc(pVia->drmFD, AGP_SIZE, 0, &agp_phys, &pVia->agpHandle) < 0) {
+        xf86DrvMsg(pScreen->myNum, X_ERROR,
+                 "[drm] drmAgpAlloc failed\n");
+        drmAgpRelease(pVia->drmFD);
+        return GL_FALSE;
+    }
+   
+    if (drmAgpBind(pVia->drmFD, pVia->agpHandle, 0) < 0) {
+        xf86DrvMsg(pScreen->myNum, X_ERROR,
+                 "[drm] drmAgpBind failed\n");
+        drmAgpFree(pVia->drmFD, pVia->agpHandle);
+        drmAgpRelease(pVia->drmFD);
+
+        return GL_FALSE;
+    }
+
+    /*
+     * Place the ring-buffer last in the AGP region, and restrict the
+     * public map not to include the buffer for security reasons.
+     */
+
+    pVia->agpSize = AGP_SIZE - AGP_CMDBUF_SIZE;
+    pVia->agpAddr = drmAgpBase(pVia->drmFD);
+    xf86DrvMsg(pScreen->myNum, X_INFO,
+                 "[drm] agpAddr = 0x%08lx\n",pVia->agpAddr);
+		 
+    pVIADRI->agp.size = pVia->agpSize;
+    if (drmAddMap(pVia->drmFD, (drm_handle_t)0,
+                 pVIADRI->agp.size, DRM_AGP, 0, 
+                 &pVIADRI->agp.handle) < 0) {
+	xf86DrvMsg(pScreen->myNum, X_ERROR,
+	    "[drm] Failed to map public agp area\n");
+        pVIADRI->agp.size = 0;
+        return GL_FALSE;
+    }  
+    /* Map AGP from kernel to Xserver - Not really needed */
+    drmMap(pVia->drmFD, pVIADRI->agp.handle,pVIADRI->agp.size, &agpaddr);
+
+    xf86DrvMsg(pScreen->myNum, X_INFO, 
+                "[drm] agpAddr = 0x%08lx\n", pVia->agpAddr);
+    xf86DrvMsg(pScreen->myNum, X_INFO, 
+                "[drm] agpSize = 0x%08lx\n", pVia->agpSize);
+    xf86DrvMsg(pScreen->myNum, X_INFO, 
+                "[drm] agp physical addr = 0x%08lx\n", agp_phys);
+
+    {
+	drm_via_agp_t agp;
+	agp.offset = 0;
+	agp.size = AGP_SIZE-AGP_CMDBUF_SIZE;
+	if (drmCommandWrite(pVia->drmFD, DRM_VIA_AGP_INIT, &agp,
+			    sizeof(drm_via_agp_t)) < 0) {
+	    drmUnmap(&agpaddr,pVia->agpSize);
+	    drmRmMap(pVia->drmFD,pVIADRI->agp.handle);
+	    drmAgpUnbind(pVia->drmFD, pVia->agpHandle);
+	    drmAgpFree(pVia->drmFD, pVia->agpHandle);
+	    drmAgpRelease(pVia->drmFD);
+	    return GL_FALSE;
+	}
+    }
+
+    return GL_TRUE;
+}
+
+static int VIADRIFBInit(DRIDriverContext * ctx, VIAPtr pVia)
+{   
+    int FBSize = pVia->FBFreeEnd-pVia->FBFreeStart;
+    int FBOffset = pVia->FBFreeStart; 
+    VIADRIPtr pVIADRI = pVia->devPrivate;
+    pVIADRI->fbOffset = FBOffset;
+    pVIADRI->fbSize = pVia->videoRambytes;
+
+    {
+	drm_via_fb_t fb;
+	fb.offset = FBOffset;
+	fb.size = FBSize;
+	
+	if (drmCommandWrite(pVia->drmFD, DRM_VIA_FB_INIT, &fb,
+			    sizeof(drm_via_fb_t)) < 0) {
+	    xf86DrvMsg(pScreen->myNum, X_ERROR,
+		       "[drm] failed to init frame buffer area\n");
+	    return GL_FALSE;
+	} else {
+	    xf86DrvMsg(pScreen->myNum, X_INFO,
+		       "[drm] FBFreeStart= 0x%08x FBFreeEnd= 0x%08x "
+		       "FBSize= 0x%08x\n",
+		       pVia->FBFreeStart, pVia->FBFreeEnd, FBSize);
+	    return GL_TRUE;	
+	}   
+    }
+}
+
+static int VIADRIPciInit(DRIDriverContext * ctx, VIAPtr pVia)
+{
+    return GL_TRUE;	
+}
+
+static int VIADRIScreenInit(DRIDriverContext * ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI;
+    int err;
+
+#if 0
+    ctx->shared.SAREASize = ((sizeof(drm_sarea_t) + 0xfff) & 0x1000);
+#else
+    if (sizeof(drm_sarea_t)+sizeof(drm_via_sarea_t) > SAREA_MAX) {
+	xf86DrvMsg(pScrn->scrnIndex, X_ERROR,
+			"Data does not fit in SAREA\n");
+	return GL_FALSE;
+    }
+    ctx->shared.SAREASize = SAREA_MAX;
+#endif
+
+    ctx->drmFD = drmOpen(VIAKernelDriverName, NULL);
+    if (ctx->drmFD < 0) {
+        fprintf(stderr, "[drm] drmOpen failed\n");
+        return 0;
+    }
+    pVia->drmFD = ctx->drmFD;
+
+    err = drmSetBusid(ctx->drmFD, ctx->pciBusID);
+    if (err < 0) {
+        fprintf(stderr, "[drm] drmSetBusid failed (%d, %s), %s\n",
+                ctx->drmFD, ctx->pciBusID, strerror(-err));
+        return 0;
+    }
+
+    err = drmAddMap(ctx->drmFD, 0, ctx->shared.SAREASize, DRM_SHM,
+                  DRM_CONTAINS_LOCK, &ctx->shared.hSAREA);
+    if (err < 0) {
+        fprintf(stderr, "[drm] drmAddMap failed\n");
+        return 0;
+    }
+    fprintf(stderr, "[drm] added %d byte SAREA at 0x%08lx\n",
+            ctx->shared.SAREASize, ctx->shared.hSAREA);
+
+    if (drmMap(ctx->drmFD,
+               ctx->shared.hSAREA,
+               ctx->shared.SAREASize,
+               (drmAddressPtr)(&ctx->pSAREA)) < 0)
+    {
+        fprintf(stderr, "[drm] drmMap failed\n");
+        return 0;
+    }
+    memset(ctx->pSAREA, 0, ctx->shared.SAREASize);
+    fprintf(stderr, "[drm] mapped SAREA 0x%08lx to %p, size %d\n",
+            ctx->shared.hSAREA, ctx->pSAREA, ctx->shared.SAREASize);
+
+    /* Need to AddMap the framebuffer and mmio regions here:
+     */
+    if (drmAddMap(ctx->drmFD,
+                  (drm_handle_t)ctx->FBStart,
+                  ctx->FBSize,
+                  DRM_FRAME_BUFFER,
+#ifndef _EMBEDDED
+                   0,
+#else
+                   DRM_READ_ONLY,
+#endif
+                   &ctx->shared.hFrameBuffer) < 0)
+    {
+        fprintf(stderr, "[drm] drmAddMap framebuffer failed\n");
+        return 0;
+    }
+
+    fprintf(stderr, "[drm] framebuffer handle = 0x%08lx\n",
+            ctx->shared.hFrameBuffer);
+
+    pVIADRI = (VIADRIPtr) CALLOC(sizeof(VIADRIRec));
+    if (!pVIADRI) {
+        drmClose(ctx->drmFD);
+        return GL_FALSE;
+    }
+    pVia->devPrivate = pVIADRI;
+    ctx->driverClientMsg = pVIADRI;
+    ctx->driverClientMsgSize = sizeof(*pVIADRI);
+
+    /* DRIScreenInit doesn't add all the common mappings.  Add additional mappings here. */
+    if (!VIADRIMapInit(ctx, pVia)) {
+	VIADRICloseScreen(ctx);
+	return GL_FALSE;
+    }
+
+    pVIADRI->regs.size = VIA_MMIO_REGSIZE;
+    pVIADRI->regs.handle = pVia->registerHandle;
+    xf86DrvMsg(pScreen->myNum, X_INFO, "[drm] mmio Registers = 0x%08lx\n",
+	pVIADRI->regs.handle);
+
+    if (drmMap(pVia->drmFD,
+               pVIADRI->regs.handle,
+               pVIADRI->regs.size,
+               (drmAddress *)&pVia->MapBase) != 0)
+    {
+        VIADRICloseScreen(ctx);
+        return GL_FALSE;
+    }
+
+    xf86DrvMsg(pScrn->scrnIndex, X_INFO, "[dri] mmio mapped.\n" );
+
+    VIAEnableMMIO(ctx);
+
+    /* Get video memory clock. */
+    VGAOUT8(0x3D4, 0x3D);
+    pVia->MemClk = (VGAIN8(0x3D5) & 0xF0) >> 4;
+    xf86DrvMsg(0, X_INFO, "[dri] MemClk (0x%x)\n", pVia->MemClk);
+
+    /* 3D rendering has noise if not enabled. */
+    VIAEnableExtendedFIFO(ctx);
+
+    VIAInitialize2DEngine(ctx);
+
+    /* Must disable MMIO or 3D won't work. */
+    VIADisableMMIO(ctx);
+
+    VIAInitialize3DEngine(ctx);
+
+    pVia->IsPCI = !VIADRIAgpInit(ctx, pVia);
+
+    if (pVia->IsPCI) {
+        VIADRIPciInit(ctx, pVia);
+	xf86DrvMsg(pScrn->scrnIndex, X_INFO, "[dri] use pci.\n" );
+    }
+    else
+        xf86DrvMsg(pScrn->scrnIndex, X_INFO, "[dri] use agp.\n" );
+
+    if (!(VIADRIFBInit(ctx, pVia))) {
+	VIADRICloseScreen(ctx);
+        xf86DrvMsg(pScrn->scrnIndex, X_ERROR, "[dri] frame buffer initialize fail .\n" );
+        return GL_FALSE;
+    }
+    
+    xf86DrvMsg(pScrn->scrnIndex, X_INFO, "[dri] frame buffer initialized.\n" );
+ 
+    return VIADRIFinishScreenInit(ctx);
+}
+
+static void
+VIADRICloseScreen(DRIDriverContext * ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI=(VIADRIPtr)pVia->devPrivate;
+
+    VIADRIRingBufferCleanup(ctx);
+
+    if (pVia->MapBase) {
+	xf86DrvMsg(pScreen->myNum, X_INFO, "[drm] Unmapping MMIO registers\n");
+        drmUnmap(pVia->MapBase, pVIADRI->regs.size);
+    }
+
+    if (pVia->agpSize) {
+	xf86DrvMsg(pScreen->myNum, X_INFO, "[drm] Freeing agp memory\n");
+        drmAgpFree(pVia->drmFD, pVia->agpHandle);
+	xf86DrvMsg(pScreen->myNum, X_INFO, "[drm] Releasing agp module\n");
+    	drmAgpRelease(pVia->drmFD);
+    }
+
+#if 0
+    if (pVia->DRIIrqEnable) 
+#endif
+        VIADRIIrqExit(ctx);
+}
+
+static int
+VIADRIFinishScreenInit(DRIDriverContext * ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    VIADRIPtr pVIADRI;
+    int err;
+
+    err = drmCreateContext(ctx->drmFD, &ctx->serverContext);
+    if (err != 0) {
+        fprintf(stderr, "%s: drmCreateContext failed %d\n", __FUNCTION__, err);
+        return GL_FALSE;
+    }
+
+    DRM_LOCK(ctx->drmFD, ctx->pSAREA, ctx->serverContext, 0);
+
+
+    if (!VIADRIKernelInit(ctx, pVia)) {
+	VIADRICloseScreen(ctx);
+	return GL_FALSE;
+    }
+    xf86DrvMsg(pScreen->myNum, X_INFO, "[dri] kernel data initialized.\n");
+
+    /* set SAREA value */
+    {
+	drm_via_sarea_t *saPriv;
+
+	saPriv=(drm_via_sarea_t*)(((char*)ctx->pSAREA) +
+                               sizeof(drm_sarea_t));
+	assert(saPriv);
+	memset(saPriv, 0, sizeof(*saPriv));
+	saPriv->ctxOwner = -1;
+    }
+    pVIADRI=(VIADRIPtr)pVia->devPrivate;
+    pVIADRI->deviceID=pVia->Chipset;  
+    pVIADRI->width=ctx->shared.virtualWidth;
+    pVIADRI->height=ctx->shared.virtualHeight;
+    pVIADRI->mem=ctx->shared.fbSize;
+    pVIADRI->bytesPerPixel= (ctx->bpp+7) / 8; 
+    pVIADRI->sarea_priv_offset = sizeof(drm_sarea_t);
+    /* TODO */
+    pVIADRI->scrnX=pVIADRI->width;
+    pVIADRI->scrnY=pVIADRI->height;
+
+    /* Initialize IRQ */
+#if 0
+    if (pVia->DRIIrqEnable) 
+#endif
+	VIADRIIrqInit(ctx);
+    
+    pVIADRI->ringBufActive = 0;
+    VIADRIRingBufferInit(ctx);
+
+    return GL_TRUE;
+}
+
+/* Initialize the kernel data structures. */
+static int VIADRIKernelInit(DRIDriverContext * ctx, VIAPtr pVia)
+{
+    drm_via_init_t drmInfo;
+    memset(&drmInfo, 0, sizeof(drm_via_init_t));
+    drmInfo.sarea_priv_offset   = sizeof(drm_sarea_t);
+    drmInfo.func = VIA_INIT_MAP;
+    drmInfo.fb_offset           = pVia->FrameBufferBase;
+    drmInfo.mmio_offset         = pVia->registerHandle;
+    if (pVia->IsPCI)
+	drmInfo.agpAddr = (u_int32_t)NULL;
+    else
+	drmInfo.agpAddr = (u_int32_t)pVia->agpAddr;
+
+    if ((drmCommandWrite(pVia->drmFD, DRM_VIA_MAP_INIT,&drmInfo,
+			     sizeof(drm_via_init_t))) < 0)
+	    return GL_FALSE;
+
+    return GL_TRUE;
+}
+/* Add a map for the MMIO registers */
+static int VIADRIMapInit(DRIDriverContext * ctx, VIAPtr pVia)
+{
+    int flags = 0;
+
+    if (drmAddMap(pVia->drmFD, pVia->MmioBase, VIA_MMIO_REGSIZE,
+		  DRM_REGISTERS, flags, &pVia->registerHandle) < 0) {
+	return GL_FALSE;
+    }
+
+    xf86DrvMsg(pScreen->myNum, X_INFO,
+	"[drm] register handle = 0x%08lx\n", pVia->registerHandle);
+
+    return GL_TRUE;
+}
+
+static int viaValidateMode(const DRIDriverContext *ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+
+    return 1;
+}
+
+static int viaPostValidateMode(const DRIDriverContext *ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+
+    return 1;
+}
+
+static void VIAEnableMMIO(DRIDriverContext * ctx)
+{
+    /*vgaHWPtr hwp = VGAHWPTR(ctx);*/
+    VIAPtr pVia = VIAPTR(ctx);
+    unsigned char val;
+
+#if 0
+    if (xf86IsPrimaryPci(pVia->PciInfo)) {
+        /* If we are primary card, we still use std vga port. If we use
+         * MMIO, system will hang in vgaHWSave when our card used in
+         * PLE and KLE (integrated Trident MVP4)
+         */
+        vgaHWSetStdFuncs(hwp);
+    }
+    else {
+        vgaHWSetMmioFuncs(hwp, pVia->MapBase, 0x8000);
+    }
+#endif
+
+    val = VGAIN8(0x3c3);
+    VGAOUT8(0x3c3, val | 0x01);
+    val = VGAIN8(0x3cc);
+    VGAOUT8(0x3c2, val | 0x01);
+
+    /* Unlock Extended IO Space */
+    VGAOUT8(0x3c4, 0x10);
+    VGAOUT8(0x3c5, 0x01);
+
+    /* Enable MMIO */
+    if(!pVia->IsSecondary) {
+	VGAOUT8(0x3c4, 0x1a);
+	val = VGAIN8(0x3c5);
+#ifdef DEBUG
+	xf86DrvMsg(pScrn->scrnIndex, X_INFO, "primary val = %x\n", val);
+#endif
+	VGAOUT8(0x3c5, val | 0x68);
+    }
+    else {
+	VGAOUT8(0x3c4, 0x1a);
+	val = VGAIN8(0x3c5);
+#ifdef DEBUG
+	xf86DrvMsg(pScrn->scrnIndex, X_INFO, "secondary val = %x\n", val);
+#endif
+	VGAOUT8(0x3c5, val | 0x38);
+    }
+
+    /* Unlock CRTC registers */
+    VGAOUT8(0x3d4, 0x47);
+    VGAOUT8(0x3d5, 0x00);
+
+    return;
+}
+
+static void VIADisableMMIO(DRIDriverContext * ctx)
+{
+    VIAPtr pVia = VIAPTR(ctx);
+    unsigned char val;
+
+    VGAOUT8(0x3c4, 0x1a);
+    val = VGAIN8(0x3c5);
+    VGAOUT8(0x3c5, val & 0x97);
+
+    return;
+}
+
+static void VIADisableExtendedFIFO(DRIDriverContext *ctx)
+{
+    VIAPtr  pVia = VIAPTR(ctx);
+    u_int32_t  dwGE230, dwGE298;
+
+    /* Cause of exit XWindow will dump back register value, others chipset no
+     * need to set extended fifo value */
+    if (pVia->Chipset == VIA_CLE266 && pVia->ChipRev < 15 &&
+        (ctx->shared.virtualWidth > 1024 || pVia->HasSecondary)) {
+        /* Turn off Extend FIFO */
+        /* 0x298[29] */
+        dwGE298 = VIAGETREG(0x298);
+        VIASETREG(0x298, dwGE298 | 0x20000000);
+        /* 0x230[21] */
+        dwGE230 = VIAGETREG(0x230);
+        VIASETREG(0x230, dwGE230 & ~0x00200000);
+        /* 0x298[29] */
+        dwGE298 = VIAGETREG(0x298);
+        VIASETREG(0x298, dwGE298 & ~0x20000000);
+    }
+}
+
+static void VIAEnableExtendedFIFO(DRIDriverContext *ctx)
+{
+    VIAPtr  pVia = VIAPTR(ctx);
+    u_int8_t   bRegTemp;
+    u_int32_t  dwGE230, dwGE298;
+
+    switch (pVia->Chipset) {
+    case VIA_CLE266:
+        if (pVia->ChipRev > 14) {  /* For 3123Cx */
+            if (pVia->HasSecondary) {  /* SAMM or DuoView case */
+                if (ctx->shared.virtualWidth >= 1024)
+    	        {
+    	            /* 3c5.16[0:5] */
+        	        VGAOUT8(0x3C4, 0x16);
+            	    bRegTemp = VGAIN8(0x3C5);
+    	            bRegTemp &= ~0x3F;
+        	        bRegTemp |= 0x1C;
+            	    VGAOUT8(0x3C5, bRegTemp);
+        	        /* 3c5.17[0:6] */
+            	    VGAOUT8(0x3C4, 0x17);
+                	bRegTemp = VGAIN8(0x3C5);
+    	            bRegTemp &= ~0x7F;
+        	        bRegTemp |= 0x3F;
+            	    VGAOUT8(0x3C5, bRegTemp);
+            	    pVia->EnableExtendedFIFO = GL_TRUE;
+    	        }
+            }
+            else   /* Single view or Simultaneoue case */
+            {
+                if (ctx->shared.virtualWidth > 1024)
+    	        {
+    	            /* 3c5.16[0:5] */
+        	        VGAOUT8(0x3C4, 0x16);
+            	    bRegTemp = VGAIN8(0x3C5);
+    	            bRegTemp &= ~0x3F;
+        	        bRegTemp |= 0x17;
+            	    VGAOUT8(0x3C5, bRegTemp);
+        	        /* 3c5.17[0:6] */
+            	    VGAOUT8(0x3C4, 0x17);
+                	bRegTemp = VGAIN8(0x3C5);
+    	            bRegTemp &= ~0x7F;
+        	        bRegTemp |= 0x2F;
+            	    VGAOUT8(0x3C5, bRegTemp);
+            	    pVia->EnableExtendedFIFO = GL_TRUE;
+    	        }
+            }
+            /* 3c5.18[0:5] */
+            VGAOUT8(0x3C4, 0x18);
+            bRegTemp = VGAIN8(0x3C5);
+            bRegTemp &= ~0x3F;
+            bRegTemp |= 0x17;
+            bRegTemp |= 0x40;  /* force the preq always higher than treq */
+            VGAOUT8(0x3C5, bRegTemp);
+        }
+        else {      /* for 3123Ax */
+            if (ctx->shared.virtualWidth > 1024 || pVia->HasSecondary) {
+                /* Turn on Extend FIFO */
+                /* 0x298[29] */
+                dwGE298 = VIAGETREG(0x298);
+                VIASETREG(0x298, dwGE298 | 0x20000000);
+                /* 0x230[21] */
+                dwGE230 = VIAGETREG(0x230);
+                VIASETREG(0x230, dwGE230 | 0x00200000);
+                /* 0x298[29] */
+                dwGE298 = VIAGETREG(0x298);
+                VIASETREG(0x298, dwGE298 & ~0x20000000);
+
+                /* 3c5.16[0:5] */
+                VGAOUT8(0x3C4, 0x16);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp |= 0x17;
+                /* bRegTemp |= 0x10; */
+                VGAOUT8(0x3C5, bRegTemp);
+                /* 3c5.17[0:6] */
+                VGAOUT8(0x3C4, 0x17);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x7F;
+                bRegTemp |= 0x2F;
+                /*bRegTemp |= 0x1F;*/
+                VGAOUT8(0x3C5, bRegTemp);
+                /* 3c5.18[0:5] */
+                VGAOUT8(0x3C4, 0x18);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp |= 0x17;
+                bRegTemp |= 0x40;  /* force the preq always higher than treq */
+                VGAOUT8(0x3C5, bRegTemp);
+          	    pVia->EnableExtendedFIFO = GL_TRUE;
+            }
+        }
+        break;
+    case VIA_KM400:
+        if (pVia->HasSecondary) {  /* SAMM or DuoView case */
+            if ((ctx->shared.virtualWidth >= 1600) &&
+                (pVia->MemClk <= VIA_MEM_DDR200)) {
+        	    /* enable CRT extendded FIFO */
+            	VGAOUT8(0x3C4, 0x17);
+                VGAOUT8(0x3C5, 0x1C);
+    	        /* revise second display queue depth and read threshold */
+        	    VGAOUT8(0x3C4, 0x16);
+            	bRegTemp = VGAIN8(0x3C5);
+    	        bRegTemp &= ~0x3F;
+    	        bRegTemp = (bRegTemp) | (0x09);
+                VGAOUT8(0x3C5, bRegTemp);
+            }
+            else {
+                /* enable CRT extendded FIFO */
+                VGAOUT8(0x3C4, 0x17);
+                VGAOUT8(0x3C5,0x3F);
+                /* revise second display queue depth and read threshold */
+                VGAOUT8(0x3C4, 0x16);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp = (bRegTemp) | (0x1C);
+                VGAOUT8(0x3C5, bRegTemp);
+            }
+            /* 3c5.18[0:5] */
+            VGAOUT8(0x3C4, 0x18);
+            bRegTemp = VGAIN8(0x3C5);
+            bRegTemp &= ~0x3F;
+            bRegTemp |= 0x17;
+            bRegTemp |= 0x40;  /* force the preq always higher than treq */
+            VGAOUT8(0x3C5, bRegTemp);
+       	    pVia->EnableExtendedFIFO = GL_TRUE;
+        }
+        else {
+            if ( (ctx->shared.virtualWidth > 1024) && (ctx->shared.virtualWidth <= 1280) )
+            {
+                /* enable CRT extendded FIFO */
+                VGAOUT8(0x3C4, 0x17);
+                VGAOUT8(0x3C5, 0x3F);
+                /* revise second display queue depth and read threshold */
+                VGAOUT8(0x3C4, 0x16);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp = (bRegTemp) | (0x17);
+                VGAOUT8(0x3C5, bRegTemp);
+           	    pVia->EnableExtendedFIFO = GL_TRUE;
+            }
+            else if ((ctx->shared.virtualWidth > 1280))
+            {
+                /* enable CRT extendded FIFO */
+                VGAOUT8(0x3C4, 0x17);
+                VGAOUT8(0x3C5, 0x3F);
+                /* revise second display queue depth and read threshold */
+                VGAOUT8(0x3C4, 0x16);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp = (bRegTemp) | (0x1C);
+                VGAOUT8(0x3C5, bRegTemp);
+           	    pVia->EnableExtendedFIFO = GL_TRUE;
+            }
+            else
+            {
+                /* enable CRT extendded FIFO */
+                VGAOUT8(0x3C4, 0x17);
+                VGAOUT8(0x3C5, 0x3F);
+                /* revise second display queue depth and read threshold */
+                VGAOUT8(0x3C4, 0x16);
+                bRegTemp = VGAIN8(0x3C5);
+                bRegTemp &= ~0x3F;
+                bRegTemp = (bRegTemp) | (0x10);
+                VGAOUT8(0x3C5, bRegTemp);
+            }
+            /* 3c5.18[0:5] */
+            VGAOUT8(0x3C4, 0x18);
+            bRegTemp = VGAIN8(0x3C5);
+            bRegTemp &= ~0x3F;
+            bRegTemp |= 0x17;
+            bRegTemp |= 0x40;  /* force the preq always higher than treq */
+            VGAOUT8(0x3C5, bRegTemp);
+        }
+        break;
+    case VIA_K8M800:
+        /*=* R1 Display FIFO depth (384 /8 -1 -> 0xbf) SR17[7:0] (8bits) *=*/
+        VGAOUT8(0x3c4, 0x17);
+        VGAOUT8(0x3c5, 0xbf);
+
+        /*=* R2 Display fetch datum threshold value (328/4 -> 0x52)
+             SR16[5:0], SR16[7] (7bits) *=*/
+        VGAOUT8(0x3c4, 0x16);
+        bRegTemp = VGAIN8(0x3c5) & ~0xBF;
+        bRegTemp |= (0x52 & 0x3F);
+        bRegTemp |= ((0x52 & 0x40) << 1);
+        VGAOUT8(0x3c5, bRegTemp);
+
+        /*=* R3 Switch to the highest agent threshold value (74 -> 0x4a)
+             SR18[5:0], SR18[7] (7bits) *=*/
+        VGAOUT8(0x3c4, 0x18);
+        bRegTemp = VGAIN8(0x3c5) & ~0xBF;
+        bRegTemp |= (0x4a & 0x3F);
+        bRegTemp |= ((0x4a & 0x40) << 1);
+        VGAOUT8(0x3c5, bRegTemp);
+#if 0
+        /*=* R4 Fetch Number for a scan line (unit: 8 bytes)
+             SR1C[7:0], SR1D[1:0] (10bits) *=*/
+        wRegTemp = (pBIOSInfo->offsetWidthByQWord >> 1) + 4;
+        VGAOUT8(0x3c4, 0x1c);
+        VGAOUT8(0x3c5, (u_int8_t)(wRegTemp & 0xFF));
+        VGAOUT8(0x3c4, 0x1d);
+        bRegTemp = VGAIN8(0x3c5) & ~0x03;
+        VGAOUT8(0x3c5, bRegTemp | ((wRegTemp & 0x300) >> 8));
+#endif
+        if (ctx->shared.virtualWidth >= 1400 && ctx->bpp == 32)
+        {
+            /*=* Max. length for a request SR22[4:0] (64/4 -> 0x10) *=*/
+            VGAOUT8(0x3c4, 0x22);
+            bRegTemp = VGAIN8(0x3c5) & ~0x1F;
+            VGAOUT8(0x3c5, bRegTemp | 0x10);
+        }
+        else
+        {
+            /*=* Max. length for a request SR22[4:0]
+                 (128/4 -> over flow 0x0) *=*/
+            VGAOUT8(0x3c4, 0x22);
+            bRegTemp = VGAIN8(0x3c5) & ~0x1F;
+            VGAOUT8(0x3c5, bRegTemp);
+        }
+        break;
+    case VIA_PM800:
+        /*=* R1 Display FIFO depth (96-1 -> 0x5f) SR17[7:0] (8bits) *=*/
+        VGAOUT8(0x3c4, 0x17);
+        VGAOUT8(0x3c5, 0x5f);
+
+        /*=* R2 Display fetch datum threshold value (32 -> 0x20)
+             SR16[5:0], SR16[7] (7bits) *=*/
+        VGAOUT8(0x3c4, 0x16);
+        bRegTemp = VGAIN8(0x3c5) & ~0xBF;
+        bRegTemp |= (0x20 & 0x3F);
+        bRegTemp |= ((0x20 & 0x40) << 1);
+        VGAOUT8(0x3c5, bRegTemp);
+
+        /*=* R3 Switch to the highest agent threshold value (16 -> 0x10)
+             SR18[5:0], SR18[7] (7bits) *=*/
+        VGAOUT8(0x3c4, 0x18);
+        bRegTemp = VGAIN8(0x3c5) & ~0xBF;
+        bRegTemp |= (0x10 & 0x3F);
+        bRegTemp |= ((0x10 & 0x40) << 1);
+        VGAOUT8(0x3c5, bRegTemp);
+#if 0
+        /*=* R4 Fetch Number for a scan line (unit: 8 bytes)
+             SR1C[7:0], SR1D[1:0] (10bits) *=*/
+        wRegTemp = (pBIOSInfo->offsetWidthByQWord >> 1) + 4;
+        VGAOUT8(0x3c4, 0x1c);
+        VGAOUT8(0x3c5, (u_int8_t)(wRegTemp & 0xFF));
+        VGAOUT8(0x3c4, 0x1d);
+        bRegTemp = VGAIN8(0x3c5) & ~0x03;
+        VGAOUT8(0x3c5, bRegTemp | ((wRegTemp & 0x300) >> 8));
+#endif
+        if (ctx->shared.virtualWidth >= 1400 && ctx->bpp == 32)
+        {
+            /*=* Max. length for a request SR22[4:0] (64/4 -> 0x10) *=*/
+            VGAOUT8(0x3c4, 0x22);
+            bRegTemp = VGAIN8(0x3c5) & ~0x1F;
+            VGAOUT8(0x3c5, bRegTemp | 0x10);
+        }
+        else
+        {
+            /*=* Max. length for a request SR22[4:0] (0x1F) *=*/
+            VGAOUT8(0x3c4, 0x22);
+            bRegTemp = VGAIN8(0x3c5) & ~0x1F;
+            VGAOUT8(0x3c5, bRegTemp | 0x1F);
+        }
+        break;
+    default:
+        break;
+    }
+}
+
+static void VIAInitialize2DEngine(DRIDriverContext *ctx)
+{
+    VIAPtr  pVia = VIAPTR(ctx);
+    u_int32_t  dwVQStartAddr, dwVQEndAddr;
+    u_int32_t  dwVQLen, dwVQStartL, dwVQEndL, dwVQStartEndH;
+    u_int32_t  dwGEMode;
+
+    /* init 2D engine regs to reset 2D engine */
+    VIASETREG(0x04, 0x0);
+    VIASETREG(0x08, 0x0);
+    VIASETREG(0x0c, 0x0);
+    VIASETREG(0x10, 0x0);
+    VIASETREG(0x14, 0x0);
+    VIASETREG(0x18, 0x0);
+    VIASETREG(0x1c, 0x0);
+    VIASETREG(0x20, 0x0);
+    VIASETREG(0x24, 0x0);
+    VIASETREG(0x28, 0x0);
+    VIASETREG(0x2c, 0x0);
+    VIASETREG(0x30, 0x0);
+    VIASETREG(0x34, 0x0);
+    VIASETREG(0x38, 0x0);
+    VIASETREG(0x3c, 0x0);
+    VIASETREG(0x40, 0x0);
+
+    VIADisableMMIO(ctx);
+
+    /* Init AGP and VQ regs */
+    VIASETREG(0x43c, 0x00100000);
+    VIASETREG(0x440, 0x00000000);
+    VIASETREG(0x440, 0x00333004);
+    VIASETREG(0x440, 0x60000000);
+    VIASETREG(0x440, 0x61000000);
+    VIASETREG(0x440, 0x62000000);
+    VIASETREG(0x440, 0x63000000);
+    VIASETREG(0x440, 0x64000000);
+    VIASETREG(0x440, 0x7D000000);
+
+    VIASETREG(0x43c, 0xfe020000);
+    VIASETREG(0x440, 0x00000000);
+
+    if (pVia->VQStart != 0) {
+        /* Enable VQ */
+        dwVQStartAddr = pVia->VQStart;
+        dwVQEndAddr = pVia->VQEnd;
+        dwVQStartL = 0x50000000 | (dwVQStartAddr & 0xFFFFFF);
+        dwVQEndL = 0x51000000 | (dwVQEndAddr & 0xFFFFFF);
+        dwVQStartEndH = 0x52000000 | ((dwVQStartAddr & 0xFF000000) >> 24) |
+                        ((dwVQEndAddr & 0xFF000000) >> 16);
+        dwVQLen = 0x53000000 | (VIA_VQ_SIZE >> 3);
+
+        VIASETREG(0x43c, 0x00fe0000);
+        VIASETREG(0x440, 0x080003fe);
+        VIASETREG(0x440, 0x0a00027c);
+        VIASETREG(0x440, 0x0b000260);
+        VIASETREG(0x440, 0x0c000274);
+        VIASETREG(0x440, 0x0d000264);
+        VIASETREG(0x440, 0x0e000000);
+        VIASETREG(0x440, 0x0f000020);
+        VIASETREG(0x440, 0x1000027e);
+        VIASETREG(0x440, 0x110002fe);
+        VIASETREG(0x440, 0x200f0060);
+
+        VIASETREG(0x440, 0x00000006);
+        VIASETREG(0x440, 0x40008c0f);
+        VIASETREG(0x440, 0x44000000);
+        VIASETREG(0x440, 0x45080c04);
+        VIASETREG(0x440, 0x46800408);
+
+        VIASETREG(0x440, dwVQStartEndH);
+        VIASETREG(0x440, dwVQStartL);
+        VIASETREG(0x440, dwVQEndL);
+        VIASETREG(0x440, dwVQLen);
+    }
+    else {
+        /* Diable VQ */
+        VIASETREG(0x43c, 0x00fe0000);
+        VIASETREG(0x440, 0x00000004);
+        VIASETREG(0x440, 0x40008c0f);
+        VIASETREG(0x440, 0x44000000);
+        VIASETREG(0x440, 0x45080c04);
+        VIASETREG(0x440, 0x46800408);
+    }
+
+    dwGEMode = 0;
+
+    switch (ctx->bpp) {
+    case 16:
+        dwGEMode |= VIA_GEM_16bpp;
+        break;
+    case 32:
+        dwGEMode |= VIA_GEM_32bpp;
+        break;
+    default:
+        dwGEMode |= VIA_GEM_8bpp;
+        break;
+    }
+
+#if 0
+    switch (ctx->shared.virtualWidth) {
+    case 800:
+        dwGEMode |= VIA_GEM_800;
+        break;
+    case 1024:
+        dwGEMode |= VIA_GEM_1024;
+        break;
+    case 1280:
+        dwGEMode |= VIA_GEM_1280;
+        break;
+    case 1600:
+        dwGEMode |= VIA_GEM_1600;
+        break;
+    case 2048:
+        dwGEMode |= VIA_GEM_2048;
+        break;
+    default:
+        dwGEMode |= VIA_GEM_640;
+        break;
+    }
+#endif
+    
+    VIAEnableMMIO(ctx);
+
+    /* Set BPP and Pitch */
+    VIASETREG(VIA_REG_GEMODE, dwGEMode);
+
+    /* Set Src and Dst base address and pitch, pitch is qword */
+    VIASETREG(VIA_REG_SRCBASE, 0x0);
+    VIASETREG(VIA_REG_DSTBASE, 0x0);
+    VIASETREG(VIA_REG_PITCH, VIA_PITCH_ENABLE |
+              ((ctx->shared.virtualWidth * ctx->bpp >> 3) >> 3) |
+              (((ctx->shared.virtualWidth * ctx->bpp >> 3) >> 3) << 16));
+}
+
+static int b3DRegsInitialized = 0;
+
+static void VIAInitialize3DEngine(DRIDriverContext *ctx)
+{
+    VIAPtr  pVia = VIAPTR(ctx);
+    int i;
+
+    if (!b3DRegsInitialized)
+    {
+
+        VIASETREG(0x43C, 0x00010000);
+
+        for (i = 0; i <= 0x7D; i++)
+        {
+            VIASETREG(0x440, (u_int32_t) i << 24);
+        }
+
+        VIASETREG(0x43C, 0x00020000);
+
+        for (i = 0; i <= 0x94; i++)
+        {
+            VIASETREG(0x440, (u_int32_t) i << 24);
+        }
+
+        VIASETREG(0x440, 0x82400000);
+
+        VIASETREG(0x43C, 0x01020000);
+
+
+        for (i = 0; i <= 0x94; i++)
+        {
+            VIASETREG(0x440, (u_int32_t) i << 24);
+        }
+
+        VIASETREG(0x440, 0x82400000);
+        VIASETREG(0x43C, 0xfe020000);
+
+        for (i = 0; i <= 0x03; i++)
+        {
+            VIASETREG(0x440, (u_int32_t) i << 24);
+        }
+
+        VIASETREG(0x43C, 0x00030000);
+
+        for (i = 0; i <= 0xff; i++)
+        {
+            VIASETREG(0x440, 0);
+        }
+        VIASETREG(0x43C, 0x00100000);
+        VIASETREG(0x440, 0x00333004);
+        VIASETREG(0x440, 0x10000002);
+        VIASETREG(0x440, 0x60000000);
+        VIASETREG(0x440, 0x61000000);
+        VIASETREG(0x440, 0x62000000);
+        VIASETREG(0x440, 0x63000000);
+        VIASETREG(0x440, 0x64000000);
+
+        VIASETREG(0x43C, 0x00fe0000);
+
+        if (pVia->ChipRev >= 3 )
+            VIASETREG(0x440,0x40008c0f);
+        else
+            VIASETREG(0x440,0x4000800f);
+
+        VIASETREG(0x440,0x44000000);
+        VIASETREG(0x440,0x45080C04);
+        VIASETREG(0x440,0x46800408);
+        VIASETREG(0x440,0x50000000);
+        VIASETREG(0x440,0x51000000);
+        VIASETREG(0x440,0x52000000);
+        VIASETREG(0x440,0x53000000);
+
+        b3DRegsInitialized = 1;
+        xf86DrvMsg(pScrn->scrnIndex, X_INFO,
+                   "3D Engine has been initialized.\n");
+    }
+
+    VIASETREG(0x43C,0x00fe0000);
+    VIASETREG(0x440,0x08000001);
+    VIASETREG(0x440,0x0A000183);
+    VIASETREG(0x440,0x0B00019F);
+    VIASETREG(0x440,0x0C00018B);
+    VIASETREG(0x440,0x0D00019B);
+    VIASETREG(0x440,0x0E000000);
+    VIASETREG(0x440,0x0F000000);
+    VIASETREG(0x440,0x10000000);
+    VIASETREG(0x440,0x11000000);
+    VIASETREG(0x440,0x20000000);
+}
+
+static int
+WaitIdleCLE266(VIAPtr pVia)
+{
+    int loop = 0;
+
+    /*mem_barrier();*/
+
+    while (!(VIAGETREG(VIA_REG_STATUS) & VIA_VR_QUEUE_BUSY) && (loop++ < MAXLOOP))
+        ;
+
+    while ((VIAGETREG(VIA_REG_STATUS) &
+          (VIA_CMD_RGTR_BUSY | VIA_2D_ENG_BUSY | VIA_3D_ENG_BUSY)) &&
+          (loop++ < MAXLOOP))
+        ;
+
+    return loop >= MAXLOOP;
+}
+
+static int viaInitFBDev(DRIDriverContext *ctx)
+{
+    VIAPtr pVia = CALLOC(sizeof(*pVia));
+
+    ctx->driverPrivate = (void *)pVia;
+
+    switch (ctx->chipset) {
+    case PCI_CHIP_CLE3122:
+    case PCI_CHIP_CLE3022:
+        pVia->Chipset = VIA_CLE266;
+        break;
+    case PCI_CHIP_VT7205:
+    case PCI_CHIP_VT3205:
+        pVia->Chipset = VIA_KM400;
+        break;
+    case PCI_CHIP_VT3204:
+    case PCI_CHIP_VT3344:
+        pVia->Chipset = VIA_K8M800;
+        break;
+    case PCI_CHIP_VT3259:
+        pVia->Chipset = VIA_PM800;
+        break;
+    default:
+        xf86DrvMsg(0, X_ERROR, "VIA: Unknown device ID (0x%x)\n", ctx->chipset);
+    }
+
+    /* _SOLO TODO XXX need to read ChipRev too */
+    pVia->ChipRev = 0;
+
+    pVia->videoRambytes = ctx->shared.fbSize;
+    pVia->MmioBase = ctx->MMIOStart;
+    pVia->FrameBufferBase = ctx->FBStart & 0xfc000000;
+
+    pVia->FBFreeStart = ctx->shared.virtualWidth * ctx->cpp *
+        ctx->shared.virtualHeight;
+
+#if 1
+    /* Alloc a second framebuffer for the second head */
+    pVia->FBFreeStart += ctx->shared.virtualWidth * ctx->cpp *
+	ctx->shared.virtualHeight;
+#endif
+
+    pVia->VQStart = pVia->FBFreeStart;
+    pVia->VQEnd = pVia->FBFreeStart + VIA_VQ_SIZE - 1;
+
+    pVia->FBFreeStart += VIA_VQ_SIZE;
+
+    pVia->FBFreeEnd = pVia->videoRambytes;
+
+    if (!VIADRIScreenInit(ctx))
+        return 0;
+
+    return 1;
+}
+
+static void viaHaltFBDev(DRIDriverContext *ctx)
+{
+    drmUnmap( ctx->pSAREA, ctx->shared.SAREASize );
+    drmClose(ctx->drmFD);
+
+    if (ctx->driverPrivate) {
+        free(ctx->driverPrivate);
+        ctx->driverPrivate = 0;
+    }
+}
+
+static int viaEngineShutdown(const DRIDriverContext *ctx)
+{
+    return 1;
+}
+
+static int viaEngineRestore(const DRIDriverContext *ctx)
+{
+    return 1;
+}
+
+const struct DRIDriverRec __driDriver =
+{
+    viaValidateMode,
+    viaPostValidateMode,
+    viaInitFBDev,
+    viaHaltFBDev,
+    viaEngineShutdown,
+    viaEngineRestore,  
+    0,
+};
+
diff --git a/src/server/via_dri.h b/src/server/via_dri.h
new file mode 100644
index 0000000..b47397d
--- /dev/null
+++ b/src/server/via_dri.h
@@ -0,0 +1,78 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * THE AUTHORS OR COPYRIGHT HOLDERS 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.
+ */
+/*
+ * Keep this file in perfect sync between the ddx and dri drivers.
+ * At least bump the VIA_DRIDDX_VERSION defines appropriately.
+ *
+ */
+#ifndef _VIA_DRI_H_
+#define _VIA_DRI_H_ 1
+
+#define VIA_MAX_DRAWABLES 256
+
+#define VIA_DRIDDX_VERSION_MAJOR  5
+#define VIA_DRIDDX_VERSION_MINOR  0
+#define VIA_DRIDDX_VERSION_PATCH  0
+
+#ifndef XFree86Server
+typedef int Bool;
+#endif
+
+typedef struct {
+    drm_handle_t handle;
+    drmSize size;
+} viaRegion, *viaRegionPtr;
+
+typedef struct {
+    viaRegion regs, agp;
+    int deviceID;
+    int width;
+    int height;
+    int mem;
+    int bytesPerPixel;
+    int priv1;
+    int priv2;
+    int fbOffset;
+    int fbSize;
+    Bool drixinerama;
+    int backOffset;
+    int depthOffset;
+    int textureOffset;
+    int textureSize;
+    int irqEnabled;
+    unsigned int scrnX, scrnY;
+    int sarea_priv_offset;
+    int ringBufActive;
+    unsigned int reg_pause_addr;
+} VIADRIRec, *VIADRIPtr;
+
+typedef struct {
+    int dummy;
+} VIAConfigPrivRec, *VIAConfigPrivPtr;
+
+typedef struct {
+    int dummy;
+} VIADRIContextRec, *VIADRIContextPtr;
+
+#endif /* _VIA_DRI_H_ */
diff --git a/src/server/via_driver.h b/src/server/via_driver.h
new file mode 100644
index 0000000..997b2e4
--- /dev/null
+++ b/src/server/via_driver.h
@@ -0,0 +1,309 @@
+/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_driver.h,v 1.7 2003/11/06 18:38:11 tsi Exp $ */
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIA_DRIVER_H
+#define _VIA_DRIVER_H
+
+#if 0 /* DEBUG is use in VIA DRI code as a flag */
+/* #define DEBUG_PRINT */
+#ifdef DEBUG_PRINT
+#define DEBUG(x) x
+#else
+#define DEBUG(x)
+#endif
+#endif
+
+#if 0
+#include "vgaHW.h"
+#include "xf86.h"
+#include "xf86Resources.h"
+#include "xf86_ansic.h"
+#include "xf86Pci.h"
+#include "xf86PciInfo.h"
+#include "xf86_OSproc.h"
+#include "compiler.h"
+#include "xf86Cursor.h"
+#include "mipointer.h"
+#include "micmap.h"
+
+#define USE_FB
+#ifdef USE_FB
+#include "fb.h"
+#else
+#include "cfb.h"
+#include "cfb16.h"
+#include "cfb32.h"
+#endif
+
+#include "xf86cmap.h"
+#include "vbe.h"
+#include "xaa.h"
+
+#include "via_regs.h"
+#include "via_bios.h"
+#include "via_gpioi2c.h"
+#include "via_priv.h"
+#include "ginfo.h"
+
+#ifdef XF86DRI
+#define _XF86DRI_SERVER_
+#include "sarea.h"
+#include "dri.h"
+#include "GL/glxint.h"
+#include "via_dri.h"
+#endif
+#else
+#include "via_regs.h"
+
+#include "GL/internal/dri_interface.h"
+#include "via_dri.h"
+#endif
+
+/* _SOLO : copied from via_bios.h */
+/* System Memory CLK */
+#define		VIA_MEM_SDR66					0x00
+#define		VIA_MEM_SDR100					0x01
+#define		VIA_MEM_SDR133					0x02
+#define		VIA_MEM_DDR200					0x03
+#define		VIA_MEM_DDR266					0x04
+#define		VIA_MEM_DDR333					0x05
+#define		VIA_MEM_DDR400					0x06
+
+#define DRIVER_NAME     "via"
+#define DRIVER_VERSION  "4.1.0"
+#define VERSION_MAJOR   4
+#define VERSION_MINOR   1
+#define PATCHLEVEL      41
+#define VIA_VERSION     ((VERSION_MAJOR<<24) | (VERSION_MINOR<<16) | PATCHLEVEL)
+
+#define VGAIN8(addr)        MMIO_IN8(pVia->MapBase+0x8000, addr)
+#define VGAIN16(addr)       MMIO_IN16(pVia->MapBase+0x8000, addr)
+#define VGAIN(addr)         MMIO_IN32(pVia->MapBase+0x8000, addr)
+
+#define VGAOUT8(addr, val)  MMIO_OUT8(pVia->MapBase+0x8000, addr, val)
+#define VGAOUT16(addr, val) MMIO_OUT16(pVia->MapBase+0x8000, addr, val)
+#define VGAOUT(addr, val)   MMIO_OUT32(pVia->MapBase+0x8000, addr, val)
+
+#define INREG(addr)         MMIO_IN32(pVia->MapBase, addr)
+#define OUTREG(addr, val)   MMIO_OUT32(pVia->MapBase, addr, val)
+#define INREG16(addr)       MMIO_IN16(pVia->MapBase, addr)
+#define OUTREG16(addr, val) MMIO_OUT16(pVia->MapBase, addr, val)
+
+#define VIA_PIXMAP_CACHE_SIZE   (256 * 1024)
+#define VIA_CURSOR_SIZE         (4 * 1024)
+#define VIA_VQ_SIZE             (256 * 1024)
+
+typedef struct {
+    unsigned int    mode, refresh, resMode;
+    int             countWidthByQWord;
+    int             offsetWidthByQWord;
+    unsigned char   SR08, SR0A, SR0F;
+
+    /*   extended Sequencer registers */
+    unsigned char   SR10, SR11, SR12, SR13,SR14,SR15,SR16;
+    unsigned char   SR17, SR18, SR19, SR1A,SR1B,SR1C,SR1D,SR1E;
+    unsigned char   SR1F, SR20, SR21, SR22,SR23,SR24,SR25,SR26;
+    unsigned char   SR27, SR28, SR29, SR2A,SR2B,SR2C,SR2D,SR2E;
+    unsigned char   SR2F, SR30, SR31, SR32,SR33,SR34,SR40,SR41;
+    unsigned char   SR42, SR43, SR44, SR45,SR46,SR47;
+
+    unsigned char   Clock;
+
+    /*   extended CRTC registers */
+    unsigned char   CR13, CR30, CR31, CR32, CR33, CR34, CR35, CR36;
+    unsigned char   CR37, CR38, CR39, CR3A, CR40, CR41, CR42, CR43;
+    unsigned char   CR44, CR45, CR46, CR47, CR48, CR49, CR4A;
+    unsigned char   CRTCRegs[83];
+    unsigned char   TVRegs[0xCF];
+    unsigned char   TVRegs2[0xCF];
+/*    unsigned char   LCDRegs[0x40];*/
+
+} VIARegRec, *VIARegPtr;
+
+
+typedef struct _VIA {
+    VIARegRec           SavedReg;
+    VIARegRec           ModeReg;
+#if 0
+    xf86CursorInfoPtr   CursorInfoRec;
+    int                 stateMode;
+    VIAModeInfoPtr      VIAModeList;
+#endif
+    int                 ModeStructInit;
+    int                 Bpp, Bpl, ScissB;
+    unsigned            PlaneMask;
+
+    unsigned long       videoRambytes;
+    int                 videoRamKbytes;
+    int                 FBFreeStart;
+    int                 FBFreeEnd;
+    int                 CursorStart;
+    int                 VQStart;
+    int                 VQEnd;
+
+    /* These are physical addresses. */
+    unsigned long       FrameBufferBase;
+    unsigned long       MmioBase;
+
+    /* These are linear addresses. */
+    unsigned char*      MapBase;
+    unsigned char*      VidMapBase;
+    unsigned char*      BltBase;
+    unsigned char*      MapBaseDense;
+    unsigned char*      FBBase;
+    unsigned char*      FBStart;
+
+    int                 PrimaryVidMapped;
+    int                 dacSpeedBpp;
+    int                 minClock, maxClock;
+    int                 MCLK, REFCLK, LCDclk;
+    double              refclk_fact;
+
+    /* Here are all the Options */
+    int                 VQEnable;
+    int                 pci_burst;
+    int                 NoPCIRetry;
+    int                 hwcursor;
+    int                 NoAccel;
+    int                 shadowFB;
+    int                 NoDDCValue;
+    int                 rotate;
+
+#if 0
+    CloseScreenProcPtr  CloseScreen;
+    pciVideoPtr         PciInfo;
+    PCITAG              PciTag;
+#endif
+    int                 Chipset;
+    int                 ChipId;
+    int                 ChipRev;
+    /*vbeInfoPtr          pVbe;*/
+    int                 EntityIndex;
+
+    /* Support for shadowFB and rotation */
+    unsigned char*      ShadowPtr;
+    int                 ShadowPitch;
+    void                (*PointerMoved)(int index, int x, int y);
+
+    /* Support for XAA acceleration */
+#if 0
+    XAAInfoRecPtr       AccelInfoRec;
+    xRectangle          Rect;
+#endif
+    u_int32_t            SavedCmd;
+    u_int32_t            SavedFgColor;
+    u_int32_t            SavedBgColor;
+    u_int32_t            SavedPattern0;
+    u_int32_t            SavedPattern1;
+    u_int32_t            SavedPatternAddr;
+
+#if 0
+    /* Support for Int10 processing */
+    xf86Int10InfoPtr    pInt10;
+
+    /* BIOS Info Ptr */
+    VIABIOSInfoPtr      pBIOSInfo;
+    VGABIOSVERPtr       pBIOSVer;
+#endif
+
+    /* Support for DGA */
+    int                 numDGAModes;
+    /*DGAModePtr          DGAModes;*/
+    int                 DGAactive;
+    int                 DGAViewportStatus;
+
+    /* The various wait handlers. */
+    int                 (*myWaitIdle)(struct _VIA*);
+
+#if 0
+    /* I2C & DDC */
+    I2CBusPtr           I2C_Port1;
+    I2CBusPtr           I2C_Port2;
+    xf86MonPtr          DDC1;
+    xf86MonPtr          DDC2;
+#endif
+
+    /* MHS */
+    int                 IsSecondary;
+    int                 HasSecondary;
+
+#if 0
+    /* Capture information */
+    VIACAPINFO     CapInfo[2];      /* 2 capture information */
+#endif
+
+/*
+    u_int32_t            Cap0_Deinterlace;
+    u_int32_t            Cap1_Deinterlace;
+
+    int                 Cap0_FieldSwap;
+    int                 NoCap0_HFilter;
+    int                 Capture_OverScanOff;
+    int                 NoMPEGHQV_VFilter;
+*/
+#ifdef XF86DRI
+    int 		directRenderingEnabled;
+    DRIInfoPtr		pDRIInfo;
+    int 		drmFD;
+    int 		numVisualConfigs;
+    __GLXvisualConfig* 	pVisualConfigs;
+    VIAConfigPrivPtr 	pVisualConfigsPriv;
+    unsigned long 	agpHandle;
+    unsigned long 	registerHandle;
+    u_int32_t            agpAddr;
+    unsigned char 	*agpBase;
+    unsigned int 	agpSize;
+    int  		IsPCI;
+    int  		drixinerama;
+#else
+    int 		drmFD;
+    unsigned long 	agpHandle;
+    unsigned long 	registerHandle;
+    unsigned long 	agpAddr;
+    unsigned char 	*agpBase;
+    unsigned int 	agpSize;
+    int  		IsPCI;
+#endif
+
+    int     V4LEnabled;
+    u_int16_t    ActiveDevice;	/* if SAMM, non-equal pBIOSInfo->ActiveDevice */
+    unsigned char       *CursorImage;
+    u_int32_t    CursorFG;
+    u_int32_t    CursorBG;
+    u_int32_t    CursorMC;
+
+    unsigned char	MemClk;
+    int 		EnableExtendedFIFO;
+    VIADRIPtr		devPrivate;
+} VIARec, *VIAPtr;
+
+
+/* Shortcuts.  These depend on a local symbol "pVia". */
+
+#define WaitIdle()      pVia->myWaitIdle(pVia)
+#define VIAPTR(p)       ((VIAPtr)((p)->driverPrivate))
+
+#endif /* _VIA_DRIVER_H */
+
diff --git a/src/server/via_priv.h b/src/server/via_priv.h
new file mode 100644
index 0000000..587531b
--- /dev/null
+++ b/src/server/via_priv.h
@@ -0,0 +1,70 @@
+/* $XFree86: xc/programs/Xserver/hw/xfree86/drivers/via/via_priv.h,v 1.3 2003/08/27 15:16:12 tsi Exp $ */
+
+#ifndef _VIA_PRIV_H_
+#define _VIA_PRIV_H_ 1
+
+//#include "ddmpeg.h"
+#include "via_common.h"
+
+#define MEM_BLOCKS		4
+
+typedef struct {
+    unsigned long   base;		/* Offset into fb */
+    int    pool;			/* Pool we drew from */
+    int    drm_fd;			/* Fd in DRM mode */
+    drm_via_mem_t drm;			/* DRM management object */
+    int    slot;			/* Pool 3 slot */
+    void  *pVia;			/* VIA driver pointer */
+    //FBLinearPtr linear;			/* X linear pool info ptr */
+} VIAMem;
+
+typedef VIAMem *VIAMemPtr;
+
+
+
+#if 0
+typedef struct  {
+    unsigned long   gdwVideoFlagTV1;
+    unsigned long   gdwVideoFlagSW;
+    unsigned long   gdwVideoFlagMPEG;
+    unsigned long   gdwAlphaEnabled;		/* For Alpha blending use*/
+
+    VIAMem SWOVMem;
+    VIAMem HQVMem;
+    VIAMem SWfbMem;
+
+    DDPIXELFORMAT DPFsrc; 
+    DDUPDATEOVERLAY UpdateOverlayBackup;    /* For HQVcontrol func use
+					    // To save MPEG updateoverlay info.*/
+
+/* device struct */
+    SWDEVICE   SWDevice;
+    SUBDEVICE   SUBDevice;
+    MPGDEVICE   MPGDevice;
+    OVERLAYRECORD   overlayRecordV1;
+    OVERLAYRECORD   overlayRecordV3;
+
+    BoxRec  AvailFBArea;
+    FBLinearPtr   SWOVlinear;
+
+    int MPEG_ON;
+    int SWVideo_ON;
+
+/*To solve the bandwidth issue */
+    unsigned long   gdwUseExtendedFIFO;
+
+/* For panning mode use */
+    int panning_old_x;
+    int panning_old_y;
+    int panning_x;
+    int panning_y;
+
+/*To solve the bandwidth issue */
+    unsigned char Save_3C4_16;
+    unsigned char Save_3C4_17;
+    unsigned char Save_3C4_18;
+
+} swovRec, *swovPtr;
+#endif
+
+#endif /* _VIA_PRIV_H_ */
diff --git a/src/server/via_regs.h b/src/server/via_regs.h
new file mode 100644
index 0000000..87e1e9d
--- /dev/null
+++ b/src/server/via_regs.h
@@ -0,0 +1,198 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+/*************************************************************************
+ *
+ *  File:       via_regs.c
+ *  Content:    The defines of Via registers
+ *
+ ************************************************************************/
+
+#ifndef _VIA_REGS_H
+#define _VIA_REGS_H
+
+#include "via_driver.h"
+
+/*#define VIA_SERIES(chip)  (chip == VIA_CLE266)*/
+
+#define PCI_VIA_VENDOR_ID       0x1106
+
+#define PCI_CHIP_CLE3122        0x3122
+#define PCI_CHIP_CLE3022        0x3022
+#define PCI_CHIP_VT3205         0x3205
+#define PCI_CHIP_VT7205         0x7205
+#define PCI_CHIP_VT3204         0x3108
+#define PCI_CHIP_VT3259         0x3118
+#define PCI_CHIP_VT3344         0x3344
+
+
+#define BIOS_BSIZE              1024
+#define BIOS_BASE               0xc0000
+
+
+#define VIA_MMIO_REGSIZE        0x9000
+#define VIA_MMIO_REGBASE        0x0
+#define VIA_MMIO_VGABASE        0x8000
+#define VIA_MMIO_BLTBASE        0x200000
+#define VIA_MMIO_BLTSIZE        0x10000
+
+
+/* defines for VIA 2D registers */
+#define VIA_REG_GECMD           0x000
+#define VIA_REG_GEMODE          0x004
+#define VIA_REG_GESTATUS        0x004       /* as same as VIA_REG_GEMODE */
+#define VIA_REG_SRCPOS          0x008
+#define VIA_REG_DSTPOS          0x00C
+#define VIA_REG_LINE_K1K2       0x008
+#define VIA_REG_LINE_XY         0x00C
+#define VIA_REG_DIMENSION       0x010       /* width and height */
+#define VIA_REG_PATADDR         0x014
+#define VIA_REG_FGCOLOR         0x018
+#define VIA_REG_DSTCOLORKEY     0x018       /* as same as VIA_REG_FG */
+#define VIA_REG_BGCOLOR         0x01C
+#define VIA_REG_SRCCOLORKEY     0x01C       /* as same as VIA_REG_BG */
+#define VIA_REG_CLIPTL          0x020       /* top and left of clipping */
+#define VIA_REG_CLIPBR          0x024       /* bottom and right of clipping */
+#define VIA_REG_OFFSET          0x028
+#define VIA_REG_LINE_ERROR      0x028
+#define VIA_REG_KEYCONTROL      0x02C       /* color key control */
+#define VIA_REG_SRCBASE         0x030
+#define VIA_REG_DSTBASE         0x034
+#define VIA_REG_PITCH           0x038       /* pitch of src and dst */
+#define VIA_REG_MONOPAT0        0x03C
+#define VIA_REG_MONOPAT1        0x040
+#define VIA_REG_COLORPAT        0x100       /* from 0x100 to 0x1ff */
+
+
+
+/* defines for VIA video registers */
+#define VIA_REG_INTERRUPT       0x200
+#define VIA_REG_CRTCSTART       0x214
+
+
+/* defines for VIA HW cursor registers */
+#define VIA_REG_CURSOR_MODE     0x2D0
+#define VIA_REG_CURSOR_POS      0x2D4
+#define VIA_REG_CURSOR_ORG      0x2D8
+#define VIA_REG_CURSOR_BG       0x2DC
+#define VIA_REG_CURSOR_FG       0x2E0
+
+
+/* defines for VIA 3D registers */
+#define VIA_REG_STATUS          0x400
+#define VIA_REG_TRANSET         0x43C
+#define VIA_REG_TRANSPACE       0x440
+
+/* VIA_REG_STATUS(0x400): Engine Status */
+#define VIA_CMD_RGTR_BUSY       0x00000080  /* Command Regulator is busy */
+#define VIA_2D_ENG_BUSY         0x00000001  /* 2D Engine is busy */
+#define VIA_3D_ENG_BUSY         0x00000002  /* 3D Engine is busy */
+#define VIA_VR_QUEUE_BUSY       0x00020000 /* Virtual Queue is busy */
+
+
+/* VIA_REG_GECMD(0x00): 2D Engine Command  */
+#define VIA_GEC_NOOP            0x00000000
+#define VIA_GEC_BLT             0x00000001
+#define VIA_GEC_LINE            0x00000005
+
+#define VIA_GEC_SRC_XY          0x00000000
+#define VIA_GEC_SRC_LINEAR      0x00000010
+#define VIA_GEC_DST_XY          0x00000000
+#define VIA_GEC_DST_LINRAT      0x00000020
+
+#define VIA_GEC_SRC_FB          0x00000000
+#define VIA_GEC_SRC_SYS         0x00000040
+#define VIA_GEC_DST_FB          0x00000000
+#define VIA_GEC_DST_SYS         0x00000080
+
+#define VIA_GEC_SRC_MONO        0x00000100  /* source is mono */
+#define VIA_GEC_PAT_MONO        0x00000200  /* pattern is mono */
+
+#define VIA_GEC_MSRC_OPAQUE     0x00000000  /* mono src is opaque */
+#define VIA_GEC_MSRC_TRANS      0x00000400  /* mono src is transparent */
+
+#define VIA_GEC_PAT_FB          0x00000000  /* pattern is in frame buffer */
+#define VIA_GEC_PAT_REG         0x00000800  /* pattern is from reg setting */
+
+#define VIA_GEC_CLIP_DISABLE    0x00000000
+#define VIA_GEC_CLIP_ENABLE     0x00001000
+
+#define VIA_GEC_FIXCOLOR_PAT    0x00002000
+
+#define VIA_GEC_INCX            0x00000000
+#define VIA_GEC_DECY            0x00004000
+#define VIA_GEC_INCY            0x00000000
+#define VIA_GEC_DECX            0x00008000
+
+#define VIA_GEC_MPAT_OPAQUE     0x00000000  /* mono pattern is opaque */
+#define VIA_GEC_MPAT_TRANS      0x00010000  /* mono pattern is transparent */
+
+#define VIA_GEC_MONO_UNPACK     0x00000000
+#define VIA_GEC_MONO_PACK       0x00020000
+#define VIA_GEC_MONO_DWORD      0x00000000
+#define VIA_GEC_MONO_WORD       0x00040000
+#define VIA_GEC_MONO_BYTE       0x00080000
+
+#define VIA_GEC_LASTPIXEL_ON    0x00000000
+#define VIA_GEC_LASTPIXEL_OFF   0x00100000
+#define VIA_GEC_X_MAJOR         0x00000000
+#define VIA_GEC_Y_MAJOR         0x00200000
+#define VIA_GEC_QUICK_START     0x00800000
+
+
+/* VIA_REG_GEMODE(0x04): GE mode */
+#define VIA_GEM_8bpp            0x00000000
+#define VIA_GEM_16bpp           0x00000100
+#define VIA_GEM_32bpp           0x00000300
+
+#define VIA_GEM_640             0x00000000   /* 640*480 */
+#define VIA_GEM_800             0x00000400   /* 800*600 */
+#define VIA_GEM_1024            0x00000800   /* 1024*768 */
+#define VIA_GEM_1280            0x00000C00   /* 1280*1024 */
+#define VIA_GEM_1600            0x00001000   /* 1600*1200 */
+#define VIA_GEM_2048            0x00001400   /* 2048*1536 */
+
+/* VIA_REG_PITCH(0x38): Pitch Setting */
+#define VIA_PITCH_ENABLE        0x80000000
+
+
+#define MAXLOOP                 0xffffff
+
+
+#define VerticalRetraceWait() \
+{ \
+    VGAOUT8(vgaCRIndex, 0x17); \
+    if (VGAIN8(vgaCRReg) & 0x80) { \
+        while ((VGAIN8(vgaIOBase + 0x0a) & 0x08) == 0x00) ; \
+        while ((VGAIN8(vgaIOBase + 0x0a) & 0x08) == 0x08) ; \
+        while ((VGAIN8(vgaIOBase + 0x0a) & 0x08) == 0x00) ; \
+    } \
+}
+
+
+#define VIASETREG(addr, data)   *(volatile unsigned int *)(pVia->MapBase + (addr)) = (data)
+#define VIAGETREG(addr)         *(volatile unsigned int *)(pVia->MapBase + (addr))
+
+
+#endif /* _VIA_REGS_H */
diff --git a/src/via_3d_reg.h b/src/via_3d_reg.h
new file mode 100644
index 0000000..77b24dc
--- /dev/null
+++ b/src/via_3d_reg.h
@@ -0,0 +1,1652 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef VIA_3D_REG_H
+#define VIA_3D_REG_H
+#define HC_REG_BASE             0x0400
+
+#define HC_ParaN_MASK           0xffffffff
+#define HC_Para_MASK            0x00ffffff
+#define HC_SubA_MASK            0xff000000
+#define HC_SubA_SHIFT           24
+/* Transmission Setting
+ */
+#define HC_REG_TRANS_SET        0x003c
+#define HC_ParaSubType_MASK     0xff000000
+#define HC_ParaType_MASK        0x00ff0000
+#define HC_ParaOS_MASK          0x0000ff00
+#define HC_ParaAdr_MASK         0x000000ff
+#define HC_ParaSubType_SHIFT    24
+#define HC_ParaType_SHIFT       16
+#define HC_ParaOS_SHIFT         8
+#define HC_ParaAdr_SHIFT        0
+
+#define HC_ParaType_CmdVdata    0x0000
+#define HC_ParaType_NotTex      0x0001
+#define HC_ParaType_Tex         0x0002
+#define HC_ParaType_Palette     0x0003
+#define HC_ParaType_PreCR       0x0010
+#define HC_ParaType_Auto        0x00fe
+
+/* Transmission Space
+ */
+#define HC_REG_Hpara0           0x0040          
+#define HC_REG_HpataAF          0x02fc          
+
+/* Read
+ */
+#define HC_REG_HREngSt          0x0000
+#define HC_REG_HRFIFOempty      0x0004
+#define HC_REG_HRFIFOfull       0x0008
+#define HC_REG_HRErr            0x000c
+#define HC_REG_FIFOstatus       0x0010
+/* HC_REG_HREngSt          0x0000
+ */
+#define HC_HDASZC_MASK          0x00010000
+#define HC_HSGEMI_MASK          0x0000f000
+#define HC_HLGEMISt_MASK        0x00000f00
+#define HC_HCRSt_MASK           0x00000080
+#define HC_HSE0St_MASK          0x00000040
+#define HC_HSE1St_MASK          0x00000020
+#define HC_HPESt_MASK           0x00000010
+#define HC_HXESt_MASK           0x00000008
+#define HC_HBESt_MASK           0x00000004
+#define HC_HE2St_MASK           0x00000002
+#define HC_HE3St_MASK           0x00000001
+/* HC_REG_HRFIFOempty      0x0004
+ */
+#define HC_HRZDempty_MASK       0x00000010
+#define HC_HRTXAempty_MASK      0x00000008
+#define HC_HRTXDempty_MASK      0x00000004
+#define HC_HWZDempty_MASK       0x00000002
+#define HC_HWCDempty_MASK       0x00000001
+/* HC_REG_HRFIFOfull       0x0008
+ */
+#define HC_HRZDfull_MASK        0x00000010
+#define HC_HRTXAfull_MASK       0x00000008
+#define HC_HRTXDfull_MASK       0x00000004
+#define HC_HWZDfull_MASK        0x00000002
+#define HC_HWCDfull_MASK        0x00000001
+/* HC_REG_HRErr            0x000c
+ */
+#define HC_HAGPCMErr_MASK       0x80000000
+#define HC_HAGPCMErrC_MASK      0x70000000
+/* HC_REG_FIFOstatus       0x0010
+ */
+#define HC_HRFIFOATall_MASK     0x80000000
+#define HC_HRFIFOATbusy_MASK    0x40000000
+#define HC_HRATFGMDo_MASK       0x00000100
+#define HC_HRATFGMDi_MASK       0x00000080
+#define HC_HRATFRZD_MASK        0x00000040
+#define HC_HRATFRTXA_MASK       0x00000020
+#define HC_HRATFRTXD_MASK       0x00000010
+#define HC_HRATFWZD_MASK        0x00000008
+#define HC_HRATFWCD_MASK        0x00000004
+#define HC_HRATTXTAG_MASK       0x00000002
+#define HC_HRATTXCH_MASK        0x00000001
+
+/* AGP Command Setting
+ */
+#define HC_SubA_HAGPBstL        0x0060
+#define HC_SubA_HAGPBendL       0x0061
+#define HC_SubA_HAGPCMNT        0x0062
+#define HC_SubA_HAGPBpL         0x0063
+#define HC_SubA_HAGPBpH         0x0064
+/* HC_SubA_HAGPCMNT        0x0062
+ */
+#define HC_HAGPCMNT_MASK        0x00800000
+#define HC_HCmdErrClr_MASK      0x00400000
+#define HC_HAGPBendH_MASK       0x0000ff00
+#define HC_HAGPBstH_MASK        0x000000ff
+#define HC_HAGPBendH_SHIFT      8
+#define HC_HAGPBstH_SHIFT       0
+/* HC_SubA_HAGPBpL         0x0063
+ */
+#define HC_HAGPBpL_MASK         0x00fffffc
+#define HC_HAGPBpID_MASK        0x00000003
+#define HC_HAGPBpID_PAUSE       0x00000000
+#define HC_HAGPBpID_JUMP        0x00000001
+#define HC_HAGPBpID_STOP        0x00000002
+/* HC_SubA_HAGPBpH         0x0064
+ */
+#define HC_HAGPBpH_MASK         0x00ffffff
+
+/* Miscellaneous Settings
+ */
+#define HC_SubA_HClipTB         0x0070
+#define HC_SubA_HClipLR         0x0071
+#define HC_SubA_HFPClipTL       0x0072
+#define HC_SubA_HFPClipBL       0x0073
+#define HC_SubA_HFPClipLL       0x0074
+#define HC_SubA_HFPClipRL       0x0075
+#define HC_SubA_HFPClipTBH      0x0076
+#define HC_SubA_HFPClipLRH      0x0077
+#define HC_SubA_HLP             0x0078
+#define HC_SubA_HLPRF           0x0079
+#define HC_SubA_HSolidCL        0x007a
+#define HC_SubA_HPixGC          0x007b
+#define HC_SubA_HSPXYOS         0x007c
+#define HC_SubA_HVertexCNT      0x007d
+
+#define HC_HClipT_MASK          0x00fff000
+#define HC_HClipT_SHIFT         12
+#define HC_HClipB_MASK          0x00000fff
+#define HC_HClipB_SHIFT         0
+#define HC_HClipL_MASK          0x00fff000
+#define HC_HClipL_SHIFT         12
+#define HC_HClipR_MASK          0x00000fff
+#define HC_HClipR_SHIFT         0
+#define HC_HFPClipBH_MASK       0x0000ff00
+#define HC_HFPClipBH_SHIFT      8
+#define HC_HFPClipTH_MASK       0x000000ff
+#define HC_HFPClipTH_SHIFT      0
+#define HC_HFPClipRH_MASK       0x0000ff00
+#define HC_HFPClipRH_SHIFT      8
+#define HC_HFPClipLH_MASK       0x000000ff
+#define HC_HFPClipLH_SHIFT      0
+#define HC_HSolidCH_MASK        0x000000ff
+#define HC_HPixGC_MASK          0x00800000
+#define HC_HSPXOS_MASK          0x00fff000
+#define HC_HSPXOS_SHIFT         12
+#define HC_HSPYOS_MASK          0x00000fff
+
+/* Command
+ * Command A
+ */
+#define HC_HCmdHeader_MASK      0xfe000000  /*0xffe00000*/
+#define HC_HE3Fire_MASK         0x00100000
+#define HC_HPMType_MASK         0x000f0000
+#define HC_HEFlag_MASK          0x0000e000
+#define HC_HShading_MASK        0x00001c00
+#define HC_HPMValidN_MASK       0x00000200
+#define HC_HPLEND_MASK          0x00000100
+#define HC_HVCycle_MASK         0x000000ff
+#define HC_HVCycle_Style_MASK   0x000000c0
+#define HC_HVCycle_ChgA_MASK    0x00000030
+#define HC_HVCycle_ChgB_MASK    0x0000000c
+#define HC_HVCycle_ChgC_MASK    0x00000003
+#define HC_HPMType_Point        0x00000000
+#define HC_HPMType_Line         0x00010000
+#define HC_HPMType_Tri          0x00020000
+#define HC_HPMType_TriWF        0x00040000
+#define HC_HEFlag_NoAA          0x00000000
+#define HC_HEFlag_ab            0x00008000
+#define HC_HEFlag_bc            0x00004000
+#define HC_HEFlag_ca            0x00002000
+#define HC_HShading_Solid       0x00000000
+#define HC_HShading_FlatA       0x00000400
+#define HC_HShading_FlatB       0x00000800
+#define HC_HShading_FlatC       0x00000c00
+#define HC_HShading_Gouraud     0x00001000
+#define HC_HVCycle_Full         0x00000000
+#define HC_HVCycle_AFP          0x00000040
+#define HC_HVCycle_One          0x000000c0
+#define HC_HVCycle_NewA         0x00000000
+#define HC_HVCycle_AA           0x00000010
+#define HC_HVCycle_AB           0x00000020
+#define HC_HVCycle_AC           0x00000030
+#define HC_HVCycle_NewB         0x00000000
+#define HC_HVCycle_BA           0x00000004
+#define HC_HVCycle_BB           0x00000008
+#define HC_HVCycle_BC           0x0000000c
+#define HC_HVCycle_NewC         0x00000000
+#define HC_HVCycle_CA           0x00000001
+#define HC_HVCycle_CB           0x00000002
+#define HC_HVCycle_CC           0x00000003
+
+/* Command B
+ */
+#define HC_HLPrst_MASK          0x00010000
+#define HC_HLLastP_MASK         0x00008000
+#define HC_HVPMSK_MASK          0x00007f80
+#define HC_HBFace_MASK          0x00000040
+#define HC_H2nd1VT_MASK         0x0000003f
+#define HC_HVPMSK_X             0x00004000
+#define HC_HVPMSK_Y             0x00002000
+#define HC_HVPMSK_Z             0x00001000
+#define HC_HVPMSK_W             0x00000800
+#define HC_HVPMSK_Cd            0x00000400
+#define HC_HVPMSK_Cs            0x00000200
+#define HC_HVPMSK_S             0x00000100
+#define HC_HVPMSK_T             0x00000080
+
+/* Enable Setting
+ */
+#define HC_SubA_HEnable         0x0000
+#define HC_HenTXEnvMap_MASK     0x00200000
+#define HC_HenVertexCNT_MASK    0x00100000
+#define HC_HenCPUDAZ_MASK       0x00080000
+#define HC_HenDASZWC_MASK       0x00040000
+#define HC_HenFBCull_MASK       0x00020000
+#define HC_HenCW_MASK           0x00010000
+#define HC_HenAA_MASK           0x00008000
+#define HC_HenST_MASK           0x00004000
+#define HC_HenZT_MASK           0x00002000
+#define HC_HenZW_MASK           0x00001000
+#define HC_HenAT_MASK           0x00000800
+#define HC_HenAW_MASK           0x00000400
+#define HC_HenSP_MASK           0x00000200
+#define HC_HenLP_MASK           0x00000100
+#define HC_HenTXCH_MASK         0x00000080
+#define HC_HenTXMP_MASK         0x00000040
+#define HC_HenTXPP_MASK         0x00000020
+#define HC_HenTXTR_MASK         0x00000010
+#define HC_HenCS_MASK           0x00000008
+#define HC_HenFOG_MASK          0x00000004
+#define HC_HenABL_MASK          0x00000002
+#define HC_HenDT_MASK           0x00000001
+
+/* Z Setting
+ */
+#define HC_SubA_HZWBBasL        0x0010
+#define HC_SubA_HZWBBasH        0x0011
+#define HC_SubA_HZWBType        0x0012
+#define HC_SubA_HZBiasL         0x0013
+#define HC_SubA_HZWBend         0x0014
+#define HC_SubA_HZWTMD          0x0015
+#define HC_SubA_HZWCDL          0x0016
+#define HC_SubA_HZWCTAGnum      0x0017
+#define HC_SubA_HZCYNum         0x0018
+#define HC_SubA_HZWCFire        0x0019
+/* HC_SubA_HZWBType
+ */
+#define HC_HZWBType_MASK        0x00800000
+#define HC_HZBiasedWB_MASK      0x00400000
+#define HC_HZONEasFF_MASK       0x00200000
+#define HC_HZOONEasFF_MASK      0x00100000
+#define HC_HZWBFM_MASK          0x00030000
+#define HC_HZWBLoc_MASK         0x0000c000
+#define HC_HZWBPit_MASK         0x00003fff
+#define HC_HZWBFM_16            0x00000000
+#define HC_HZWBFM_32            0x00020000
+#define HC_HZWBFM_24            0x00030000
+#define HC_HZWBLoc_Local        0x00000000
+#define HC_HZWBLoc_SyS          0x00004000
+/* HC_SubA_HZWBend
+ */
+#define HC_HZWBend_MASK         0x00ffe000
+#define HC_HZBiasH_MASK         0x000000ff
+#define HC_HZWBend_SHIFT        10
+/* HC_SubA_HZWTMD
+ */
+#define HC_HZWTMD_MASK          0x00070000
+#define HC_HEBEBias_MASK        0x00007f00
+#define HC_HZNF_MASK            0x000000ff
+#define HC_HZWTMD_NeverPass     0x00000000
+#define HC_HZWTMD_LT            0x00010000
+#define HC_HZWTMD_EQ            0x00020000
+#define HC_HZWTMD_LE            0x00030000
+#define HC_HZWTMD_GT            0x00040000
+#define HC_HZWTMD_NE            0x00050000
+#define HC_HZWTMD_GE            0x00060000
+#define HC_HZWTMD_AllPass       0x00070000
+#define HC_HEBEBias_SHIFT       8
+/* HC_SubA_HZWCDL          0x0016 
+ */
+#define HC_HZWCDL_MASK          0x00ffffff
+/* HC_SubA_HZWCTAGnum      0x0017 
+ */
+#define HC_HZWCTAGnum_MASK      0x00ff0000
+#define HC_HZWCTAGnum_SHIFT     16
+#define HC_HZWCDH_MASK          0x000000ff
+#define HC_HZWCDH_SHIFT         0
+/* HC_SubA_HZCYNum         0x0018
+ */
+#define HC_HZCYNum_MASK         0x00030000
+#define HC_HZCYNum_SHIFT        16
+#define HC_HZWCQWnum_MASK       0x00003fff
+#define HC_HZWCQWnum_SHIFT      0
+/* HC_SubA_HZWCFire        0x0019
+ */
+#define HC_ZWCFire_MASK         0x00010000
+#define HC_HZWCQWnumLast_MASK   0x00003fff
+#define HC_HZWCQWnumLast_SHIFT  0
+
+/* Stencil Setting
+ */
+#define HC_SubA_HSTREF          0x0023
+#define HC_SubA_HSTMD           0x0024
+/* HC_SubA_HSBFM
+ */
+#define HC_HSBFM_MASK           0x00030000
+#define HC_HSBLoc_MASK          0x0000c000
+#define HC_HSBPit_MASK          0x00003fff
+/* HC_SubA_HSTREF
+ */
+#define HC_HSTREF_MASK          0x00ff0000
+#define HC_HSTOPMSK_MASK        0x0000ff00
+#define HC_HSTBMSK_MASK         0x000000ff
+#define HC_HSTREF_SHIFT         16
+#define HC_HSTOPMSK_SHIFT       8
+/* HC_SubA_HSTMD
+ */
+#define HC_HSTMD_MASK           0x00070000
+#define HC_HSTOPSF_MASK         0x000001c0
+#define HC_HSTOPSPZF_MASK       0x00000038
+#define HC_HSTOPSPZP_MASK       0x00000007
+#define HC_HSTMD_NeverPass      0x00000000
+#define HC_HSTMD_LT             0x00010000
+#define HC_HSTMD_EQ             0x00020000
+#define HC_HSTMD_LE             0x00030000
+#define HC_HSTMD_GT             0x00040000
+#define HC_HSTMD_NE             0x00050000
+#define HC_HSTMD_GE             0x00060000
+#define HC_HSTMD_AllPass        0x00070000
+#define HC_HSTOPSF_KEEP         0x00000000
+#define HC_HSTOPSF_ZERO         0x00000040
+#define HC_HSTOPSF_REPLACE      0x00000080
+#define HC_HSTOPSF_INCRSAT      0x000000c0
+#define HC_HSTOPSF_DECRSAT      0x00000100
+#define HC_HSTOPSF_INVERT       0x00000140
+#define HC_HSTOPSF_INCR         0x00000180
+#define HC_HSTOPSF_DECR         0x000001c0
+#define HC_HSTOPSPZF_KEEP       0x00000000
+#define HC_HSTOPSPZF_ZERO       0x00000008
+#define HC_HSTOPSPZF_REPLACE    0x00000010
+#define HC_HSTOPSPZF_INCRSAT    0x00000018
+#define HC_HSTOPSPZF_DECRSAT    0x00000020
+#define HC_HSTOPSPZF_INVERT     0x00000028
+#define HC_HSTOPSPZF_INCR       0x00000030
+#define HC_HSTOPSPZF_DECR       0x00000038
+#define HC_HSTOPSPZP_KEEP       0x00000000
+#define HC_HSTOPSPZP_ZERO       0x00000001
+#define HC_HSTOPSPZP_REPLACE    0x00000002
+#define HC_HSTOPSPZP_INCRSAT    0x00000003
+#define HC_HSTOPSPZP_DECRSAT    0x00000004
+#define HC_HSTOPSPZP_INVERT     0x00000005
+#define HC_HSTOPSPZP_INCR       0x00000006
+#define HC_HSTOPSPZP_DECR       0x00000007
+
+/* Alpha Setting
+ */
+#define HC_SubA_HABBasL         0x0030
+#define HC_SubA_HABBasH         0x0031
+#define HC_SubA_HABFM           0x0032
+#define HC_SubA_HATMD           0x0033
+#define HC_SubA_HABLCsat        0x0034
+#define HC_SubA_HABLCop         0x0035
+#define HC_SubA_HABLAsat        0x0036
+#define HC_SubA_HABLAop         0x0037
+#define HC_SubA_HABLRCa         0x0038
+#define HC_SubA_HABLRFCa        0x0039
+#define HC_SubA_HABLRCbias      0x003a
+#define HC_SubA_HABLRCb         0x003b
+#define HC_SubA_HABLRFCb        0x003c
+#define HC_SubA_HABLRAa         0x003d
+#define HC_SubA_HABLRAb         0x003e
+/* HC_SubA_HABFM
+ */
+#define HC_HABFM_MASK           0x00030000
+#define HC_HABLoc_MASK          0x0000c000
+#define HC_HABPit_MASK          0x000007ff
+/* HC_SubA_HATMD
+ */
+#define HC_HATMD_MASK           0x00000700
+#define HC_HATREF_MASK          0x000000ff
+#define HC_HATMD_NeverPass      0x00000000
+#define HC_HATMD_LT             0x00000100
+#define HC_HATMD_EQ             0x00000200
+#define HC_HATMD_LE             0x00000300
+#define HC_HATMD_GT             0x00000400
+#define HC_HATMD_NE             0x00000500
+#define HC_HATMD_GE             0x00000600
+#define HC_HATMD_AllPass        0x00000700
+/* HC_SubA_HABLCsat
+ */
+#define HC_HABLCsat_MASK        0x00010000
+#define HC_HABLCa_MASK          0x0000fc00
+#define HC_HABLCa_C_MASK        0x0000c000
+#define HC_HABLCa_OPC_MASK      0x00003c00
+#define HC_HABLFCa_MASK         0x000003f0
+#define HC_HABLFCa_C_MASK       0x00000300
+#define HC_HABLFCa_OPC_MASK     0x000000f0
+#define HC_HABLCbias_MASK       0x0000000f
+#define HC_HABLCbias_C_MASK     0x00000008
+#define HC_HABLCbias_OPC_MASK   0x00000007
+/*-- Define the input color.
+ */
+#define HC_XC_Csrc              0x00000000
+#define HC_XC_Cdst              0x00000001
+#define HC_XC_Asrc              0x00000002
+#define HC_XC_Adst              0x00000003
+#define HC_XC_Fog               0x00000004
+#define HC_XC_HABLRC            0x00000005
+#define HC_XC_minSrcDst         0x00000006
+#define HC_XC_maxSrcDst         0x00000007
+#define HC_XC_mimAsrcInvAdst    0x00000008
+#define HC_XC_OPC               0x00000000
+#define HC_XC_InvOPC            0x00000010
+#define HC_XC_OPCp5             0x00000020
+/*-- Define the input Alpha
+ */
+#define HC_XA_OPA               0x00000000
+#define HC_XA_InvOPA            0x00000010
+#define HC_XA_OPAp5             0x00000020
+#define HC_XA_0                 0x00000000
+#define HC_XA_Asrc              0x00000001
+#define HC_XA_Adst              0x00000002
+#define HC_XA_Fog               0x00000003
+#define HC_XA_minAsrcFog        0x00000004
+#define HC_XA_minAsrcAdst       0x00000005
+#define HC_XA_maxAsrcFog        0x00000006
+#define HC_XA_maxAsrcAdst       0x00000007
+#define HC_XA_HABLRA            0x00000008
+#define HC_XA_minAsrcInvAdst    0x00000008
+#define HC_XA_HABLFRA           0x00000009
+/*--
+ */
+#define HC_HABLCa_OPC           (HC_XC_OPC << 10)
+#define HC_HABLCa_InvOPC        (HC_XC_InvOPC << 10)
+#define HC_HABLCa_OPCp5         (HC_XC_OPCp5 << 10)
+#define HC_HABLCa_Csrc          (HC_XC_Csrc << 10)
+#define HC_HABLCa_Cdst          (HC_XC_Cdst << 10)
+#define HC_HABLCa_Asrc          (HC_XC_Asrc << 10)
+#define HC_HABLCa_Adst          (HC_XC_Adst << 10)
+#define HC_HABLCa_Fog           (HC_XC_Fog << 10)
+#define HC_HABLCa_HABLRCa       (HC_XC_HABLRC << 10)
+#define HC_HABLCa_minSrcDst     (HC_XC_minSrcDst << 10)
+#define HC_HABLCa_maxSrcDst     (HC_XC_maxSrcDst << 10)
+#define HC_HABLFCa_OPC              (HC_XC_OPC << 4)
+#define HC_HABLFCa_InvOPC           (HC_XC_InvOPC << 4)
+#define HC_HABLFCa_OPCp5            (HC_XC_OPCp5 << 4)
+#define HC_HABLFCa_Csrc             (HC_XC_Csrc << 4)
+#define HC_HABLFCa_Cdst             (HC_XC_Cdst << 4)
+#define HC_HABLFCa_Asrc             (HC_XC_Asrc << 4)
+#define HC_HABLFCa_Adst             (HC_XC_Adst << 4)
+#define HC_HABLFCa_Fog              (HC_XC_Fog << 4)
+#define HC_HABLFCa_HABLRCa          (HC_XC_HABLRC << 4)
+#define HC_HABLFCa_minSrcDst        (HC_XC_minSrcDst << 4)
+#define HC_HABLFCa_maxSrcDst        (HC_XC_maxSrcDst << 4)
+#define HC_HABLFCa_mimAsrcInvAdst   (HC_XC_mimAsrcInvAdst << 4)
+#define HC_HABLCbias_HABLRCbias 0x00000000
+#define HC_HABLCbias_Asrc       0x00000001
+#define HC_HABLCbias_Adst       0x00000002
+#define HC_HABLCbias_Fog        0x00000003
+#define HC_HABLCbias_Cin        0x00000004
+/* HC_SubA_HABLCop         0x0035
+ */
+#define HC_HABLdot_MASK         0x00010000
+#define HC_HABLCop_MASK         0x00004000
+#define HC_HABLCb_MASK          0x00003f00
+#define HC_HABLCb_C_MASK        0x00003000
+#define HC_HABLCb_OPC_MASK      0x00000f00
+#define HC_HABLFCb_MASK         0x000000fc
+#define HC_HABLFCb_C_MASK       0x000000c0
+#define HC_HABLFCb_OPC_MASK     0x0000003c
+#define HC_HABLCshift_MASK      0x00000003
+#define HC_HABLCb_OPC           (HC_XC_OPC << 8)
+#define HC_HABLCb_InvOPC        (HC_XC_InvOPC << 8)
+#define HC_HABLCb_OPCp5         (HC_XC_OPCp5 << 8)
+#define HC_HABLCb_Csrc          (HC_XC_Csrc << 8)
+#define HC_HABLCb_Cdst          (HC_XC_Cdst << 8)
+#define HC_HABLCb_Asrc          (HC_XC_Asrc << 8)
+#define HC_HABLCb_Adst          (HC_XC_Adst << 8)
+#define HC_HABLCb_Fog           (HC_XC_Fog << 8)
+#define HC_HABLCb_HABLRCa       (HC_XC_HABLRC << 8)
+#define HC_HABLCb_minSrcDst     (HC_XC_minSrcDst << 8)
+#define HC_HABLCb_maxSrcDst     (HC_XC_maxSrcDst << 8)
+#define HC_HABLFCb_OPC              (HC_XC_OPC << 2)
+#define HC_HABLFCb_InvOPC           (HC_XC_InvOPC << 2)
+#define HC_HABLFCb_OPCp5            (HC_XC_OPCp5 << 2)
+#define HC_HABLFCb_Csrc             (HC_XC_Csrc << 2)
+#define HC_HABLFCb_Cdst             (HC_XC_Cdst << 2)
+#define HC_HABLFCb_Asrc             (HC_XC_Asrc << 2)
+#define HC_HABLFCb_Adst             (HC_XC_Adst << 2)
+#define HC_HABLFCb_Fog              (HC_XC_Fog << 2)
+#define HC_HABLFCb_HABLRCb          (HC_XC_HABLRC << 2)
+#define HC_HABLFCb_minSrcDst        (HC_XC_minSrcDst << 2)
+#define HC_HABLFCb_maxSrcDst        (HC_XC_maxSrcDst << 2)
+#define HC_HABLFCb_mimAsrcInvAdst   (HC_XC_mimAsrcInvAdst << 2)
+/* HC_SubA_HABLAsat        0x0036
+ */
+#define HC_HABLAsat_MASK        0x00010000
+#define HC_HABLAa_MASK          0x0000fc00
+#define HC_HABLAa_A_MASK        0x0000c000
+#define HC_HABLAa_OPA_MASK      0x00003c00
+#define HC_HABLFAa_MASK         0x000003f0
+#define HC_HABLFAa_A_MASK       0x00000300
+#define HC_HABLFAa_OPA_MASK     0x000000f0
+#define HC_HABLAbias_MASK       0x0000000f
+#define HC_HABLAbias_A_MASK     0x00000008
+#define HC_HABLAbias_OPA_MASK   0x00000007
+#define HC_HABLAa_OPA           (HC_XA_OPA << 10)
+#define HC_HABLAa_InvOPA        (HC_XA_InvOPA << 10)
+#define HC_HABLAa_OPAp5         (HC_XA_OPAp5 << 10)
+#define HC_HABLAa_0             (HC_XA_0 << 10)
+#define HC_HABLAa_Asrc          (HC_XA_Asrc << 10)
+#define HC_HABLAa_Adst          (HC_XA_Adst << 10)
+#define HC_HABLAa_Fog           (HC_XA_Fog << 10)
+#define HC_HABLAa_minAsrcFog    (HC_XA_minAsrcFog << 10)
+#define HC_HABLAa_minAsrcAdst   (HC_XA_minAsrcAdst << 10)
+#define HC_HABLAa_maxAsrcFog    (HC_XA_maxAsrcFog << 10)
+#define HC_HABLAa_maxAsrcAdst   (HC_XA_maxAsrcAdst << 10)
+#define HC_HABLAa_HABLRA        (HC_XA_HABLRA << 10)
+#define HC_HABLFAa_OPA          (HC_XA_OPA << 4)
+#define HC_HABLFAa_InvOPA       (HC_XA_InvOPA << 4)
+#define HC_HABLFAa_OPAp5        (HC_XA_OPAp5 << 4)
+#define HC_HABLFAa_0            (HC_XA_0 << 4)
+#define HC_HABLFAa_Asrc         (HC_XA_Asrc << 4)
+#define HC_HABLFAa_Adst         (HC_XA_Adst << 4)
+#define HC_HABLFAa_Fog          (HC_XA_Fog << 4)
+#define HC_HABLFAa_minAsrcFog   (HC_XA_minAsrcFog << 4)
+#define HC_HABLFAa_minAsrcAdst  (HC_XA_minAsrcAdst << 4)
+#define HC_HABLFAa_maxAsrcFog   (HC_XA_maxAsrcFog << 4)
+#define HC_HABLFAa_maxAsrcAdst  (HC_XA_maxAsrcAdst << 4)
+#define HC_HABLFAa_minAsrcInvAdst   (HC_XA_minAsrcInvAdst << 4)
+#define HC_HABLFAa_HABLFRA          (HC_XA_HABLFRA << 4)
+#define HC_HABLAbias_HABLRAbias 0x00000000
+#define HC_HABLAbias_Asrc       0x00000001
+#define HC_HABLAbias_Adst       0x00000002
+#define HC_HABLAbias_Fog        0x00000003
+#define HC_HABLAbias_Aaa        0x00000004
+/* HC_SubA_HABLAop         0x0037
+ */
+#define HC_HABLAop_MASK         0x00004000
+#define HC_HABLAb_MASK          0x00003f00
+#define HC_HABLAb_OPA_MASK      0x00000f00
+#define HC_HABLFAb_MASK         0x000000fc
+#define HC_HABLFAb_OPA_MASK     0x0000003c
+#define HC_HABLAshift_MASK      0x00000003
+#define HC_HABLAb_OPA           (HC_XA_OPA << 8)
+#define HC_HABLAb_InvOPA        (HC_XA_InvOPA << 8)
+#define HC_HABLAb_OPAp5         (HC_XA_OPAp5 << 8)
+#define HC_HABLAb_0             (HC_XA_0 << 8)
+#define HC_HABLAb_Asrc          (HC_XA_Asrc << 8)
+#define HC_HABLAb_Adst          (HC_XA_Adst << 8)
+#define HC_HABLAb_Fog           (HC_XA_Fog << 8)
+#define HC_HABLAb_minAsrcFog    (HC_XA_minAsrcFog << 8)
+#define HC_HABLAb_minAsrcAdst   (HC_XA_minAsrcAdst << 8)
+#define HC_HABLAb_maxAsrcFog    (HC_XA_maxAsrcFog << 8)
+#define HC_HABLAb_maxAsrcAdst   (HC_XA_maxAsrcAdst << 8)
+#define HC_HABLAb_HABLRA        (HC_XA_HABLRA << 8)
+#define HC_HABLFAb_OPA          (HC_XA_OPA << 2)
+#define HC_HABLFAb_InvOPA       (HC_XA_InvOPA << 2)
+#define HC_HABLFAb_OPAp5        (HC_XA_OPAp5 << 2)
+#define HC_HABLFAb_0            (HC_XA_0 << 2)
+#define HC_HABLFAb_Asrc         (HC_XA_Asrc << 2)
+#define HC_HABLFAb_Adst         (HC_XA_Adst << 2)
+#define HC_HABLFAb_Fog          (HC_XA_Fog << 2)
+#define HC_HABLFAb_minAsrcFog   (HC_XA_minAsrcFog << 2)
+#define HC_HABLFAb_minAsrcAdst  (HC_XA_minAsrcAdst << 2)
+#define HC_HABLFAb_maxAsrcFog   (HC_XA_maxAsrcFog << 2)
+#define HC_HABLFAb_maxAsrcAdst  (HC_XA_maxAsrcAdst << 2)
+#define HC_HABLFAb_minAsrcInvAdst   (HC_XA_minAsrcInvAdst << 2)
+#define HC_HABLFAb_HABLFRA          (HC_XA_HABLFRA << 2)
+/* HC_SubA_HABLRAa         0x003d
+ */
+#define HC_HABLRAa_MASK         0x00ff0000
+#define HC_HABLRFAa_MASK        0x0000ff00
+#define HC_HABLRAbias_MASK      0x000000ff
+#define HC_HABLRAa_SHIFT        16
+#define HC_HABLRFAa_SHIFT       8
+/* HC_SubA_HABLRAb         0x003e
+ */
+#define HC_HABLRAb_MASK         0x0000ff00
+#define HC_HABLRFAb_MASK        0x000000ff
+#define HC_HABLRAb_SHIFT        8
+
+/* Destination Setting
+ */
+#define HC_SubA_HDBBasL         0x0040
+#define HC_SubA_HDBBasH         0x0041
+#define HC_SubA_HDBFM           0x0042
+#define HC_SubA_HFBBMSKL        0x0043
+#define HC_SubA_HROP            0x0044
+/* HC_SubA_HDBFM           0x0042
+ */
+#define HC_HDBFM_MASK           0x001f0000
+#define HC_HDBLoc_MASK          0x0000c000
+#define HC_HDBPit_MASK          0x00003fff
+#define HC_HDBFM_RGB555         0x00000000
+#define HC_HDBFM_RGB565         0x00010000
+#define HC_HDBFM_ARGB4444       0x00020000
+#define HC_HDBFM_ARGB1555       0x00030000
+#define HC_HDBFM_BGR555         0x00040000
+#define HC_HDBFM_BGR565         0x00050000
+#define HC_HDBFM_ABGR4444       0x00060000
+#define HC_HDBFM_ABGR1555       0x00070000
+#define HC_HDBFM_ARGB0888       0x00080000
+#define HC_HDBFM_ARGB8888       0x00090000
+#define HC_HDBFM_ABGR0888       0x000a0000
+#define HC_HDBFM_ABGR8888       0x000b0000
+#define HC_HDBLoc_Local         0x00000000
+#define HC_HDBLoc_Sys           0x00004000
+/* HC_SubA_HROP            0x0044
+ */
+#define HC_HROP_MASK            0x00000f00
+#define HC_HFBBMSKH_MASK        0x000000ff
+#define HC_HROP_BLACK           0x00000000
+#define HC_HROP_DPon            0x00000100
+#define HC_HROP_DPna            0x00000200
+#define HC_HROP_Pn              0x00000300
+#define HC_HROP_PDna            0x00000400
+#define HC_HROP_Dn              0x00000500
+#define HC_HROP_DPx             0x00000600
+#define HC_HROP_DPan            0x00000700
+#define HC_HROP_DPa             0x00000800
+#define HC_HROP_DPxn            0x00000900
+#define HC_HROP_D               0x00000a00
+#define HC_HROP_DPno            0x00000b00
+#define HC_HROP_P               0x00000c00
+#define HC_HROP_PDno            0x00000d00
+#define HC_HROP_DPo             0x00000e00
+#define HC_HROP_WHITE           0x00000f00
+
+/* Fog Setting
+ */
+#define HC_SubA_HFogLF          0x0050
+#define HC_SubA_HFogCL          0x0051
+#define HC_SubA_HFogCH          0x0052
+#define HC_SubA_HFogStL         0x0053
+#define HC_SubA_HFogStH         0x0054
+#define HC_SubA_HFogOOdMF       0x0055
+#define HC_SubA_HFogOOdEF       0x0056
+#define HC_SubA_HFogEndL        0x0057
+#define HC_SubA_HFogDenst       0x0058
+/* HC_SubA_FogLF           0x0050
+ */
+#define HC_FogLF_MASK           0x00000010
+#define HC_FogEq_MASK           0x00000008
+#define HC_FogMD_MASK           0x00000007
+#define HC_FogMD_LocalFog        0x00000000
+#define HC_FogMD_LinearFog       0x00000002
+#define HC_FogMD_ExponentialFog  0x00000004
+#define HC_FogMD_Exponential2Fog 0x00000005
+/* #define HC_FogMD_FogTable       0x00000003 */
+
+/* HC_SubA_HFogDenst        0x0058
+ */
+#define HC_FogDenst_MASK        0x001fff00
+#define HC_FogEndL_MASK         0x000000ff
+
+/* Texture subtype definitions
+ */
+#define HC_SubType_Tex0         0x00000000
+#define HC_SubType_Tex1         0x00000001
+#define HC_SubType_TexGeneral   0x000000fe
+
+/* Attribute of texture n
+ */
+#define HC_SubA_HTXnL0BasL      0x0000
+#define HC_SubA_HTXnL1BasL      0x0001
+#define HC_SubA_HTXnL2BasL      0x0002
+#define HC_SubA_HTXnL3BasL      0x0003
+#define HC_SubA_HTXnL4BasL      0x0004
+#define HC_SubA_HTXnL5BasL      0x0005
+#define HC_SubA_HTXnL6BasL      0x0006
+#define HC_SubA_HTXnL7BasL      0x0007
+#define HC_SubA_HTXnL8BasL      0x0008
+#define HC_SubA_HTXnL9BasL      0x0009
+#define HC_SubA_HTXnLaBasL      0x000a
+#define HC_SubA_HTXnLbBasL      0x000b
+#define HC_SubA_HTXnLcBasL      0x000c
+#define HC_SubA_HTXnLdBasL      0x000d
+#define HC_SubA_HTXnLeBasL      0x000e
+#define HC_SubA_HTXnLfBasL      0x000f
+#define HC_SubA_HTXnL10BasL     0x0010
+#define HC_SubA_HTXnL11BasL     0x0011
+#define HC_SubA_HTXnL012BasH    0x0020
+#define HC_SubA_HTXnL345BasH    0x0021
+#define HC_SubA_HTXnL678BasH    0x0022
+#define HC_SubA_HTXnL9abBasH    0x0023
+#define HC_SubA_HTXnLcdeBasH    0x0024
+#define HC_SubA_HTXnLf1011BasH  0x0025
+#define HC_SubA_HTXnL0Pit       0x002b
+#define HC_SubA_HTXnL1Pit       0x002c
+#define HC_SubA_HTXnL2Pit       0x002d
+#define HC_SubA_HTXnL3Pit       0x002e
+#define HC_SubA_HTXnL4Pit       0x002f
+#define HC_SubA_HTXnL5Pit       0x0030
+#define HC_SubA_HTXnL6Pit       0x0031
+#define HC_SubA_HTXnL7Pit       0x0032
+#define HC_SubA_HTXnL8Pit       0x0033
+#define HC_SubA_HTXnL9Pit       0x0034
+#define HC_SubA_HTXnLaPit       0x0035
+#define HC_SubA_HTXnLbPit       0x0036
+#define HC_SubA_HTXnLcPit       0x0037
+#define HC_SubA_HTXnLdPit       0x0038
+#define HC_SubA_HTXnLePit       0x0039
+#define HC_SubA_HTXnLfPit       0x003a
+#define HC_SubA_HTXnL10Pit      0x003b
+#define HC_SubA_HTXnL11Pit      0x003c
+#define HC_SubA_HTXnL0_5WE      0x004b
+#define HC_SubA_HTXnL6_bWE      0x004c
+#define HC_SubA_HTXnLc_11WE     0x004d
+#define HC_SubA_HTXnL0_5HE      0x0051
+#define HC_SubA_HTXnL6_bHE      0x0052
+#define HC_SubA_HTXnLc_11HE     0x0053
+#define HC_SubA_HTXnL0OS        0x0077
+#define HC_SubA_HTXnTB          0x0078
+#define HC_SubA_HTXnMPMD        0x0079
+#define HC_SubA_HTXnCLODu       0x007a
+#define HC_SubA_HTXnFM          0x007b
+#define HC_SubA_HTXnTRCH        0x007c
+#define HC_SubA_HTXnTRCL        0x007d
+#define HC_SubA_HTXnTBC         0x007e
+#define HC_SubA_HTXnTRAH        0x007f
+#define HC_SubA_HTXnTBLCsat     0x0080
+#define HC_SubA_HTXnTBLCop      0x0081
+#define HC_SubA_HTXnTBLMPfog    0x0082
+#define HC_SubA_HTXnTBLAsat     0x0083
+#define HC_SubA_HTXnTBLRCa      0x0085
+#define HC_SubA_HTXnTBLRCb      0x0086
+#define HC_SubA_HTXnTBLRCc      0x0087
+#define HC_SubA_HTXnTBLRCbias   0x0088
+#define HC_SubA_HTXnTBLRAa      0x0089
+#define HC_SubA_HTXnTBLRFog     0x008a
+#define HC_SubA_HTXnBumpM00     0x0090
+#define HC_SubA_HTXnBumpM01     0x0091
+#define HC_SubA_HTXnBumpM10     0x0092
+#define HC_SubA_HTXnBumpM11     0x0093
+#define HC_SubA_HTXnLScale      0x0094
+#define HC_SubA_HTXSMD          0x0000
+/* HC_SubA_HTXnL012BasH    0x0020
+ */
+#define HC_HTXnL0BasH_MASK      0x000000ff
+#define HC_HTXnL1BasH_MASK      0x0000ff00
+#define HC_HTXnL2BasH_MASK      0x00ff0000
+#define HC_HTXnL1BasH_SHIFT     8
+#define HC_HTXnL2BasH_SHIFT     16
+/* HC_SubA_HTXnL345BasH    0x0021
+ */
+#define HC_HTXnL3BasH_MASK      0x000000ff
+#define HC_HTXnL4BasH_MASK      0x0000ff00
+#define HC_HTXnL5BasH_MASK      0x00ff0000
+#define HC_HTXnL4BasH_SHIFT     8
+#define HC_HTXnL5BasH_SHIFT     16
+/* HC_SubA_HTXnL678BasH    0x0022
+ */
+#define HC_HTXnL6BasH_MASK      0x000000ff
+#define HC_HTXnL7BasH_MASK      0x0000ff00
+#define HC_HTXnL8BasH_MASK      0x00ff0000
+#define HC_HTXnL7BasH_SHIFT     8
+#define HC_HTXnL8BasH_SHIFT     16
+/* HC_SubA_HTXnL9abBasH    0x0023
+ */
+#define HC_HTXnL9BasH_MASK      0x000000ff
+#define HC_HTXnLaBasH_MASK      0x0000ff00
+#define HC_HTXnLbBasH_MASK      0x00ff0000
+#define HC_HTXnLaBasH_SHIFT     8
+#define HC_HTXnLbBasH_SHIFT     16
+/* HC_SubA_HTXnLcdeBasH    0x0024
+ */
+#define HC_HTXnLcBasH_MASK      0x000000ff
+#define HC_HTXnLdBasH_MASK      0x0000ff00
+#define HC_HTXnLeBasH_MASK      0x00ff0000
+#define HC_HTXnLdBasH_SHIFT     8
+#define HC_HTXnLeBasH_SHIFT     16
+/* HC_SubA_HTXnLcdeBasH    0x0025
+ */
+#define HC_HTXnLfBasH_MASK      0x000000ff
+#define HC_HTXnL10BasH_MASK      0x0000ff00
+#define HC_HTXnL11BasH_MASK      0x00ff0000
+#define HC_HTXnL10BasH_SHIFT     8
+#define HC_HTXnL11BasH_SHIFT     16
+/* HC_SubA_HTXnL0Pit       0x002b
+ */
+#define HC_HTXnLnPit_MASK       0x00003fff
+#define HC_HTXnEnPit_MASK       0x00080000
+#define HC_HTXnLnPitE_MASK      0x00f00000
+#define HC_HTXnLnPitE_SHIFT     20
+/* HC_SubA_HTXnL0_5WE      0x004b
+ */
+#define HC_HTXnL0WE_MASK        0x0000000f
+#define HC_HTXnL1WE_MASK        0x000000f0
+#define HC_HTXnL2WE_MASK        0x00000f00
+#define HC_HTXnL3WE_MASK        0x0000f000
+#define HC_HTXnL4WE_MASK        0x000f0000
+#define HC_HTXnL5WE_MASK        0x00f00000
+#define HC_HTXnL1WE_SHIFT       4
+#define HC_HTXnL2WE_SHIFT       8
+#define HC_HTXnL3WE_SHIFT       12
+#define HC_HTXnL4WE_SHIFT       16
+#define HC_HTXnL5WE_SHIFT       20
+/* HC_SubA_HTXnL6_bWE      0x004c
+ */
+#define HC_HTXnL6WE_MASK        0x0000000f
+#define HC_HTXnL7WE_MASK        0x000000f0
+#define HC_HTXnL8WE_MASK        0x00000f00
+#define HC_HTXnL9WE_MASK        0x0000f000
+#define HC_HTXnLaWE_MASK        0x000f0000
+#define HC_HTXnLbWE_MASK        0x00f00000
+#define HC_HTXnL7WE_SHIFT       4
+#define HC_HTXnL8WE_SHIFT       8
+#define HC_HTXnL9WE_SHIFT       12
+#define HC_HTXnLaWE_SHIFT       16
+#define HC_HTXnLbWE_SHIFT       20
+/* HC_SubA_HTXnLc_11WE      0x004d
+ */
+#define HC_HTXnLcWE_MASK        0x0000000f
+#define HC_HTXnLdWE_MASK        0x000000f0
+#define HC_HTXnLeWE_MASK        0x00000f00
+#define HC_HTXnLfWE_MASK        0x0000f000
+#define HC_HTXnL10WE_MASK       0x000f0000
+#define HC_HTXnL11WE_MASK       0x00f00000
+#define HC_HTXnLdWE_SHIFT       4
+#define HC_HTXnLeWE_SHIFT       8
+#define HC_HTXnLfWE_SHIFT       12
+#define HC_HTXnL10WE_SHIFT      16
+#define HC_HTXnL11WE_SHIFT      20
+/* HC_SubA_HTXnL0_5HE      0x0051
+ */
+#define HC_HTXnL0HE_MASK        0x0000000f
+#define HC_HTXnL1HE_MASK        0x000000f0
+#define HC_HTXnL2HE_MASK        0x00000f00
+#define HC_HTXnL3HE_MASK        0x0000f000
+#define HC_HTXnL4HE_MASK        0x000f0000
+#define HC_HTXnL5HE_MASK        0x00f00000
+#define HC_HTXnL1HE_SHIFT       4
+#define HC_HTXnL2HE_SHIFT       8
+#define HC_HTXnL3HE_SHIFT       12
+#define HC_HTXnL4HE_SHIFT       16
+#define HC_HTXnL5HE_SHIFT       20
+/* HC_SubA_HTXnL6_bHE      0x0052
+ */
+#define HC_HTXnL6HE_MASK        0x0000000f
+#define HC_HTXnL7HE_MASK        0x000000f0
+#define HC_HTXnL8HE_MASK        0x00000f00
+#define HC_HTXnL9HE_MASK        0x0000f000
+#define HC_HTXnLaHE_MASK        0x000f0000
+#define HC_HTXnLbHE_MASK        0x00f00000
+#define HC_HTXnL7HE_SHIFT       4
+#define HC_HTXnL8HE_SHIFT       8
+#define HC_HTXnL9HE_SHIFT       12
+#define HC_HTXnLaHE_SHIFT       16
+#define HC_HTXnLbHE_SHIFT       20
+/* HC_SubA_HTXnLc_11HE      0x0053
+ */
+#define HC_HTXnLcHE_MASK        0x0000000f
+#define HC_HTXnLdHE_MASK        0x000000f0
+#define HC_HTXnLeHE_MASK        0x00000f00
+#define HC_HTXnLfHE_MASK        0x0000f000
+#define HC_HTXnL10HE_MASK       0x000f0000
+#define HC_HTXnL11HE_MASK       0x00f00000
+#define HC_HTXnLdHE_SHIFT       4
+#define HC_HTXnLeHE_SHIFT       8
+#define HC_HTXnLfHE_SHIFT       12
+#define HC_HTXnL10HE_SHIFT      16
+#define HC_HTXnL11HE_SHIFT      20
+/* HC_SubA_HTXnL0OS        0x0077
+ */
+#define HC_HTXnL0OS_MASK        0x003ff000
+#define HC_HTXnLVmax_MASK       0x00000fc0
+#define HC_HTXnLVmin_MASK       0x0000003f
+#define HC_HTXnL0OS_SHIFT       12
+#define HC_HTXnLVmax_SHIFT      6
+/* HC_SubA_HTXnTB          0x0078
+ */
+#define HC_HTXnTB_MASK          0x00f00000
+#define HC_HTXnFLSe_MASK        0x0000e000
+#define HC_HTXnFLSs_MASK        0x00001c00
+#define HC_HTXnFLTe_MASK        0x00000380
+#define HC_HTXnFLTs_MASK        0x00000070
+#define HC_HTXnFLDs_MASK        0x0000000f
+#define HC_HTXnTB_NoTB          0x00000000
+#define HC_HTXnTB_TBC_S         0x00100000
+#define HC_HTXnTB_TBC_T         0x00200000
+#define HC_HTXnTB_TB_S          0x00400000
+#define HC_HTXnTB_TB_T          0x00800000
+
+/* The "S" in FLS? means the S texture coordinate, and a "T" means the T
+ * texture coordinage.  The "e" in FL?e means the magnification ("enlarge")
+ * mode, and the "s" in FL?s means the minification ("shrink") mode.
+ *
+ * The "D" in FLD? means the intermipmap level mode.  That means that the
+ * GL_*_MIPMAP_LINEAR modes get FLDs_Linear, and the GL_*_MIPMAP_NEAREST modes
+ * get FLDs_Nearest.
+ */
+#define HC_HTXnFLSe_Nearest     0x00000000
+#define HC_HTXnFLSe_Linear      0x00002000
+#define HC_HTXnFLSe_NonLinear   0x00004000
+#define HC_HTXnFLSe_Sharp       0x00008000
+#define HC_HTXnFLSe_Flat_Gaussian_Cubic 0x0000c000
+#define HC_HTXnFLSs_Nearest     0x00000000
+#define HC_HTXnFLSs_Linear      0x00000400
+#define HC_HTXnFLSs_NonLinear   0x00000800
+#define HC_HTXnFLSs_Flat_Gaussian_Cubic 0x00001800
+#define HC_HTXnFLTe_Nearest     0x00000000
+#define HC_HTXnFLTe_Linear      0x00000080
+#define HC_HTXnFLTe_NonLinear   0x00000100
+#define HC_HTXnFLTe_Sharp       0x00000180
+#define HC_HTXnFLTe_Flat_Gaussian_Cubic 0x00000300
+#define HC_HTXnFLTs_Nearest     0x00000000
+#define HC_HTXnFLTs_Linear      0x00000010
+#define HC_HTXnFLTs_NonLinear   0x00000020
+#define HC_HTXnFLTs_Flat_Gaussian_Cubic 0x00000060
+#define HC_HTXnFLDs_Tex0        0x00000000
+#define HC_HTXnFLDs_Nearest     0x00000001
+#define HC_HTXnFLDs_Linear      0x00000002
+#define HC_HTXnFLDs_NonLinear   0x00000003
+#define HC_HTXnFLDs_Dither      0x00000004
+#define HC_HTXnFLDs_ConstLOD    0x00000005
+#define HC_HTXnFLDs_Ani         0x00000006
+#define HC_HTXnFLDs_AniDither   0x00000007
+
+/* HC_SubA_HTXnMPMD        0x0079
+ */
+#define HC_HTXnMPMD_SMASK       0x00070000
+#define HC_HTXnMPMD_TMASK       0x00380000
+#define HC_HTXnLODDTf_MASK      0x00000007
+#define HC_HTXnXY2ST_MASK       0x00000008
+#define HC_HTXnMPMD_Tsingle     0x00000000
+#define HC_HTXnMPMD_Tclamp      0x00080000
+#define HC_HTXnMPMD_Trepeat     0x00100000
+#define HC_HTXnMPMD_Tmirror     0x00180000
+#define HC_HTXnMPMD_Twrap       0x00200000
+#define HC_HTXnMPMD_Ssingle     0x00000000
+#define HC_HTXnMPMD_Sclamp      0x00010000
+#define HC_HTXnMPMD_Srepeat     0x00020000
+#define HC_HTXnMPMD_Smirror     0x00030000
+#define HC_HTXnMPMD_Swrap       0x00040000
+/* HC_SubA_HTXnCLODu       0x007a
+ */
+#define HC_HTXnCLODu_MASK       0x000ffc00
+#define HC_HTXnCLODd_MASK       0x000003ff
+#define HC_HTXnCLODu_SHIFT      10
+/* HC_SubA_HTXnFM          0x007b
+ */
+#define HC_HTXnFM_MASK          0x00ff0000
+#define HC_HTXnLoc_MASK         0x00000003
+#define HC_HTXnFM_INDEX         0x00000000 /*  0 << 19 */
+#define HC_HTXnFM_Intensity     0x00080000 /*  1 << 19 */
+#define HC_HTXnFM_Lum           0x00100000 /*  2 << 19 */
+#define HC_HTXnFM_Alpha         0x00180000 /*  3 << 19 */
+#define HC_HTXnFM_DX            0x00280000 /*  5 << 19 */
+#define HC_HTXnFM_BUMPMAP       0x00380000 /*  7 << 19 */
+#define HC_HTXnFM_ARGB16        0x00880000 /* 17 << 19 */
+#define HC_HTXnFM_ARGB32        0x00980000 /* 19 << 19 */
+#define HC_HTXnFM_ABGR16        0x00a80000 /* 21 << 19 */
+#define HC_HTXnFM_ABGR32        0x00b80000 /* 23 << 19 */
+#define HC_HTXnFM_RGBA16        0x00c80000 /* 25 << 19 */
+#define HC_HTXnFM_RGBA32        0x00d80000 /* 27 << 19 */
+#define HC_HTXnFM_BGRA16        0x00e80000 /* 29 << 19 */
+#define HC_HTXnFM_BGRA32        0x00f80000 /* 31 << 19 */
+#define HC_HTXnFM_Index1        (HC_HTXnFM_INDEX     | 0x00000000)
+#define HC_HTXnFM_Index2        (HC_HTXnFM_INDEX     | 0x00010000)
+#define HC_HTXnFM_Index4        (HC_HTXnFM_INDEX     | 0x00020000)
+#define HC_HTXnFM_Index8        (HC_HTXnFM_INDEX     | 0x00030000)
+#define HC_HTXnFM_T1            (HC_HTXnFM_Intensity | 0x00000000)
+#define HC_HTXnFM_T2            (HC_HTXnFM_Intensity | 0x00010000)
+#define HC_HTXnFM_T4            (HC_HTXnFM_Intensity | 0x00020000)
+#define HC_HTXnFM_T8            (HC_HTXnFM_Intensity | 0x00030000)
+#define HC_HTXnFM_L1            (HC_HTXnFM_Lum       | 0x00000000)
+#define HC_HTXnFM_L2            (HC_HTXnFM_Lum       | 0x00010000)
+#define HC_HTXnFM_L4            (HC_HTXnFM_Lum       | 0x00020000)
+#define HC_HTXnFM_L8            (HC_HTXnFM_Lum       | 0x00030000)
+#define HC_HTXnFM_AL44          (HC_HTXnFM_Lum       | 0x00040000)
+#define HC_HTXnFM_AL88          (HC_HTXnFM_Lum       | 0x00050000)
+#define HC_HTXnFM_A1            (HC_HTXnFM_Alpha     | 0x00000000)
+#define HC_HTXnFM_A2            (HC_HTXnFM_Alpha     | 0x00010000)
+#define HC_HTXnFM_A4            (HC_HTXnFM_Alpha     | 0x00020000)
+#define HC_HTXnFM_A8            (HC_HTXnFM_Alpha     | 0x00030000)
+#define HC_HTXnFM_DX1           (HC_HTXnFM_DX        | 0x00010000)
+#define HC_HTXnFM_DX23          (HC_HTXnFM_DX        | 0x00020000)
+#define HC_HTXnFM_DX45          (HC_HTXnFM_DX        | 0x00030000)
+#define HC_HTXnFM_RGB555        (HC_HTXnFM_ARGB16    | 0x00000000)
+#define HC_HTXnFM_RGB565        (HC_HTXnFM_ARGB16    | 0x00010000)
+#define HC_HTXnFM_ARGB1555      (HC_HTXnFM_ARGB16    | 0x00020000)
+#define HC_HTXnFM_ARGB4444      (HC_HTXnFM_ARGB16    | 0x00030000)
+#define HC_HTXnFM_ARGB0888      (HC_HTXnFM_ARGB32    | 0x00000000)
+#define HC_HTXnFM_ARGB8888      (HC_HTXnFM_ARGB32    | 0x00010000)
+#define HC_HTXnFM_BGR555        (HC_HTXnFM_ABGR16    | 0x00000000)
+#define HC_HTXnFM_BGR565        (HC_HTXnFM_ABGR16    | 0x00010000)
+#define HC_HTXnFM_ABGR1555      (HC_HTXnFM_ABGR16    | 0x00020000)
+#define HC_HTXnFM_ABGR4444      (HC_HTXnFM_ABGR16    | 0x00030000)
+#define HC_HTXnFM_ABGR0888      (HC_HTXnFM_ABGR32    | 0x00000000)
+#define HC_HTXnFM_ABGR8888      (HC_HTXnFM_ABGR32    | 0x00010000)
+#define HC_HTXnFM_RGBA5550      (HC_HTXnFM_RGBA16    | 0x00000000)
+#define HC_HTXnFM_RGBA5551      (HC_HTXnFM_RGBA16    | 0x00020000)
+#define HC_HTXnFM_RGBA4444      (HC_HTXnFM_RGBA16    | 0x00030000)
+#define HC_HTXnFM_RGBA8880      (HC_HTXnFM_RGBA32    | 0x00000000)
+#define HC_HTXnFM_RGBA8888      (HC_HTXnFM_RGBA32    | 0x00010000)
+#define HC_HTXnFM_BGRA5550      (HC_HTXnFM_BGRA16    | 0x00000000)
+#define HC_HTXnFM_BGRA5551      (HC_HTXnFM_BGRA16    | 0x00020000)
+#define HC_HTXnFM_BGRA4444      (HC_HTXnFM_BGRA16    | 0x00030000)
+#define HC_HTXnFM_BGRA8880      (HC_HTXnFM_BGRA32    | 0x00000000)
+#define HC_HTXnFM_BGRA8888      (HC_HTXnFM_BGRA32    | 0x00010000)
+#define HC_HTXnFM_VU88          (HC_HTXnFM_BUMPMAP   | 0x00000000)
+#define HC_HTXnFM_LVU655        (HC_HTXnFM_BUMPMAP   | 0x00010000)
+#define HC_HTXnFM_LVU888        (HC_HTXnFM_BUMPMAP   | 0x00020000)
+#define HC_HTXnLoc_Local        0x00000000
+#define HC_HTXnLoc_Sys          0x00000002
+#define HC_HTXnLoc_AGP          0x00000003
+/* HC_SubA_HTXnTRAH        0x007f
+ */
+#define HC_HTXnTRAH_MASK        0x00ff0000
+#define HC_HTXnTRAL_MASK        0x0000ff00
+#define HC_HTXnTBA_MASK         0x000000ff
+#define HC_HTXnTRAH_SHIFT       16
+#define HC_HTXnTRAL_SHIFT       8
+
+/*-- Define the input texture, for below
+ */
+#define HC_XTC_TOPC             0x00000000
+#define HC_XTC_InvTOPC          0x00000010
+#define HC_XTC_TOPCp5           0x00000020
+#define HC_XTC_Cbias            0x00000000
+#define HC_XTC_InvCbias         0x00000010
+#define HC_XTC_0                0x00000000
+#define HC_XTC_Dif              0x00000001
+#define HC_XTC_Spec             0x00000002
+#define HC_XTC_Tex              0x00000003
+#define HC_XTC_Cur              0x00000004
+#define HC_XTC_Adif             0x00000005
+#define HC_XTC_Fog              0x00000006
+#define HC_XTC_Atex             0x00000007
+#define HC_XTC_Acur             0x00000008
+#define HC_XTC_HTXnTBLRC        0x00000009
+#define HC_XTC_Ctexnext         0x0000000a 
+/** HC_SubA_HTXnTBLCsat     0x0080 
+ */
+#define HC_HTXnTBLCsat_MASK     0x00800000
+#define HC_HTXnTBLCa_MASK       0x000fc000
+#define HC_HTXnTBLCb_MASK       0x00001f80
+#define HC_HTXnTBLCc_MASK       0x0000003f
+#define HC_HTXnTBLCa_TOPC       (HC_XTC_TOPC << 14)
+#define HC_HTXnTBLCa_InvTOPC    (HC_XTC_InvTOPC << 14)
+#define HC_HTXnTBLCa_TOPCp5     (HC_XTC_TOPCp5 << 14)
+#define HC_HTXnTBLCa_0          (HC_XTC_0 << 14)
+#define HC_HTXnTBLCa_Dif        (HC_XTC_Dif << 14)
+#define HC_HTXnTBLCa_Spec       (HC_XTC_Spec << 14)
+#define HC_HTXnTBLCa_Tex        (HC_XTC_Tex << 14)
+#define HC_HTXnTBLCa_Cur        (HC_XTC_Cur << 14)
+#define HC_HTXnTBLCa_Adif       (HC_XTC_Adif << 14)
+#define HC_HTXnTBLCa_Fog        (HC_XTC_Fog << 14)
+#define HC_HTXnTBLCa_Atex       (HC_XTC_Atex << 14)
+#define HC_HTXnTBLCa_Acur       (HC_XTC_Acur << 14)
+#define HC_HTXnTBLCa_HTXnTBLRC  (HC_XTC_HTXnTBLRC << 14)
+#define HC_HTXnTBLCa_Ctexnext   (HC_XTC_Ctexnext << 14) 
+#define HC_HTXnTBLCb_TOPC       (HC_XTC_TOPC << 7)
+#define HC_HTXnTBLCb_InvTOPC    (HC_XTC_InvTOPC << 7)
+#define HC_HTXnTBLCb_TOPCp5     (HC_XTC_TOPCp5 << 7)
+#define HC_HTXnTBLCb_0          (HC_XTC_0 << 7)
+#define HC_HTXnTBLCb_Dif        (HC_XTC_Dif << 7)
+#define HC_HTXnTBLCb_Spec       (HC_XTC_Spec << 7)
+#define HC_HTXnTBLCb_Tex        (HC_XTC_Tex << 7)
+#define HC_HTXnTBLCb_Cur        (HC_XTC_Cur << 7)
+#define HC_HTXnTBLCb_Adif       (HC_XTC_Adif << 7)
+#define HC_HTXnTBLCb_Fog        (HC_XTC_Fog << 7)
+#define HC_HTXnTBLCb_Atex       (HC_XTC_Atex << 7)
+#define HC_HTXnTBLCb_Acur       (HC_XTC_Acur << 7)
+#define HC_HTXnTBLCb_HTXnTBLRC  (HC_XTC_HTXnTBLRC << 7)
+#define HC_HTXnTBLCb_Ctexnext   (HC_XTC_Ctexnext << 7) 
+#define HC_HTXnTBLCc_TOPC       (HC_XTC_TOPC << 0)
+#define HC_HTXnTBLCc_InvTOPC    (HC_XTC_InvTOPC << 0)
+#define HC_HTXnTBLCc_TOPCp5     (HC_XTC_TOPCp5 << 0)
+#define HC_HTXnTBLCc_0          (HC_XTC_0 << 0)
+#define HC_HTXnTBLCc_Dif        (HC_XTC_Dif << 0)
+#define HC_HTXnTBLCc_Spec       (HC_XTC_Spec << 0)
+#define HC_HTXnTBLCc_Tex        (HC_XTC_Tex << 0)
+#define HC_HTXnTBLCc_Cur        (HC_XTC_Cur << 0)
+#define HC_HTXnTBLCc_Adif       (HC_XTC_Adif << 0)
+#define HC_HTXnTBLCc_Fog        (HC_XTC_Fog << 0)
+#define HC_HTXnTBLCc_Atex       (HC_XTC_Atex << 0)
+#define HC_HTXnTBLCc_Acur       (HC_XTC_Acur << 0)
+#define HC_HTXnTBLCc_HTXnTBLRC  (HC_XTC_HTXnTBLRC << 0)
+#define HC_HTXnTBLCc_Ctexnext   (HC_XTC_Ctexnext << 0) 
+/* HC_SubA_HTXnTBLCop      0x0081
+ */
+#define HC_HTXnTBLdot_MASK      0x00c00000
+#define HC_HTXnTBLCop_MASK      0x00380000
+#define HC_HTXnTBLCbias_MASK    0x0007c000
+#define HC_HTXnTBLCshift_MASK   0x00001800
+#define HC_HTXnTBLAop_MASK      0x00000380
+#define HC_HTXnTBLAbias_MASK    0x00000078
+#define HC_HTXnTBLAshift_MASK   0x00000003
+#define HC_HTXnTBLDOT3          0x00800000
+#define HC_HTXnTBLDOT4          0x00c00000
+#define HC_HTXnTBLCop_Add       0x00000000
+#define HC_HTXnTBLCop_Sub       0x00080000
+#define HC_HTXnTBLCop_Min       0x00100000
+#define HC_HTXnTBLCop_Max       0x00180000
+#define HC_HTXnTBLCop_Mask      0x00200000
+#define HC_HTXnTBLCbias_Cbias           (HC_XTC_Cbias << 14)
+#define HC_HTXnTBLCbias_InvCbias        (HC_XTC_InvCbias << 14)
+#define HC_HTXnTBLCbias_0               (HC_XTC_0 << 14)
+#define HC_HTXnTBLCbias_Dif             (HC_XTC_Dif << 14)
+#define HC_HTXnTBLCbias_Spec            (HC_XTC_Spec << 14)
+#define HC_HTXnTBLCbias_Tex             (HC_XTC_Tex << 14)
+#define HC_HTXnTBLCbias_Cur             (HC_XTC_Cur << 14)
+#define HC_HTXnTBLCbias_Adif            (HC_XTC_Adif << 14)
+#define HC_HTXnTBLCbias_Fog             (HC_XTC_Fog << 14)
+#define HC_HTXnTBLCbias_Atex            (HC_XTC_Atex << 14)
+#define HC_HTXnTBLCbias_Acur            (HC_XTC_Acur << 14)
+#define HC_HTXnTBLCbias_HTXnTBLRC       (HC_XTC_HTXnTBLRC << 14)
+#define HC_HTXnTBLCshift_1      0x00000000
+#define HC_HTXnTBLCshift_2      0x00000800
+#define HC_HTXnTBLCshift_No     0x00001000
+#define HC_HTXnTBLCshift_DotP   0x00001800
+#define HC_HTXnTBLAop_Add       0x00000000
+#define HC_HTXnTBLAop_Sub       0x00000080
+#define HC_HTXnTBLAop_Min       0x00000100
+#define HC_HTXnTBLAop_Max       0x00000180
+#define HC_HTXnTBLAop_Mask      0x00000200
+#define HC_HTXnTBLAbias_Inv             0x00000040
+#define HC_HTXnTBLAbias_Adif            0x00000000
+#define HC_HTXnTBLAbias_Fog             0x00000008
+#define HC_HTXnTBLAbias_Acur            0x00000010
+#define HC_HTXnTBLAbias_HTXnTBLRAbias   0x00000018
+#define HC_HTXnTBLAbias_Atex            0x00000020
+#define HC_HTXnTBLAshift_1      0x00000000
+#define HC_HTXnTBLAshift_2      0x00000001
+#define HC_HTXnTBLAshift_No     0x00000002
+#define HC_HTXnTBLAshift_DotP   0x00000003
+/* HC_SubA_HTXnTBLMPFog    0x0082
+ */
+#define HC_HTXnTBLMPfog_MASK    0x00e00000
+#define HC_HTXnTBLMPfog_0       0x00000000
+#define HC_HTXnTBLMPfog_Adif    0x00200000
+#define HC_HTXnTBLMPfog_Fog     0x00400000
+#define HC_HTXnTBLMPfog_Atex    0x00600000
+#define HC_HTXnTBLMPfog_Acur    0x00800000
+#define HC_HTXnTBLMPfog_GHTXnTBLRFog    0x00a00000
+/* HC_SubA_HTXnTBLAsat     0x0083
+ *-- Define the texture alpha input.
+ */
+#define HC_XTA_TOPA             0x00000000
+#define HC_XTA_InvTOPA          0x00000008
+#define HC_XTA_TOPAp5           0x00000010
+#define HC_XTA_Adif             0x00000000
+#define HC_XTA_Fog              0x00000001
+#define HC_XTA_Acur             0x00000002
+#define HC_XTA_HTXnTBLRA        0x00000003
+#define HC_XTA_Atex             0x00000004
+#define HC_XTA_Atexnext         0x00000005 
+/*--
+ */
+#define HC_HTXnTBLAsat_MASK     0x00800000
+#define HC_HTXnTBLAMB_MASK      0x00700000
+#define HC_HTXnTBLAa_MASK       0x0007c000
+#define HC_HTXnTBLAb_MASK       0x00000f80
+#define HC_HTXnTBLAc_MASK       0x0000001f
+#define HC_HTXnTBLAMB_SHIFT     20
+#define HC_HTXnTBLAa_TOPA       (HC_XTA_TOPA << 14)
+#define HC_HTXnTBLAa_InvTOPA    (HC_XTA_InvTOPA << 14)
+#define HC_HTXnTBLAa_TOPAp5     (HC_XTA_TOPAp5 << 14)
+#define HC_HTXnTBLAa_Adif       (HC_XTA_Adif << 14)
+#define HC_HTXnTBLAa_Fog        (HC_XTA_Fog << 14)
+#define HC_HTXnTBLAa_Acur       (HC_XTA_Acur << 14)
+#define HC_HTXnTBLAa_HTXnTBLRA  (HC_XTA_HTXnTBLRA << 14)
+#define HC_HTXnTBLAa_Atex       (HC_XTA_Atex << 14)
+#define HC_HTXnTBLAa_Atexnext   (HC_XTA_Atexnext << 14) 
+#define HC_HTXnTBLAb_TOPA       (HC_XTA_TOPA << 7)
+#define HC_HTXnTBLAb_InvTOPA    (HC_XTA_InvTOPA << 7)
+#define HC_HTXnTBLAb_TOPAp5     (HC_XTA_TOPAp5 << 7)
+#define HC_HTXnTBLAb_Adif       (HC_XTA_Adif << 7)
+#define HC_HTXnTBLAb_Fog        (HC_XTA_Fog << 7)
+#define HC_HTXnTBLAb_Acur       (HC_XTA_Acur << 7)
+#define HC_HTXnTBLAb_HTXnTBLRA  (HC_XTA_HTXnTBLRA << 7)
+#define HC_HTXnTBLAb_Atex       (HC_XTA_Atex << 7)
+#define HC_HTXnTBLAb_Atexnext   (HC_XTA_Atexnext << 7) 
+#define HC_HTXnTBLAc_TOPA       (HC_XTA_TOPA << 0)
+#define HC_HTXnTBLAc_InvTOPA    (HC_XTA_InvTOPA << 0)
+#define HC_HTXnTBLAc_TOPAp5     (HC_XTA_TOPAp5 << 0)
+#define HC_HTXnTBLAc_Adif       (HC_XTA_Adif << 0)
+#define HC_HTXnTBLAc_Fog        (HC_XTA_Fog << 0)
+#define HC_HTXnTBLAc_Acur       (HC_XTA_Acur << 0)
+#define HC_HTXnTBLAc_HTXnTBLRA  (HC_XTA_HTXnTBLRA << 0)
+#define HC_HTXnTBLAc_Atex       (HC_XTA_Atex << 0)
+#define HC_HTXnTBLAc_Atexnext   (HC_XTA_Atexnext << 0) 
+/* HC_SubA_HTXnTBLRAa      0x0089
+ */
+#define HC_HTXnTBLRAa_MASK      0x00ff0000
+#define HC_HTXnTBLRAb_MASK      0x0000ff00
+#define HC_HTXnTBLRAc_MASK      0x000000ff
+#define HC_HTXnTBLRAa_SHIFT     16
+#define HC_HTXnTBLRAb_SHIFT     8
+#define HC_HTXnTBLRAc_SHIFT     0
+/* HC_SubA_HTXnTBLRFog     0x008a
+ */
+#define HC_HTXnTBLRFog_MASK     0x0000ff00
+#define HC_HTXnTBLRAbias_MASK   0x000000ff
+#define HC_HTXnTBLRFog_SHIFT    8
+#define HC_HTXnTBLRAbias_SHIFT  0
+/* HC_SubA_HTXnLScale      0x0094
+ */
+#define HC_HTXnLScale_MASK      0x0007fc00
+#define HC_HTXnLOff_MASK        0x000001ff
+#define HC_HTXnLScale_SHIFT     10
+/* HC_SubA_HTXSMD          0x0000
+ */
+#define HC_HTXSMD_MASK          0x00000080
+#define HC_HTXTMD_MASK          0x00000040
+#define HC_HTXNum_MASK          0x00000038
+#define HC_HTXTRMD_MASK         0x00000006
+#define HC_HTXCHCLR_MASK        0x00000001
+#define HC_HTXNum_SHIFT         3
+
+/* Texture Palette n
+ */
+#define HC_SubType_TexPalette0  0x00000000
+#define HC_SubType_TexPalette1  0x00000001
+#define HC_SubType_FogTable     0x00000010
+#define HC_SubType_Stipple      0x00000014
+/* HC_SubA_TexPalette0     0x0000
+ */
+#define HC_HTPnA_MASK           0xff000000
+#define HC_HTPnR_MASK           0x00ff0000
+#define HC_HTPnG_MASK           0x0000ff00
+#define HC_HTPnB_MASK           0x000000ff
+/* HC_SubA_FogTable        0x0010
+ */
+#define HC_HFPn3_MASK           0xff000000
+#define HC_HFPn2_MASK           0x00ff0000
+#define HC_HFPn1_MASK           0x0000ff00
+#define HC_HFPn_MASK            0x000000ff
+#define HC_HFPn3_SHIFT          24
+#define HC_HFPn2_SHIFT          16
+#define HC_HFPn1_SHIFT          8
+
+/* Auto Testing & Security
+ */
+#define HC_SubA_HenFIFOAT       0x0000
+#define HC_SubA_HFBDrawFirst    0x0004
+#define HC_SubA_HFBBasL         0x0005
+#define HC_SubA_HFBDst          0x0006
+/* HC_SubA_HenFIFOAT       0x0000
+ */
+#define HC_HenFIFOAT_MASK       0x00000020
+#define HC_HenGEMILock_MASK     0x00000010
+#define HC_HenFBASwap_MASK      0x00000008
+#define HC_HenOT_MASK           0x00000004
+#define HC_HenCMDQ_MASK         0x00000002
+#define HC_HenTXCTSU_MASK       0x00000001
+/* HC_SubA_HFBDrawFirst    0x0004
+ */
+#define HC_HFBDrawFirst_MASK    0x00000800
+#define HC_HFBQueue_MASK        0x00000400
+#define HC_HFBLock_MASK         0x00000200
+#define HC_HEOF_MASK            0x00000100
+#define HC_HFBBasH_MASK         0x000000ff
+
+/* GEMI Setting
+ */
+#define HC_SubA_HTArbRCM        0x0008
+#define HC_SubA_HTArbRZ         0x000a
+#define HC_SubA_HTArbWZ         0x000b
+#define HC_SubA_HTArbRTX        0x000c
+#define HC_SubA_HTArbRCW        0x000d
+#define HC_SubA_HTArbE2         0x000e
+#define HC_SubA_HArbRQCM        0x0010
+#define HC_SubA_HArbWQCM        0x0011
+#define HC_SubA_HGEMITout       0x0020
+#define HC_SubA_HFthRTXD        0x0040
+#define HC_SubA_HFthRTXA        0x0044
+#define HC_SubA_HCMDQstL        0x0050
+#define HC_SubA_HCMDQendL       0x0051
+#define HC_SubA_HCMDQLen        0x0052
+/* HC_SubA_HTArbRCM        0x0008
+ */
+#define HC_HTArbRCM_MASK        0x0000ffff
+/* HC_SubA_HTArbRZ         0x000a
+ */
+#define HC_HTArbRZ_MASK         0x0000ffff
+/* HC_SubA_HTArbWZ         0x000b
+ */
+#define HC_HTArbWZ_MASK         0x0000ffff
+/* HC_SubA_HTArbRTX        0x000c
+ */
+#define HC_HTArbRTX_MASK        0x0000ffff
+/* HC_SubA_HTArbRCW        0x000d
+ */
+#define HC_HTArbRCW_MASK        0x0000ffff
+/* HC_SubA_HTArbE2         0x000e
+ */
+#define HC_HTArbE2_MASK         0x0000ffff
+/* HC_SubA_HArbRQCM        0x0010
+ */
+#define HC_HTArbRQCM_MASK       0x0000ffff
+/* HC_SubA_HArbWQCM        0x0011
+ */
+#define HC_HArbWQCM_MASK        0x0000ffff
+/* HC_SubA_HGEMITout       0x0020
+ */
+#define HC_HGEMITout_MASK       0x000f0000
+#define HC_HNPArbZC_MASK        0x0000ffff
+#define HC_HGEMITout_SHIFT      16
+/* HC_SubA_HFthRTXD        0x0040
+ */
+#define HC_HFthRTXD_MASK        0x00ff0000
+#define HC_HFthRZD_MASK         0x0000ff00
+#define HC_HFthWZD_MASK         0x000000ff
+#define HC_HFthRTXD_SHIFT       16
+#define HC_HFthRZD_SHIFT        8
+/* HC_SubA_HFthRTXA        0x0044
+ */
+#define HC_HFthRTXA_MASK        0x000000ff
+
+/******************************************************************************
+** Define the Halcyon Internal register access constants. For simulator only.
+******************************************************************************/
+#define HC_SIMA_HAGPBstL        0x0000
+#define HC_SIMA_HAGPBendL       0x0001
+#define HC_SIMA_HAGPCMNT        0x0002
+#define HC_SIMA_HAGPBpL         0x0003
+#define HC_SIMA_HAGPBpH         0x0004
+#define HC_SIMA_HClipTB         0x0005
+#define HC_SIMA_HClipLR         0x0006
+#define HC_SIMA_HFPClipTL       0x0007
+#define HC_SIMA_HFPClipBL       0x0008
+#define HC_SIMA_HFPClipLL       0x0009
+#define HC_SIMA_HFPClipRL       0x000a
+#define HC_SIMA_HFPClipTBH      0x000b
+#define HC_SIMA_HFPClipLRH      0x000c
+#define HC_SIMA_HLP             0x000d
+#define HC_SIMA_HLPRF           0x000e
+#define HC_SIMA_HSolidCL        0x000f
+#define HC_SIMA_HPixGC          0x0010
+#define HC_SIMA_HSPXYOS         0x0011
+#define HC_SIMA_HCmdA           0x0012
+#define HC_SIMA_HCmdB           0x0013
+#define HC_SIMA_HEnable         0x0014
+#define HC_SIMA_HZWBBasL        0x0015
+#define HC_SIMA_HZWBBasH        0x0016
+#define HC_SIMA_HZWBType        0x0017
+#define HC_SIMA_HZBiasL         0x0018
+#define HC_SIMA_HZWBend         0x0019
+#define HC_SIMA_HZWTMD          0x001a
+#define HC_SIMA_HZWCDL          0x001b
+#define HC_SIMA_HZWCTAGnum      0x001c
+#define HC_SIMA_HZCYNum         0x001d
+#define HC_SIMA_HZWCFire        0x001e
+/* #define HC_SIMA_HSBBasL         0x001d */
+/* #define HC_SIMA_HSBBasH         0x001e */
+/* #define HC_SIMA_HSBFM           0x001f */
+#define HC_SIMA_HSTREF          0x0020
+#define HC_SIMA_HSTMD           0x0021
+#define HC_SIMA_HABBasL         0x0022
+#define HC_SIMA_HABBasH         0x0023
+#define HC_SIMA_HABFM           0x0024
+#define HC_SIMA_HATMD           0x0025
+#define HC_SIMA_HABLCsat        0x0026
+#define HC_SIMA_HABLCop         0x0027
+#define HC_SIMA_HABLAsat        0x0028
+#define HC_SIMA_HABLAop         0x0029
+#define HC_SIMA_HABLRCa         0x002a
+#define HC_SIMA_HABLRFCa        0x002b
+#define HC_SIMA_HABLRCbias      0x002c
+#define HC_SIMA_HABLRCb         0x002d
+#define HC_SIMA_HABLRFCb        0x002e
+#define HC_SIMA_HABLRAa         0x002f
+#define HC_SIMA_HABLRAb         0x0030
+#define HC_SIMA_HDBBasL         0x0031
+#define HC_SIMA_HDBBasH         0x0032
+#define HC_SIMA_HDBFM           0x0033
+#define HC_SIMA_HFBBMSKL        0x0034
+#define HC_SIMA_HROP            0x0035
+#define HC_SIMA_HFogLF          0x0036
+#define HC_SIMA_HFogCL          0x0037
+#define HC_SIMA_HFogCH          0x0038
+#define HC_SIMA_HFogStL         0x0039
+#define HC_SIMA_HFogStH         0x003a
+#define HC_SIMA_HFogOOdMF       0x003b
+#define HC_SIMA_HFogOOdEF       0x003c
+#define HC_SIMA_HFogEndL        0x003d
+#define HC_SIMA_HFogDenst       0x003e
+/*---- start of texture 0 setting ----
+ */
+#define HC_SIMA_HTX0L0BasL      0x0040
+#define HC_SIMA_HTX0L1BasL      0x0041
+#define HC_SIMA_HTX0L2BasL      0x0042
+#define HC_SIMA_HTX0L3BasL      0x0043
+#define HC_SIMA_HTX0L4BasL      0x0044
+#define HC_SIMA_HTX0L5BasL      0x0045
+#define HC_SIMA_HTX0L6BasL      0x0046
+#define HC_SIMA_HTX0L7BasL      0x0047
+#define HC_SIMA_HTX0L8BasL      0x0048
+#define HC_SIMA_HTX0L9BasL      0x0049
+#define HC_SIMA_HTX0LaBasL      0x004a
+#define HC_SIMA_HTX0LbBasL      0x004b
+#define HC_SIMA_HTX0LcBasL      0x004c
+#define HC_SIMA_HTX0LdBasL      0x004d
+#define HC_SIMA_HTX0LeBasL      0x004e
+#define HC_SIMA_HTX0LfBasL      0x004f
+#define HC_SIMA_HTX0L10BasL     0x0050
+#define HC_SIMA_HTX0L11BasL     0x0051
+#define HC_SIMA_HTX0L012BasH    0x0052
+#define HC_SIMA_HTX0L345BasH    0x0053
+#define HC_SIMA_HTX0L678BasH    0x0054
+#define HC_SIMA_HTX0L9abBasH    0x0055
+#define HC_SIMA_HTX0LcdeBasH    0x0056
+#define HC_SIMA_HTX0Lf1011BasH  0x0057
+#define HC_SIMA_HTX0L0Pit       0x0058
+#define HC_SIMA_HTX0L1Pit       0x0059
+#define HC_SIMA_HTX0L2Pit       0x005a
+#define HC_SIMA_HTX0L3Pit       0x005b
+#define HC_SIMA_HTX0L4Pit       0x005c
+#define HC_SIMA_HTX0L5Pit       0x005d
+#define HC_SIMA_HTX0L6Pit       0x005e
+#define HC_SIMA_HTX0L7Pit       0x005f
+#define HC_SIMA_HTX0L8Pit       0x0060
+#define HC_SIMA_HTX0L9Pit       0x0061
+#define HC_SIMA_HTX0LaPit       0x0062
+#define HC_SIMA_HTX0LbPit       0x0063
+#define HC_SIMA_HTX0LcPit       0x0064
+#define HC_SIMA_HTX0LdPit       0x0065
+#define HC_SIMA_HTX0LePit       0x0066
+#define HC_SIMA_HTX0LfPit       0x0067
+#define HC_SIMA_HTX0L10Pit      0x0068
+#define HC_SIMA_HTX0L11Pit      0x0069
+#define HC_SIMA_HTX0L0_5WE      0x006a
+#define HC_SIMA_HTX0L6_bWE      0x006b
+#define HC_SIMA_HTX0Lc_11WE     0x006c
+#define HC_SIMA_HTX0L0_5HE      0x006d
+#define HC_SIMA_HTX0L6_bHE      0x006e
+#define HC_SIMA_HTX0Lc_11HE     0x006f
+#define HC_SIMA_HTX0L0OS        0x0070
+#define HC_SIMA_HTX0TB          0x0071
+#define HC_SIMA_HTX0MPMD        0x0072
+#define HC_SIMA_HTX0CLODu       0x0073
+#define HC_SIMA_HTX0FM          0x0074
+#define HC_SIMA_HTX0TRCH        0x0075
+#define HC_SIMA_HTX0TRCL        0x0076
+#define HC_SIMA_HTX0TBC         0x0077
+#define HC_SIMA_HTX0TRAH        0x0078
+#define HC_SIMA_HTX0TBLCsat     0x0079
+#define HC_SIMA_HTX0TBLCop      0x007a
+#define HC_SIMA_HTX0TBLMPfog    0x007b
+#define HC_SIMA_HTX0TBLAsat     0x007c
+#define HC_SIMA_HTX0TBLRCa      0x007d
+#define HC_SIMA_HTX0TBLRCb      0x007e
+#define HC_SIMA_HTX0TBLRCc      0x007f
+#define HC_SIMA_HTX0TBLRCbias   0x0080
+#define HC_SIMA_HTX0TBLRAa      0x0081
+#define HC_SIMA_HTX0TBLRFog     0x0082
+#define HC_SIMA_HTX0BumpM00     0x0083
+#define HC_SIMA_HTX0BumpM01     0x0084
+#define HC_SIMA_HTX0BumpM10     0x0085
+#define HC_SIMA_HTX0BumpM11     0x0086
+#define HC_SIMA_HTX0LScale      0x0087
+/*---- end of texture 0 setting ----      0x008f
+ */
+#define HC_SIMA_TX0TX1_OFF      0x0050
+/*---- start of texture 1 setting ----
+ */
+#define HC_SIMA_HTX1L0BasL      (HC_SIMA_HTX0L0BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L1BasL      (HC_SIMA_HTX0L1BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L2BasL      (HC_SIMA_HTX0L2BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L3BasL      (HC_SIMA_HTX0L3BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L4BasL      (HC_SIMA_HTX0L4BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L5BasL      (HC_SIMA_HTX0L5BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L6BasL      (HC_SIMA_HTX0L6BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L7BasL      (HC_SIMA_HTX0L7BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L8BasL      (HC_SIMA_HTX0L8BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L9BasL      (HC_SIMA_HTX0L9BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LaBasL      (HC_SIMA_HTX0LaBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LbBasL      (HC_SIMA_HTX0LbBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LcBasL      (HC_SIMA_HTX0LcBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LdBasL      (HC_SIMA_HTX0LdBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LeBasL      (HC_SIMA_HTX0LeBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LfBasL      (HC_SIMA_HTX0LfBasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L10BasL     (HC_SIMA_HTX0L10BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L11BasL     (HC_SIMA_HTX0L11BasL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L012BasH    (HC_SIMA_HTX0L012BasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L345BasH    (HC_SIMA_HTX0L345BasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L678BasH    (HC_SIMA_HTX0L678BasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L9abBasH    (HC_SIMA_HTX0L9abBasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LcdeBasH    (HC_SIMA_HTX0LcdeBasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1Lf1011BasH  (HC_SIMA_HTX0Lf1011BasH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L0Pit       (HC_SIMA_HTX0L0Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L1Pit       (HC_SIMA_HTX0L1Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L2Pit       (HC_SIMA_HTX0L2Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L3Pit       (HC_SIMA_HTX0L3Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L4Pit       (HC_SIMA_HTX0L4Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L5Pit       (HC_SIMA_HTX0L5Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L6Pit       (HC_SIMA_HTX0L6Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L7Pit       (HC_SIMA_HTX0L7Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L8Pit       (HC_SIMA_HTX0L8Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L9Pit       (HC_SIMA_HTX0L9Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LaPit       (HC_SIMA_HTX0LaPit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LbPit       (HC_SIMA_HTX0LbPit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LcPit       (HC_SIMA_HTX0LcPit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LdPit       (HC_SIMA_HTX0LdPit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LePit       (HC_SIMA_HTX0LePit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LfPit       (HC_SIMA_HTX0LfPit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L10Pit      (HC_SIMA_HTX0L10Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L11Pit      (HC_SIMA_HTX0L11Pit + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L0_5WE      (HC_SIMA_HTX0L0_5WE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L6_bWE      (HC_SIMA_HTX0L6_bWE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1Lc_11WE     (HC_SIMA_HTX0Lc_11WE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L0_5HE      (HC_SIMA_HTX0L0_5HE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L6_bHE      (HC_SIMA_HTX0L6_bHE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1Lc_11HE      (HC_SIMA_HTX0Lc_11HE + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1L0OS        (HC_SIMA_HTX0L0OS + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TB          (HC_SIMA_HTX0TB + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1MPMD        (HC_SIMA_HTX0MPMD + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1CLODu       (HC_SIMA_HTX0CLODu + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1FM          (HC_SIMA_HTX0FM + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TRCH        (HC_SIMA_HTX0TRCH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TRCL        (HC_SIMA_HTX0TRCL + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBC         (HC_SIMA_HTX0TBC + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TRAH        (HC_SIMA_HTX0TRAH + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LTC         (HC_SIMA_HTX0LTC + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LTA         (HC_SIMA_HTX0LTA + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLCsat     (HC_SIMA_HTX0TBLCsat + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLCop      (HC_SIMA_HTX0TBLCop + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLMPfog    (HC_SIMA_HTX0TBLMPfog + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLAsat     (HC_SIMA_HTX0TBLAsat + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRCa      (HC_SIMA_HTX0TBLRCa + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRCb      (HC_SIMA_HTX0TBLRCb + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRCc      (HC_SIMA_HTX0TBLRCc + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRCbias   (HC_SIMA_HTX0TBLRCbias + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRAa      (HC_SIMA_HTX0TBLRAa + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1TBLRFog     (HC_SIMA_HTX0TBLRFog + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1BumpM00     (HC_SIMA_HTX0BumpM00 + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1BumpM01     (HC_SIMA_HTX0BumpM01 + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1BumpM10     (HC_SIMA_HTX0BumpM10 + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1BumpM11     (HC_SIMA_HTX0BumpM11 + HC_SIMA_TX0TX1_OFF)
+#define HC_SIMA_HTX1LScale      (HC_SIMA_HTX0LScale + HC_SIMA_TX0TX1_OFF)
+/*---- end of texture 1 setting ---- 0xaf
+ */
+#define HC_SIMA_HTXSMD          0x00b0
+#define HC_SIMA_HenFIFOAT       0x00b1
+#define HC_SIMA_HFBDrawFirst    0x00b2
+#define HC_SIMA_HFBBasL         0x00b3
+#define HC_SIMA_HTArbRCM        0x00b4
+#define HC_SIMA_HTArbRZ         0x00b5
+#define HC_SIMA_HTArbWZ         0x00b6
+#define HC_SIMA_HTArbRTX        0x00b7
+#define HC_SIMA_HTArbRCW        0x00b8
+#define HC_SIMA_HTArbE2         0x00b9
+#define HC_SIMA_HGEMITout       0x00ba
+#define HC_SIMA_HFthRTXD        0x00bb
+#define HC_SIMA_HFthRTXA        0x00bc
+/* Define the texture palette 0
+ */
+#define HC_SIMA_HTP0            0x0100
+#define HC_SIMA_HTP1            0x0200
+#define HC_SIMA_FOGTABLE        0x0300
+#define HC_SIMA_STIPPLE         0x0400
+#define HC_SIMA_HE3Fire         0x0440
+#define HC_SIMA_TRANS_SET       0x0441
+#define HC_SIMA_HREngSt         0x0442
+#define HC_SIMA_HRFIFOempty     0x0443
+#define HC_SIMA_HRFIFOfull      0x0444
+#define HC_SIMA_HRErr           0x0445
+#define HC_SIMA_FIFOstatus      0x0446
+
+/******************************************************************************
+** Define the AGP command header.
+******************************************************************************/
+#define HC_ACMD_MASK            0xfe000000      
+#define HC_ACMD_SUB_MASK        0x0c000000      
+#define HC_ACMD_HCmdA           0xee000000      
+#define HC_ACMD_HCmdB           0xec000000      
+#define HC_ACMD_HCmdC           0xea000000      
+#define HC_ACMD_H1              0xf0000000      
+#define HC_ACMD_H2              0xf2000000      
+#define HC_ACMD_H3              0xf4000000      
+#define HC_ACMD_H4              0xf6000000      
+
+#define HC_ACMD_H1IO_MASK       0x000001ff
+#define HC_ACMD_H2IO1_MASK      0x001ff000      
+#define HC_ACMD_H2IO2_MASK      0x000001ff
+#define HC_ACMD_H2IO1_SHIFT     12              
+#define HC_ACMD_H2IO2_SHIFT     0
+#define HC_ACMD_H3IO_MASK       0x000001ff
+#define HC_ACMD_H3COUNT_MASK    0x01fff000      
+#define HC_ACMD_H3COUNT_SHIFT   12              
+#define HC_ACMD_H4ID_MASK       0x000001ff
+#define HC_ACMD_H4COUNT_MASK    0x01fffe00
+#define HC_ACMD_H4COUNT_SHIFT   9
+
+/********************************************************************************
+** Define Header 
+********************************************************************************/
+#define HC_HEADER2		0xF210F110
+
+/********************************************************************************
+** Define Dummy Value 
+********************************************************************************/
+#define HC_DUMMY		0xCCCCCCCC
+/********************************************************************************
+** Define for DMA use 
+********************************************************************************/
+#define HALCYON_HEADER2     0XF210F110
+#define HALCYON_FIRECMD     0XEE100000  
+#define HALCYON_FIREMASK    0XFFF00000
+#define HALCYON_CMDB        0XEC000000  
+#define HALCYON_CMDBMASK    0XFFFE0000
+#define HALCYON_SUB_ADDR0   0X00000000
+#define HALCYON_HEADER1MASK 0XFFFFFF00
+#define HALCYON_HEADER1     0XF0000000
+#define HC_SubA_HAGPBstL        0x0060
+#define HC_SubA_HAGPBendL       0x0061
+#define HC_SubA_HAGPCMNT        0x0062
+#define HC_SubA_HAGPBpL         0x0063
+#define HC_SubA_HAGPBpH         0x0064
+#define HC_HAGPCMNT_MASK        0x00800000
+#define HC_HCmdErrClr_MASK      0x00400000
+#define HC_HAGPBendH_MASK       0x0000ff00
+#define HC_HAGPBstH_MASK        0x000000ff
+#define HC_HAGPBendH_SHIFT      8
+#define HC_HAGPBstH_SHIFT       0
+#define HC_HAGPBpL_MASK         0x00fffffc
+#define HC_HAGPBpID_MASK        0x00000003
+#define HC_HAGPBpID_PAUSE       0x00000000
+#define HC_HAGPBpID_JUMP        0x00000001
+#define HC_HAGPBpID_STOP        0x00000002
+#define HC_HAGPBpH_MASK         0x00ffffff
+
+#endif
diff --git a/src/via_context.c b/src/via_context.c
new file mode 100644
index 0000000..7c73877
--- /dev/null
+++ b/src/via_context.c
@@ -0,0 +1,956 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+/**
+ * \file via_context.c
+ * 
+ * \author John Sheng (presumably of either VIA Technologies or S3 Graphics)
+ * \author Others at VIA Technologies?
+ * \author Others at S3 Graphics?
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "matrix.h"
+#include "state.h"
+#include "simple_list.h"
+#include "extensions.h"
+#include "framebuffer.h"
+#include "renderbuffer.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/tnl.h"
+#include "vbo/vbo.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "drivers/common/driverfuncs.h"
+
+#include "via_screen.h"
+#include "via_dri.h"
+
+#include "via_state.h"
+#include "via_tex.h"
+#include "via_span.h"
+#include "via_tris.h"
+#include "via_ioctl.h"
+#include "via_fb.h"
+
+#include <stdio.h>
+#include "macros.h"
+#include "drirenderbuffer.h"
+
+#define need_GL_ARB_multisample
+#define need_GL_ARB_point_parameters
+#define need_GL_ARB_vertex_buffer_object
+#define need_GL_EXT_fog_coord
+#define need_GL_EXT_secondary_color
+#include "extension_helper.h"
+
+#define DRIVER_DATE	"20060710"
+
+#include "vblank.h"
+#include "utils.h"
+
+GLuint VIA_DEBUG = 0;
+
+/**
+ * Return various strings for \c glGetString.
+ *
+ * \sa glGetString
+ */
+static const GLubyte *viaGetString(GLcontext *ctx, GLenum name)
+{
+   static char buffer[128];
+   unsigned   offset;
+
+
+   switch (name) {
+   case GL_VENDOR:
+      return (GLubyte *)"VIA Technology";
+
+   case GL_RENDERER: {
+      static const char * const chipset_names[] = {
+	 "UniChrome",
+	 "CastleRock (CLE266)",
+	 "UniChrome (KM400)",
+	 "UniChrome (K8M800)",
+	 "UniChrome (PM8x0/CN400)",
+      };
+      struct via_context *vmesa = VIA_CONTEXT(ctx);
+      unsigned id = vmesa->viaScreen->deviceID;
+
+      offset = driGetRendererString( buffer, 
+				     chipset_names[(id > VIA_PM800) ? 0 : id],
+				     DRIVER_DATE, 0 );
+      return (GLubyte *)buffer;
+   }
+
+   default:
+      return NULL;
+   }
+}
+
+
+/**
+ * Calculate a width that satisfies the hardware's alignment requirements.
+ * On the Unichrome hardware, each scanline must be aligned to a multiple of
+ * 16 pixels.
+ *
+ * \param width  Minimum buffer width, in pixels.
+ * 
+ * \returns A pixel width that meets the alignment requirements.
+ */
+static __inline__ unsigned
+buffer_align( unsigned width )
+{
+    return (width + 0x0f) & ~0x0f;
+}
+
+
+static void
+viaDeleteRenderbuffer(struct gl_renderbuffer *rb)
+{
+   /* Don't free() since we're contained in via_context struct. */
+}
+
+static GLboolean
+viaRenderbufferStorage(GLcontext *ctx, struct gl_renderbuffer *rb,
+                       GLenum internalFormat, GLuint width, GLuint height)
+{
+   rb->Width = width;
+   rb->Height = height;
+   rb->InternalFormat = internalFormat;
+   return GL_TRUE;
+}
+
+
+static void
+viaInitRenderbuffer(struct via_renderbuffer *vrb, GLenum format,
+		    __DRIdrawablePrivate *dPriv)
+{
+   const GLuint name = 0;
+   struct gl_renderbuffer *rb = & vrb->Base;
+
+   vrb->dPriv = dPriv;
+   _mesa_init_renderbuffer(rb, name);
+
+   /* Make sure we're using a null-valued GetPointer routine */
+   assert(rb->GetPointer(NULL, rb, 0, 0) == NULL);
+
+   rb->InternalFormat = format;
+
+   if (format == GL_RGBA) {
+      /* Color */
+      rb->_BaseFormat = GL_RGBA;
+      rb->DataType = GL_UNSIGNED_BYTE;
+   }
+   else if (format == GL_DEPTH_COMPONENT16) {
+      /* Depth */
+      rb->_BaseFormat = GL_DEPTH_COMPONENT;
+      /* we always Get/Put 32-bit Z values */
+      rb->DataType = GL_UNSIGNED_INT;
+   }
+   else if (format == GL_DEPTH_COMPONENT24) {
+      /* Depth */
+      rb->_BaseFormat = GL_DEPTH_COMPONENT;
+      /* we always Get/Put 32-bit Z values */
+      rb->DataType = GL_UNSIGNED_INT;
+   }
+   else {
+      /* Stencil */
+      ASSERT(format == GL_STENCIL_INDEX8_EXT);
+      rb->_BaseFormat = GL_STENCIL_INDEX;
+      rb->DataType = GL_UNSIGNED_BYTE;
+   }
+
+   rb->Delete = viaDeleteRenderbuffer;
+   rb->AllocStorage = viaRenderbufferStorage;
+}
+
+
+/**
+ * Calculate the framebuffer parameters for all buffers (front, back, depth,
+ * and stencil) associated with the specified context.
+ * 
+ * \warning
+ * This function also calls \c AllocateBuffer to actually allocate the
+ * buffers.
+ * 
+ * \sa AllocateBuffer
+ */
+static GLboolean
+calculate_buffer_parameters(struct via_context *vmesa,
+			    struct gl_framebuffer *fb,
+			    __DRIdrawablePrivate *dPriv)
+{
+   const unsigned shift = vmesa->viaScreen->bitsPerPixel / 16;
+   const unsigned extra = 32;
+   unsigned w;
+   unsigned h;
+
+   /* Normally, the renderbuffer would be added to the framebuffer just once
+    * when the framebuffer was created.  The VIA driver is a bit funny
+    * though in that the front/back/depth renderbuffers are in the per-context
+    * state!
+    * That should be fixed someday.
+    */
+
+   if (!vmesa->front.Base.InternalFormat) {
+      /* do one-time init for the renderbuffers */
+      viaInitRenderbuffer(&vmesa->front, GL_RGBA, dPriv);
+      viaSetSpanFunctions(&vmesa->front, &fb->Visual);
+      _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &vmesa->front.Base);
+
+      if (fb->Visual.doubleBufferMode) {
+         viaInitRenderbuffer(&vmesa->back, GL_RGBA, dPriv);
+         viaSetSpanFunctions(&vmesa->back, &fb->Visual);
+         _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &vmesa->back.Base);
+      }
+
+      if (vmesa->glCtx->Visual.depthBits > 0) {
+         viaInitRenderbuffer(&vmesa->depth,
+                             (vmesa->glCtx->Visual.depthBits == 16
+                              ? GL_DEPTH_COMPONENT16 : GL_DEPTH_COMPONENT24),
+			     dPriv);
+         viaSetSpanFunctions(&vmesa->depth, &fb->Visual);
+         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &vmesa->depth.Base);
+      }
+
+      if (vmesa->glCtx->Visual.stencilBits > 0) {
+         viaInitRenderbuffer(&vmesa->stencil, GL_STENCIL_INDEX8_EXT,
+			     dPriv);
+         viaSetSpanFunctions(&vmesa->stencil, &fb->Visual);
+         _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &vmesa->stencil.Base);
+      }
+   }
+
+   assert(vmesa->front.Base.InternalFormat);
+   assert(vmesa->front.Base.AllocStorage);
+   if (fb->Visual.doubleBufferMode) {
+      assert(vmesa->back.Base.AllocStorage);
+   }
+   if (fb->Visual.depthBits) {
+      assert(vmesa->depth.Base.AllocStorage);
+   }
+
+
+   /* Allocate front-buffer */
+   if (vmesa->drawType == GLX_PBUFFER_BIT) {
+      w = vmesa->driDrawable->w;
+      h = vmesa->driDrawable->h;
+
+      vmesa->front.bpp = vmesa->viaScreen->bitsPerPixel;
+      vmesa->front.pitch = buffer_align( w ) << shift; /* bytes, not pixels */
+      vmesa->front.size = vmesa->front.pitch * h;
+
+      if (vmesa->front.map)
+	 via_free_draw_buffer(vmesa, &vmesa->front);
+      if (!via_alloc_draw_buffer(vmesa, &vmesa->front))
+	 return GL_FALSE;
+
+   } else {
+      w = vmesa->viaScreen->width;
+      h = vmesa->viaScreen->height;
+
+      vmesa->front.bpp = vmesa->viaScreen->bitsPerPixel;
+      vmesa->front.pitch = buffer_align( w ) << shift; /* bytes, not pixels */
+      vmesa->front.size = vmesa->front.pitch * h;
+      if (getenv("ALTERNATE_SCREEN")) 
+        vmesa->front.offset = vmesa->front.size;
+      else
+      	vmesa->front.offset = 0;
+      vmesa->front.map = (char *) vmesa->driScreen->pFB;
+   }
+
+
+   /* Allocate back-buffer */
+   if (vmesa->hasBack) {
+      vmesa->back.bpp = vmesa->viaScreen->bitsPerPixel;
+      vmesa->back.pitch = (buffer_align( vmesa->driDrawable->w ) << shift);
+      vmesa->back.pitch += extra;
+      vmesa->back.pitch = MIN2(vmesa->back.pitch, vmesa->front.pitch);
+      vmesa->back.size = vmesa->back.pitch * vmesa->driDrawable->h;
+      if (vmesa->back.map)
+	 via_free_draw_buffer(vmesa, &vmesa->back);
+      if (!via_alloc_draw_buffer(vmesa, &vmesa->back))
+	 return GL_FALSE;
+   }
+   else {
+      if (vmesa->back.map)
+	 via_free_draw_buffer(vmesa, &vmesa->back);
+      (void) memset( &vmesa->back, 0, sizeof( vmesa->back ) );
+   }
+
+
+   /* Allocate depth-buffer */
+   if ( vmesa->hasStencil || vmesa->hasDepth ) {
+      vmesa->depth.bpp = vmesa->depthBits;
+      if (vmesa->depth.bpp == 24)
+	 vmesa->depth.bpp = 32;
+
+      vmesa->depth.pitch = (buffer_align( vmesa->driDrawable->w ) * 
+			    (vmesa->depth.bpp/8)) + extra;
+      vmesa->depth.size = vmesa->depth.pitch * vmesa->driDrawable->h;
+
+      if (vmesa->depth.map)
+	 via_free_draw_buffer(vmesa, &vmesa->depth);
+      if (!via_alloc_draw_buffer(vmesa, &vmesa->depth)) {
+	 return GL_FALSE;
+      }
+   }
+   else {
+      if (vmesa->depth.map)
+   	 via_free_draw_buffer(vmesa, &vmesa->depth);
+      (void) memset( & vmesa->depth, 0, sizeof( vmesa->depth ) );
+   }
+
+   /* stencil buffer is same as depth buffer */
+   vmesa->stencil.handle = vmesa->depth.handle;
+   vmesa->stencil.size = vmesa->depth.size;
+   vmesa->stencil.offset = vmesa->depth.offset;
+   vmesa->stencil.index = vmesa->depth.index;
+   vmesa->stencil.pitch = vmesa->depth.pitch;
+   vmesa->stencil.bpp = vmesa->depth.bpp;
+   vmesa->stencil.map = vmesa->depth.map;
+   vmesa->stencil.orig = vmesa->depth.orig;
+   vmesa->stencil.origMap = vmesa->depth.origMap;
+
+   if( vmesa->viaScreen->width == vmesa->driDrawable->w && 
+       vmesa->viaScreen->height == vmesa->driDrawable->h ) {
+      vmesa->doPageFlip = vmesa->allowPageFlip;
+      if (vmesa->hasBack) {
+         assert(vmesa->back.pitch == vmesa->front.pitch);
+      }
+   }
+   else
+      vmesa->doPageFlip = GL_FALSE;
+
+   return GL_TRUE;
+}
+
+
+void viaReAllocateBuffers(GLcontext *ctx, GLframebuffer *drawbuffer,
+                          GLuint width, GLuint height)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    calculate_buffer_parameters(vmesa, drawbuffer, vmesa->driDrawable);
+
+    _mesa_resize_framebuffer(ctx, drawbuffer, width, height);
+}
+
+/* Extension strings exported by the Unichrome driver.
+ */
+const struct dri_extension card_extensions[] =
+{
+    { "GL_ARB_multisample",                GL_ARB_multisample_functions },
+    { "GL_ARB_multitexture",               NULL },
+    { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
+    { "GL_ARB_texture_env_add",            NULL },
+    { "GL_ARB_texture_env_combine",        NULL },
+/*    { "GL_ARB_texture_env_dot3",           NULL }, */
+    { "GL_ARB_texture_mirrored_repeat",    NULL },
+    { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
+    { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+    { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
+    { "GL_EXT_stencil_wrap",               NULL },
+    { "GL_EXT_texture_env_combine",        NULL },
+/*    { "GL_EXT_texture_env_dot3",           NULL }, */
+    { "GL_EXT_texture_lod_bias",           NULL },
+    { "GL_NV_blend_square",                NULL },
+    { NULL,                                NULL }
+};
+
+extern const struct tnl_pipeline_stage _via_fastrender_stage;
+extern const struct tnl_pipeline_stage _via_render_stage;
+
+static const struct tnl_pipeline_stage *via_pipeline[] = {
+    &_tnl_vertex_transform_stage,
+    &_tnl_normal_transform_stage,
+    &_tnl_lighting_stage,
+    &_tnl_fog_coordinate_stage,
+    &_tnl_texgen_stage,
+    &_tnl_texture_transform_stage,
+    /* REMOVE: point attenuation stage */
+#if 1
+    &_via_fastrender_stage,     /* ADD: unclipped rastersetup-to-dma */
+#endif
+    &_tnl_render_stage,
+    0,
+};
+
+
+static const struct dri_debug_control debug_control[] =
+{
+    { "fall",  DEBUG_FALLBACKS },
+    { "tex",   DEBUG_TEXTURE },
+    { "ioctl", DEBUG_IOCTL },
+    { "prim",  DEBUG_PRIMS },
+    { "vert",  DEBUG_VERTS },
+    { "state", DEBUG_STATE },
+    { "verb",  DEBUG_VERBOSE },
+    { "dri",   DEBUG_DRI },
+    { "dma",   DEBUG_DMA },
+    { "san",   DEBUG_SANITY },
+    { "sync",  DEBUG_SYNC },
+    { "sleep", DEBUG_SLEEP },
+    { "pix",   DEBUG_PIXEL },
+    { "2d",    DEBUG_2D },
+    { NULL,    0 }
+};
+
+
+static GLboolean
+AllocateDmaBuffer(struct via_context *vmesa)
+{
+    if (vmesa->dma)
+        via_free_dma_buffer(vmesa);
+    
+    if (!via_alloc_dma_buffer(vmesa))
+        return GL_FALSE;
+
+    vmesa->dmaLow = 0;
+    vmesa->dmaCliprectAddr = ~0;
+    return GL_TRUE;
+}
+
+static void
+FreeBuffer(struct via_context *vmesa)
+{
+    if (vmesa->front.map && vmesa->drawType == GLX_PBUFFER_BIT)
+	via_free_draw_buffer(vmesa, &vmesa->front);
+
+    if (vmesa->back.map)
+        via_free_draw_buffer(vmesa, &vmesa->back);
+
+    if (vmesa->depth.map)
+        via_free_draw_buffer(vmesa, &vmesa->depth);
+
+    if (vmesa->breadcrumb.map)
+        via_free_draw_buffer(vmesa, &vmesa->breadcrumb);
+
+    if (vmesa->dma)
+        via_free_dma_buffer(vmesa);
+}
+
+
+GLboolean
+viaCreateContext(const __GLcontextModes *visual,
+                 __DRIcontextPrivate *driContextPriv,
+                 void *sharedContextPrivate)
+{
+    GLcontext *ctx, *shareCtx;
+    struct via_context *vmesa;
+    __DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+    viaScreenPrivate *viaScreen = (viaScreenPrivate *)sPriv->private;
+    drm_via_sarea_t *saPriv = (drm_via_sarea_t *)
+        (((GLubyte *)sPriv->pSAREA) + viaScreen->sareaPrivOffset);
+    struct dd_function_table functions;
+
+    /* Allocate via context */
+    vmesa = (struct via_context *) CALLOC_STRUCT(via_context);
+    if (!vmesa) {
+        return GL_FALSE;
+    }
+
+    /* Parse configuration files.
+     */
+    driParseConfigFiles (&vmesa->optionCache, &viaScreen->optionCache,
+			 sPriv->myNum, "unichrome");
+
+    /* pick back buffer */
+    vmesa->hasBack = visual->doubleBufferMode;
+
+    switch(visual->depthBits) {
+    case 0:			
+       vmesa->hasDepth = GL_FALSE;
+       vmesa->depthBits = 0; 
+       vmesa->depth_max = 1.0;
+       break;
+    case 16:
+       vmesa->hasDepth = GL_TRUE;
+       vmesa->depthBits = visual->depthBits;
+       vmesa->have_hw_stencil = GL_FALSE;
+       vmesa->depth_max = (GLfloat)0xffff;
+       vmesa->depth_clear_mask = 0xf << 28;
+       vmesa->ClearDepth = 0xffff;
+       vmesa->polygon_offset_scale = 1.0 / vmesa->depth_max;
+       break;
+    case 24:
+       vmesa->hasDepth = GL_TRUE;
+       vmesa->depthBits = visual->depthBits;
+       vmesa->depth_max = (GLfloat) 0xffffff;
+       vmesa->depth_clear_mask = 0xe << 28;
+       vmesa->ClearDepth = 0xffffff00;
+
+       assert(visual->haveStencilBuffer);
+       assert(visual->stencilBits == 8);
+
+       vmesa->have_hw_stencil = GL_TRUE;
+       vmesa->stencilBits = visual->stencilBits;
+       vmesa->stencil_clear_mask = 0x1 << 28;
+       vmesa->polygon_offset_scale = 2.0 / vmesa->depth_max;
+       break;
+    case 32:
+       vmesa->hasDepth = GL_TRUE;
+       vmesa->depthBits = visual->depthBits;
+       assert(!visual->haveStencilBuffer);
+       vmesa->have_hw_stencil = GL_FALSE;
+       vmesa->depth_max = (GLfloat)0xffffffff;
+       vmesa->depth_clear_mask = 0xf << 28;
+       vmesa->ClearDepth = 0xffffffff;
+       vmesa->polygon_offset_scale = 2.0 / vmesa->depth_max;
+       break;
+    default:
+       assert(0); 
+       break;
+    }
+
+    make_empty_list(&vmesa->freed_tex_buffers);
+    make_empty_list(&vmesa->tex_image_list[VIA_MEM_VIDEO]);
+    make_empty_list(&vmesa->tex_image_list[VIA_MEM_AGP]);
+    make_empty_list(&vmesa->tex_image_list[VIA_MEM_SYSTEM]);
+
+    _mesa_init_driver_functions(&functions);
+    viaInitTextureFuncs(&functions);
+
+    /* Allocate the Mesa context */
+    if (sharedContextPrivate)
+        shareCtx = ((struct via_context *) sharedContextPrivate)->glCtx;
+    else
+        shareCtx = NULL;
+
+    vmesa->glCtx = _mesa_create_context(visual, shareCtx, &functions,
+					(void*) vmesa);
+    
+    vmesa->shareCtx = shareCtx;
+    
+    if (!vmesa->glCtx) {
+        FREE(vmesa);
+        return GL_FALSE;
+    }
+    driContextPriv->driverPrivate = vmesa;
+
+    ctx = vmesa->glCtx;
+
+    if (driQueryOptionb(&vmesa->optionCache, "excess_mipmap"))
+        ctx->Const.MaxTextureLevels = 11;
+    else
+        ctx->Const.MaxTextureLevels = 10;
+
+    ctx->Const.MaxTextureUnits = 2;
+    ctx->Const.MaxTextureImageUnits = ctx->Const.MaxTextureUnits;
+    ctx->Const.MaxTextureCoordUnits = ctx->Const.MaxTextureUnits;
+
+    ctx->Const.MinLineWidth = 1.0;
+    ctx->Const.MinLineWidthAA = 1.0;
+    ctx->Const.MaxLineWidth = 1.0;
+    ctx->Const.MaxLineWidthAA = 1.0;
+    ctx->Const.LineWidthGranularity = 1.0;
+
+    ctx->Const.MinPointSize = 1.0;
+    ctx->Const.MinPointSizeAA = 1.0;
+    ctx->Const.MaxPointSize = 1.0;
+    ctx->Const.MaxPointSizeAA = 1.0;
+    ctx->Const.PointSizeGranularity = 1.0;
+
+    ctx->Driver.GetString = viaGetString;
+
+    ctx->DriverCtx = (void *)vmesa;
+    vmesa->glCtx = ctx;
+
+    /* Initialize the software rasterizer and helper modules.
+     */
+    _swrast_CreateContext(ctx);
+    _vbo_CreateContext(ctx);
+    _tnl_CreateContext(ctx);
+    _swsetup_CreateContext(ctx);
+
+    /* Install the customized pipeline:
+     */
+    _tnl_destroy_pipeline(ctx);
+    _tnl_install_pipeline(ctx, via_pipeline);
+
+    /* Configure swrast and T&L to match hardware characteristics:
+     */
+    _swrast_allow_pixel_fog(ctx, GL_FALSE);
+    _swrast_allow_vertex_fog(ctx, GL_TRUE);
+    _tnl_allow_pixel_fog(ctx, GL_FALSE);
+    _tnl_allow_vertex_fog(ctx, GL_TRUE);
+
+/*     vmesa->display = dpy; */
+    vmesa->display = sPriv->display;
+    
+    vmesa->hHWContext = driContextPriv->hHWContext;
+    vmesa->driFd = sPriv->fd;
+    vmesa->driHwLock = &sPriv->pSAREA->lock;
+
+    vmesa->viaScreen = viaScreen;
+    vmesa->driScreen = sPriv;
+    vmesa->sarea = saPriv;
+
+    vmesa->renderIndex = ~0;
+    vmesa->setupIndex = ~0;
+    vmesa->hwPrimitive = GL_POLYGON+1;
+
+    /* KW: Hardwire this.  Was previously set bogusly in
+     * viaCreateBuffer.  Needs work before PBUFFER can be used:
+     */
+    vmesa->drawType = GLX_WINDOW_BIT;
+
+
+    _math_matrix_ctr(&vmesa->ViewportMatrix);
+
+    /* Do this early, before VIA_FLUSH_DMA can be called:
+     */
+    if (!AllocateDmaBuffer(vmesa)) {
+	fprintf(stderr ,"AllocateDmaBuffer fail\n");
+	FreeBuffer(vmesa);
+        FREE(vmesa);
+        return GL_FALSE;
+    }
+
+    /* Allocate a small piece of fb memory for synchronization:
+     */
+    vmesa->breadcrumb.bpp = 32;
+    vmesa->breadcrumb.pitch = buffer_align( 64 ) << 2;
+    vmesa->breadcrumb.size = vmesa->breadcrumb.pitch;
+
+    if (!via_alloc_draw_buffer(vmesa, &vmesa->breadcrumb)) {
+        fprintf(stderr ,"AllocateDmaBuffer fail\n");
+        FreeBuffer(vmesa);
+        FREE(vmesa);
+        return GL_FALSE;
+    }
+
+    driInitExtensions( ctx, card_extensions, GL_TRUE );
+    viaInitStateFuncs(ctx);
+    viaInitTriFuncs(ctx);
+    viaInitSpanFuncs(ctx);
+    viaInitIoctlFuncs(ctx);
+    viaInitState(ctx);
+        
+    if (getenv("VIA_DEBUG"))
+       VIA_DEBUG = driParseDebugString( getenv( "VIA_DEBUG" ),
+					debug_control );
+
+    if (getenv("VIA_NO_RAST") ||
+        driQueryOptionb(&vmesa->optionCache, "no_rast"))
+       FALLBACK(vmesa, VIA_FALLBACK_USER_DISABLE, 1);
+
+    vmesa->vblank_flags =
+       vmesa->viaScreen->irqEnabled ?
+        driGetDefaultVBlankFlags(&vmesa->optionCache) : VBLANK_FLAG_NO_IRQ;
+
+    if (getenv("VIA_PAGEFLIP"))
+       vmesa->allowPageFlip = 1;
+
+    (*dri_interface->getUST)( &vmesa->swap_ust );
+
+
+    vmesa->regMMIOBase = (GLuint *)((unsigned long)viaScreen->reg);
+    vmesa->pnGEMode = (GLuint *)((unsigned long)viaScreen->reg + 0x4);
+    vmesa->regEngineStatus = (GLuint *)((unsigned long)viaScreen->reg + 0x400);
+    vmesa->regTranSet = (GLuint *)((unsigned long)viaScreen->reg + 0x43C);
+    vmesa->regTranSpace = (GLuint *)((unsigned long)viaScreen->reg + 0x440);
+    vmesa->agpBase = viaScreen->agpBase;
+
+
+    return GL_TRUE;
+}
+
+void
+viaDestroyContext(__DRIcontextPrivate *driContextPriv)
+{
+    GET_CURRENT_CONTEXT(ctx);
+    struct via_context *vmesa = 
+       (struct via_context *)driContextPriv->driverPrivate;
+    struct via_context *current = ctx ? VIA_CONTEXT(ctx) : NULL;
+    assert(vmesa); /* should never be null */
+
+    /* check if we're deleting the currently bound context */
+    if (vmesa == current) {
+      VIA_FLUSH_DMA(vmesa);
+      _mesa_make_current(NULL, NULL, NULL);
+    }
+
+    if (vmesa) {
+        viaWaitIdle(vmesa, GL_FALSE);
+	if (vmesa->doPageFlip) {
+	   LOCK_HARDWARE(vmesa);
+	   if (vmesa->pfCurrentOffset != 0) {
+	      fprintf(stderr, "%s - reset pf\n", __FUNCTION__);
+	      viaResetPageFlippingLocked(vmesa);
+	   }
+	   UNLOCK_HARDWARE(vmesa);
+	}
+	
+	_swsetup_DestroyContext(vmesa->glCtx);
+        _tnl_DestroyContext(vmesa->glCtx);
+        _vbo_DestroyContext(vmesa->glCtx);
+        _swrast_DestroyContext(vmesa->glCtx);
+        /* free the Mesa context */
+	_mesa_destroy_context(vmesa->glCtx);
+	/* release our data */
+	FreeBuffer(vmesa);
+
+	assert (is_empty_list(&vmesa->tex_image_list[VIA_MEM_AGP]));
+	assert (is_empty_list(&vmesa->tex_image_list[VIA_MEM_VIDEO]));
+	assert (is_empty_list(&vmesa->tex_image_list[VIA_MEM_SYSTEM]));
+	assert (is_empty_list(&vmesa->freed_tex_buffers));
+
+	driDestroyOptionCache(&vmesa->optionCache);
+
+	FREE(vmesa);
+    }
+}
+
+
+void viaXMesaWindowMoved(struct via_context *vmesa)
+{
+   __DRIdrawablePrivate *const drawable = vmesa->driDrawable;
+   __DRIdrawablePrivate *const readable = vmesa->driReadable;
+   struct via_renderbuffer * draw_buffer;
+   struct via_renderbuffer * read_buffer;
+   GLuint bytePerPixel = vmesa->viaScreen->bitsPerPixel >> 3;
+
+   if (!drawable)
+      return;
+
+   draw_buffer =  (struct via_renderbuffer *) drawable->driverPrivate;
+   read_buffer =  (struct via_renderbuffer *) readable->driverPrivate;
+   
+   switch (vmesa->glCtx->DrawBuffer->_ColorDrawBufferMask[0]) {
+   case BUFFER_BIT_BACK_LEFT: 
+      if (drawable->numBackClipRects == 0) {
+	 vmesa->numClipRects = drawable->numClipRects;
+	 vmesa->pClipRects = drawable->pClipRects;
+      } 
+      else {
+	 vmesa->numClipRects = drawable->numBackClipRects;
+	 vmesa->pClipRects = drawable->pBackClipRects;
+      }
+      break;
+   case BUFFER_BIT_FRONT_LEFT:
+      vmesa->numClipRects = drawable->numClipRects;
+      vmesa->pClipRects = drawable->pClipRects;
+      break;
+   default:
+      vmesa->numClipRects = 0;
+      break;
+   }
+
+   if ((draw_buffer->drawW != drawable->w) 
+       || (draw_buffer->drawH != drawable->h)) {
+      calculate_buffer_parameters(vmesa, vmesa->glCtx->DrawBuffer,
+				  drawable);
+   }
+
+   draw_buffer->drawX = drawable->x;
+   draw_buffer->drawY = drawable->y;
+   draw_buffer->drawW = drawable->w;
+   draw_buffer->drawH = drawable->h;
+
+   if (drawable != readable) {
+      if ((read_buffer->drawW != readable->w) 
+	  || (read_buffer->drawH != readable->h)) {
+	 calculate_buffer_parameters(vmesa, vmesa->glCtx->ReadBuffer,
+				     readable);
+      }
+
+      read_buffer->drawX = readable->x;
+      read_buffer->drawY = readable->y;
+      read_buffer->drawW = readable->w;
+      read_buffer->drawH = readable->h;
+   }
+
+   vmesa->front.orig = (vmesa->front.offset + 
+			draw_buffer->drawY * vmesa->front.pitch + 
+			draw_buffer->drawX * bytePerPixel);
+
+   vmesa->front.origMap = (vmesa->front.map + 
+			draw_buffer->drawY * vmesa->front.pitch + 
+			draw_buffer->drawX * bytePerPixel);
+
+   vmesa->back.orig = (vmesa->back.offset +
+			draw_buffer->drawY * vmesa->back.pitch +
+			draw_buffer->drawX * bytePerPixel);
+
+   vmesa->back.origMap = (vmesa->back.map +
+			draw_buffer->drawY * vmesa->back.pitch +
+			draw_buffer->drawX * bytePerPixel);
+
+   vmesa->depth.orig = (vmesa->depth.offset +
+			draw_buffer->drawY * vmesa->depth.pitch +
+			draw_buffer->drawX * bytePerPixel);   
+
+   vmesa->depth.origMap = (vmesa->depth.map +
+			draw_buffer->drawY * vmesa->depth.pitch +
+			draw_buffer->drawX * bytePerPixel);
+
+   viaCalcViewport(vmesa->glCtx);
+}
+
+GLboolean
+viaUnbindContext(__DRIcontextPrivate *driContextPriv)
+{
+    return GL_TRUE;
+}
+
+GLboolean
+viaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+               __DRIdrawablePrivate *driDrawPriv,
+               __DRIdrawablePrivate *driReadPriv)
+{
+    if (VIA_DEBUG & DEBUG_DRI) {
+	fprintf(stderr, "driContextPriv = %016lx\n", (unsigned long)driContextPriv);
+	fprintf(stderr, "driDrawPriv = %016lx\n", (unsigned long)driDrawPriv);    
+	fprintf(stderr, "driReadPriv = %016lx\n", (unsigned long)driReadPriv);
+    }	
+
+    if (driContextPriv) {
+        struct via_context *vmesa = 
+	   (struct via_context *)driContextPriv->driverPrivate;
+	GLcontext *ctx = vmesa->glCtx;
+        struct gl_framebuffer *drawBuffer, *readBuffer;
+
+        drawBuffer = (GLframebuffer *)driDrawPriv->driverPrivate;
+        readBuffer = (GLframebuffer *)driReadPriv->driverPrivate;
+
+	if (vmesa->driDrawable != driDrawPriv) {
+	   driDrawableInitVBlank(driDrawPriv, vmesa->vblank_flags,
+				 &vmesa->vbl_seq);
+	}
+
+       if ((vmesa->driDrawable != driDrawPriv)
+	   || (vmesa->driReadable != driReadPriv)) {
+	  vmesa->driDrawable = driDrawPriv;
+	  vmesa->driReadable = driReadPriv;
+
+	  if ((drawBuffer->Width != driDrawPriv->w) 
+	      || (drawBuffer->Height != driDrawPriv->h)) {
+	     _mesa_resize_framebuffer(ctx, drawBuffer,
+				      driDrawPriv->w, driDrawPriv->h);
+	     drawBuffer->Initialized = GL_TRUE;
+	  }
+
+	  if (!calculate_buffer_parameters(vmesa, drawBuffer, driDrawPriv)) {
+	     return GL_FALSE;
+	  }
+
+	  if (driDrawPriv != driReadPriv) {
+	     if ((readBuffer->Width != driReadPriv->w)
+		 || (readBuffer->Height != driReadPriv->h)) {
+		_mesa_resize_framebuffer(ctx, readBuffer,
+					 driReadPriv->w, driReadPriv->h);
+		readBuffer->Initialized = GL_TRUE;
+	     }
+
+	     if (!calculate_buffer_parameters(vmesa, readBuffer, driReadPriv)) {
+		return GL_FALSE;
+	     }
+	  }
+       }
+
+        _mesa_make_current(vmesa->glCtx, drawBuffer, readBuffer);
+
+	ctx->Driver.DrawBuffer( ctx, ctx->Color.DrawBuffer[0] );
+	   
+        viaXMesaWindowMoved(vmesa);
+	ctx->Driver.Scissor(vmesa->glCtx,
+			    vmesa->glCtx->Scissor.X,
+			    vmesa->glCtx->Scissor.Y,
+			    vmesa->glCtx->Scissor.Width,
+			    vmesa->glCtx->Scissor.Height);
+    }
+    else {
+        _mesa_make_current(NULL, NULL, NULL);
+    }
+        
+    return GL_TRUE;
+}
+
+void viaGetLock(struct via_context *vmesa, GLuint flags)
+{
+    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;
+    __DRIscreenPrivate *sPriv = vmesa->driScreen;
+
+    drmGetLock(vmesa->driFd, vmesa->hHWContext, flags);
+
+    DRI_VALIDATE_DRAWABLE_INFO(sPriv, dPriv);
+    if (dPriv != vmesa->driReadable) {
+	DRI_VALIDATE_DRAWABLE_INFO(sPriv, vmesa->driReadable);
+    }
+
+    if (vmesa->sarea->ctxOwner != vmesa->hHWContext) {
+       vmesa->sarea->ctxOwner = vmesa->hHWContext;
+       vmesa->newEmitState = ~0;
+    }
+
+    if (vmesa->lastStamp != dPriv->lastStamp) {
+       viaXMesaWindowMoved(vmesa);
+       driUpdateFramebufferSize(vmesa->glCtx, dPriv);
+       vmesa->newEmitState = ~0;
+       vmesa->lastStamp = dPriv->lastStamp;
+    }
+
+    if (vmesa->doPageFlip &&
+	vmesa->pfCurrentOffset != vmesa->sarea->pfCurrentOffset) {
+       fprintf(stderr, "%s - reset pf\n", __FUNCTION__);
+       viaResetPageFlippingLocked(vmesa);
+    }
+}
+
+
+void
+viaSwapBuffers(__DRIdrawablePrivate *drawablePrivate)
+{
+    __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *)drawablePrivate;
+
+    if (dPriv && 
+	dPriv->driContextPriv && 
+	dPriv->driContextPriv->driverPrivate) {
+        struct via_context *vmesa = 
+	   (struct via_context *)dPriv->driContextPriv->driverPrivate;
+        GLcontext *ctx = vmesa->glCtx;
+
+	_mesa_notifySwapBuffers(ctx);
+
+        if (ctx->Visual.doubleBufferMode) {
+            if (vmesa->doPageFlip) {
+                viaPageFlip(dPriv);
+            }
+            else {
+                viaCopyBuffer(dPriv);
+            }
+        }
+	else
+	    VIA_FLUSH_DMA(vmesa);
+    }
+    else {
+        _mesa_problem(NULL, "viaSwapBuffers: drawable has no context!\n");
+    }
+}
diff --git a/src/via_context.h b/src/via_context.h
new file mode 100644
index 0000000..fecd278
--- /dev/null
+++ b/src/via_context.h
@@ -0,0 +1,423 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+
+#ifndef _VIACONTEXT_H
+#define _VIACONTEXT_H
+
+#include "dri_util.h"
+
+#include "mtypes.h"
+#include "drm.h"
+#include "mm.h"
+#include "tnl/t_vertex.h"
+
+#include "via_screen.h"
+#include "via_tex.h"
+#include "via_drm.h"
+
+struct via_context;
+
+/* Chip tags.  These are used to group the adapters into
+ * related families.
+ */
+enum VIACHIPTAGS {
+    VIA_UNKNOWN = 0,
+    VIA_CLE266,
+    VIA_KM400,
+    VIA_K8M800,
+    VIA_PM800,
+    VIA_LAST
+};
+
+#define VIA_FALLBACK_TEXTURE           	0x1
+#define VIA_FALLBACK_DRAW_BUFFER       	0x2
+#define VIA_FALLBACK_READ_BUFFER       	0x4
+#define VIA_FALLBACK_COLORMASK         	0x8
+#define VIA_FALLBACK_SPECULAR          	0x20
+#define VIA_FALLBACK_LOGICOP           	0x40
+#define VIA_FALLBACK_RENDERMODE        	0x80
+#define VIA_FALLBACK_STENCIL           	0x100
+#define VIA_FALLBACK_BLEND_EQ          	0x200
+#define VIA_FALLBACK_BLEND_FUNC        	0x400
+#define VIA_FALLBACK_USER_DISABLE      	0x800
+#define VIA_FALLBACK_PROJ_TEXTURE      	0x1000
+#define VIA_FALLBACK_POLY_STIPPLE	0x2000
+
+#define VIA_DMA_BUFSIZ                  4096
+#define VIA_DMA_HIGHWATER               (VIA_DMA_BUFSIZ - 128)
+
+#define VIA_NO_CLIPRECTS 0x1
+
+
+/* Use the templated vertex formats:
+ */
+#define TAG(x) via##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+typedef void (*via_tri_func)(struct via_context *, viaVertex *, viaVertex *,
+                             viaVertex *);
+typedef void (*via_line_func)(struct via_context *, viaVertex *, viaVertex *);
+typedef void (*via_point_func)(struct via_context *, viaVertex *);
+
+/**
+ * Derived from gl_renderbuffer.
+ */
+struct via_renderbuffer {
+   struct gl_renderbuffer Base;  /* must be first! */
+   drm_handle_t handle;
+   drmSize size;
+   unsigned long offset;
+   unsigned long index;
+   GLuint pitch;
+   GLuint bpp;
+   char *map;
+   GLuint orig;		/* The drawing origin, 
+			 * at (drawX,drawY) in screen space.
+			 */
+   char *origMap;
+
+   int drawX;                   /* origin of drawable in draw buffer */
+   int drawY;    
+   int drawW;                  
+   int drawH;    
+
+   __DRIdrawablePrivate *dPriv;
+};
+
+
+#define VIA_MAX_TEXLEVELS	10
+
+struct via_tex_buffer {
+   struct via_tex_buffer *next, *prev;
+   struct via_texture_image *image;
+   unsigned long index;
+   unsigned long offset;
+   GLuint size;
+   GLuint memType;    
+   unsigned char *bufAddr;
+   GLuint texBase;
+   GLuint lastUsed;
+};
+
+
+
+struct via_texture_image {
+   struct gl_texture_image image;
+   struct via_tex_buffer *texMem;
+   GLint pitchLog2;
+};
+
+struct via_texture_object {
+   struct gl_texture_object obj; /* The "parent" object */
+
+   GLuint texelBytes;
+   GLuint memType;
+
+   GLuint regTexFM;
+   GLuint regTexWidthLog2[2];
+   GLuint regTexHeightLog2[2];
+   GLuint regTexBaseH[4];
+   struct {
+      GLuint baseL;
+      GLuint pitchLog2;
+   } regTexBaseAndPitch[12];
+
+   GLint firstLevel, lastLevel;  /* upload tObj->Image[first .. lastLevel] */
+};              
+
+
+
+struct via_context {
+   GLint refcount;   
+   GLcontext *glCtx;
+   GLcontext *shareCtx;
+
+   /* XXX These don't belong here.  They should be per-drawable state. */
+   struct via_renderbuffer front;
+   struct via_renderbuffer back;
+   struct via_renderbuffer depth;
+   struct via_renderbuffer stencil; /* mirrors depth */
+   struct via_renderbuffer breadcrumb;
+
+   GLboolean hasBack;
+   GLboolean hasDepth;
+   GLboolean hasStencil;
+   GLboolean hasAccum;
+   GLuint    depthBits;
+   GLuint    stencilBits;
+
+   GLboolean have_hw_stencil;
+   GLuint ClearDepth;
+   GLuint depth_clear_mask;
+   GLuint stencil_clear_mask;
+   GLfloat depth_max;
+   GLfloat polygon_offset_scale;
+
+   GLubyte    *dma;
+   viaRegion tex;
+    
+   /* Bit flag to keep 0track of fallbacks.
+    */
+   GLuint Fallback;
+
+   /* State for via_tris.c.
+    */
+   GLuint newState;            /* _NEW_* flags */
+   GLuint newEmitState;            /* _NEW_* flags */
+   GLuint newRenderState;            /* _NEW_* flags */
+
+   struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
+   GLuint vertex_attr_count;
+
+   GLuint setupIndex;
+   GLuint renderIndex;
+   GLmatrix ViewportMatrix;
+   GLenum renderPrimitive;
+   GLenum hwPrimitive;
+   GLenum hwShadeModel;
+   unsigned char *verts;
+
+   /* drmBufPtr dma_buffer;
+    */
+   GLuint dmaLow;
+   GLuint dmaCliprectAddr;
+   GLuint dmaLastPrim;
+   GLboolean useAgp;
+   
+
+   /* Fallback rasterization functions 
+    */
+   via_point_func drawPoint;
+   via_line_func drawLine;
+   via_tri_func drawTri;
+
+   /* Hardware register
+    */
+   GLuint regCmdA_End;
+   GLuint regCmdB;
+
+   GLuint regEnable;
+   GLuint regHFBBMSKL;
+   GLuint regHROP;
+
+   GLuint regHZWTMD;
+   GLuint regHSTREF;
+   GLuint regHSTMD;
+
+   GLuint regHATMD;
+   GLuint regHABLCsat;
+   GLuint regHABLCop;
+   GLuint regHABLAsat;
+   GLuint regHABLAop;
+   GLuint regHABLRCa;
+   GLuint regHABLRFCa;
+   GLuint regHABLRCbias;
+   GLuint regHABLRCb;
+   GLuint regHABLRFCb;
+   GLuint regHABLRAa;
+   GLuint regHABLRAb;
+   GLuint regHFogLF;
+   GLuint regHFogCL;
+   GLuint regHFogCH;
+
+   GLuint regHLP;
+   GLuint regHLPRF;
+   
+   GLuint regHTXnCLOD[2];
+   GLuint regHTXnTB[2];
+   GLuint regHTXnMPMD[2];
+   GLuint regHTXnTBLCsat[2];
+   GLuint regHTXnTBLCop[2];
+   GLuint regHTXnTBLMPfog[2];
+   GLuint regHTXnTBLAsat[2];
+   GLuint regHTXnTBLRCb[2];
+   GLuint regHTXnTBLRAa[2];
+   GLuint regHTXnTBLRFog[2];
+   GLuint regHTXnTBLRCa[2];
+   GLuint regHTXnTBLRCc[2];
+   GLuint regHTXnTBLRCbias[2];
+   GLuint regHTXnTBC[2];
+   GLuint regHTXnTRAH[2];
+
+   int vertexSize;
+   int hwVertexSize;
+   GLboolean ptexHack;
+   int coloroffset;
+   int specoffset;
+
+   GLint lastStamp;
+
+   GLuint ClearColor;
+   GLuint ClearMask;
+
+   /* DRI stuff
+    */
+   GLboolean doPageFlip;
+
+   struct via_renderbuffer *drawBuffer;
+
+   GLuint numClipRects;         /* cliprects for that buffer */
+   drm_clip_rect_t *pClipRects;
+
+   GLboolean scissor;
+   drm_clip_rect_t drawRect;
+   drm_clip_rect_t scissorRect;
+
+   drm_context_t hHWContext;
+   drm_hw_lock_t *driHwLock;
+   int driFd;
+   __DRInativeDisplay *display;
+
+   /**
+    * DRI drawable bound to this context for drawing.
+    */
+   __DRIdrawablePrivate	*driDrawable;
+
+   /**
+    * DRI drawable bound to this context for reading.
+    */
+   __DRIdrawablePrivate	*driReadable;
+
+   __DRIscreenPrivate *driScreen;
+   viaScreenPrivate *viaScreen;
+   drm_via_sarea_t *sarea;
+   volatile GLuint* regMMIOBase;
+   volatile GLuint* pnGEMode;
+   volatile GLuint* regEngineStatus;
+   volatile GLuint* regTranSet;
+   volatile GLuint* regTranSpace;
+   GLuint agpBase;
+   GLuint drawType;
+
+   GLuint nDoneFirstFlip;
+   GLuint agpFullCount;
+
+   GLboolean clearTexCache;
+   GLboolean thrashing;
+
+   /* Configuration cache
+    */
+   driOptionCache optionCache;
+
+   GLuint vblank_flags;
+   GLuint vbl_seq;
+
+   int64_t swap_ust;
+   int64_t swap_missed_ust;
+
+   GLuint swap_count;
+   GLuint swap_missed_count;
+
+
+   GLuint pfCurrentOffset;
+   GLboolean allowPageFlip;
+
+   GLuint lastBreadcrumbRead;
+   GLuint lastBreadcrumbWrite;
+   GLuint lastSwap[2];
+   GLuint lastDma;
+   
+   GLuint total_alloc[VIA_MEM_SYSTEM+1];
+
+   struct via_tex_buffer tex_image_list[VIA_MEM_SYSTEM+1];
+   struct via_tex_buffer freed_tex_buffers;
+   
+};
+
+
+
+#define VIA_CONTEXT(ctx)   ((struct via_context *)(ctx->DriverCtx))
+
+
+
+/* Lock the hardware and validate our state.  
+ */
+#define LOCK_HARDWARE(vmesa)                                	\
+	do {                                                    \
+    	    char __ret = 0;                                     \
+    	    DRM_CAS(vmesa->driHwLock, vmesa->hHWContext,        \
+        	(DRM_LOCK_HELD|vmesa->hHWContext), __ret);      \
+    	    if (__ret)                                          \
+        	viaGetLock(vmesa, 0);                           \
+	} while (0)
+
+
+/* Release the kernel lock.
+ */
+#define UNLOCK_HARDWARE(vmesa)                                  	\
+	DRM_UNLOCK(vmesa->driFd, vmesa->driHwLock, vmesa->hHWContext);	
+
+	
+
+extern GLuint VIA_DEBUG;
+
+#define DEBUG_TEXTURE	0x1
+#define DEBUG_STATE	0x2
+#define DEBUG_IOCTL	0x4
+#define DEBUG_PRIMS	0x8
+#define DEBUG_VERTS	0x10
+#define DEBUG_FALLBACKS	0x20
+#define DEBUG_VERBOSE	0x40
+#define DEBUG_DRI       0x80
+#define DEBUG_DMA       0x100
+#define DEBUG_SANITY    0x200
+#define DEBUG_SYNC      0x400
+#define DEBUG_SLEEP     0x800
+#define DEBUG_PIXEL     0x1000
+#define DEBUG_2D        0x2000
+
+
+extern void viaGetLock(struct via_context *vmesa, GLuint flags);
+extern void viaLock(struct via_context *vmesa, GLuint flags);
+extern void viaUnLock(struct via_context *vmesa, GLuint flags);
+extern void viaEmitHwStateLocked(struct via_context *vmesa);
+extern void viaEmitScissorValues(struct via_context *vmesa, int box_nr, int emit);
+extern void viaXMesaSetBackClipRects(struct via_context *vmesa);
+extern void viaXMesaSetFrontClipRects(struct via_context *vmesa);
+extern void viaReAllocateBuffers(GLcontext *ctx, GLframebuffer *drawbuffer, GLuint width, GLuint height);
+extern void viaXMesaWindowMoved(struct via_context *vmesa);
+
+extern GLboolean viaTexCombineState(struct via_context *vmesa,
+				    const struct gl_tex_env_combine_state * combine, 
+				    unsigned unit );
+
+/* Via hw already adjusted for GL pixel centers:
+ */
+#define SUBPIXEL_X 0
+#define SUBPIXEL_Y 0
+
+/* TODO XXX _SOLO temp defines to make code compilable */
+#ifndef GLX_PBUFFER_BIT
+#define GLX_PBUFFER_BIT        0x00000004
+#endif
+#ifndef GLX_WINDOW_BIT
+#define GLX_WINDOW_BIT 0x00000001
+#endif
+#ifndef VERT_BIT_CLIP
+#define VERT_BIT_CLIP       0x1000000
+#endif
+
+#endif
diff --git a/src/via_drmclient.h b/src/via_drmclient.h
new file mode 100644
index 0000000..7beff9a
--- /dev/null
+++ b/src/via_drmclient.h
@@ -0,0 +1,29 @@
+/*
+ *
+ * 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
+ * THE AUTHOR(S) AND/OR THE COPYRIGHT HOLDER(S) 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.
+ */
+#ifndef _VIA_DRMCLIENT_H_
+#define _VIA_DRMCLIENT_H_
+
+#define VIA_DMA_BUF_ORDER		12
+#define VIA_DMA_BUF_SZ 		        (1 << VIA_DMA_BUF_ORDER)
+#define VIA_DMA_BUF_NR                  256
+
+#endif				/* _VIA_DRMCLIENT_H_ */
diff --git a/src/via_fb.c b/src/via_fb.c
new file mode 100644
index 0000000..48c7ed4
--- /dev/null
+++ b/src/via_fb.c
@@ -0,0 +1,243 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include <assert.h>
+
+#include "via_context.h"
+#include "via_ioctl.h"
+#include "via_fb.h"
+#include "xf86drm.h"
+#include "imports.h"
+#include "simple_list.h"
+#include <sys/ioctl.h>
+
+GLboolean
+via_alloc_draw_buffer(struct via_context *vmesa, struct via_renderbuffer *buf)
+{
+   drm_via_mem_t mem;
+   mem.context = vmesa->hHWContext;
+   mem.size = buf->size;
+   mem.type = VIA_MEM_VIDEO;
+   mem.offset = 0;
+   mem.index = 0;
+
+   if (ioctl(vmesa->driFd, DRM_IOCTL_VIA_ALLOCMEM, &mem)) 
+      return GL_FALSE;
+    
+    
+   buf->offset = mem.offset;
+   buf->map = (char *)vmesa->driScreen->pFB + mem.offset;
+   buf->index = mem.index;
+   return GL_TRUE;
+}
+
+void
+via_free_draw_buffer(struct via_context *vmesa, struct via_renderbuffer *buf)
+{
+   drm_via_mem_t mem;
+
+   if (!vmesa) return;
+
+   mem.context = vmesa->hHWContext;
+   mem.index = buf->index;
+   mem.type = VIA_MEM_VIDEO;
+   mem.offset = buf->offset;
+   mem.size = buf->size;
+
+   ioctl(vmesa->driFd, DRM_IOCTL_VIA_FREEMEM, &mem);
+   buf->map = NULL;
+}
+
+
+GLboolean
+via_alloc_dma_buffer(struct via_context *vmesa)
+{
+   drm_via_dma_init_t init;
+
+   vmesa->dma = (GLubyte *) malloc(VIA_DMA_BUFSIZ);
+    
+   /*
+    * Check whether AGP DMA has been initialized.
+    */
+   memset(&init, 0, sizeof(init));
+   init.func = VIA_DMA_INITIALIZED;
+
+   vmesa->useAgp = 
+     ( 0 == drmCommandWrite(vmesa->driFd, DRM_VIA_DMA_INIT, 
+			     &init, sizeof(init)));
+   if (VIA_DEBUG & DEBUG_DMA) {
+      if (vmesa->useAgp) 
+         fprintf(stderr, "unichrome_dri.so: Using AGP.\n");
+      else
+         fprintf(stderr, "unichrome_dri.so: Using PCI.\n");
+   }
+      
+   return ((vmesa->dma) ? GL_TRUE : GL_FALSE);
+}
+
+void
+via_free_dma_buffer(struct via_context *vmesa)
+{
+    if (!vmesa) return;
+    free(vmesa->dma);
+    vmesa->dma = 0;
+} 
+
+
+/* These functions now allocate and free the via_tex_buffer struct as well:
+ */
+struct via_tex_buffer *
+via_alloc_texture(struct via_context *vmesa,
+		  GLuint size,
+		  GLuint memType)
+{
+   struct via_tex_buffer *t = CALLOC_STRUCT(via_tex_buffer);
+   
+   if (!t)
+      goto cleanup;
+
+   t->size = size;
+   t->memType = memType;
+   insert_at_tail(&vmesa->tex_image_list[memType], t);
+
+   if (t->memType == VIA_MEM_AGP || 
+       t->memType == VIA_MEM_VIDEO) {
+      drm_via_mem_t fb;
+
+      fb.context = vmesa->hHWContext;
+      fb.size = t->size;
+      fb.type = t->memType;
+      fb.offset = 0;
+      fb.index = 0;
+
+      if (ioctl(vmesa->driFd, DRM_IOCTL_VIA_ALLOCMEM, &fb) != 0 || 
+	  fb.index == 0) 
+	 goto cleanup;
+
+      if (0)
+	 fprintf(stderr, "offset %lx index %lx\n", fb.offset, fb.index);
+
+      t->offset = fb.offset;
+      t->index = fb.index;
+      
+      if (t->memType == VIA_MEM_AGP) {
+	 t->bufAddr = (GLubyte *)((unsigned long)vmesa->viaScreen->agpLinearStart +
+				  fb.offset); 	
+	 t->texBase = vmesa->agpBase + fb.offset;
+      }
+      else {
+	 t->bufAddr = (GLubyte *)((unsigned long)vmesa->driScreen->pFB + fb.offset);
+	 t->texBase = fb.offset;
+      }
+
+      vmesa->total_alloc[t->memType] += t->size;
+      return t;
+   }
+   else if (t->memType == VIA_MEM_SYSTEM) {
+      
+      t->bufAddr = _mesa_malloc(t->size);      
+      if (!t->bufAddr)
+	 goto cleanup;
+
+      vmesa->total_alloc[t->memType] += t->size;
+      return t;
+   }
+
+ cleanup:
+   if (t) {
+      remove_from_list(t);
+      FREE(t);
+   }
+
+   return NULL;
+}
+
+
+static void
+via_do_free_texture(struct via_context *vmesa, struct via_tex_buffer *t)
+{
+   drm_via_mem_t fb;
+
+   remove_from_list( t );
+
+   vmesa->total_alloc[t->memType] -= t->size;
+
+   fb.context = vmesa->hHWContext;
+   fb.index = t->index;
+   fb.offset = t->offset;
+   fb.type = t->memType;
+   fb.size = t->size;
+
+   if (ioctl(vmesa->driFd, DRM_IOCTL_VIA_FREEMEM, &fb)) {
+      fprintf(stderr, "via_free_texture fail\n");
+   }
+
+   FREE(t);
+}
+
+
+/* Release textures which were potentially still being referenced by
+ * hardware at the time when they were originally freed.
+ */
+void 
+via_release_pending_textures( struct via_context *vmesa )
+{
+   struct via_tex_buffer *s, *tmp;
+   
+   foreach_s( s, tmp, &vmesa->freed_tex_buffers ) {
+      if (!VIA_GEQ_WRAP(s->lastUsed, vmesa->lastBreadcrumbRead)) {
+	 if (VIA_DEBUG & DEBUG_TEXTURE)
+	    fprintf(stderr, "%s: release tex sz %d lastUsed %x\n",
+		    __FUNCTION__, s->size, s->lastUsed); 
+	 via_do_free_texture(vmesa, s);
+      }
+   }
+}
+      
+
+
+void
+via_free_texture(struct via_context *vmesa, struct via_tex_buffer *t)
+{
+   if (!t) {
+      return;
+   }
+   else if (t->memType == VIA_MEM_SYSTEM) {
+      remove_from_list(t);
+      vmesa->total_alloc[t->memType] -= t->size;
+      _mesa_free(t->bufAddr);
+      _mesa_free(t);
+   }
+   else if (t->index && viaCheckBreadcrumb(vmesa, t->lastUsed)) {
+      via_do_free_texture( vmesa, t );
+   }
+   else {
+      /* Close current breadcrumb so that we can free this eventually:
+       */
+      if (t->lastUsed == vmesa->lastBreadcrumbWrite) 
+	 viaEmitBreadcrumb(vmesa);
+
+      move_to_tail( &vmesa->freed_tex_buffers, t );
+   }
+}
diff --git a/src/via_fb.h b/src/via_fb.h
new file mode 100644
index 0000000..2d329ac
--- /dev/null
+++ b/src/via_fb.h
@@ -0,0 +1,42 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIAFB_INC
+#define _VIAFB_INC
+
+#include "via_context.h"
+
+extern GLboolean via_alloc_draw_buffer(struct via_context *vmesa, struct via_renderbuffer *buf);
+extern GLboolean via_alloc_dma_buffer(struct via_context *vmesa);
+
+struct via_tex_buffer *
+via_alloc_texture(struct via_context *vmesa,
+		  GLuint size,
+		  GLuint memType);
+
+extern void via_free_draw_buffer(struct via_context *vmesa, struct via_renderbuffer *buf);
+extern void via_free_dma_buffer(struct via_context *vmesa);
+extern void via_free_texture(struct via_context *vmesa, struct via_tex_buffer *t);
+void via_release_pending_textures( struct via_context *vmesa );
+#endif
diff --git a/src/via_ioctl.c b/src/via_ioctl.c
new file mode 100644
index 0000000..4a733fb
--- /dev/null
+++ b/src/via_ioctl.c
@@ -0,0 +1,981 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+#include <stdio.h>
+#include <unistd.h>
+
+#include "glheader.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "dd.h"
+#include "swrast/swrast.h"
+
+#include "mm.h"
+#include "via_context.h"
+#include "via_tris.h"
+#include "via_ioctl.h"
+#include "via_state.h"
+#include "via_fb.h"
+#include "via_3d_reg.h"
+
+#include "vblank.h"
+#include "drm.h"
+#include "xf86drm.h"
+#include <sys/ioctl.h>
+#include <errno.h>
+
+
+#define VIA_REG_STATUS          0x400
+#define VIA_REG_GEMODE          0x004
+#define VIA_REG_SRCBASE         0x030
+#define VIA_REG_DSTBASE         0x034
+#define VIA_REG_PITCH           0x038      
+#define VIA_REG_SRCCOLORKEY     0x01C      
+#define VIA_REG_KEYCONTROL      0x02C       
+#define VIA_REG_SRCPOS          0x008
+#define VIA_REG_DSTPOS          0x00C
+#define VIA_REG_GECMD           0x000
+#define VIA_REG_DIMENSION       0x010       /* width and height */
+#define VIA_REG_FGCOLOR         0x018
+
+#define VIA_GEM_8bpp            0x00000000
+#define VIA_GEM_16bpp           0x00000100
+#define VIA_GEM_32bpp           0x00000300
+#define VIA_GEC_BLT             0x00000001
+#define VIA_PITCH_ENABLE        0x80000000
+#define VIA_GEC_INCX            0x00000000
+#define VIA_GEC_DECY            0x00004000
+#define VIA_GEC_INCY            0x00000000
+#define VIA_GEC_DECX            0x00008000
+#define VIA_GEC_FIXCOLOR_PAT    0x00002000
+
+
+#define VIA_BLIT_CLEAR 0x00
+#define VIA_BLIT_COPY 0xCC
+#define VIA_BLIT_FILL 0xF0
+#define VIA_BLIT_SET 0xFF
+
+static void dump_dma( struct via_context *vmesa )
+{
+   GLuint i;
+   GLuint *data = (GLuint *)vmesa->dma;
+   for (i = 0; i < vmesa->dmaLow; i += 16) {
+      fprintf(stderr, "%04x:   ", i);
+      fprintf(stderr, "%08x  ", *data++);
+      fprintf(stderr, "%08x  ", *data++);
+      fprintf(stderr, "%08x  ", *data++);
+      fprintf(stderr, "%08x\n", *data++);
+   }
+   fprintf(stderr, "******************************************\n");
+}
+
+
+
+void viaCheckDma(struct via_context *vmesa, GLuint bytes)
+{
+    VIA_FINISH_PRIM( vmesa );
+    if (vmesa->dmaLow + bytes > VIA_DMA_HIGHWATER) {
+	viaFlushDma(vmesa);
+    }
+}
+
+
+
+#define SetReg2DAGP(nReg, nData) do {		\
+    OUT_RING( ((nReg) >> 2) | 0xF0000000 );	\
+    OUT_RING( nData );				\
+} while (0)
+
+
+static void viaBlit(struct via_context *vmesa, GLuint bpp,
+		    GLuint srcBase, GLuint srcPitch, 
+		    GLuint dstBase, GLuint dstPitch,
+		    GLuint w, GLuint h, 
+		    GLuint blitMode, 
+		    GLuint color, GLuint nMask ) 
+{
+
+    GLuint dwGEMode, srcX, dstX, cmd;
+    RING_VARS;
+
+    if (VIA_DEBUG & DEBUG_2D)
+       fprintf(stderr, 
+	       "%s bpp %d src %x/%x dst %x/%x w %d h %d "
+	       " mode: %x color: 0x%08x mask 0x%08x\n",
+	       __FUNCTION__, bpp, srcBase, srcPitch, dstBase,
+	       dstPitch, w,h, blitMode, color, nMask);
+
+
+    if (!w || !h)
+        return;
+
+    switch (bpp) {
+    case 16:
+        dwGEMode = VIA_GEM_16bpp;
+	srcX = (srcBase & 0x1f) >> 1;
+	dstX = (dstBase & 0x1f) >> 1;
+        break;
+    case 32:
+        dwGEMode = VIA_GEM_32bpp;
+	srcX = (srcBase & 0x1f) >> 2;
+	dstX = (dstBase & 0x1f) >> 2;
+	break;
+    default:
+        return;
+    }
+
+    switch(blitMode) {
+    case VIA_BLIT_FILL:
+	cmd = VIA_GEC_BLT | VIA_GEC_FIXCOLOR_PAT | (VIA_BLIT_FILL << 24);
+	break;
+    case VIA_BLIT_COPY:
+	cmd = VIA_GEC_BLT | (VIA_BLIT_COPY << 24);
+	break;
+    default:
+        return;
+    }	
+
+    BEGIN_RING(22);
+    SetReg2DAGP( VIA_REG_GEMODE, dwGEMode);
+    SetReg2DAGP( VIA_REG_FGCOLOR, color);
+    SetReg2DAGP( 0x2C, nMask);
+    SetReg2DAGP( VIA_REG_SRCBASE, (srcBase & ~0x1f) >> 3);
+    SetReg2DAGP( VIA_REG_DSTBASE, (dstBase & ~0x1f) >> 3);
+    SetReg2DAGP( VIA_REG_PITCH, VIA_PITCH_ENABLE |
+	       (srcPitch >> 3) | ((dstPitch >> 3) << 16));
+    SetReg2DAGP( VIA_REG_SRCPOS, srcX);
+    SetReg2DAGP( VIA_REG_DSTPOS, dstX);
+    SetReg2DAGP( VIA_REG_DIMENSION, (((h - 1) << 16) | (w - 1)));
+    SetReg2DAGP( VIA_REG_GECMD, cmd);
+    SetReg2DAGP( 0x2C, 0x00000000);
+    ADVANCE_RING();
+}
+
+static void viaFillBuffer(struct via_context *vmesa,
+			  struct via_renderbuffer *buffer,
+			  drm_clip_rect_t *pbox,
+			  int nboxes,
+			  GLuint pixel,
+			  GLuint mask)
+{
+   GLuint bytePerPixel = buffer->bpp >> 3;
+   GLuint i;
+
+   for (i = 0; i < nboxes ; i++) {        
+      int x = pbox[i].x1 - buffer->drawX;
+      int y = pbox[i].y1 - buffer->drawY;
+      int w = pbox[i].x2 - pbox[i].x1;
+      int h = pbox[i].y2 - pbox[i].y1;
+
+      int offset = (buffer->offset + 
+		    y * buffer->pitch + 
+		    x * bytePerPixel);
+
+      viaBlit(vmesa,
+	      buffer->bpp, 
+	      offset, buffer->pitch,
+	      offset, buffer->pitch, 
+	      w, h,
+	      VIA_BLIT_FILL, pixel, mask); 
+   }
+}
+
+
+
+static void viaClear(GLcontext *ctx, GLbitfield mask)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   __DRIdrawablePrivate *dPriv = vmesa->driDrawable;
+   struct via_renderbuffer *const vrb = 
+     (struct via_renderbuffer *) dPriv->driverPrivate;
+   int flag = 0;
+   GLuint i = 0;
+   GLuint clear_depth_mask = 0xf << 28;
+   GLuint clear_depth = 0;
+
+   VIA_FLUSH_DMA(vmesa);
+
+   if (mask & BUFFER_BIT_FRONT_LEFT) {
+      flag |= VIA_FRONT;
+      mask &= ~BUFFER_BIT_FRONT_LEFT;
+   }
+
+   if (mask & BUFFER_BIT_BACK_LEFT) {
+      flag |= VIA_BACK;	
+      mask &= ~BUFFER_BIT_BACK_LEFT;
+   }
+
+   if (mask & BUFFER_BIT_DEPTH) {
+      flag |= VIA_DEPTH;
+      clear_depth = (GLuint)(ctx->Depth.Clear * vmesa->ClearDepth);
+      clear_depth_mask &= ~vmesa->depth_clear_mask;
+      mask &= ~BUFFER_BIT_DEPTH;
+   }
+    
+   if (mask & BUFFER_BIT_STENCIL) {
+      if (vmesa->have_hw_stencil) {
+	 if ((ctx->Stencil.WriteMask[0] & 0xff) == 0xff) {
+	    flag |= VIA_DEPTH;
+	    clear_depth &= ~0xff;
+	    clear_depth |= (ctx->Stencil.Clear & 0xff);
+	    clear_depth_mask &= ~vmesa->stencil_clear_mask;
+	    mask &= ~BUFFER_BIT_STENCIL;
+	 }
+	 else {
+	    if (VIA_DEBUG & DEBUG_2D)
+	       fprintf(stderr, "Clear stencil writemask %x\n", 
+		       ctx->Stencil.WriteMask[0]);
+	 }
+      }
+   }
+
+   /* 16bpp doesn't support masked clears */
+   if (vmesa->viaScreen->bytesPerPixel == 2 &&
+       vmesa->ClearMask & 0xf0000000) {
+      if (flag & VIA_FRONT)
+         mask |= BUFFER_BIT_FRONT_LEFT;
+      if (flag & VIA_BACK)
+         mask |= BUFFER_BIT_BACK_LEFT;
+      flag &= ~(VIA_FRONT | VIA_BACK);
+   }
+    
+   if (flag) {
+      drm_clip_rect_t *boxes, *tmp_boxes = 0;
+      int nr = 0;
+      GLint cx, cy, cw, ch;
+      GLboolean all;
+
+      LOCK_HARDWARE(vmesa);
+	    
+      /* get region after locking: */
+      cx = ctx->DrawBuffer->_Xmin;
+      cy = ctx->DrawBuffer->_Ymin;
+      cw = ctx->DrawBuffer->_Xmax - cx;
+      ch = ctx->DrawBuffer->_Ymax - cy;
+      all = (cw == ctx->DrawBuffer->Width && ch == ctx->DrawBuffer->Height);
+
+      /* flip top to bottom */
+      cy = dPriv->h - cy - ch;
+      cx += vrb->drawX;
+      cy += vrb->drawY;
+        
+      if (!all) {
+	 drm_clip_rect_t *b = vmesa->pClipRects;	 
+	 
+	 boxes = tmp_boxes = 
+	    (drm_clip_rect_t *)malloc(vmesa->numClipRects * 
+				      sizeof(drm_clip_rect_t)); 
+	 if (!boxes) {
+	    UNLOCK_HARDWARE(vmesa);
+	    return;
+	 }
+
+	 for (; i < vmesa->numClipRects; i++) {
+	    GLint x = b[i].x1;
+	    GLint y = b[i].y1;
+	    GLint w = b[i].x2 - x;
+	    GLint h = b[i].y2 - y;
+
+	    if (x < cx) w -= cx - x, x = cx;
+	    if (y < cy) h -= cy - y, y = cy;
+	    if (x + w > cx + cw) w = cx + cw - x;
+	    if (y + h > cy + ch) h = cy + ch - y;
+	    if (w <= 0) continue;
+	    if (h <= 0) continue;
+
+	    boxes[nr].x1 = x;
+	    boxes[nr].y1 = y;
+	    boxes[nr].x2 = x + w;
+	    boxes[nr].y2 = y + h;
+	    nr++;
+	 }
+      }
+      else {
+	 boxes = vmesa->pClipRects;
+	 nr = vmesa->numClipRects;
+      }
+	    
+      if (flag & VIA_FRONT) {
+	 viaFillBuffer(vmesa, &vmesa->front, boxes, nr, vmesa->ClearColor,
+		       vmesa->ClearMask);
+      } 
+		
+      if (flag & VIA_BACK) {
+	 viaFillBuffer(vmesa, &vmesa->back, boxes, nr, vmesa->ClearColor, 
+		       vmesa->ClearMask);
+      }
+
+      if (flag & VIA_DEPTH) {
+	 viaFillBuffer(vmesa, &vmesa->depth, boxes, nr, clear_depth,
+		       clear_depth_mask);
+      }		
+
+      viaFlushDmaLocked(vmesa, VIA_NO_CLIPRECTS);
+      UNLOCK_HARDWARE(vmesa);
+
+      if (tmp_boxes)
+	 free(tmp_boxes);
+   }
+   
+   if (mask)
+      _swrast_Clear(ctx, mask);
+}
+
+
+
+
+static void viaDoSwapBuffers(struct via_context *vmesa,
+			     drm_clip_rect_t *b,
+			     GLuint nbox)
+{    
+   GLuint bytePerPixel = vmesa->viaScreen->bitsPerPixel >> 3;
+   struct via_renderbuffer *front = &vmesa->front;
+   struct via_renderbuffer *back = &vmesa->back;
+   GLuint i;
+        
+   for (i = 0; i < nbox; i++, b++) {        
+      GLint x = b->x1 - back->drawX;
+      GLint y = b->y1 - back->drawY;
+      GLint w = b->x2 - b->x1;
+      GLint h = b->y2 - b->y1;
+	
+      GLuint src = back->offset + y * back->pitch + x * bytePerPixel;
+      GLuint dest = front->offset + y * front->pitch + x * bytePerPixel;
+
+      viaBlit(vmesa, 
+	      bytePerPixel << 3, 
+	      src, back->pitch,
+	      dest, front->pitch,
+	      w, h,
+	      VIA_BLIT_COPY, 0, 0); 
+   }
+
+   viaFlushDmaLocked(vmesa, VIA_NO_CLIPRECTS); /* redundant */
+}
+
+
+static void viaEmitBreadcrumbLocked( struct via_context *vmesa )
+{
+   struct via_renderbuffer *buffer = &vmesa->breadcrumb;
+   GLuint value = vmesa->lastBreadcrumbWrite + 1;
+
+   if (VIA_DEBUG & DEBUG_IOCTL) 
+      fprintf(stderr, "%s %d\n", __FUNCTION__, value);
+
+   assert(!vmesa->dmaLow);
+
+   viaBlit(vmesa,
+	   buffer->bpp, 
+	   buffer->offset, buffer->pitch,
+	   buffer->offset, buffer->pitch, 
+	   1, 1,
+	   VIA_BLIT_FILL, value, 0); 
+
+   viaFlushDmaLocked(vmesa, VIA_NO_CLIPRECTS); /* often redundant */
+   vmesa->lastBreadcrumbWrite = value;
+}
+
+void viaEmitBreadcrumb( struct via_context *vmesa )
+{
+   LOCK_HARDWARE(vmesa);
+   if (vmesa->dmaLow) 
+      viaFlushDmaLocked(vmesa, 0);
+
+   viaEmitBreadcrumbLocked( vmesa );
+   UNLOCK_HARDWARE(vmesa);
+}
+
+static GLboolean viaCheckIdle( struct via_context *vmesa )
+{
+   if ((vmesa->regEngineStatus[0] & 0xFFFEFFFF) == 0x00020000) {
+      return GL_TRUE;
+   }
+   return GL_FALSE;
+}
+
+
+GLboolean viaCheckBreadcrumb( struct via_context *vmesa, GLuint value )
+{
+   GLuint *buf = (GLuint *)vmesa->breadcrumb.map; 
+   vmesa->lastBreadcrumbRead = *buf;
+
+   if (VIA_DEBUG & DEBUG_IOCTL) 
+      fprintf(stderr, "%s %d < %d: %d\n", __FUNCTION__, value, 
+	      vmesa->lastBreadcrumbRead,
+	      !VIA_GEQ_WRAP(value, vmesa->lastBreadcrumbRead));
+
+   return !VIA_GEQ_WRAP(value, vmesa->lastBreadcrumbRead);
+}
+
+static void viaWaitBreadcrumb( struct via_context *vmesa, GLuint value )
+{
+   if (VIA_DEBUG & DEBUG_IOCTL) 
+      fprintf(stderr, "%s %d\n", __FUNCTION__, value);
+
+   assert(!VIA_GEQ_WRAP(value, vmesa->lastBreadcrumbWrite));
+
+   while (!viaCheckBreadcrumb( vmesa, value )) {
+      viaSwapOutWork( vmesa );
+      via_release_pending_textures( vmesa );
+   }
+}
+
+
+void viaWaitIdle( struct via_context *vmesa, GLboolean light )
+{
+   VIA_FLUSH_DMA(vmesa);
+
+   if (VIA_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s lastDma %d lastBreadcrumbWrite %d\n",
+	      __FUNCTION__, vmesa->lastDma, vmesa->lastBreadcrumbWrite);
+
+   /* Need to emit a new breadcrumb?
+    */
+   if (vmesa->lastDma == vmesa->lastBreadcrumbWrite) {
+      LOCK_HARDWARE(vmesa);
+      viaEmitBreadcrumbLocked( vmesa );
+      UNLOCK_HARDWARE(vmesa);
+   }
+
+   /* Need to wait?
+    */
+   if (VIA_GEQ_WRAP(vmesa->lastDma, vmesa->lastBreadcrumbRead)) 
+      viaWaitBreadcrumb( vmesa, vmesa->lastDma );
+
+   if (light) return;
+
+   LOCK_HARDWARE(vmesa);
+   while(!viaCheckIdle(vmesa))
+      ;
+   UNLOCK_HARDWARE(vmesa);
+   via_release_pending_textures(vmesa);
+}
+
+
+void viaWaitIdleLocked( struct via_context *vmesa, GLboolean light )
+{
+   if (vmesa->dmaLow) 
+      viaFlushDmaLocked(vmesa, 0);
+
+   if (VIA_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, "%s lastDma %d lastBreadcrumbWrite %d\n",
+	      __FUNCTION__, vmesa->lastDma, vmesa->lastBreadcrumbWrite);
+
+   /* Need to emit a new breadcrumb?
+    */
+   if (vmesa->lastDma == vmesa->lastBreadcrumbWrite) {
+      viaEmitBreadcrumbLocked( vmesa );
+   }
+
+   /* Need to wait?
+    */
+   if (vmesa->lastDma >= vmesa->lastBreadcrumbRead) 
+      viaWaitBreadcrumb( vmesa, vmesa->lastDma );
+
+   if (light) return;
+
+   while(!viaCheckIdle(vmesa))
+      ;
+
+   via_release_pending_textures(vmesa);
+}
+
+
+
+/* Wait for command stream to be processed *and* the next vblank to
+ * occur.  Equivalent to calling WAIT_IDLE() and then WaitVBlank,
+ * except that WAIT_IDLE() will spin the CPU polling, while this is
+ * IRQ driven.
+ */
+static void viaWaitIdleVBlank( const __DRIdrawablePrivate *dPriv, 
+			       struct via_context *vmesa,
+			       GLuint value )
+{
+   GLboolean missed_target;
+
+   VIA_FLUSH_DMA(vmesa); 
+
+   if (!value)
+      return;
+
+   do {
+      if (value < vmesa->lastBreadcrumbRead ||
+	  vmesa->thrashing)
+	 viaSwapOutWork(vmesa);
+
+      driWaitForVBlank( dPriv, & vmesa->vbl_seq, 
+			vmesa->vblank_flags, & missed_target );
+      if ( missed_target ) {
+	 vmesa->swap_missed_count++;
+	 (*dri_interface->getUST)( &vmesa->swap_missed_ust );
+      }
+   } 
+   while (!viaCheckBreadcrumb(vmesa, value));	 
+
+   vmesa->thrashing = 0;	/* reset flag on swap */
+   vmesa->swap_count++;   
+   via_release_pending_textures( vmesa );
+}
+
+
+
+static void viaDoPageFlipLocked(struct via_context *vmesa, GLuint offset)
+{
+   RING_VARS;
+
+   if (VIA_DEBUG & DEBUG_2D)
+      fprintf(stderr, "%s %x\n", __FUNCTION__, offset);
+
+   if (!vmesa->nDoneFirstFlip) {
+      vmesa->nDoneFirstFlip = GL_TRUE;
+      BEGIN_RING(4);
+      OUT_RING(HALCYON_HEADER2);
+      OUT_RING(0x00fe0000);
+      OUT_RING(0x0000000e);
+      OUT_RING(0x0000000e);
+      ADVANCE_RING();
+   }
+
+   BEGIN_RING(4);
+   OUT_RING( HALCYON_HEADER2 );
+   OUT_RING( 0x00fe0000 );
+   OUT_RING((HC_SubA_HFBBasL << 24) | (offset & 0xFFFFF8) | 0x2);
+   OUT_RING((HC_SubA_HFBDrawFirst << 24) |
+	    ((offset & 0xFF000000) >> 24) | 0x0100);
+   ADVANCE_RING();
+
+   vmesa->pfCurrentOffset = vmesa->sarea->pfCurrentOffset = offset;
+
+   viaFlushDmaLocked(vmesa, VIA_NO_CLIPRECTS); /* often redundant */
+}
+
+void viaResetPageFlippingLocked(struct via_context *vmesa)
+{
+   if (VIA_DEBUG & DEBUG_2D)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   viaDoPageFlipLocked( vmesa, 0 );
+
+   if (vmesa->front.offset != 0) {
+      struct via_renderbuffer buffer_tmp;
+      memcpy(&buffer_tmp, &vmesa->back, sizeof(struct via_renderbuffer));
+      memcpy(&vmesa->back, &vmesa->front, sizeof(struct via_renderbuffer));
+      memcpy(&vmesa->front, &buffer_tmp, sizeof(struct via_renderbuffer));
+   }
+
+   assert(vmesa->front.offset == 0);
+   vmesa->doPageFlip = vmesa->allowPageFlip = 0;
+}
+
+
+/*
+ * Copy the back buffer to the front buffer. 
+ */
+void viaCopyBuffer(const __DRIdrawablePrivate *dPriv)
+{
+   struct via_context *vmesa = 
+      (struct via_context *)dPriv->driContextPriv->driverPrivate;
+
+   if (VIA_DEBUG & DEBUG_IOCTL)
+      fprintf(stderr, 
+	      "%s: lastSwap[1] %d lastSwap[0] %d lastWrite %d lastRead %d\n",
+	      __FUNCTION__,
+	      vmesa->lastSwap[1], 
+	      vmesa->lastSwap[0], 
+	      vmesa->lastBreadcrumbWrite,
+	      vmesa->lastBreadcrumbRead);
+
+   VIA_FLUSH_DMA(vmesa);
+
+   if (vmesa->vblank_flags == VBLANK_FLAG_SYNC &&
+       vmesa->lastBreadcrumbWrite > 1)
+      viaWaitIdleVBlank(dPriv, vmesa, vmesa->lastBreadcrumbWrite-1);
+   else
+      viaWaitIdleVBlank(dPriv, vmesa, vmesa->lastSwap[1]);
+
+   LOCK_HARDWARE(vmesa);
+
+   /* Catch and cleanup situation where we were pageflipping but have
+    * stopped.
+    */
+   if (dPriv->numClipRects && vmesa->sarea->pfCurrentOffset != 0) {
+      viaResetPageFlippingLocked(vmesa);
+      UNLOCK_HARDWARE(vmesa);
+      return;
+   }
+
+   viaDoSwapBuffers(vmesa, dPriv->pClipRects, dPriv->numClipRects);
+   vmesa->lastSwap[1] = vmesa->lastSwap[0];
+   vmesa->lastSwap[0] = vmesa->lastBreadcrumbWrite;
+   viaEmitBreadcrumbLocked(vmesa);
+   UNLOCK_HARDWARE(vmesa);
+
+   (*dri_interface->getUST)( &vmesa->swap_ust );
+}
+
+
+void viaPageFlip(const __DRIdrawablePrivate *dPriv)
+{
+    struct via_context *vmesa = 
+       (struct via_context *)dPriv->driContextPriv->driverPrivate;
+    struct via_renderbuffer buffer_tmp;
+
+    VIA_FLUSH_DMA(vmesa);
+   if (vmesa->vblank_flags == VBLANK_FLAG_SYNC &&
+       vmesa->lastBreadcrumbWrite > 1)
+      viaWaitIdleVBlank(dPriv, vmesa, vmesa->lastBreadcrumbWrite - 1);
+   else
+      viaWaitIdleVBlank(dPriv, vmesa, vmesa->lastSwap[0]);
+
+    LOCK_HARDWARE(vmesa);
+    viaDoPageFlipLocked(vmesa, vmesa->back.offset);
+    vmesa->lastSwap[1] = vmesa->lastSwap[0];
+    vmesa->lastSwap[0] = vmesa->lastBreadcrumbWrite;
+    viaEmitBreadcrumbLocked(vmesa);
+    UNLOCK_HARDWARE(vmesa);
+
+    (*dri_interface->getUST)( &vmesa->swap_ust );
+
+
+    /* KW: FIXME: When buffers are freed, could free frontbuffer by
+     * accident:
+     */
+    memcpy(&buffer_tmp, &vmesa->back, sizeof(struct via_renderbuffer));
+    memcpy(&vmesa->back, &vmesa->front, sizeof(struct via_renderbuffer));
+    memcpy(&vmesa->front, &buffer_tmp, sizeof(struct via_renderbuffer));
+}
+
+
+
+
+#define VIA_CMDBUF_MAX_LAG 50000
+
+static int fire_buffer(struct via_context *vmesa)
+{
+   drm_via_cmdbuffer_t bufI;
+   int ret;
+
+   bufI.buf = (char *)vmesa->dma;
+   bufI.size = vmesa->dmaLow;
+
+   if (vmesa->useAgp) {
+      drm_via_cmdbuf_size_t bSiz;
+
+      /* Do the CMDBUF_SIZE ioctl:
+       */
+      bSiz.func = VIA_CMDBUF_LAG;
+      bSiz.wait = 1;
+      bSiz.size = VIA_CMDBUF_MAX_LAG;
+      do {
+	 ret = drmCommandWriteRead(vmesa->driFd, DRM_VIA_CMDBUF_SIZE, 
+				   &bSiz, sizeof(bSiz));
+      } while (ret == -EAGAIN);
+      if (ret) {
+	 UNLOCK_HARDWARE(vmesa);
+	 fprintf(stderr, "%s: DRM_VIA_CMDBUF_SIZE returned %d\n",
+		 __FUNCTION__, ret);
+	 abort();
+	 return ret;
+      }
+
+      /* Actually fire the buffer:
+       */
+      do {
+	 ret = drmCommandWrite(vmesa->driFd, DRM_VIA_CMDBUFFER, 
+			       &bufI, sizeof(bufI));
+      } while (ret == -EAGAIN);
+      if (ret) {
+	 UNLOCK_HARDWARE(vmesa);
+	 fprintf(stderr, "%s: DRM_VIA_CMDBUFFER returned %d\n",
+		 __FUNCTION__, ret);
+	 abort();
+	 /* If this fails, the original code fell back to the PCI path. 
+	  */
+      }
+      else 
+	 return 0;
+
+      /* Fall through to PCI handling?!?
+       */
+      viaWaitIdleLocked(vmesa, GL_FALSE);
+   }
+	    
+   ret = drmCommandWrite(vmesa->driFd, DRM_VIA_PCICMD, &bufI, sizeof(bufI));
+   if (ret) {
+      UNLOCK_HARDWARE(vmesa);
+      dump_dma(vmesa);
+      fprintf(stderr, "%s: DRM_VIA_PCICMD returned %d\n", __FUNCTION__, ret); 
+      abort();
+   }
+
+   return ret;
+}
+
+
+/* Inserts the surface addresss and active cliprects one at a time
+ * into the head of the DMA buffer being flushed.  Fires the buffer
+ * for each cliprect.
+ */
+static void via_emit_cliprect(struct via_context *vmesa,
+			      drm_clip_rect_t *b) 
+{
+   struct via_renderbuffer *buffer = vmesa->drawBuffer;
+   GLuint *vb = (GLuint *)(vmesa->dma + vmesa->dmaCliprectAddr);
+
+   GLuint format = (vmesa->viaScreen->bitsPerPixel == 0x20 
+		    ? HC_HDBFM_ARGB8888 
+		    : HC_HDBFM_RGB565);
+
+   GLuint pitch = buffer->pitch;
+   GLuint offset = buffer->offset;
+
+   if (0)
+      fprintf(stderr, "emit cliprect for box %d,%d %d,%d\n", 
+	      b->x1, b->y1, b->x2, b->y2);
+
+   vb[0] = HC_HEADER2;
+   vb[1] = (HC_ParaType_NotTex << 16);
+
+   if (vmesa->driDrawable->w == 0 || vmesa->driDrawable->h == 0) {
+      vb[2] = (HC_SubA_HClipTB << 24) | 0x0;
+      vb[3] = (HC_SubA_HClipLR << 24) | 0x0;
+   }
+   else {
+      vb[2] = (HC_SubA_HClipTB << 24) | (b->y1 << 12) | b->y2;
+      vb[3] = (HC_SubA_HClipLR << 24) | (b->x1 << 12) | b->x2;
+   }
+	    
+   vb[4] = (HC_SubA_HDBBasL << 24) | (offset & 0xFFFFFF);
+   vb[5] = (HC_SubA_HDBBasH << 24) | ((offset & 0xFF000000) >> 24); 
+
+   vb[6] = (HC_SubA_HSPXYOS << 24);
+   vb[7] = (HC_SubA_HDBFM << 24) | HC_HDBLoc_Local | format | pitch;
+}
+
+
+
+static int intersect_rect(drm_clip_rect_t *out,
+                          drm_clip_rect_t *a,
+                          drm_clip_rect_t *b)
+{
+    *out = *a;
+    
+    if (0)
+       fprintf(stderr, "intersect %d,%d %d,%d and %d,%d %d,%d\n", 
+	       a->x1, a->y1, a->x2, a->y2,
+	       b->x1, b->y1, b->x2, b->y2);
+
+    if (b->x1 > out->x1) out->x1 = b->x1;
+    if (b->x2 < out->x2) out->x2 = b->x2;
+    if (out->x1 >= out->x2) return 0;
+
+    if (b->y1 > out->y1) out->y1 = b->y1;
+    if (b->y2 < out->y2) out->y2 = b->y2;
+    if (out->y1 >= out->y2) return 0;
+
+    return 1;
+}
+
+void viaFlushDmaLocked(struct via_context *vmesa, GLuint flags)
+{
+   int i;
+   RING_VARS;
+
+   if (VIA_DEBUG & (DEBUG_IOCTL|DEBUG_DMA))
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (*(GLuint *)vmesa->driHwLock != (DRM_LOCK_HELD|vmesa->hHWContext) &&
+       *(GLuint *)vmesa->driHwLock != 
+       (DRM_LOCK_HELD|DRM_LOCK_CONT|vmesa->hHWContext)) {
+      fprintf(stderr, "%s called without lock held\n", __FUNCTION__);
+      abort();
+   }
+
+   if (vmesa->dmaLow == 0) {
+      return;
+   }
+
+   assert(vmesa->dmaLastPrim == 0);
+
+   /* viaFinishPrimitive can add up to 8 bytes beyond VIA_DMA_HIGHWATER:
+    */
+   if (vmesa->dmaLow > VIA_DMA_HIGHWATER + 8) {
+      fprintf(stderr, "buffer overflow in Flush Prims = %d\n",vmesa->dmaLow);
+      abort();
+   }
+
+   switch (vmesa->dmaLow & 0x1F) {	
+   case 8:
+      BEGIN_RING_NOCHECK( 6 );
+      OUT_RING( HC_HEADER2 );
+      OUT_RING( (HC_ParaType_NotTex << 16) );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      ADVANCE_RING();
+      break;
+   case 16:
+      BEGIN_RING_NOCHECK( 4 );
+      OUT_RING( HC_HEADER2 );
+      OUT_RING( (HC_ParaType_NotTex << 16) );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      ADVANCE_RING();
+      break;    
+   case 24:
+      BEGIN_RING_NOCHECK( 10 );
+      OUT_RING( HC_HEADER2 );
+      OUT_RING( (HC_ParaType_NotTex << 16) );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );
+      OUT_RING( HC_DUMMY );	
+      ADVANCE_RING();
+      break;    
+   case 0:
+      break;
+   default:
+      if (VIA_DEBUG & DEBUG_IOCTL)
+	 fprintf(stderr, "%s: unaligned value for vmesa->dmaLow: %x\n",
+		 __FUNCTION__, vmesa->dmaLow);
+   }
+
+   vmesa->lastDma = vmesa->lastBreadcrumbWrite;
+
+   if (VIA_DEBUG & DEBUG_DMA)
+      dump_dma( vmesa );
+
+   if (flags & VIA_NO_CLIPRECTS) {
+      if (0) fprintf(stderr, "%s VIA_NO_CLIPRECTS\n", __FUNCTION__);
+      assert(vmesa->dmaCliprectAddr == ~0);
+      fire_buffer( vmesa );
+   }
+   else if (vmesa->dmaCliprectAddr == ~0) {
+      /* Contains only state.  Could just dump the packet?
+       */
+      if (0) fprintf(stderr, "%s: no dmaCliprectAddr\n", __FUNCTION__);
+      if (0) fire_buffer( vmesa );
+   }
+   else if (vmesa->numClipRects) {
+      drm_clip_rect_t *pbox = vmesa->pClipRects;
+      __DRIdrawablePrivate *dPriv = vmesa->driDrawable;
+      struct via_renderbuffer *const vrb = 
+	(struct via_renderbuffer *) dPriv->driverPrivate;
+
+      for (i = 0; i < vmesa->numClipRects; i++) {
+	 drm_clip_rect_t b;
+
+	 b.x1 = pbox[i].x1;
+	 b.x2 = pbox[i].x2;
+	 b.y1 = pbox[i].y1;
+	 b.y2 = pbox[i].y2;
+
+	 if (vmesa->scissor &&
+	     !intersect_rect(&b, &b, &vmesa->scissorRect)) 
+	    continue;
+
+	 via_emit_cliprect(vmesa, &b);
+
+	 if (fire_buffer(vmesa) != 0) {
+	    dump_dma( vmesa );
+	    goto done;
+	 }
+      }
+   } else {
+      if (0) fprintf(stderr, "%s: no cliprects\n", __FUNCTION__);
+      UNLOCK_HARDWARE(vmesa);
+      sched_yield();
+      LOCK_HARDWARE(vmesa);
+   }
+
+ done:
+   /* Reset vmesa vars:
+    */
+   vmesa->dmaLow = 0;
+   vmesa->dmaCliprectAddr = ~0;
+   vmesa->newEmitState = ~0;
+}
+
+void viaWrapPrimitive( struct via_context *vmesa )
+{
+   GLenum renderPrimitive = vmesa->renderPrimitive;
+   GLenum hwPrimitive = vmesa->hwPrimitive;
+
+   if (VIA_DEBUG & DEBUG_PRIMS) fprintf(stderr, "%s\n", __FUNCTION__);
+   
+   if (vmesa->dmaLastPrim)
+      viaFinishPrimitive( vmesa );
+   
+   viaFlushDma(vmesa);
+
+   if (renderPrimitive != GL_POLYGON + 1)
+      viaRasterPrimitive( vmesa->glCtx,
+			  renderPrimitive,
+			  hwPrimitive );
+
+}
+
+void viaFlushDma(struct via_context *vmesa)
+{
+   if (vmesa->dmaLow) {
+      assert(!vmesa->dmaLastPrim);
+
+      LOCK_HARDWARE(vmesa); 
+      viaFlushDmaLocked(vmesa, 0);
+      UNLOCK_HARDWARE(vmesa);
+   }
+}
+
+static void viaFlush(GLcontext *ctx)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    VIA_FLUSH_DMA(vmesa);
+}
+
+static void viaFinish(GLcontext *ctx)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    VIA_FLUSH_DMA(vmesa);
+    viaWaitIdle(vmesa, GL_FALSE);
+}
+
+static void viaClearStencil(GLcontext *ctx,  int s)
+{
+    return;
+}
+
+void viaInitIoctlFuncs(GLcontext *ctx)
+{
+    ctx->Driver.Flush = viaFlush;
+    ctx->Driver.Clear = viaClear;
+    ctx->Driver.Finish = viaFinish;
+    ctx->Driver.ClearStencil = viaClearStencil;
+}
+
+
+
diff --git a/src/via_ioctl.h b/src/via_ioctl.h
new file mode 100644
index 0000000..a81b427
--- /dev/null
+++ b/src/via_ioctl.h
@@ -0,0 +1,133 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIAIOCTL_H
+#define _VIAIOCTL_H
+
+#include "via_context.h"
+
+
+void viaFinishPrimitive(struct via_context *vmesa);
+void viaFlushDma(struct via_context *vmesa);
+void viaFlushDmaLocked(struct via_context *vmesa, GLuint flags);
+
+void viaInitIoctlFuncs(GLcontext *ctx);
+void viaCopyBuffer(const __DRIdrawablePrivate *dpriv);
+void viaPageFlip(const __DRIdrawablePrivate *dpriv);
+void viaCheckDma(struct via_context *vmesa, GLuint bytes);
+void viaResetPageFlippingLocked(struct via_context *vmesa);
+void viaWaitIdle(struct via_context *vmesa, GLboolean light);
+void viaWaitIdleLocked(struct via_context *vmesa, GLboolean light);
+
+GLboolean viaCheckBreadcrumb( struct via_context *vmesa, GLuint value );
+void viaEmitBreadcrumb( struct via_context *vmesa );
+
+
+#define VIA_FINISH_PRIM(vmesa) do {		\
+   if (vmesa->dmaLastPrim)			\
+      viaFinishPrimitive( vmesa );		\
+} while (0)
+
+#define VIA_FLUSH_DMA(vmesa) do {		\
+   VIA_FINISH_PRIM(vmesa);			\
+   if (vmesa->dmaLow) 		\
+      viaFlushDma(vmesa);			\
+} while (0)
+    
+
+void viaWrapPrimitive( struct via_context *vmesa );
+
+static __inline__ GLuint *viaAllocDma(struct via_context *vmesa, int bytes)
+{
+   if (vmesa->dmaLow + bytes > VIA_DMA_HIGHWATER) {
+      viaFlushDma(vmesa);
+   }
+
+   {
+      GLuint *start = (GLuint *)(vmesa->dma + vmesa->dmaLow);
+      vmesa->dmaLow += bytes;
+      return start;
+   }
+}
+
+
+static GLuint __inline__ *viaExtendPrimitive(struct via_context *vmesa, int bytes)
+{
+   if (vmesa->dmaLow + bytes > VIA_DMA_HIGHWATER) {
+      viaWrapPrimitive(vmesa);
+   }
+
+   {
+      GLuint *start = (GLuint *)(vmesa->dma + vmesa->dmaLow);
+      vmesa->dmaLow += bytes;
+      return start;
+   }
+}
+
+
+
+
+#define RING_VARS GLuint *_vb = 0, _nr, _x;
+
+#define BEGIN_RING(n) do {				\
+   if (_vb != 0) abort();				\
+   _vb = viaAllocDma(vmesa, (n) * sizeof(GLuint));	\
+   _nr = (n);						\
+   _x = 0;						\
+} while (0)
+
+#define BEGIN_RING_NOCHECK(n) do {			\
+   if (_vb != 0) abort();				\
+   _vb = (GLuint *)(vmesa->dma + vmesa->dmaLow);	\
+   vmesa->dmaLow += (n) * sizeof(GLuint);		\
+   _nr = (n);						\
+   _x = 0;						\
+} while (0)
+
+#define OUT_RING(n) _vb[_x++] = (n)
+
+#define ADVANCE_RING() do {			\
+   if (_x != _nr) abort(); 			\
+   _vb = 0;						\
+} while (0)
+
+#define ADVANCE_RING_VARIABLE() do {			\
+   if (_x > _nr) abort();				\
+   vmesa->dmaLow -= (_nr - _x) * sizeof(GLuint);	\
+   _vb = 0;						\
+} while (0)
+
+
+#define QWORD_PAD_RING() do {			\
+   if (vmesa->dmaLow & 0x4) {			\
+      BEGIN_RING(1);				\
+      OUT_RING(HC_DUMMY);			\
+      ADVANCE_RING();				\
+   }						\
+} while (0)
+
+#define VIA_GEQ_WRAP(left, right) \
+    (((left) - (right)) < ( 1 << 23))
+      
+#endif
diff --git a/src/via_memcpy.c b/src/via_memcpy.c
new file mode 100644
index 0000000..d7b05bc
--- /dev/null
+++ b/src/via_memcpy.c
@@ -0,0 +1,138 @@
+/*
+ * Copyright (C) 2004 Thomas Hellstrom, 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
+ * THE CODE SUPPLIER(S) 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.
+ */
+
+/* Thomas' orginal gutted for mesa by Keith Whitwell
+ */
+
+#include "via_tex.h"
+
+#if defined( USE_SSE_ASM )
+
+#define SSE_PREFETCH "  prefetchnta "
+#define FENCE __asm__ __volatile__ ("sfence":::"memory");
+
+
+#define PREFETCH1(arch_prefetch,from)			\
+    __asm__ __volatile__ (				\
+			  "1:  " arch_prefetch "(%0)\n"	\
+			  arch_prefetch "32(%0)\n"	\
+			  arch_prefetch "64(%0)\n"	\
+			  arch_prefetch "96(%0)\n"	\
+			  arch_prefetch "128(%0)\n"	\
+			  arch_prefetch "160(%0)\n"	\
+			  arch_prefetch "192(%0)\n"	\
+			  arch_prefetch "256(%0)\n"	\
+			  arch_prefetch "288(%0)\n"	\
+			  "2:\n"			\
+			  : : "r" (from) );
+
+
+
+#define small_memcpy(to,from,n)						\
+    {									\
+	__asm__ __volatile__(						\
+			     "movl %2,%%ecx\n\t"			\
+                             "sarl $2,%%ecx\n\t"			\
+			     "rep ; movsl\n\t"				\
+			     "testb $2,%b2\n\t"				\
+			     "je 1f\n\t"				\
+			     "movsw\n"					\
+			     "1:\ttestb $1,%b2\n\t"			\
+			     "je 2f\n\t"				\
+			     "movsb\n"					\
+			     "2:"					\
+			     :"=&D" (to), "=&S" (from)			\
+			     :"q" (n),"0" ((long) to),"1" ((long) from) \
+			     : "%ecx","memory");			\
+    }
+
+
+#define SSE_CPY(prefetch,from,to,dummy,lcnt)				\
+    if ((unsigned long) from & 15)			 {		\
+	__asm__ __volatile__ (						\
+			      "1:\n"					\
+                              prefetch "320(%1)\n"			\
+			      "  movups (%1), %%xmm0\n"			\
+			      "  movups 16(%1), %%xmm1\n"		\
+			      "  movntps %%xmm0, (%0)\n"		\
+			      "  movntps %%xmm1, 16(%0)\n"		\
+                              prefetch "352(%1)\n"			\
+			      "  movups 32(%1), %%xmm2\n"		\
+			      "  movups 48(%1), %%xmm3\n"		\
+			      "  movntps %%xmm2, 32(%0)\n"		\
+			      "  movntps %%xmm3, 48(%0)\n"		\
+			      "  addl $64,%0\n"				\
+			      "  addl $64,%1\n"				\
+			      "  decl %2\n"				\
+			      "  jne 1b\n"				\
+			      :"=&D"(to), "=&S"(from), "=&r"(dummy)	\
+			      :"0" (to), "1" (from), "2" (lcnt): "memory"); \
+    } else {								\
+	__asm__ __volatile__ (						\
+			      "2:\n"					\
+			      prefetch "320(%1)\n"			\
+			      "  movaps (%1), %%xmm0\n"			\
+			      "  movaps 16(%1), %%xmm1\n"		\
+			      "  movntps %%xmm0, (%0)\n"		\
+			      "  movntps %%xmm1, 16(%0)\n"		\
+                              prefetch "352(%1)\n"			\
+			      "  movaps 32(%1), %%xmm2\n"		\
+			      "  movaps 48(%1), %%xmm3\n"		\
+			      "  movntps %%xmm2, 32(%0)\n"		\
+			      "  movntps %%xmm3, 48(%0)\n"		\
+			      "  addl $64,%0\n"				\
+			      "  addl $64,%1\n"				\
+			      "  decl %2\n"				\
+			      "  jne 2b\n"				\
+			      :"=&D"(to), "=&S"(from), "=&r"(dummy)	\
+			      :"0" (to), "1" (from), "2" (lcnt): "memory"); \
+    }
+
+
+
+/*
+ */
+void via_sse_memcpy(void *to,
+		    const void *from,
+		    size_t sz)
+
+{
+   int dummy;
+   int lcnt = sz >> 6;
+   int rest = sz & 63;
+
+   PREFETCH1(SSE_PREFETCH,from);
+
+   if (lcnt > 5) {
+      lcnt -= 5;
+      SSE_CPY(SSE_PREFETCH,from,to,dummy,lcnt);
+      lcnt = 5;
+   }
+   if (lcnt) {
+      SSE_CPY("#",from,to,dummy,lcnt);
+   }
+   if (rest) small_memcpy(to, from, rest);
+   FENCE;
+}
+
+#endif /* defined( USE_SSE_ASM ) */
diff --git a/src/via_render.c b/src/via_render.c
new file mode 100644
index 0000000..d7e373d
--- /dev/null
+++ b/src/via_render.c
@@ -0,0 +1,131 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+
+/*
+ * Render unclipped vertex buffers by emitting vertices directly to
+ * dma buffers.  Use strip/fan hardware acceleration where possible.
+ *
+ */
+#include "glheader.h"
+#include "context.h"
+#include "macros.h"
+#include "mtypes.h"
+
+#include "tnl/t_context.h"
+
+#include "via_context.h"
+#include "via_tris.h"
+#include "via_state.h"
+#include "via_ioctl.h"
+
+/*
+ * Render unclipped vertex buffers by emitting vertices directly to
+ * dma buffers.  Use strip/fan hardware primitives where possible.
+ * Try to simulate missing primitives with indexed vertices.
+ */
+#define HAVE_POINTS      1
+#define HAVE_LINES       1
+#define HAVE_LINE_STRIPS 1
+#define HAVE_LINE_LOOP   1
+#define HAVE_TRIANGLES   1
+#define HAVE_TRI_STRIPS  1
+#define HAVE_TRI_STRIP_1 0  
+#define HAVE_TRI_FANS    1
+#define HAVE_POLYGONS    1
+#define HAVE_QUADS       0
+#define HAVE_QUAD_STRIPS 0
+
+#define HAVE_ELTS        0
+
+#define LOCAL_VARS struct via_context *vmesa = VIA_CONTEXT(ctx)
+#define INIT(prim) do {					\
+   viaRasterPrimitive(ctx, prim, prim);	\
+} while (0)
+#define GET_CURRENT_VB_MAX_VERTS() \
+    ((VIA_DMA_BUF_SZ - (512 + (int)vmesa->dmaLow)) / (vmesa->vertexSize * 4))
+#define GET_SUBSEQUENT_VB_MAX_VERTS() \
+    (VIA_DMA_BUF_SZ - 512) / (vmesa->vertexSize * 4)
+
+#define ALLOC_VERTS( nr ) \
+    viaExtendPrimitive( vmesa, (nr) * vmesa->vertexSize * 4)
+
+#define EMIT_VERTS(ctx, j, nr, buf) \
+    _tnl_emit_vertices_to_buffer(ctx, j, (j)+(nr), buf )  
+    
+#define FLUSH() VIA_FINISH_PRIM( vmesa )
+
+#define TAG(x) via_fast##x
+#include "tnl_dd/t_dd_dmatmp.h"
+#undef TAG
+#undef LOCAL_VARS
+#undef INIT
+
+/**********************************************************************/
+/*                          Fast Render pipeline stage                */
+/**********************************************************************/
+static GLboolean via_run_fastrender(GLcontext *ctx,
+                                    struct tnl_pipeline_stage *stage)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *VB = &tnl->vb;
+    GLuint i;
+    
+
+    tnl->Driver.Render.Start(ctx);
+    
+    if (VB->ClipOrMask || 
+	vmesa->renderIndex != 0 || 
+	!via_fastvalidate_render( ctx, VB )) {
+	tnl->Driver.Render.Finish(ctx);
+        return GL_TRUE;
+    }
+
+    tnl->clipspace.new_inputs |= VERT_BIT_POS;
+
+    for (i = 0; i < VB->PrimitiveCount; ++i) {
+        GLuint mode = _tnl_translate_prim(&VB->Primitive[i]);
+        GLuint start = VB->Primitive[i].start;
+        GLuint length = VB->Primitive[i].count;
+        if (length)
+            via_fastrender_tab_verts[mode & PRIM_MODE_MASK](ctx, start, start+length, mode);
+    }
+
+    tnl->Driver.Render.Finish(ctx);
+
+    return GL_FALSE;            /* finished the pipe */
+}
+
+const struct tnl_pipeline_stage _via_fastrender_stage =
+{
+    "via fast render",
+    NULL,
+    NULL,
+    NULL,
+    NULL,
+    via_run_fastrender           /* run */
+};
+
+
diff --git a/src/via_screen.c b/src/via_screen.c
new file mode 100644
index 0000000..90f76be
--- /dev/null
+++ b/src/via_screen.c
@@ -0,0 +1,499 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include <stdio.h>
+
+#include "utils.h"
+#include "dri_util.h"
+#include "glheader.h"
+#include "context.h"
+#include "framebuffer.h"
+#include "renderbuffer.h"
+#include "matrix.h"
+#include "simple_list.h"
+#include "vblank.h"
+
+#include "via_state.h"
+#include "via_tex.h"
+#include "via_span.h"
+#include "via_tris.h"
+#include "via_ioctl.h"
+#include "via_screen.h"
+#include "via_fb.h"
+#include "via_dri.h"
+
+#include "GL/internal/dri_interface.h"
+#include "drirenderbuffer.h"
+
+#include "xmlpool.h"
+
+const char __driConfigOptions[] =
+DRI_CONF_BEGIN
+    DRI_CONF_SECTION_PERFORMANCE
+        DRI_CONF_VBLANK_MODE(DRI_CONF_VBLANK_DEF_INTERVAL_0)
+    DRI_CONF_SECTION_END
+    DRI_CONF_SECTION_QUALITY
+        DRI_CONF_EXCESS_MIPMAP(false)
+    DRI_CONF_SECTION_END
+    DRI_CONF_SECTION_DEBUG
+        DRI_CONF_NO_RAST(false)
+    DRI_CONF_SECTION_END
+DRI_CONF_END;
+static const GLuint __driNConfigOptions = 3;
+
+extern const struct dri_extension card_extensions[];
+
+static int getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo );
+
+static drmBufMapPtr via_create_empty_buffers(void)
+{
+    drmBufMapPtr retval;
+
+    retval = (drmBufMapPtr)MALLOC(sizeof(drmBufMap));
+    if (retval == NULL) return NULL;
+    memset(retval, 0, sizeof(drmBufMap));
+
+    retval->list = (drmBufPtr)MALLOC(sizeof(drmBuf) * VIA_DMA_BUF_NR);
+    if (retval->list == NULL) {
+       FREE(retval);
+       return NULL;
+    }
+    memset(retval->list, 0, sizeof(drmBuf) * VIA_DMA_BUF_NR);
+    return retval;
+}
+
+static void via_free_empty_buffers( drmBufMapPtr bufs )
+{
+   if (bufs && bufs->list)
+      FREE(bufs->list);
+
+   if (bufs)
+      FREE(bufs);
+}
+
+
+static GLboolean
+viaInitDriver(__DRIscreenPrivate *sPriv)
+{
+    viaScreenPrivate *viaScreen;
+    VIADRIPtr gDRIPriv = (VIADRIPtr)sPriv->pDevPriv;
+    PFNGLXSCRENABLEEXTENSIONPROC glx_enable_extension =
+      (PFNGLXSCRENABLEEXTENSIONPROC) (*dri_interface->getProcAddress("glxEnableExtension"));
+    void * const psc = sPriv->psc->screenConfigs;
+
+    if (sPriv->devPrivSize != sizeof(VIADRIRec)) {
+      fprintf(stderr,"\nERROR!  sizeof(VIADRIRec) does not match passed size from device driver\n");
+      return GL_FALSE;
+    }
+
+    /* Allocate the private area */
+    viaScreen = (viaScreenPrivate *) CALLOC(sizeof(viaScreenPrivate));
+    if (!viaScreen) {
+        __driUtilMessage("viaInitDriver: alloc viaScreenPrivate struct failed");
+        return GL_FALSE;
+    }
+
+    /* parse information in __driConfigOptions */
+    driParseOptionInfo (&viaScreen->optionCache,
+			__driConfigOptions, __driNConfigOptions);
+
+
+    viaScreen->driScrnPriv = sPriv;
+    sPriv->private = (void *)viaScreen;
+
+    viaScreen->deviceID = gDRIPriv->deviceID;
+    viaScreen->width = gDRIPriv->width;
+    viaScreen->height = gDRIPriv->height;
+    viaScreen->mem = gDRIPriv->mem;
+    viaScreen->bitsPerPixel = gDRIPriv->bytesPerPixel * 8;
+    viaScreen->bytesPerPixel = gDRIPriv->bytesPerPixel;
+    viaScreen->fbOffset = 0;
+    viaScreen->fbSize = gDRIPriv->fbSize;
+    viaScreen->irqEnabled = gDRIPriv->irqEnabled;
+
+    if (VIA_DEBUG & DEBUG_DRI) {
+	fprintf(stderr, "deviceID = %08x\n", viaScreen->deviceID);
+	fprintf(stderr, "width = %08x\n", viaScreen->width);	
+	fprintf(stderr, "height = %08x\n", viaScreen->height);	
+	fprintf(stderr, "cpp = %08x\n", viaScreen->cpp);	
+	fprintf(stderr, "fbOffset = %08x\n", viaScreen->fbOffset);	
+    }
+
+    viaScreen->bufs = via_create_empty_buffers();
+    if (viaScreen->bufs == NULL) {
+        __driUtilMessage("viaInitDriver: via_create_empty_buffers() failed");
+        FREE(viaScreen);
+        return GL_FALSE;
+    }
+
+    if (drmMap(sPriv->fd,
+               gDRIPriv->regs.handle,
+               gDRIPriv->regs.size,
+               &viaScreen->reg) != 0) {
+        FREE(viaScreen);
+        sPriv->private = NULL;
+        __driUtilMessage("viaInitDriver: drmMap regs failed");
+        return GL_FALSE;
+    }
+
+    if (gDRIPriv->agp.size) {
+        if (drmMap(sPriv->fd,
+                   gDRIPriv->agp.handle,
+                   gDRIPriv->agp.size,
+	           (drmAddress *)&viaScreen->agpLinearStart) != 0) {
+	    drmUnmap(viaScreen->reg, gDRIPriv->regs.size);
+	    FREE(viaScreen);
+	    sPriv->private = NULL;
+	    __driUtilMessage("viaInitDriver: drmMap agp failed");
+	    return GL_FALSE;
+	}
+
+	viaScreen->agpBase = drmAgpBase(sPriv->fd);
+    } else
+	viaScreen->agpLinearStart = 0;
+
+    viaScreen->sareaPrivOffset = gDRIPriv->sarea_priv_offset;
+
+    if ( glx_enable_extension != NULL ) {
+       if ( viaScreen->irqEnabled ) {
+	  (*glx_enable_extension)( psc, "GLX_SGI_swap_control" );
+	  (*glx_enable_extension)( psc, "GLX_SGI_video_sync" );
+	  (*glx_enable_extension)( psc, "GLX_MESA_swap_control" );
+       }
+
+       (*glx_enable_extension)( psc, "GLX_SGI_make_current_read" );
+       (*glx_enable_extension)( psc, "GLX_MESA_swap_frame_usage" );
+    }
+
+    return GL_TRUE;
+}
+
+static void
+viaDestroyScreen(__DRIscreenPrivate *sPriv)
+{
+    viaScreenPrivate *viaScreen = (viaScreenPrivate *)sPriv->private;
+    VIADRIPtr gDRIPriv = (VIADRIPtr)sPriv->pDevPriv;
+
+    drmUnmap(viaScreen->reg, gDRIPriv->regs.size);
+    if (gDRIPriv->agp.size)
+        drmUnmap(viaScreen->agpLinearStart, gDRIPriv->agp.size);
+
+    via_free_empty_buffers(viaScreen->bufs);
+
+    driDestroyOptionInfo(&viaScreen->optionCache);
+
+    FREE(viaScreen);
+    sPriv->private = NULL;
+}
+
+
+static GLboolean
+viaCreateBuffer(__DRIscreenPrivate *driScrnPriv,
+                __DRIdrawablePrivate *driDrawPriv,
+                const __GLcontextModes *mesaVis,
+                GLboolean isPixmap)
+{
+    viaScreenPrivate *screen = (viaScreenPrivate *) driScrnPriv->private;
+
+    GLboolean swStencil = (mesaVis->stencilBits > 0 && 
+			   mesaVis->depthBits != 24);
+    GLboolean swAccum = mesaVis->accumRedBits > 0;
+
+    if (isPixmap) {
+       /* KW: This needs work, disabled for now:
+	*/
+#if 0
+	driDrawPriv->driverPrivate = (void *)
+            _mesa_create_framebuffer(mesaVis,
+                                     GL_FALSE,	/* software depth buffer? */
+                                     swStencil,
+                                     mesaVis->accumRedBits > 0,
+                                     GL_FALSE 	/* s/w alpha planes */);
+
+        return (driDrawPriv->driverPrivate != NULL);
+#endif
+	return GL_FALSE;
+    }
+    else {
+      struct gl_framebuffer *fb = _mesa_create_framebuffer(mesaVis);
+
+      /* The front color, back color and depth renderbuffers are
+       * set up later in calculate_buffer_parameters().
+       * Only create/connect software-based buffers here.
+       */
+
+#if 000
+      /* This code _should_ be put to use.  We have to move the
+       * viaRenderbuffer members out of the via_context structure.
+       * Those members should just be the renderbuffers hanging off the
+       * gl_framebuffer object.
+       */
+      /* XXX check/fix the offset/pitch parameters! */
+      {
+         driRenderbuffer *frontRb
+            = driNewRenderbuffer(GL_RGBA, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(frontRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &frontRb->Base);
+      }
+
+      if (mesaVis->doubleBufferMode) {
+         driRenderbuffer *backRb
+            = driNewRenderbuffer(GL_RGBA, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(backRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &backRb->Base);
+      }
+
+      if (mesaVis->depthBits == 16) {
+         driRenderbuffer *depthRb
+            = driNewRenderbuffer(GL_DEPTH_COMPONENT16, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(depthRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
+      }
+      else if (mesaVis->depthBits == 24) {
+         driRenderbuffer *depthRb
+            = driNewRenderbuffer(GL_DEPTH_COMPONENT24, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(depthRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
+      }
+      else if (mesaVis->depthBits == 32) {
+         driRenderbuffer *depthRb
+            = driNewRenderbuffer(GL_DEPTH_COMPONENT32, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(depthRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_DEPTH, &depthRb->Base);
+      }
+
+      if (mesaVis->stencilBits > 0 && !swStencil) {
+         driRenderbuffer *stencilRb
+            = driNewRenderbuffer(GL_STENCIL_INDEX8_EXT, NULL,
+                                 screen->bytesPerPixel,
+                                 0, screen->width, driDrawPriv);
+         viaSetSpanFunctions(stencilRb, mesaVis);
+         _mesa_add_renderbuffer(fb, BUFFER_STENCIL, &stencilRb->Base);
+      }
+#endif
+
+      _mesa_add_soft_renderbuffers(fb,
+                                   GL_FALSE, /* color */
+                                   GL_FALSE, /* depth */
+                                   swStencil,
+                                   swAccum,
+                                   GL_FALSE, /* alpha */
+                                   GL_FALSE /* aux */);
+      driDrawPriv->driverPrivate = (void *) fb;
+
+      return (driDrawPriv->driverPrivate != NULL);
+   }
+}
+
+
+static void
+viaDestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
+{
+   _mesa_unreference_framebuffer((GLframebuffer **)(&(driDrawPriv->driverPrivate)));
+}
+
+
+
+static struct __DriverAPIRec viaAPI = {
+   .InitDriver      = viaInitDriver,
+   .DestroyScreen   = viaDestroyScreen,
+   .CreateContext   = viaCreateContext,
+   .DestroyContext  = viaDestroyContext,
+   .CreateBuffer    = viaCreateBuffer,
+   .DestroyBuffer   = viaDestroyBuffer,
+   .SwapBuffers     = viaSwapBuffers,
+   .MakeCurrent     = viaMakeCurrent,
+   .UnbindContext   = viaUnbindContext,
+   .GetSwapInfo     = getSwapInfo,
+   .GetMSC          = driGetMSC32,
+   .WaitForMSC      = driWaitForMSC32,
+   .WaitForSBC      = NULL,
+   .SwapBuffersMSC  = NULL
+};
+
+
+static __GLcontextModes *
+viaFillInModes( unsigned pixel_bits, GLboolean have_back_buffer )
+{
+    __GLcontextModes * modes;
+    __GLcontextModes * m;
+    unsigned num_modes;
+    const unsigned back_buffer_factor = (have_back_buffer) ? 2 : 1;
+    GLenum fb_format;
+    GLenum fb_type;
+
+    /* Right now GLX_SWAP_COPY_OML isn't supported, but it would be easy
+     * enough to add support.  Basically, if a context is created with an
+     * fbconfig where the swap method is GLX_SWAP_COPY_OML, pageflipping
+     * will never be used.
+     */
+    static const GLenum back_buffer_modes[] = {
+	GLX_NONE, GLX_SWAP_UNDEFINED_OML /*, GLX_SWAP_COPY_OML */
+    };
+
+    /* The 32-bit depth-buffer mode isn't supported yet, so don't actually
+     * enable it.
+     */
+    static const u_int8_t depth_bits_array[4]   = { 0, 16, 24, 32 };
+    static const u_int8_t stencil_bits_array[4] = { 0,  0,  8,  0 };
+    const unsigned depth_buffer_factor = 3;
+
+
+    num_modes = depth_buffer_factor * back_buffer_factor * 4;
+
+    if ( pixel_bits == 16 ) {
+        fb_format = GL_RGB;
+        fb_type = GL_UNSIGNED_SHORT_5_6_5;
+    }
+    else {
+        fb_format = GL_BGRA;
+        fb_type = GL_UNSIGNED_INT_8_8_8_8_REV;
+    }
+
+    modes = (*dri_interface->createContextModes)( num_modes, sizeof( __GLcontextModes ) );
+    m = modes;
+    if ( ! driFillInModes( & m, fb_format, fb_type,
+			   depth_bits_array, stencil_bits_array, 
+			   depth_buffer_factor,
+			   back_buffer_modes, back_buffer_factor,
+			   GLX_TRUE_COLOR ) ) {
+	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
+		 __func__, __LINE__ );
+	return NULL;
+    }
+
+    if ( ! driFillInModes( & m, fb_format, fb_type,
+			   depth_bits_array, stencil_bits_array, 
+			   depth_buffer_factor,
+			   back_buffer_modes, back_buffer_factor,
+			   GLX_DIRECT_COLOR ) ) {
+	fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
+		 __func__, __LINE__ );
+	return NULL;
+    }
+
+    return modes;
+}
+
+
+/**
+ * This is the bootstrap function for the driver.  libGL supplies all of the
+ * requisite information about the system, and the driver initializes itself.
+ * This routine also fills in the linked list pointed to by \c driver_modes
+ * with the \c __GLcontextModes that the driver can support for windows or
+ * pbuffers.
+ * 
+ * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on 
+ *         failure.
+ */
+PUBLIC
+void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn,
+			     __DRIscreen *psc,
+			     const __GLcontextModes * modes,
+			     const __DRIversion * ddx_version,
+			     const __DRIversion * dri_version,
+			     const __DRIversion * drm_version,
+			     const __DRIframebuffer * frame_buffer,
+			     drmAddress pSAREA, int fd, 
+			     int internal_api_version,
+			     const __DRIinterfaceMethods * interface,
+			     __GLcontextModes ** driver_modes )
+			     
+{
+   __DRIscreenPrivate *psp;
+   static const __DRIversion ddx_expected = { VIA_DRIDDX_VERSION_MAJOR,
+                                              VIA_DRIDDX_VERSION_MINOR,
+                                              VIA_DRIDDX_VERSION_PATCH };
+   static const __DRIversion dri_expected = { 4, 0, 0 };
+   static const __DRIversion drm_expected = { 2, 3, 0 };
+   static const char *driver_name = "Unichrome";
+
+   dri_interface = interface;
+
+   if ( ! driCheckDriDdxDrmVersions2( driver_name,
+				      dri_version, & dri_expected,
+				      ddx_version, & ddx_expected,
+				      drm_version, & drm_expected) ) {
+      return NULL;
+   }
+      
+   psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL,
+				  ddx_version, dri_version, drm_version,
+				  frame_buffer, pSAREA, fd,
+				  internal_api_version, &viaAPI);
+   if ( psp != NULL ) {
+      VIADRIPtr dri_priv = (VIADRIPtr) psp->pDevPriv;
+      *driver_modes = viaFillInModes( dri_priv->bytesPerPixel * 8,
+				      GL_TRUE );
+
+      /* Calling driInitExtensions here, with a NULL context pointer, does not actually
+       * enable the extensions.  It just makes sure that all the dispatch offsets for all
+       * the extensions that *might* be enables are known.  This is needed because the
+       * dispatch offsets need to be known when _mesa_context_create is called, but we can't
+       * enable the extensions until we have a context pointer.
+       *
+       * Hello chicken.  Hello egg.  How are you two today?
+       */
+      driInitExtensions( NULL, card_extensions, GL_FALSE );
+   }
+
+   return (void *) psp;
+}
+
+
+/**
+ * Get information about previous buffer swaps.
+ */
+static int
+getSwapInfo( __DRIdrawablePrivate *dPriv, __DRIswapInfo * sInfo )
+{
+   struct via_context *vmesa;
+
+   if ( (dPriv == NULL) || (dPriv->driContextPriv == NULL)
+	|| (dPriv->driContextPriv->driverPrivate == NULL)
+	|| (sInfo == NULL) ) {
+      return -1;
+   }
+
+   vmesa = (struct via_context *) dPriv->driContextPriv->driverPrivate;
+   sInfo->swap_count = vmesa->swap_count;
+   sInfo->swap_ust = vmesa->swap_ust;
+   sInfo->swap_missed_count = vmesa->swap_missed_count;
+
+   sInfo->swap_missed_usage = (sInfo->swap_missed_count != 0)
+       ? driCalculateSwapUsage( dPriv, 0, vmesa->swap_missed_ust )
+       : 0.0;
+
+   return 0;
+}
diff --git a/src/via_screen.h b/src/via_screen.h
new file mode 100644
index 0000000..84aa5ae
--- /dev/null
+++ b/src/via_screen.h
@@ -0,0 +1,95 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIAINIT_H
+#define _VIAINIT_H
+
+#include <sys/time.h>
+#include "dri_util.h"
+#include "via_dri.h"
+#include "xmlconfig.h"
+
+typedef struct {
+    viaRegion regs;
+    viaRegion agp;
+    int deviceID;
+    int width;
+    int height;
+    int mem;
+
+    int cpp;                    
+    int bitsPerPixel;
+    int bytesPerPixel;
+    int fbFormat;
+    int fbOffset;
+    int fbSize;
+    
+    int fbStride;
+
+    int backOffset;
+    int depthOffset;
+
+    int backPitch;
+    int backPitchBits;
+
+    int textureOffset;
+    int textureSize;
+    int logTextureGranularity;
+    
+    drmAddress reg;
+    drmAddress agpLinearStart;
+    GLuint agpBase;
+
+    __DRIscreenPrivate *driScrnPriv;
+    drmBufMapPtr bufs;
+    unsigned int sareaPrivOffset;
+    /*=* John Sheng [2003.12.9] Tuxracer & VQ *=*/
+    int VQEnable;
+    int irqEnabled;
+
+    /* Configuration cache with default values for all contexts */
+    driOptionCache optionCache;
+} viaScreenPrivate;
+
+
+extern GLboolean
+viaCreateContext(const __GLcontextModes *mesaVis,
+                 __DRIcontextPrivate *driContextPriv,
+                 void *sharedContextPrivate);
+
+extern void
+viaDestroyContext(__DRIcontextPrivate *driContextPriv);
+
+extern GLboolean
+viaUnbindContext(__DRIcontextPrivate *driContextPriv);
+
+extern GLboolean
+viaMakeCurrent(__DRIcontextPrivate *driContextPriv,
+               __DRIdrawablePrivate *driDrawPriv,
+               __DRIdrawablePrivate *driReadPriv);
+
+extern void
+viaSwapBuffers(__DRIdrawablePrivate *drawablePrivate);
+
+#endif
diff --git a/src/via_span.c b/src/via_span.c
new file mode 100644
index 0000000..3a16dad
--- /dev/null
+++ b/src/via_span.c
@@ -0,0 +1,195 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include "glheader.h"
+#include "macros.h"
+#include "mtypes.h"
+#include "colormac.h"
+#include "via_context.h"
+#include "via_span.h"
+#include "via_ioctl.h"
+#include "swrast/swrast.h"
+
+#define DBG 0
+
+#define Y_FLIP(_y) (height - _y - 1)
+
+#define HW_LOCK() 
+
+#define HW_UNLOCK()
+
+#undef LOCAL_VARS
+#define LOCAL_VARS                                                   	\
+    struct via_renderbuffer *vrb = (struct via_renderbuffer *) rb;   	\
+    __DRIdrawablePrivate *dPriv = vrb->dPriv;                           \
+    GLuint pitch = vrb->pitch;                                          \
+    GLuint height = dPriv->h;                                        	\
+    GLint p = 0;							\
+    char *buf = (char *)(vrb->origMap);					\
+    (void) p;
+
+/* ================================================================
+ * Color buffer
+ */
+
+/* 16 bit, RGB565 color spanline and pixel functions
+ */
+#define GET_PTR(_x, _y) (buf + (_x) * 2 + (_y) * pitch)
+#define SPANTMP_PIXEL_FMT GL_RGB
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
+
+#define TAG(x)    via##x##_565
+#define TAG2(x,y) via##x##_565##y
+#include "spantmp2.h"
+
+
+/* 32 bit, ARGB8888 color spanline and pixel functions
+ */
+#define GET_PTR(_x, _y) (buf + (_x) * 4 + (_y) * pitch)
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    via##x##_8888
+#define TAG2(x,y) via##x##_8888##y
+#include "spantmp2.h"
+
+
+/* 16 bit depthbuffer functions.
+ */
+#define LOCAL_DEPTH_VARS                                            \
+    struct via_renderbuffer *vrb = (struct via_renderbuffer *) rb;  \
+    __DRIdrawablePrivate *dPriv = vrb->dPriv;                       \
+    GLuint depth_pitch = vrb->pitch;                                \
+    GLuint height = dPriv->h;                                       \
+    char *buf = (char *)(vrb->map)
+
+#define LOCAL_STENCIL_VARS LOCAL_DEPTH_VARS 
+
+
+#define WRITE_DEPTH(_x, _y, d)                      \
+    *(GLushort *)(buf + (_x) * 2 + (_y) * depth_pitch) = d;
+
+#define READ_DEPTH(d, _x, _y)                       \
+    d = *(volatile GLushort *)(buf + (_x) * 2 + (_y) * depth_pitch);
+
+#define TAG(x) via##x##_z16
+#include "depthtmp.h"
+
+/* 32 bit depthbuffer functions.
+ */
+#define WRITE_DEPTH(_x, _y, d)                      \
+    *(GLuint *)(buf + (_x) * 4 + (_y) * depth_pitch) = d;
+
+#define READ_DEPTH(d, _x, _y)                       \
+    d = *(volatile GLuint *)(buf + (_x) * 4 + (_y) * depth_pitch);
+
+#define TAG(x) via##x##_z32
+#include "depthtmp.h"
+
+
+
+/* 24/8 bit interleaved depth/stencil functions
+ */
+#define WRITE_DEPTH( _x, _y, d ) {			\
+   GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch);	\
+   tmp &= 0x000000ff;					\
+   tmp |= ((d)<<8);				\
+   *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) = tmp;		\
+}
+
+#define READ_DEPTH( d, _x, _y )		\
+   d = (*(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch)) >> 8;
+
+
+#define TAG(x) via##x##_z24_s8
+#include "depthtmp.h"
+
+#define WRITE_STENCIL( _x, _y, d ) {			\
+   GLuint tmp = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch);	\
+   tmp &= 0xffffff00;					\
+   tmp |= (d);					\
+   *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) = tmp;		\
+}
+
+#define READ_STENCIL( d, _x, _y )			\
+   d = *(GLuint *)(buf + (_x)*4 + (_y)*depth_pitch) & 0xff;
+
+#define TAG(x) via##x##_z24_s8
+#include "stenciltmp.h"
+
+
+
+
+/* Move locking out to get reasonable span performance.
+ */
+void viaSpanRenderStart( GLcontext *ctx )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);     
+   viaWaitIdle(vmesa, GL_FALSE);
+   LOCK_HARDWARE(vmesa);
+}
+
+void viaSpanRenderFinish( GLcontext *ctx )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   _swrast_flush( ctx );
+   UNLOCK_HARDWARE( vmesa );
+}
+
+void viaInitSpanFuncs(GLcontext *ctx)
+{
+    struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
+    swdd->SpanRenderStart = viaSpanRenderStart;
+    swdd->SpanRenderFinish = viaSpanRenderFinish; 
+}
+
+
+
+/**
+ * Plug in the Get/Put routines for the given driRenderbuffer.
+ */
+void
+viaSetSpanFunctions(struct via_renderbuffer *vrb, const GLvisual *vis)
+{
+   if (vrb->Base.InternalFormat == GL_RGBA) {
+      if (vis->redBits == 5 && vis->greenBits == 6 && vis->blueBits == 5) {
+         viaInitPointers_565(&vrb->Base);
+      }
+      else {
+         viaInitPointers_8888(&vrb->Base);
+      }
+   }
+   else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT16) {
+      viaInitDepthPointers_z16(&vrb->Base);
+   }
+   else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT24) {
+      viaInitDepthPointers_z24_s8(&vrb->Base);
+   }
+   else if (vrb->Base.InternalFormat == GL_DEPTH_COMPONENT32) {
+      viaInitDepthPointers_z32(&vrb->Base);
+   }
+   else if (vrb->Base.InternalFormat == GL_STENCIL_INDEX8_EXT) {
+      viaInitStencilPointers_z24_s8(&vrb->Base);
+   }
+}
diff --git a/src/via_span.h b/src/via_span.h
new file mode 100644
index 0000000..3dca0d5
--- /dev/null
+++ b/src/via_span.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIA_SPAN_H
+#define _VIA_SPAN_H
+
+extern void viaInitSpanFuncs(GLcontext *ctx);
+extern void viaSpanRenderStart( GLcontext *ctx );
+extern void viaSpanRenderFinish( GLcontext *ctx );
+
+extern void
+viaSetSpanFunctions(struct via_renderbuffer *vrb, const GLvisual *vis);
+
+#endif
diff --git a/src/via_state.c b/src/via_state.c
new file mode 100644
index 0000000..bccdbe9
--- /dev/null
+++ b/src/via_state.c
@@ -0,0 +1,1527 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include <stdio.h>
+
+#include "glheader.h"
+#include "context.h"
+#include "macros.h"
+#include "colormac.h"
+#include "enums.h"
+#include "dd.h"
+
+#include "mm.h"
+#include "via_context.h"
+#include "via_state.h"
+#include "via_tex.h"
+#include "via_tris.h"
+#include "via_ioctl.h"
+#include "via_3d_reg.h"
+
+#include "swrast/swrast.h"
+#include "vbo/vbo.h"
+#include "tnl/tnl.h"
+#include "swrast_setup/swrast_setup.h"
+
+#include "tnl/t_pipeline.h"
+
+
+static GLuint ROP[16] = {
+    HC_HROP_BLACK,    /* GL_CLEAR           0                      	*/
+    HC_HROP_DPa,      /* GL_AND             s & d                  	*/
+    HC_HROP_PDna,     /* GL_AND_REVERSE     s & ~d  			*/
+    HC_HROP_P,        /* GL_COPY            s                       	*/
+    HC_HROP_DPna,     /* GL_AND_INVERTED    ~s & d                      */
+    HC_HROP_D,        /* GL_NOOP            d  		                */
+    HC_HROP_DPx,      /* GL_XOR             s ^ d                       */
+    HC_HROP_DPo,      /* GL_OR              s | d                       */
+    HC_HROP_DPon,     /* GL_NOR             ~(s | d)                    */
+    HC_HROP_DPxn,     /* GL_EQUIV           ~(s ^ d)                    */
+    HC_HROP_Dn,       /* GL_INVERT          ~d                       	*/
+    HC_HROP_PDno,     /* GL_OR_REVERSE      s | ~d                      */
+    HC_HROP_Pn,       /* GL_COPY_INVERTED   ~s                       	*/
+    HC_HROP_DPno,     /* GL_OR_INVERTED     ~s | d                      */
+    HC_HROP_DPan,     /* GL_NAND            ~(s & d)                    */
+    HC_HROP_WHITE     /* GL_SET             1                       	*/
+};
+
+/*
+ * Compute the 'S5.5' lod bias factor from the floating point OpenGL bias.
+ */
+static GLuint viaComputeLodBias(GLfloat bias)
+{
+   int b = (int) (bias * 32.0);
+   if (b > 511)
+      b = 511;
+   else if (b < -512)
+      b = -512;
+   return (GLuint) b;
+}
+
+void viaEmitState(struct via_context *vmesa)
+{
+   GLcontext *ctx = vmesa->glCtx;
+   GLuint i = 0;
+   GLuint j = 0;
+   RING_VARS;
+
+   viaCheckDma(vmesa, 0x110);
+    
+   BEGIN_RING(5);
+   OUT_RING( HC_HEADER2 );
+   OUT_RING( (HC_ParaType_NotTex << 16) );
+   OUT_RING( ((HC_SubA_HEnable << 24) | vmesa->regEnable) );
+   OUT_RING( ((HC_SubA_HFBBMSKL << 24) | vmesa->regHFBBMSKL) );    
+   OUT_RING( ((HC_SubA_HROP << 24) | vmesa->regHROP) );        
+   ADVANCE_RING();
+    
+   if (vmesa->have_hw_stencil) {
+      GLuint pitch, format, offset;
+	
+      format = HC_HZWBFM_24;	    	
+      offset = vmesa->depth.offset;
+      pitch = vmesa->depth.pitch;
+	
+      BEGIN_RING(6);
+      OUT_RING( (HC_SubA_HZWBBasL << 24) | (offset & 0xFFFFFF) );
+      OUT_RING( (HC_SubA_HZWBBasH << 24) | ((offset & 0xFF000000) >> 24) );	
+      OUT_RING( (HC_SubA_HZWBType << 24) | HC_HDBLoc_Local | HC_HZONEasFF_MASK |
+	         format | pitch );            
+      OUT_RING( (HC_SubA_HZWTMD << 24) | vmesa->regHZWTMD );
+      OUT_RING( (HC_SubA_HSTREF << 24) | vmesa->regHSTREF );
+      OUT_RING( (HC_SubA_HSTMD << 24) | vmesa->regHSTMD );
+      ADVANCE_RING();
+   }
+   else if (vmesa->hasDepth) {
+      GLuint pitch, format, offset;
+	
+      if (vmesa->depthBits == 16) {
+	 format = HC_HZWBFM_16;
+      }	    
+      else {
+	 format = HC_HZWBFM_32;
+      }
+	    
+	    
+      offset = vmesa->depth.offset;
+      pitch = vmesa->depth.pitch;
+	
+      BEGIN_RING(4);
+      OUT_RING( (HC_SubA_HZWBBasL << 24) | (offset & 0xFFFFFF) );
+      OUT_RING( (HC_SubA_HZWBBasH << 24) | ((offset & 0xFF000000) >> 24) );
+      OUT_RING( (HC_SubA_HZWBType << 24) | HC_HDBLoc_Local | HC_HZONEasFF_MASK |
+	         format | pitch );
+      OUT_RING( (HC_SubA_HZWTMD << 24) | vmesa->regHZWTMD );
+      ADVANCE_RING();
+   }
+    
+   if (ctx->Color.AlphaEnabled) {
+      BEGIN_RING(1);
+      OUT_RING( (HC_SubA_HATMD << 24) | vmesa->regHATMD );
+      ADVANCE_RING();
+      i++;
+   }   
+
+   if (ctx->Color.BlendEnabled) {
+      BEGIN_RING(11);
+      OUT_RING( (HC_SubA_HABLCsat << 24) | vmesa->regHABLCsat );
+      OUT_RING( (HC_SubA_HABLCop  << 24) | vmesa->regHABLCop ); 
+      OUT_RING( (HC_SubA_HABLAsat << 24) | vmesa->regHABLAsat );        
+      OUT_RING( (HC_SubA_HABLAop  << 24) | vmesa->regHABLAop ); 
+      OUT_RING( (HC_SubA_HABLRCa  << 24) | vmesa->regHABLRCa ); 
+      OUT_RING( (HC_SubA_HABLRFCa << 24) | vmesa->regHABLRFCa );        
+      OUT_RING( (HC_SubA_HABLRCbias << 24) | vmesa->regHABLRCbias ); 
+      OUT_RING( (HC_SubA_HABLRCb  << 24) | vmesa->regHABLRCb ); 
+      OUT_RING( (HC_SubA_HABLRFCb << 24) | vmesa->regHABLRFCb );        
+      OUT_RING( (HC_SubA_HABLRAa  << 24) | vmesa->regHABLRAa ); 
+      OUT_RING( (HC_SubA_HABLRAb  << 24) | vmesa->regHABLRAb ); 
+      ADVANCE_RING();
+   }
+    
+   if (ctx->Fog.Enabled) {
+      BEGIN_RING(3);
+      OUT_RING( (HC_SubA_HFogLF << 24) | vmesa->regHFogLF ); 
+      OUT_RING( (HC_SubA_HFogCL << 24) | vmesa->regHFogCL ); 
+      OUT_RING( (HC_SubA_HFogCH << 24) | vmesa->regHFogCH ); 
+      ADVANCE_RING();
+   }
+    
+   if (ctx->Line.StippleFlag) {
+      BEGIN_RING(2);
+      OUT_RING( (HC_SubA_HLP << 24) | ctx->Line.StipplePattern ); 
+      OUT_RING( (HC_SubA_HLPRF << 24) | ctx->Line.StippleFactor );
+      ADVANCE_RING();
+   }
+
+   BEGIN_RING(1);
+   OUT_RING( (HC_SubA_HPixGC << 24) | 0x0 ); 
+   ADVANCE_RING();
+    
+   QWORD_PAD_RING();
+
+
+   if (ctx->Texture._EnabledUnits) {
+    
+      struct gl_texture_unit *texUnit0 = &ctx->Texture.Unit[0];
+      struct gl_texture_unit *texUnit1 = &ctx->Texture.Unit[1];
+
+      {
+	 GLuint nDummyValue = 0;
+
+	 BEGIN_RING( 8 );
+	 OUT_RING( HC_HEADER2 );
+	 OUT_RING( (HC_ParaType_Tex << 16) | (HC_SubType_TexGeneral << 24) );
+
+	 if (texUnit0->Enabled && texUnit1->Enabled) {
+	    nDummyValue = (HC_SubA_HTXSMD << 24) | (1 << 3);                
+	 }
+	 else {
+	    nDummyValue = (HC_SubA_HTXSMD << 24) | 0;
+	 }
+
+	 if (vmesa->clearTexCache) {
+	    vmesa->clearTexCache = 0;
+	    OUT_RING( nDummyValue | HC_HTXCHCLR_MASK );
+	    OUT_RING( nDummyValue );
+	 }
+	 else {
+	    OUT_RING( nDummyValue );
+	    OUT_RING( nDummyValue );
+	 }
+
+	 OUT_RING( HC_HEADER2 );
+	 OUT_RING( HC_ParaType_NotTex << 16 );
+	 OUT_RING( (HC_SubA_HEnable << 24) | vmesa->regEnable );
+	 OUT_RING( (HC_SubA_HEnable << 24) | vmesa->regEnable );
+	 ADVANCE_RING();
+      }
+
+      if (texUnit0->Enabled) {
+	 struct gl_texture_object *texObj = texUnit0->_Current;
+	 struct via_texture_object *t = (struct via_texture_object *)texObj;
+	 GLuint numLevels = t->lastLevel - t->firstLevel + 1;
+	 if (VIA_DEBUG & DEBUG_STATE) {
+	    fprintf(stderr, "texture0 enabled\n");
+	 }		
+	 if (numLevels == 8) {
+	    BEGIN_RING(27);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (0 << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexWidthLog2[1] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[1] );
+	    OUT_RING( t->regTexBaseH[0] );
+	    OUT_RING( t->regTexBaseH[1] );
+	    OUT_RING( t->regTexBaseH[2] );
+	    OUT_RING( t->regTexBaseAndPitch[0].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[0].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[1].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[1].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[2].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[2].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[3].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[3].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[4].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[4].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[5].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[5].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[6].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[6].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[7].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[7].pitchLog2 );
+	    ADVANCE_RING();
+	 }
+	 else if (numLevels > 1) {
+
+	    BEGIN_RING(12 + numLevels * 2);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (0 << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+		
+	    if (numLevels > 6) {
+	       OUT_RING( t->regTexWidthLog2[1] );
+	       OUT_RING( t->regTexHeightLog2[1] );
+	    }
+                
+	    OUT_RING( t->regTexBaseH[0] );
+		
+	    if (numLevels > 3) {
+	       OUT_RING( t->regTexBaseH[1] );
+	    }
+	    if (numLevels > 6) {
+	       OUT_RING( t->regTexBaseH[2] );
+	    }
+	    if (numLevels > 9)  {
+	       OUT_RING( t->regTexBaseH[3] );
+	    }
+
+	    for (j = 0; j < numLevels; j++) {
+	       OUT_RING( t->regTexBaseAndPitch[j].baseL );
+	       OUT_RING( t->regTexBaseAndPitch[j].pitchLog2 );
+	    }
+
+	    ADVANCE_RING_VARIABLE();
+	 }
+	 else {
+
+	    BEGIN_RING(9);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (0 << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+	    OUT_RING( t->regTexBaseH[0] );
+	    OUT_RING( t->regTexBaseAndPitch[0].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[0].pitchLog2 );
+	    ADVANCE_RING();
+	 }
+
+	 BEGIN_RING(14);
+	 OUT_RING( (HC_SubA_HTXnTB << 24) | vmesa->regHTXnTB[0] );
+	 OUT_RING( (HC_SubA_HTXnMPMD << 24) | vmesa->regHTXnMPMD[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLCsat << 24) | vmesa->regHTXnTBLCsat[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLCop << 24) | vmesa->regHTXnTBLCop[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLMPfog << 24) | vmesa->regHTXnTBLMPfog[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLAsat << 24) | vmesa->regHTXnTBLAsat[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCb << 24) | vmesa->regHTXnTBLRCb[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRAa << 24) | vmesa->regHTXnTBLRAa[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRFog << 24) | vmesa->regHTXnTBLRFog[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCa << 24) | vmesa->regHTXnTBLRCa[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCc << 24) | vmesa->regHTXnTBLRCc[0] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCbias << 24) | vmesa->regHTXnTBLRCbias[0] );
+	 OUT_RING( (HC_SubA_HTXnTBC << 24) | vmesa->regHTXnTBC[0] );
+	 OUT_RING( (HC_SubA_HTXnTRAH << 24) | vmesa->regHTXnTRAH[0] );
+/* 	 OUT_RING( (HC_SubA_HTXnCLODu << 24) | vmesa->regHTXnCLOD[0] ); */
+	 ADVANCE_RING();
+
+	 /* KW:  This test never succeeds:
+	  */
+	 if (t->regTexFM == HC_HTXnFM_Index8) {
+	    const struct gl_color_table *table = &texObj->Palette;
+	    const GLfloat *tableF = table->TableF;
+
+	    BEGIN_RING(2 + table->Size);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Palette << 16) | (0 << 24) );
+	    for (j = 0; j < table->Size; j++) 
+	       OUT_RING( tableF[j] );
+	    ADVANCE_RING();
+	       
+	 }
+
+	 QWORD_PAD_RING();
+      }
+	
+      if (texUnit1->Enabled) {
+	 struct gl_texture_object *texObj = texUnit1->_Current;
+	 struct via_texture_object *t = (struct via_texture_object *)texObj;
+	 GLuint numLevels = t->lastLevel - t->firstLevel + 1;
+	 int texunit = (texUnit0->Enabled ? 1 : 0);
+	 if (VIA_DEBUG & DEBUG_STATE) {
+	    fprintf(stderr, "texture1 enabled\n");
+	 }		
+	 if (numLevels == 8) {
+	    BEGIN_RING(27);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (texunit << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexWidthLog2[1] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[1] );
+	    OUT_RING( t->regTexBaseH[0] );
+	    OUT_RING( t->regTexBaseH[1] );
+	    OUT_RING( t->regTexBaseH[2] );
+	    OUT_RING( t->regTexBaseAndPitch[0].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[0].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[1].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[1].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[2].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[2].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[3].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[3].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[4].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[4].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[5].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[5].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[6].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[6].pitchLog2 );
+	    OUT_RING( t->regTexBaseAndPitch[7].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[7].pitchLog2 );
+	    ADVANCE_RING();
+	 }
+	 else if (numLevels > 1) {
+	    BEGIN_RING(12 + numLevels * 2);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (texunit << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+		
+	    if (numLevels > 6) {
+	       OUT_RING( t->regTexWidthLog2[1] );
+	       OUT_RING( t->regTexHeightLog2[1] );
+	       i += 2;
+	    }
+                
+	    OUT_RING( t->regTexBaseH[0] );
+		
+	    if (numLevels > 3) { 
+	       OUT_RING( t->regTexBaseH[1] );
+	    }
+	    if (numLevels > 6) {
+	       OUT_RING( t->regTexBaseH[2] );
+	    }
+	    if (numLevels > 9)  {
+	       OUT_RING( t->regTexBaseH[3] );
+	    }
+		
+	    for (j = 0; j < numLevels; j++) {
+	       OUT_RING( t->regTexBaseAndPitch[j].baseL );
+	       OUT_RING( t->regTexBaseAndPitch[j].pitchLog2 );
+	    }
+	    ADVANCE_RING_VARIABLE();
+	 }
+	 else {
+	    BEGIN_RING(9);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Tex << 16) |  (texunit << 24) );
+	    OUT_RING( t->regTexFM );
+	    OUT_RING( (HC_SubA_HTXnL0OS << 24) |
+	       ((t->lastLevel) << HC_HTXnLVmax_SHIFT) | t->firstLevel );
+	    OUT_RING( t->regTexWidthLog2[0] );
+	    OUT_RING( t->regTexHeightLog2[0] );
+	    OUT_RING( t->regTexBaseH[0] );
+	    OUT_RING( t->regTexBaseAndPitch[0].baseL );
+	    OUT_RING( t->regTexBaseAndPitch[0].pitchLog2 );
+	    ADVANCE_RING();
+	 }
+
+	 BEGIN_RING(14);
+	 OUT_RING( (HC_SubA_HTXnTB << 24) | vmesa->regHTXnTB[1] );
+	 OUT_RING( (HC_SubA_HTXnMPMD << 24) | vmesa->regHTXnMPMD[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLCsat << 24) | vmesa->regHTXnTBLCsat[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLCop << 24) | vmesa->regHTXnTBLCop[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLMPfog << 24) | vmesa->regHTXnTBLMPfog[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLAsat << 24) | vmesa->regHTXnTBLAsat[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCb << 24) | vmesa->regHTXnTBLRCb[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRAa << 24) | vmesa->regHTXnTBLRAa[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRFog << 24) | vmesa->regHTXnTBLRFog[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCa << 24) | vmesa->regHTXnTBLRCa[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCc << 24) | vmesa->regHTXnTBLRCc[1] );
+	 OUT_RING( (HC_SubA_HTXnTBLRCbias << 24) | vmesa->regHTXnTBLRCbias[1] );
+	 OUT_RING( (HC_SubA_HTXnTBC << 24) | vmesa->regHTXnTBC[1] );
+	 OUT_RING( (HC_SubA_HTXnTRAH << 24) | vmesa->regHTXnTRAH[1] );
+/* 	 OUT_RING( (HC_SubA_HTXnCLODu << 24) | vmesa->regHTXnCLOD[1] ); */
+	 ADVANCE_RING();
+
+	 /* KW:  This test never succeeds:
+	  */
+	 if (t->regTexFM == HC_HTXnFM_Index8) {
+	    const struct gl_color_table *table = &texObj->Palette;
+	    const GLfloat *tableF = table->TableF;
+
+	    BEGIN_RING(2 + table->Size);
+	    OUT_RING( HC_HEADER2 );
+	    OUT_RING( (HC_ParaType_Palette << 16) | (texunit << 24) );
+	    for (j = 0; j < table->Size; j++) {
+	       OUT_RING( tableF[j] );
+	    }
+	    ADVANCE_RING();
+	 }
+
+	 QWORD_PAD_RING();
+      }
+   }
+    
+#if 0
+   /* Polygon stipple is broken - for certain stipple values,
+    * eg. 0xf0f0f0f0, the hardware will refuse to accept the stipple.
+    * Coincidentally, conform generates just such a stipple.
+    */
+   if (ctx->Polygon.StippleFlag) {
+      GLuint *stipple = &ctx->PolygonStipple[0];
+      __DRIdrawablePrivate *dPriv = vmesa->driDrawable;
+      struct via_renderbuffer *const vrb = 
+	(struct via_renderbuffer *) dPriv->driverPrivate;
+      GLint i;
+        
+      BEGIN_RING(38);
+      OUT_RING( HC_HEADER2 );             
+
+      OUT_RING( ((HC_ParaType_Palette << 16) | (HC_SubType_Stipple << 24)) );
+      for (i = 31; i >= 0; i--) {
+	 GLint j;
+	 GLuint k = 0;
+
+	 /* Need to flip bits left to right:
+	  */
+	 for (j = 0 ; j < 32; j++)
+	    if (stipple[i] & (1<<j))
+	       k |= 1 << (31-j);
+
+	 OUT_RING( k );     
+      }
+
+      OUT_RING( HC_HEADER2 );                     
+      OUT_RING( (HC_ParaType_NotTex << 16) );
+      OUT_RING( (HC_SubA_HSPXYOS << 24) );
+      OUT_RING( (HC_SubA_HSPXYOS << 24) );
+
+      ADVANCE_RING();
+   }
+#endif
+   
+   vmesa->newEmitState = 0;
+}
+
+
+static __inline__ GLuint viaPackColor(GLuint bpp,
+                                      GLubyte r, GLubyte g,
+                                      GLubyte b, GLubyte a)
+{
+    switch (bpp) {
+    case 16:
+        return PACK_COLOR_565(r, g, b);
+    case 32:
+        return PACK_COLOR_8888(a, r, g, b);        
+    default:
+       assert(0);
+       return 0;
+   }
+}
+
+static void viaBlendEquationSeparate(GLcontext *ctx,
+				     GLenum rgbMode, 
+				     GLenum aMode)
+{
+    if (VIA_DEBUG & DEBUG_STATE) 
+       fprintf(stderr, "%s in\n", __FUNCTION__);
+
+    /* GL_EXT_blend_equation_separate not supported */
+    ASSERT(rgbMode == aMode);
+
+    /* Can only do GL_ADD equation in hardware */
+    FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_BLEND_EQ, 
+	     rgbMode != GL_FUNC_ADD_EXT);
+
+    /* BlendEquation sets ColorLogicOpEnabled in an unexpected
+     * manner.
+     */
+    FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_LOGICOP,
+             (ctx->Color.ColorLogicOpEnabled &&
+              ctx->Color.LogicOp != GL_COPY));
+}
+
+static void viaBlendFunc(GLcontext *ctx, GLenum sfactor, GLenum dfactor)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    GLboolean fallback = GL_FALSE;
+    if (VIA_DEBUG & DEBUG_STATE) 
+       fprintf(stderr, "%s in\n", __FUNCTION__);
+
+    switch (ctx->Color.BlendSrcRGB) {
+    case GL_SRC_ALPHA_SATURATE:  
+    case GL_CONSTANT_COLOR:
+    case GL_ONE_MINUS_CONSTANT_COLOR:
+    case GL_CONSTANT_ALPHA:
+    case GL_ONE_MINUS_CONSTANT_ALPHA:
+        fallback = GL_TRUE;
+        break;
+    default:
+        break;
+    }
+
+    switch (ctx->Color.BlendDstRGB) {
+    case GL_CONSTANT_COLOR:
+    case GL_ONE_MINUS_CONSTANT_COLOR:
+    case GL_CONSTANT_ALPHA:
+    case GL_ONE_MINUS_CONSTANT_ALPHA:
+        fallback = GL_TRUE;
+        break;
+    default:
+        break;
+    }
+
+    FALLBACK(vmesa, VIA_FALLBACK_BLEND_FUNC, fallback);
+}
+
+/* Shouldn't be called as the extension is disabled.
+ */
+static void viaBlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB,
+                                 GLenum dfactorRGB, GLenum sfactorA,
+                                 GLenum dfactorA)
+{
+    if (dfactorRGB != dfactorA || sfactorRGB != sfactorA) {
+        _mesa_error(ctx, GL_INVALID_OPERATION, "glBlendEquation (disabled)");
+    }
+
+    viaBlendFunc(ctx, sfactorRGB, dfactorRGB);
+}
+
+
+
+
+/* =============================================================
+ * Hardware clipping
+ */
+static void viaScissor(GLcontext *ctx, GLint x, GLint y,
+                       GLsizei w, GLsizei h)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    if (!vmesa->driDrawable)
+       return;
+
+    if (VIA_DEBUG & DEBUG_STATE)
+       fprintf(stderr, "%s %d,%d %dx%d, drawH %d\n", __FUNCTION__, 
+	       x,y,w,h, vmesa->driDrawable->h);
+
+    if (vmesa->scissor) {
+        VIA_FLUSH_DMA(vmesa); /* don't pipeline cliprect changes */
+    }
+
+    vmesa->scissorRect.x1 = x;
+    vmesa->scissorRect.y1 = vmesa->driDrawable->h - y - h;
+    vmesa->scissorRect.x2 = x + w;
+    vmesa->scissorRect.y2 = vmesa->driDrawable->h - y;
+}
+
+static void viaEnable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+   switch (cap) {
+   case GL_SCISSOR_TEST:
+      VIA_FLUSH_DMA(vmesa);
+      vmesa->scissor = state;
+      break;
+   default:
+      break;
+   }
+}
+
+
+
+/* Fallback to swrast for select and feedback.
+ */
+static void viaRenderMode(GLcontext *ctx, GLenum mode)
+{
+    FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_RENDERMODE, (mode != GL_RENDER));
+}
+
+
+static void viaDrawBuffer(GLcontext *ctx, GLenum mode)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+   if (VIA_DEBUG & (DEBUG_DRI|DEBUG_STATE)) 
+      fprintf(stderr, "%s in\n", __FUNCTION__);
+
+   if (!ctx->DrawBuffer)
+      return;
+
+   switch ( ctx->DrawBuffer->_ColorDrawBufferMask[0] ) {
+   case BUFFER_BIT_FRONT_LEFT:
+      VIA_FLUSH_DMA(vmesa);
+      vmesa->drawBuffer = &vmesa->front;
+      FALLBACK(vmesa, VIA_FALLBACK_DRAW_BUFFER, GL_FALSE);
+      break;
+   case BUFFER_BIT_BACK_LEFT:
+      VIA_FLUSH_DMA(vmesa);
+      vmesa->drawBuffer = &vmesa->back;
+      FALLBACK(vmesa, VIA_FALLBACK_DRAW_BUFFER, GL_FALSE);
+      break;
+   default:
+      FALLBACK(vmesa, VIA_FALLBACK_DRAW_BUFFER, GL_TRUE);
+      return;
+   }
+
+
+   viaXMesaWindowMoved(vmesa);
+}
+
+static void viaClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    GLubyte pcolor[4];
+    CLAMPED_FLOAT_TO_UBYTE(pcolor[0], color[0]);
+    CLAMPED_FLOAT_TO_UBYTE(pcolor[1], color[1]);
+    CLAMPED_FLOAT_TO_UBYTE(pcolor[2], color[2]);
+    CLAMPED_FLOAT_TO_UBYTE(pcolor[3], color[3]);
+    vmesa->ClearColor = viaPackColor(vmesa->viaScreen->bitsPerPixel,
+                                     pcolor[0], pcolor[1],
+                                     pcolor[2], pcolor[3]);
+}
+
+#define WRITEMASK_ALPHA_SHIFT 31
+#define WRITEMASK_RED_SHIFT   30
+#define WRITEMASK_GREEN_SHIFT 29
+#define WRITEMASK_BLUE_SHIFT  28
+
+static void viaColorMask(GLcontext *ctx,
+			 GLboolean r, GLboolean g,
+			 GLboolean b, GLboolean a)
+{
+   struct via_context *vmesa = VIA_CONTEXT( ctx );
+
+   if (VIA_DEBUG & DEBUG_STATE)
+      fprintf(stderr, "%s r(%d) g(%d) b(%d) a(%d)\n", __FUNCTION__, r, g, b, a);
+
+   vmesa->ClearMask = (((!r) << WRITEMASK_RED_SHIFT) |
+		       ((!g) << WRITEMASK_GREEN_SHIFT) |
+		       ((!b) << WRITEMASK_BLUE_SHIFT) |
+		       ((!a) << WRITEMASK_ALPHA_SHIFT));
+}
+
+
+
+/* This hardware just isn't capable of private back buffers without
+ * glitches and/or a hefty locking scheme.
+ */
+void viaCalcViewport(GLcontext *ctx)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    __DRIdrawablePrivate *dPriv = vmesa->driDrawable;
+    struct via_renderbuffer *const vrb = 
+      (struct via_renderbuffer *) dPriv->driverPrivate;
+    const GLfloat *v = ctx->Viewport._WindowMap.m;
+    GLfloat *m = vmesa->ViewportMatrix.m;
+    
+    m[MAT_SX] =   v[MAT_SX];
+    m[MAT_TX] =   v[MAT_TX] + vrb->drawX + SUBPIXEL_X;
+    m[MAT_SY] = - v[MAT_SY];
+    m[MAT_TY] = - v[MAT_TY] + vrb->drawY + SUBPIXEL_Y + vrb->drawH;
+    m[MAT_SZ] =   v[MAT_SZ] * (1.0 / vmesa->depth_max);
+    m[MAT_TZ] =   v[MAT_TZ] * (1.0 / vmesa->depth_max);
+}
+
+static void viaViewport(GLcontext *ctx,
+                        GLint x, GLint y,
+                        GLsizei width, GLsizei height)
+{
+    viaCalcViewport(ctx);
+}
+
+static void viaDepthRange(GLcontext *ctx,
+                          GLclampd nearval, GLclampd farval)
+{
+    viaCalcViewport(ctx);
+}
+
+void viaInitState(GLcontext *ctx)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    vmesa->regCmdB = HC_ACMD_HCmdB;
+    vmesa->regEnable = HC_HenCW_MASK;
+
+   /* Mesa should do this for us:
+    */
+
+   ctx->Driver.BlendEquationSeparate( ctx, 
+				      ctx->Color.BlendEquationRGB,
+				      ctx->Color.BlendEquationA);
+
+   ctx->Driver.BlendFuncSeparate( ctx,
+				  ctx->Color.BlendSrcRGB,
+				  ctx->Color.BlendDstRGB,
+				  ctx->Color.BlendSrcA,
+				  ctx->Color.BlendDstA);
+
+   ctx->Driver.Scissor( ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			ctx->Scissor.Width, ctx->Scissor.Height );
+
+   ctx->Driver.DrawBuffer( ctx, ctx->Color.DrawBuffer[0] );
+}
+
+/**
+ * Convert S and T texture coordinate wrap modes to hardware bits.
+ */
+static u_int32_t
+get_wrap_mode( GLenum sWrap, GLenum tWrap )
+{
+    u_int32_t v = 0;
+
+
+    switch( sWrap ) {
+    case GL_REPEAT:
+	v |= HC_HTXnMPMD_Srepeat;
+	break;
+    case GL_CLAMP:
+    case GL_CLAMP_TO_EDGE:
+	v |= HC_HTXnMPMD_Sclamp;
+	break;
+    case GL_MIRRORED_REPEAT:
+	v |= HC_HTXnMPMD_Smirror;
+	break;
+    }
+
+    switch( tWrap ) {
+    case GL_REPEAT:
+	v |= HC_HTXnMPMD_Trepeat;
+	break;
+    case GL_CLAMP:
+    case GL_CLAMP_TO_EDGE:
+	v |= HC_HTXnMPMD_Tclamp;
+	break;
+    case GL_MIRRORED_REPEAT:
+	v |= HC_HTXnMPMD_Tmirror;
+	break;
+    }
+    
+    return v;
+}
+
+static u_int32_t
+get_minmag_filter( GLenum min, GLenum mag )
+{
+    u_int32_t v = 0;
+
+    switch (min) {
+    case GL_NEAREST:
+        v = HC_HTXnFLSs_Nearest |
+            HC_HTXnFLTs_Nearest;
+        break;
+    case GL_LINEAR:
+        v = HC_HTXnFLSs_Linear |
+            HC_HTXnFLTs_Linear;
+        break;
+    case GL_NEAREST_MIPMAP_NEAREST:
+        v = HC_HTXnFLSs_Nearest |
+            HC_HTXnFLTs_Nearest;
+        v |= HC_HTXnFLDs_Nearest;
+        break;
+    case GL_LINEAR_MIPMAP_NEAREST:
+        v = HC_HTXnFLSs_Linear |
+            HC_HTXnFLTs_Linear;
+        v |= HC_HTXnFLDs_Nearest;
+        break;
+    case GL_NEAREST_MIPMAP_LINEAR:
+        v = HC_HTXnFLSs_Nearest |
+            HC_HTXnFLTs_Nearest;
+        v |= HC_HTXnFLDs_Linear;
+        break;
+    case GL_LINEAR_MIPMAP_LINEAR:
+        v = HC_HTXnFLSs_Linear |
+            HC_HTXnFLTs_Linear;
+        v |= HC_HTXnFLDs_Linear;
+        break;
+    default:
+        break;
+    }
+
+    switch (mag) {
+    case GL_LINEAR:
+        v |= HC_HTXnFLSe_Linear |
+             HC_HTXnFLTe_Linear;
+	break;
+    case GL_NEAREST:
+        v |= HC_HTXnFLSe_Nearest |
+             HC_HTXnFLTe_Nearest;
+	break;
+    default:
+        break;
+    }
+
+    return v;
+}
+
+
+static GLboolean viaChooseTextureState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    struct gl_texture_unit *texUnit0 = &ctx->Texture.Unit[0];
+    struct gl_texture_unit *texUnit1 = &ctx->Texture.Unit[1];
+
+    if (texUnit0->_ReallyEnabled || texUnit1->_ReallyEnabled) {
+        vmesa->regEnable |= HC_HenTXMP_MASK | HC_HenTXCH_MASK | HC_HenTXPP_MASK;
+
+        if (texUnit0->_ReallyEnabled) {
+            struct gl_texture_object *texObj = texUnit0->_Current;
+   
+	    vmesa->regHTXnTB[0] = get_minmag_filter( texObj->MinFilter,
+						    texObj->MagFilter );
+
+	    vmesa->regHTXnMPMD[0] &= ~(HC_HTXnMPMD_SMASK | HC_HTXnMPMD_TMASK);
+	    vmesa->regHTXnMPMD[0] |= get_wrap_mode( texObj->WrapS,
+						   texObj->WrapT );
+
+	    vmesa->regHTXnTB[0] &= ~(HC_HTXnTB_TBC_S | HC_HTXnTB_TBC_T);
+            if (texObj->Image[0][texObj->BaseLevel]->Border > 0) {
+	       vmesa->regHTXnTB[0] |= (HC_HTXnTB_TBC_S | HC_HTXnTB_TBC_T);
+	       vmesa->regHTXnTBC[0] = 
+		  PACK_COLOR_888(FLOAT_TO_UBYTE(texObj->BorderColor[0]),
+				 FLOAT_TO_UBYTE(texObj->BorderColor[1]),
+				 FLOAT_TO_UBYTE(texObj->BorderColor[2]));
+	       vmesa->regHTXnTRAH[0] = FLOAT_TO_UBYTE(texObj->BorderColor[3]);
+            }
+
+	    if (texUnit0->LodBias != 0.0f) {
+	       GLuint b = viaComputeLodBias(texUnit0->LodBias);
+	       vmesa->regHTXnTB[0] &= ~HC_HTXnFLDs_MASK;
+	       vmesa->regHTXnTB[0] |= HC_HTXnFLDs_ConstLOD;
+	       vmesa->regHTXnCLOD[0] = (b&0x1f) | (((~b)&0x1f)<<10); /* FIXME */
+	    }
+
+	    if (!viaTexCombineState( vmesa, texUnit0->_CurrentCombine, 0 )) {
+	       if (VIA_DEBUG & DEBUG_TEXTURE)
+		  fprintf(stderr, "viaTexCombineState failed for unit 0\n");
+	       return GL_FALSE;
+	    }
+        }
+
+        if (texUnit1->_ReallyEnabled) {
+            struct gl_texture_object *texObj = texUnit1->_Current;
+
+	    vmesa->regHTXnTB[1] = get_minmag_filter( texObj->MinFilter,
+						    texObj->MagFilter );
+	    vmesa->regHTXnMPMD[1] &= ~(HC_HTXnMPMD_SMASK | HC_HTXnMPMD_TMASK);
+	    vmesa->regHTXnMPMD[1] |= get_wrap_mode( texObj->WrapS,
+						   texObj->WrapT );
+
+	    vmesa->regHTXnTB[1] &= ~(HC_HTXnTB_TBC_S | HC_HTXnTB_TBC_T);
+            if (texObj->Image[0][texObj->BaseLevel]->Border > 0) {
+	       vmesa->regHTXnTB[1] |= (HC_HTXnTB_TBC_S | HC_HTXnTB_TBC_T);
+	       vmesa->regHTXnTBC[1] = 
+		  PACK_COLOR_888(FLOAT_TO_UBYTE(texObj->BorderColor[0]),
+				 FLOAT_TO_UBYTE(texObj->BorderColor[1]),
+				 FLOAT_TO_UBYTE(texObj->BorderColor[2]));
+	       vmesa->regHTXnTRAH[1] = FLOAT_TO_UBYTE(texObj->BorderColor[3]);
+            }
+
+
+	    if (texUnit1->LodBias != 0.0f) {
+	       GLuint b = viaComputeLodBias(texUnit1->LodBias);
+	       vmesa->regHTXnTB[1] &= ~HC_HTXnFLDs_MASK;
+	       vmesa->regHTXnTB[1] |= HC_HTXnFLDs_ConstLOD;
+	       vmesa->regHTXnCLOD[1] = (b&0x1f) | (((~b)&0x1f)<<10); /* FIXME */
+	    }
+
+	    if (!viaTexCombineState( vmesa, texUnit1->_CurrentCombine, 1 )) {
+	       if (VIA_DEBUG & DEBUG_TEXTURE)
+		  fprintf(stderr, "viaTexCombineState failed for unit 1\n");
+	       return GL_FALSE;
+	    }
+        }
+    }
+    else {
+        vmesa->regEnable &= ~(HC_HenTXMP_MASK | HC_HenTXCH_MASK | 
+			      HC_HenTXPP_MASK);
+    }
+    
+    return GL_TRUE;
+}
+
+static void viaChooseColorState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    GLenum s = ctx->Color.BlendSrcRGB;
+    GLenum d = ctx->Color.BlendDstRGB;
+
+    /* The HW's blending equation is:
+     * (Ca * FCa + Cbias + Cb * FCb) << Cshift
+     */
+
+    if (ctx->Color.BlendEnabled) {
+        vmesa->regEnable |= HC_HenABL_MASK;
+        /* Ca  -- always from source color.
+         */
+        vmesa->regHABLCsat = HC_HABLCsat_MASK | HC_HABLCa_OPC | HC_HABLCa_Csrc;
+        /* Aa  -- always from source alpha.
+         */
+        vmesa->regHABLAsat = HC_HABLAsat_MASK | HC_HABLAa_OPA | HC_HABLAa_Asrc;
+        /* FCa -- depend on following condition.
+         * FAa -- depend on following condition.
+         */
+        switch (s) {
+        case GL_ZERO:
+            /* (0, 0, 0, 0)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_HABLRCa;
+            vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_HABLFRA;
+            vmesa->regHABLRFCa = 0x0;
+            vmesa->regHABLRAa = 0x0;
+            break;
+        case GL_ONE:
+            /* (1, 1, 1, 1)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_InvOPC | HC_HABLFCa_HABLRCa;
+            vmesa->regHABLAsat |= HC_HABLFAa_InvOPA | HC_HABLFAa_HABLFRA;
+            vmesa->regHABLRFCa = 0x0;
+            vmesa->regHABLRAa = 0x0;
+            break;
+        case GL_SRC_COLOR:
+            /* (Rs, Gs, Bs, As)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_Csrc;
+            vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_Asrc;
+            break;
+        case GL_ONE_MINUS_SRC_COLOR:
+            /* (1, 1, 1, 1) - (Rs, Gs, Bs, As)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_InvOPC | HC_HABLFCa_Csrc;
+            vmesa->regHABLAsat |= HC_HABLFAa_InvOPA | HC_HABLFAa_Asrc;
+            break;
+        case GL_DST_COLOR:
+            /* (Rd, Gd, Bd, Ad)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_Cdst;
+            vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_Adst;
+            break;
+        case GL_ONE_MINUS_DST_COLOR:
+            /* (1, 1, 1, 1) - (Rd, Gd, Bd, Ad)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_InvOPC | HC_HABLFCa_Cdst;
+            vmesa->regHABLAsat |= HC_HABLFAa_InvOPA | HC_HABLFAa_Adst;
+            break;
+        case GL_SRC_ALPHA:
+            /* (As, As, As, As)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_Asrc;
+            vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_Asrc;
+            break;
+        case GL_ONE_MINUS_SRC_ALPHA:
+            /* (1, 1, 1, 1) - (As, As, As, As)
+             */
+            vmesa->regHABLCsat |= HC_HABLFCa_InvOPC | HC_HABLFCa_Asrc;
+            vmesa->regHABLAsat |= HC_HABLFAa_InvOPA | HC_HABLFAa_Asrc;
+            break;
+        case GL_DST_ALPHA:
+            {
+                if (vmesa->viaScreen->bitsPerPixel == 16) {
+                    /* (1, 1, 1, 1)
+                     */
+                    vmesa->regHABLCsat |= (HC_HABLFCa_InvOPC | 
+					   HC_HABLFCa_HABLRCa);
+                    vmesa->regHABLAsat |= (HC_HABLFAa_InvOPA | 
+					   HC_HABLFAa_HABLFRA);
+                    vmesa->regHABLRFCa = 0x0;
+                    vmesa->regHABLRAa = 0x0;
+                }
+                else {
+                    /* (Ad, Ad, Ad, Ad)
+                     */
+                    vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_Adst;
+                    vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_Adst;
+                }
+            }
+            break;
+        case GL_ONE_MINUS_DST_ALPHA:
+            {
+                if (vmesa->viaScreen->bitsPerPixel == 16) {
+                    /* (1, 1, 1, 1) - (1, 1, 1, 1) = (0, 0, 0, 0)
+                     */
+                    vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_HABLRCa;
+                    vmesa->regHABLAsat |= HC_HABLFAa_OPA | HC_HABLFAa_HABLFRA;
+                    vmesa->regHABLRFCa = 0x0;
+                    vmesa->regHABLRAa = 0x0;
+                }
+                else {
+                    /* (1, 1, 1, 1) - (Ad, Ad, Ad, Ad)
+                     */
+                    vmesa->regHABLCsat |= HC_HABLFCa_InvOPC | HC_HABLFCa_Adst;
+                    vmesa->regHABLAsat |= HC_HABLFAa_InvOPA | HC_HABLFAa_Adst;
+                }
+            }
+            break;
+        case GL_SRC_ALPHA_SATURATE:
+            {
+                if (vmesa->viaScreen->bitsPerPixel == 16) {
+                    /* (f, f, f, 1), f = min(As, 1 - Ad) = min(As, 1 - 1) = 0
+                     * So (f, f, f, 1) = (0, 0, 0, 1)
+                     */
+                    vmesa->regHABLCsat |= HC_HABLFCa_OPC | HC_HABLFCa_HABLRCa;
+                    vmesa->regHABLAsat |= (HC_HABLFAa_InvOPA | 
+					   HC_HABLFAa_HABLFRA);
+                    vmesa->regHABLRFCa = 0x0;
+                    vmesa->regHABLRAa = 0x0;
+                }
+                else {
+                    /* (f, f, f, 1), f = min(As, 1 - Ad)
+                     */
+                    vmesa->regHABLCsat |= (HC_HABLFCa_OPC | 
+					   HC_HABLFCa_mimAsrcInvAdst);
+                    vmesa->regHABLAsat |= (HC_HABLFAa_InvOPA | 
+					   HC_HABLFAa_HABLFRA);
+                    vmesa->regHABLRFCa = 0x0;
+                    vmesa->regHABLRAa = 0x0;
+                }
+            }
+            break;
+        }
+
+        /* Op is add.
+         */
+
+        /* bias is 0.
+         */
+        vmesa->regHABLCsat |= HC_HABLCbias_HABLRCbias;
+        vmesa->regHABLAsat |= HC_HABLAbias_HABLRAbias;
+
+        /* Cb  -- always from destination color.
+         */
+        vmesa->regHABLCop = HC_HABLCb_OPC | HC_HABLCb_Cdst;
+        /* Ab  -- always from destination alpha.
+         */
+        vmesa->regHABLAop = HC_HABLAb_OPA | HC_HABLAb_Adst;
+        /* FCb -- depend on following condition.
+         */
+        switch (d) {
+        case GL_ZERO:
+            /* (0, 0, 0, 0)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_HABLRCb;
+            vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_HABLFRA;
+            vmesa->regHABLRFCb = 0x0;
+            vmesa->regHABLRAb = 0x0;
+            break;
+        case GL_ONE:
+            /* (1, 1, 1, 1)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_HABLRCb;
+            vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_HABLFRA;
+            vmesa->regHABLRFCb = 0x0;
+            vmesa->regHABLRAb = 0x0;
+            break;
+        case GL_SRC_COLOR:
+            /* (Rs, Gs, Bs, As)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_Csrc;
+            vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_Asrc;
+            break;
+        case GL_ONE_MINUS_SRC_COLOR:
+            /* (1, 1, 1, 1) - (Rs, Gs, Bs, As)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_Csrc;
+            vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_Asrc;
+            break;
+        case GL_DST_COLOR:
+            /* (Rd, Gd, Bd, Ad)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_Cdst;
+            vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_Adst;
+            break;
+        case GL_ONE_MINUS_DST_COLOR:
+            /* (1, 1, 1, 1) - (Rd, Gd, Bd, Ad)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_Cdst;
+            vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_Adst;
+            break;
+        case GL_SRC_ALPHA:
+            /* (As, As, As, As)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_Asrc;
+            vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_Asrc;
+            break;
+        case GL_ONE_MINUS_SRC_ALPHA:
+            /* (1, 1, 1, 1) - (As, As, As, As)
+             */
+            vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_Asrc;
+            vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_Asrc;
+            break;
+        case GL_DST_ALPHA:
+            {
+                if (vmesa->viaScreen->bitsPerPixel == 16) {
+                    /* (1, 1, 1, 1)
+                     */
+                    vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_HABLRCb;
+                    vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_HABLFRA;
+                    vmesa->regHABLRFCb = 0x0;
+                    vmesa->regHABLRAb = 0x0;
+                }
+                else {
+                    /* (Ad, Ad, Ad, Ad)
+                     */
+                    vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_Adst;
+                    vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_Adst;
+                }
+            }
+            break;
+        case GL_ONE_MINUS_DST_ALPHA:
+            {
+                if (vmesa->viaScreen->bitsPerPixel == 16) {
+                    /* (1, 1, 1, 1) - (1, 1, 1, 1) = (0, 0, 0, 0)
+                     */
+                    vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_HABLRCb;
+                    vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_HABLFRA;
+                    vmesa->regHABLRFCb = 0x0;
+                    vmesa->regHABLRAb = 0x0;
+                }
+                else {
+                    /* (1, 1, 1, 1) - (Ad, Ad, Ad, Ad)
+                     */
+                    vmesa->regHABLCop |= HC_HABLFCb_InvOPC | HC_HABLFCb_Adst;
+                    vmesa->regHABLAop |= HC_HABLFAb_InvOPA | HC_HABLFAb_Adst;
+                }
+            }
+            break;
+        default:
+            vmesa->regHABLCop |= HC_HABLFCb_OPC | HC_HABLFCb_HABLRCb;
+            vmesa->regHABLAop |= HC_HABLFAb_OPA | HC_HABLFAb_HABLFRA;
+            vmesa->regHABLRFCb = 0x0;
+            vmesa->regHABLRAb = 0x0;
+            break;
+        }
+
+        if (vmesa->viaScreen->bitsPerPixel <= 16)
+            vmesa->regEnable &= ~HC_HenDT_MASK;
+
+    }
+    else {
+        vmesa->regEnable &= (~HC_HenABL_MASK);
+    }
+
+    if (ctx->Color.AlphaEnabled) {
+        vmesa->regEnable |= HC_HenAT_MASK;
+        vmesa->regHATMD = FLOAT_TO_UBYTE(ctx->Color.AlphaRef) |
+            ((ctx->Color.AlphaFunc - GL_NEVER) << 8);
+    }
+    else {
+        vmesa->regEnable &= (~HC_HenAT_MASK);
+    }
+
+    if (ctx->Color.DitherFlag && (vmesa->viaScreen->bitsPerPixel < 32)) {
+        if (ctx->Color.BlendEnabled) {
+            vmesa->regEnable &= ~HC_HenDT_MASK;
+        }
+        else {
+            vmesa->regEnable |= HC_HenDT_MASK;
+        }
+    }
+
+
+    vmesa->regEnable &= ~HC_HenDT_MASK;
+
+    if (ctx->Color.ColorLogicOpEnabled) 
+        vmesa->regHROP = ROP[ctx->Color.LogicOp & 0xF];
+    else
+        vmesa->regHROP = HC_HROP_P;
+
+    vmesa->regHFBBMSKL = PACK_COLOR_888(ctx->Color.ColorMask[0],
+					ctx->Color.ColorMask[1],
+					ctx->Color.ColorMask[2]);
+    vmesa->regHROP |= ctx->Color.ColorMask[3];
+
+    if (ctx->Color.ColorMask[3])
+        vmesa->regEnable |= HC_HenAW_MASK;
+    else
+        vmesa->regEnable &= ~HC_HenAW_MASK;
+}
+
+static void viaChooseFogState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    if (ctx->Fog.Enabled) {
+        GLubyte r, g, b, a;
+
+        vmesa->regEnable |= HC_HenFOG_MASK;
+
+        /* Use fog equation 0 (OpenGL's default) & local fog.
+         */
+        vmesa->regHFogLF = 0x0;
+
+        r = (GLubyte)(ctx->Fog.Color[0] * 255.0F);
+        g = (GLubyte)(ctx->Fog.Color[1] * 255.0F);
+        b = (GLubyte)(ctx->Fog.Color[2] * 255.0F);
+        a = (GLubyte)(ctx->Fog.Color[3] * 255.0F);
+        vmesa->regHFogCL = (r << 16) | (g << 8) | b;
+        vmesa->regHFogCH = a;
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenFOG_MASK;
+    }
+}
+
+static void viaChooseDepthState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    if (ctx->Depth.Test) {
+        vmesa->regEnable |= HC_HenZT_MASK;
+        if (ctx->Depth.Mask)
+            vmesa->regEnable |= HC_HenZW_MASK;
+        else
+            vmesa->regEnable &= (~HC_HenZW_MASK);
+	vmesa->regHZWTMD = (ctx->Depth.Func - GL_NEVER) << 16;
+	
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenZT_MASK;
+        
+        /*=* [DBG] racer : can't display cars in car selection menu *=*/
+	/*if (ctx->Depth.Mask)
+            vmesa->regEnable |= HC_HenZW_MASK;
+        else
+            vmesa->regEnable &= (~HC_HenZW_MASK);*/
+	vmesa->regEnable &= (~HC_HenZW_MASK);
+    }
+}
+
+static void viaChooseLineState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    if (ctx->Line.StippleFlag) {
+        vmesa->regEnable |= HC_HenLP_MASK;
+        vmesa->regHLP = ctx->Line.StipplePattern;
+        vmesa->regHLPRF = ctx->Line.StippleFactor;
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenLP_MASK;
+    }
+}
+
+static void viaChoosePolygonState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+#if 0
+    /* Polygon stipple is broken - see via_state.c
+     */
+    if (ctx->Polygon.StippleFlag) {
+        vmesa->regEnable |= HC_HenSP_MASK;
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenSP_MASK;
+    }
+#else
+    FALLBACK(vmesa, VIA_FALLBACK_POLY_STIPPLE, 
+	     ctx->Polygon.StippleFlag);
+#endif
+
+    if (ctx->Polygon.CullFlag) {
+        vmesa->regEnable |= HC_HenFBCull_MASK;
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenFBCull_MASK;
+    }
+}
+
+static void viaChooseStencilState(GLcontext *ctx) 
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    
+    if (ctx->Stencil.Enabled) {
+        GLuint temp;
+
+        vmesa->regEnable |= HC_HenST_MASK;
+        temp = (ctx->Stencil.Ref[0] & 0xFF) << HC_HSTREF_SHIFT;
+        temp |= 0xFF << HC_HSTOPMSK_SHIFT;
+        temp |= (ctx->Stencil.ValueMask[0] & 0xFF);
+        vmesa->regHSTREF = temp;
+
+        temp = (ctx->Stencil.Function[0] - GL_NEVER) << 16;
+
+        switch (ctx->Stencil.FailFunc[0]) {
+        case GL_KEEP:
+            temp |= HC_HSTOPSF_KEEP;
+            break;
+        case GL_ZERO:
+            temp |= HC_HSTOPSF_ZERO;
+            break;
+        case GL_REPLACE:
+            temp |= HC_HSTOPSF_REPLACE;
+            break;
+        case GL_INVERT:
+            temp |= HC_HSTOPSF_INVERT;
+            break;
+        case GL_INCR:
+            temp |= HC_HSTOPSF_INCR;
+            break;
+        case GL_DECR:
+            temp |= HC_HSTOPSF_DECR;
+            break;
+        }
+
+        switch (ctx->Stencil.ZFailFunc[0]) {
+        case GL_KEEP:
+            temp |= HC_HSTOPSPZF_KEEP;
+            break;
+        case GL_ZERO:
+            temp |= HC_HSTOPSPZF_ZERO;
+            break;
+        case GL_REPLACE:
+            temp |= HC_HSTOPSPZF_REPLACE;
+            break;
+        case GL_INVERT:
+            temp |= HC_HSTOPSPZF_INVERT;
+            break;
+        case GL_INCR:
+            temp |= HC_HSTOPSPZF_INCR;
+            break;
+        case GL_DECR:
+            temp |= HC_HSTOPSPZF_DECR;
+            break;
+        }
+
+        switch (ctx->Stencil.ZPassFunc[0]) {
+        case GL_KEEP:
+            temp |= HC_HSTOPSPZP_KEEP;
+            break;
+        case GL_ZERO:
+            temp |= HC_HSTOPSPZP_ZERO;
+            break;
+        case GL_REPLACE:
+            temp |= HC_HSTOPSPZP_REPLACE;
+            break;
+        case GL_INVERT:
+            temp |= HC_HSTOPSPZP_INVERT;
+            break;
+        case GL_INCR:
+            temp |= HC_HSTOPSPZP_INCR;
+            break;
+        case GL_DECR:
+            temp |= HC_HSTOPSPZP_DECR;
+            break;
+        }
+        vmesa->regHSTMD = temp;
+    }
+    else {
+        vmesa->regEnable &= ~HC_HenST_MASK;
+    }
+}
+
+
+
+static void viaChooseTriangle(GLcontext *ctx) 
+{       
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    if (ctx->Polygon.CullFlag == GL_TRUE) {
+        switch (ctx->Polygon.CullFaceMode) {
+        case GL_FRONT:
+            if (ctx->Polygon.FrontFace == GL_CCW)
+                vmesa->regCmdB |= HC_HBFace_MASK;
+            else
+                vmesa->regCmdB &= ~HC_HBFace_MASK;
+            break;
+        case GL_BACK:
+            if (ctx->Polygon.FrontFace == GL_CW)
+                vmesa->regCmdB |= HC_HBFace_MASK;
+            else
+                vmesa->regCmdB &= ~HC_HBFace_MASK;
+            break;
+        case GL_FRONT_AND_BACK:
+            return;
+        }
+    }
+}
+
+void viaValidateState( GLcontext *ctx )
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    if (vmesa->newState & _NEW_TEXTURE) {
+       GLboolean ok = (viaChooseTextureState(ctx) &&
+		       viaUpdateTextureState(ctx));
+
+       FALLBACK(vmesa, VIA_FALLBACK_TEXTURE, !ok);
+    }
+
+    if (vmesa->newState & _NEW_COLOR)
+        viaChooseColorState(ctx);
+
+    if (vmesa->newState & _NEW_DEPTH)
+        viaChooseDepthState(ctx);
+
+    if (vmesa->newState & _NEW_FOG)
+        viaChooseFogState(ctx);
+
+    if (vmesa->newState & _NEW_LINE)
+        viaChooseLineState(ctx);
+
+    if (vmesa->newState & (_NEW_POLYGON | _NEW_POLYGONSTIPPLE)) {
+        viaChoosePolygonState(ctx);
+	viaChooseTriangle(ctx);
+    }
+
+    if ((vmesa->newState & _NEW_STENCIL) && vmesa->have_hw_stencil)
+        viaChooseStencilState(ctx);
+    
+    if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR)
+        vmesa->regEnable |= HC_HenCS_MASK;
+    else
+        vmesa->regEnable &= ~HC_HenCS_MASK;
+
+    if (ctx->Point.SmoothFlag ||
+	ctx->Line.SmoothFlag ||
+	ctx->Polygon.SmoothFlag)
+        vmesa->regEnable |= HC_HenAA_MASK;
+    else 
+        vmesa->regEnable &= ~HC_HenAA_MASK;
+
+    vmesa->newEmitState |= vmesa->newState;
+    vmesa->newState = 0;
+}
+
+static void viaInvalidateState(GLcontext *ctx, GLuint newState)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+    VIA_FINISH_PRIM( vmesa );
+    vmesa->newState |= newState;
+
+    _swrast_InvalidateState(ctx, newState);
+    _swsetup_InvalidateState(ctx, newState);
+    _vbo_InvalidateState(ctx, newState);
+    _tnl_InvalidateState(ctx, newState);
+}
+
+void viaInitStateFuncs(GLcontext *ctx)
+{
+    /* Callbacks for internal Mesa events.
+     */
+    ctx->Driver.UpdateState = viaInvalidateState;
+
+    /* API callbacks
+     */
+    ctx->Driver.BlendEquationSeparate = viaBlendEquationSeparate;
+    ctx->Driver.BlendFuncSeparate = viaBlendFuncSeparate;
+    ctx->Driver.ClearColor = viaClearColor;
+    ctx->Driver.ColorMask = viaColorMask;
+    ctx->Driver.DrawBuffer = viaDrawBuffer;
+    ctx->Driver.RenderMode = viaRenderMode;
+    ctx->Driver.Scissor = viaScissor;
+    ctx->Driver.DepthRange = viaDepthRange;
+    ctx->Driver.Viewport = viaViewport;
+    ctx->Driver.Enable = viaEnable;
+
+    /* XXX this should go away */
+    ctx->Driver.ResizeBuffers = viaReAllocateBuffers;
+}
diff --git a/src/via_state.h b/src/via_state.h
new file mode 100644
index 0000000..065ec57
--- /dev/null
+++ b/src/via_state.h
@@ -0,0 +1,39 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIA_STATE_H
+#define _VIA_STATE_H
+
+#include "via_context.h"
+
+extern void viaInitState(GLcontext *ctx);
+extern void viaInitStateFuncs(GLcontext *ctx);
+extern void viaCalcViewport(GLcontext *ctx);
+extern void viaValidateState(GLcontext *ctx);
+extern void viaEmitState(struct via_context *vmesa);
+
+extern void viaFallback(struct via_context *vmesa, GLuint bit, GLboolean mode);
+#define FALLBACK(vmesa, bit, mode) viaFallback(vmesa, bit, mode)
+
+#endif
diff --git a/src/via_tex.c b/src/via_tex.c
new file mode 100644
index 0000000..0261a3f
--- /dev/null
+++ b/src/via_tex.c
@@ -0,0 +1,988 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+
+#include <stdlib.h>
+#include <stdio.h>
+
+#include "glheader.h"
+#include "macros.h"
+#include "mtypes.h"
+#include "enums.h"
+#include "colortab.h"
+#include "convolve.h"
+#include "context.h"
+#include "mipmap.h"
+#include "simple_list.h"
+#include "texcompress.h"
+#include "texformat.h"
+#include "texobj.h"
+#include "texstore.h"
+
+#include "mm.h"
+#include "via_context.h"
+#include "via_fb.h"
+#include "via_tex.h"
+#include "via_state.h"
+#include "via_ioctl.h"
+#include "via_3d_reg.h"
+
+static const struct gl_texture_format *
+viaChooseTexFormat( GLcontext *ctx, GLint internalFormat,
+		    GLenum format, GLenum type )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   const GLboolean do32bpt = ( vmesa->viaScreen->bitsPerPixel == 32
+/* 			       && vmesa->viaScreen->textureSize > 4*1024*1024 */
+      );
+
+
+   switch ( internalFormat ) {
+   case 4:
+   case GL_RGBA:
+   case GL_COMPRESSED_RGBA:
+      if ( format == GL_BGRA ) {
+	 if ( type == GL_UNSIGNED_INT_8_8_8_8_REV ||
+	      type == GL_UNSIGNED_BYTE ) {
+	    return &_mesa_texformat_argb8888;
+	 }
+         else if ( type == GL_UNSIGNED_SHORT_4_4_4_4_REV ) {
+            return &_mesa_texformat_argb4444;
+	 }
+         else if ( type == GL_UNSIGNED_SHORT_1_5_5_5_REV ) {
+	    return &_mesa_texformat_argb1555;
+	 }
+      }
+      else if ( type == GL_UNSIGNED_BYTE ||
+		type == GL_UNSIGNED_INT_8_8_8_8_REV ||
+		type == GL_UNSIGNED_INT_8_8_8_8 ) {
+	 return &_mesa_texformat_argb8888;
+      }
+      return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_argb4444;
+
+   case 3:
+   case GL_RGB:
+   case GL_COMPRESSED_RGB:
+      if ( format == GL_RGB && type == GL_UNSIGNED_SHORT_5_6_5 ) {
+	 return &_mesa_texformat_rgb565;
+      }
+      else if ( type == GL_UNSIGNED_BYTE ) {
+	 return &_mesa_texformat_argb8888;
+      }
+      return do32bpt ? &_mesa_texformat_argb8888 : &_mesa_texformat_rgb565;
+
+   case GL_RGBA8:
+   case GL_RGB10_A2:
+   case GL_RGBA12:
+   case GL_RGBA16:
+      return &_mesa_texformat_argb8888;
+
+   case GL_RGBA4:
+   case GL_RGBA2:
+      return &_mesa_texformat_argb4444;
+
+   case GL_RGB5_A1:
+      return &_mesa_texformat_argb1555;
+
+   case GL_RGB8:
+   case GL_RGB10:
+   case GL_RGB12:
+   case GL_RGB16:
+      return &_mesa_texformat_argb8888;
+
+   case GL_RGB5:
+   case GL_RGB4:
+   case GL_R3_G3_B2:
+      return &_mesa_texformat_rgb565;
+
+   case GL_ALPHA:
+   case GL_ALPHA4:
+   case GL_ALPHA8:
+   case GL_ALPHA12:
+   case GL_ALPHA16:
+   case GL_COMPRESSED_ALPHA:
+      return &_mesa_texformat_a8;
+
+   case 1:
+   case GL_LUMINANCE:
+   case GL_LUMINANCE4:
+   case GL_LUMINANCE8:
+   case GL_LUMINANCE12:
+   case GL_LUMINANCE16:
+   case GL_COMPRESSED_LUMINANCE:
+      return &_mesa_texformat_l8;
+
+   case 2:
+   case GL_LUMINANCE_ALPHA:
+   case GL_LUMINANCE4_ALPHA4:
+   case GL_LUMINANCE6_ALPHA2:
+   case GL_LUMINANCE8_ALPHA8:
+   case GL_LUMINANCE12_ALPHA4:
+   case GL_LUMINANCE12_ALPHA12:
+   case GL_LUMINANCE16_ALPHA16:
+   case GL_COMPRESSED_LUMINANCE_ALPHA:
+      return &_mesa_texformat_al88;
+
+   case GL_INTENSITY:
+   case GL_INTENSITY4:
+   case GL_INTENSITY8:
+   case GL_INTENSITY12:
+   case GL_INTENSITY16:
+   case GL_COMPRESSED_INTENSITY:
+      return &_mesa_texformat_i8;
+
+   case GL_YCBCR_MESA:
+      if (type == GL_UNSIGNED_SHORT_8_8_MESA ||
+	  type == GL_UNSIGNED_BYTE)
+         return &_mesa_texformat_ycbcr;
+      else
+         return &_mesa_texformat_ycbcr_rev;
+
+   case GL_COMPRESSED_RGB_FXT1_3DFX:
+      return &_mesa_texformat_rgb_fxt1;
+   case GL_COMPRESSED_RGBA_FXT1_3DFX:
+      return &_mesa_texformat_rgba_fxt1;
+
+   case GL_RGB_S3TC:
+   case GL_RGB4_S3TC:
+   case GL_COMPRESSED_RGB_S3TC_DXT1_EXT:
+      return &_mesa_texformat_rgb_dxt1;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT1_EXT:
+      return &_mesa_texformat_rgba_dxt1;
+
+   case GL_RGBA_S3TC:
+   case GL_RGBA4_S3TC:
+   case GL_COMPRESSED_RGBA_S3TC_DXT3_EXT:
+      return &_mesa_texformat_rgba_dxt3;
+
+   case GL_COMPRESSED_RGBA_S3TC_DXT5_EXT:
+      return &_mesa_texformat_rgba_dxt5;
+
+   case GL_COLOR_INDEX:	
+   case GL_COLOR_INDEX1_EXT:	
+   case GL_COLOR_INDEX2_EXT:	
+   case GL_COLOR_INDEX4_EXT:	
+   case GL_COLOR_INDEX8_EXT:	
+   case GL_COLOR_INDEX12_EXT:	    
+   case GL_COLOR_INDEX16_EXT:
+      return &_mesa_texformat_ci8;    
+
+   default:
+      fprintf(stderr, "unexpected texture format %s in %s\n", 
+	      _mesa_lookup_enum_by_nr(internalFormat),
+	      __FUNCTION__);
+      return NULL;
+   }
+
+   return NULL; /* never get here */
+}
+
+static int logbase2(int n)
+{
+   GLint i = 1;
+   GLint log2 = 0;
+
+   while (n > i) {
+      i *= 2;
+      log2++;
+   }
+
+   return log2;
+}
+
+static const char *get_memtype_name( GLint memType )
+{
+   static const char *names[] = {
+      "VIA_MEM_VIDEO",
+      "VIA_MEM_AGP",
+      "VIA_MEM_SYSTEM",
+      "VIA_MEM_MIXED",
+      "VIA_MEM_UNKNOWN"
+   };
+
+   return names[memType];
+}
+
+
+static GLboolean viaMoveTexBuffers( struct via_context *vmesa,
+				    struct via_tex_buffer **buffers,
+				    GLuint nr,
+				    GLint newMemType )
+{
+   struct via_tex_buffer *newTexBuf[VIA_MAX_TEXLEVELS];
+   GLint i;
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s to %s\n",
+	      __FUNCTION__,
+	      get_memtype_name(newMemType));
+
+   memset(newTexBuf, 0, sizeof(newTexBuf));
+
+   /* First do all the allocations (or fail):
+    */ 
+   for (i = 0; i < nr; i++) {    
+      if (buffers[i]->memType != newMemType) {	 
+
+	 /* Don't allow uploads in a thrash state.  Should try and
+	  * catch this earlier.
+	  */
+	 if (vmesa->thrashing && newMemType != VIA_MEM_SYSTEM)
+	    goto cleanup;
+
+	 newTexBuf[i] = via_alloc_texture(vmesa, 
+					  buffers[i]->size,
+					  newMemType);
+	 if (!newTexBuf[i]) 
+	    goto cleanup;
+      }
+   }
+
+
+   /* Now copy all the image data and free the old texture memory.
+    */
+   for (i = 0; i < nr; i++) {    
+      if (newTexBuf[i]) {
+	 memcpy(newTexBuf[i]->bufAddr,
+		buffers[i]->bufAddr, 
+		buffers[i]->size);
+
+	 newTexBuf[i]->image = buffers[i]->image;
+	 newTexBuf[i]->image->texMem = newTexBuf[i];
+	 newTexBuf[i]->image->image.Data = newTexBuf[i]->bufAddr;
+	 via_free_texture(vmesa, buffers[i]);
+      }
+   }
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s - success\n", __FUNCTION__);
+
+   return GL_TRUE;
+
+ cleanup:
+   /* Release any allocations made prior to failure:
+    */
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s - failed\n", __FUNCTION__);
+
+   for (i = 0; i < nr; i++) {    
+      if (newTexBuf[i]) {
+	 via_free_texture(vmesa, newTexBuf[i]);
+      }
+   }
+   
+   return GL_FALSE;   
+}
+
+
+static GLboolean viaMoveTexObject( struct via_context *vmesa,
+				   struct via_texture_object *viaObj,
+				   GLint newMemType )
+{   
+   struct via_texture_image **viaImage = 
+      (struct via_texture_image **)&viaObj->obj.Image[0][0];
+   struct via_tex_buffer *buffers[VIA_MAX_TEXLEVELS];
+   GLuint i, nr = 0;
+
+   for (i = viaObj->firstLevel; i <= viaObj->lastLevel; i++)
+      buffers[nr++] = viaImage[i]->texMem;
+
+   if (viaMoveTexBuffers( vmesa, &buffers[0], nr, newMemType )) {
+      viaObj->memType = newMemType;
+      return GL_TRUE;
+   }
+
+   return GL_FALSE;
+}
+
+
+
+static GLboolean viaSwapInTexObject( struct via_context *vmesa,
+				     struct via_texture_object *viaObj )
+{
+   const struct via_texture_image *baseImage = 
+      (struct via_texture_image *)viaObj->obj.Image[0][viaObj->obj.BaseLevel]; 
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (baseImage->texMem->memType != VIA_MEM_SYSTEM) 
+      return viaMoveTexObject( vmesa, viaObj, baseImage->texMem->memType );
+
+   return (viaMoveTexObject( vmesa, viaObj, VIA_MEM_AGP ) ||
+	   viaMoveTexObject( vmesa, viaObj, VIA_MEM_VIDEO ));
+}
+
+
+/* This seems crude, but it asks a fairly pertinent question and gives
+ * an accurate answer:
+ */
+static GLboolean viaIsTexMemLow( struct via_context *vmesa,
+				 GLuint heap )
+{
+   struct via_tex_buffer *buf =  via_alloc_texture(vmesa, 512 * 1024, heap );
+   if (!buf)
+      return GL_TRUE;
+   
+   via_free_texture(vmesa, buf);
+   return GL_FALSE;
+}
+
+
+/* Speculatively move texture images which haven't been used in a
+ * while back to system memory. 
+ * 
+ * TODO: only do this when texture memory is low.
+ * 
+ * TODO: use dma.
+ *
+ * TODO: keep the fb/agp version hanging around and use the local
+ * version as backing store, so re-upload might be avoided.
+ *
+ * TODO: do this properly in the kernel...
+ */
+GLboolean viaSwapOutWork( struct via_context *vmesa )
+{
+   struct via_tex_buffer *s, *tmp;
+   GLuint done = 0;
+   GLuint heap, target;
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s VID %d AGP %d SYS %d\n", __FUNCTION__,
+	      vmesa->total_alloc[VIA_MEM_VIDEO],
+	      vmesa->total_alloc[VIA_MEM_AGP],
+	      vmesa->total_alloc[VIA_MEM_SYSTEM]);
+
+   
+   for (heap = VIA_MEM_VIDEO; heap <= VIA_MEM_AGP; heap++) {
+      GLuint nr = 0, sz = 0;
+
+      if (vmesa->thrashing) {
+ 	 if (VIA_DEBUG & DEBUG_TEXTURE)
+	    fprintf(stderr, "Heap %d: trash flag\n", heap);
+	 target = 1*1024*1024;
+      }
+      else if (viaIsTexMemLow(vmesa, heap)) {
+ 	 if (VIA_DEBUG & DEBUG_TEXTURE)
+	    fprintf(stderr, "Heap %d: low memory\n", heap);
+	 target = 64*1024;
+      }
+      else {
+ 	 if (VIA_DEBUG & DEBUG_TEXTURE)
+	    fprintf(stderr, "Heap %d: nothing to do\n", heap);
+	 continue;
+      }
+
+      foreach_s( s, tmp, &vmesa->tex_image_list[heap] ) {
+	 if (s->lastUsed < vmesa->lastSwap[1]) {
+	    struct via_texture_object *viaObj = 
+	       (struct via_texture_object *) s->image->image.TexObject;
+
+	    if (VIA_DEBUG & DEBUG_TEXTURE)
+	       fprintf(stderr, 
+		       "back copy tex sz %d, lastUsed %d lastSwap %d\n", 
+		       s->size, s->lastUsed, vmesa->lastSwap[1]);
+
+	    if (viaMoveTexBuffers( vmesa, &s, 1, VIA_MEM_SYSTEM )) {
+	       viaObj->memType = VIA_MEM_MIXED;
+	       done += s->size;
+	    }
+	    else {
+	       if (VIA_DEBUG & DEBUG_TEXTURE)
+		  fprintf(stderr, "Failed to back copy texture!\n");
+	       sz += s->size;
+	    }
+	 }
+	 else {
+	    nr ++;
+	    sz += s->size;
+	 }
+
+	 if (done > target) {
+	    vmesa->thrashing = GL_FALSE; /* might not get set otherwise? */
+	    return GL_TRUE;
+	 }
+      }
+
+      assert(sz == vmesa->total_alloc[heap]);
+	 
+      if (VIA_DEBUG & DEBUG_TEXTURE)
+	 fprintf(stderr, "Heap %d: nr %d tot sz %d\n", heap, nr, sz);
+   }
+
+   
+   return done != 0;
+}
+
+
+
+/* Basically, just collect the image dimensions and addresses for each
+ * image and update the texture object state accordingly.
+ */
+static GLboolean viaSetTexImages(GLcontext *ctx,
+				 struct gl_texture_object *texObj)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   struct via_texture_object *viaObj = (struct via_texture_object *)texObj;
+   const struct via_texture_image *baseImage = 
+      (struct via_texture_image *)texObj->Image[0][texObj->BaseLevel];
+   GLint firstLevel, lastLevel, numLevels;
+   GLuint texFormat;
+   GLint w, h, p;
+   GLint i, j = 0, k = 0, l = 0, m = 0;
+   GLuint texBase;
+   GLuint basH = 0;
+   GLuint widthExp = 0;
+   GLuint heightExp = 0;    
+
+   switch (baseImage->image.TexFormat->MesaFormat) {
+   case MESA_FORMAT_ARGB8888:
+      texFormat = HC_HTXnFM_ARGB8888;
+      break;
+   case MESA_FORMAT_ARGB4444:
+      texFormat = HC_HTXnFM_ARGB4444; 
+      break;
+   case MESA_FORMAT_RGB565:
+      texFormat = HC_HTXnFM_RGB565;   
+      break;
+   case MESA_FORMAT_ARGB1555:
+      texFormat = HC_HTXnFM_ARGB1555;   
+      break;
+   case MESA_FORMAT_RGB888:
+      texFormat = HC_HTXnFM_ARGB0888;
+      break;
+   case MESA_FORMAT_L8:
+      texFormat = HC_HTXnFM_L8;       
+      break;
+   case MESA_FORMAT_I8:
+      texFormat = HC_HTXnFM_T8;       
+      break;
+   case MESA_FORMAT_CI8:
+      texFormat = HC_HTXnFM_Index8;   
+      break;
+   case MESA_FORMAT_AL88:
+      texFormat = HC_HTXnFM_AL88;     
+      break;
+   case MESA_FORMAT_A8:
+      texFormat = HC_HTXnFM_A8;     
+      break;
+   default:
+      _mesa_problem(vmesa->glCtx, "Bad texture format in viaSetTexImages");
+      return GL_FALSE;
+   }
+
+   /* Compute which mipmap levels we really want to send to the hardware.
+    * This depends on the base image size, GL_TEXTURE_MIN_LOD,
+    * GL_TEXTURE_MAX_LOD, GL_TEXTURE_BASE_LEVEL, and GL_TEXTURE_MAX_LEVEL.
+    * Yes, this looks overly complicated, but it's all needed.
+    */
+   if (texObj->MinFilter == GL_LINEAR || texObj->MinFilter == GL_NEAREST) {
+      firstLevel = lastLevel = texObj->BaseLevel;
+   }
+   else {
+      firstLevel = texObj->BaseLevel + (GLint)(texObj->MinLod + 0.5);
+      firstLevel = MAX2(firstLevel, texObj->BaseLevel);
+      lastLevel = texObj->BaseLevel + (GLint)(texObj->MaxLod + 0.5);
+      lastLevel = MAX2(lastLevel, texObj->BaseLevel);
+      lastLevel = MIN2(lastLevel, texObj->BaseLevel + baseImage->image.MaxLog2);
+      lastLevel = MIN2(lastLevel, texObj->MaxLevel);
+      lastLevel = MAX2(firstLevel, lastLevel);     /* need at least one level */
+   }
+
+   numLevels = lastLevel - firstLevel + 1;
+
+   /* The hardware supports only 10 mipmap levels; ignore higher levels.
+    */
+   if ((numLevels > 10) && (ctx->Const.MaxTextureLevels > 10)) {
+       lastLevel -= numLevels - 10;
+       numLevels = 10;
+   }
+
+   /* save these values, check if they effect the residency of the
+    * texture:
+    */
+   if (viaObj->firstLevel != firstLevel ||
+       viaObj->lastLevel != lastLevel) {
+      viaObj->firstLevel = firstLevel;
+      viaObj->lastLevel = lastLevel;
+      viaObj->memType = VIA_MEM_MIXED;
+   }
+
+   if (VIA_DEBUG & DEBUG_TEXTURE & 0)
+      fprintf(stderr, "%s, current memType: %s\n",
+	      __FUNCTION__,
+	      get_memtype_name(viaObj->memType));
+
+   
+   if (viaObj->memType == VIA_MEM_MIXED ||
+       viaObj->memType == VIA_MEM_SYSTEM) {
+      if (!viaSwapInTexObject(vmesa, viaObj)) {
+ 	 if (VIA_DEBUG & DEBUG_TEXTURE) 
+	    if (!vmesa->thrashing)
+	       fprintf(stderr, "Thrashing flag set for frame %d\n", 
+		       vmesa->swap_count);
+	 vmesa->thrashing = GL_TRUE;
+	 return GL_FALSE;
+      }
+   }
+
+   if (viaObj->memType == VIA_MEM_AGP)
+      viaObj->regTexFM = (HC_SubA_HTXnFM << 24) | HC_HTXnLoc_AGP | texFormat;
+   else
+      viaObj->regTexFM = (HC_SubA_HTXnFM << 24) | HC_HTXnLoc_Local | texFormat;
+
+
+   for (i = 0; i < numLevels; i++) {    
+      struct via_texture_image *viaImage = 
+	 (struct via_texture_image *)texObj->Image[0][firstLevel + i];
+
+      w = viaImage->image.WidthLog2;
+      h = viaImage->image.HeightLog2;
+      p = viaImage->pitchLog2;
+
+      assert(viaImage->texMem->memType == viaObj->memType);
+
+      texBase = viaImage->texMem->texBase;
+      if (!texBase) {
+	 if (VIA_DEBUG & DEBUG_TEXTURE)
+	    fprintf(stderr, "%s: no texBase[%d]\n", __FUNCTION__, i); 
+	 return GL_FALSE;
+      }
+
+      /* Image has to remain resident until the coming fence is retired.
+       */
+      move_to_head( &vmesa->tex_image_list[viaImage->texMem->memType],
+		    viaImage->texMem );
+      viaImage->texMem->lastUsed = vmesa->lastBreadcrumbWrite;
+
+
+      viaObj->regTexBaseAndPitch[i].baseL = 
+	 ((HC_SubA_HTXnL0BasL + i) << 24) | (texBase & 0xFFFFFF);
+
+      viaObj->regTexBaseAndPitch[i].pitchLog2 = 
+	 ((HC_SubA_HTXnL0Pit + i) << 24) | (p << 20);
+					      
+					      
+      /* The base high bytes for each 3 levels are packed
+       * together into one register:
+       */
+      j = i / 3;
+      k = 3 - (i % 3);
+      basH |= ((texBase & 0xFF000000) >> (k << 3));
+      if (k == 1) {
+	 viaObj->regTexBaseH[j] = ((j + HC_SubA_HTXnL012BasH) << 24) | basH;
+	 basH = 0;
+      }
+            
+      /* Likewise, sets of 6 log2width and log2height values are
+       * packed into individual registers:
+       */
+      l = i / 6;
+      m = i % 6;
+      widthExp |= (((GLuint)w & 0xF) << (m << 2));
+      heightExp |= (((GLuint)h & 0xF) << (m << 2));
+      if (m == 5) {
+	 viaObj->regTexWidthLog2[l] = 
+	    (l + HC_SubA_HTXnL0_5WE) << 24 | widthExp;
+	 viaObj->regTexHeightLog2[l] = 
+	    (l + HC_SubA_HTXnL0_5HE) << 24 | heightExp;
+	 widthExp = 0;
+	 heightExp = 0;
+      }
+      if (w) w--;
+      if (h) h--;
+      if (p) p--;                                           
+   }
+        
+   if (k != 1) {
+      viaObj->regTexBaseH[j] = ((j + HC_SubA_HTXnL012BasH) << 24) | basH;      
+   }
+   if (m != 5) {
+      viaObj->regTexWidthLog2[l] = (l + HC_SubA_HTXnL0_5WE) << 24 | widthExp;
+      viaObj->regTexHeightLog2[l] = (l + HC_SubA_HTXnL0_5HE) << 24 | heightExp;
+   }
+
+   return GL_TRUE;
+}
+
+
+GLboolean viaUpdateTextureState( GLcontext *ctx )
+{
+   struct gl_texture_unit *texUnit = ctx->Texture.Unit;
+   GLuint i;
+
+   for (i = 0; i < 2; i++) {   
+      if (texUnit[i]._ReallyEnabled == TEXTURE_2D_BIT || 
+	  texUnit[i]._ReallyEnabled == TEXTURE_1D_BIT) {
+
+	 if (!viaSetTexImages(ctx, texUnit[i]._Current)) 
+	    return GL_FALSE;
+      }
+      else if (texUnit[i]._ReallyEnabled) {
+	 return GL_FALSE;
+      } 
+   }
+   
+   return GL_TRUE;
+}
+
+
+
+
+
+				 
+
+
+static void viaTexImage(GLcontext *ctx, 
+			GLint dims,
+			GLenum target, GLint level,
+			GLint internalFormat,
+			GLint width, GLint height, GLint border,
+			GLenum format, GLenum type, const void *pixels,
+			const struct gl_pixelstore_attrib *packing,
+			struct gl_texture_object *texObj,
+			struct gl_texture_image *texImage)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   GLint postConvWidth = width;
+   GLint postConvHeight = height;
+   GLint texelBytes, sizeInBytes;
+   struct via_texture_object *viaObj = (struct via_texture_object *)texObj;
+   struct via_texture_image *viaImage = (struct via_texture_image *)texImage;
+   int heaps[3], nheaps, i;
+
+   if (!is_empty_list(&vmesa->freed_tex_buffers)) {
+      viaCheckBreadcrumb(vmesa, 0);
+      via_release_pending_textures(vmesa);
+   }
+
+   if (ctx->_ImageTransferState & IMAGE_CONVOLUTION_BIT) {
+      _mesa_adjust_image_for_convolution(ctx, dims, &postConvWidth,
+                                         &postConvHeight);
+   }
+
+   /* choose the texture format */
+   texImage->TexFormat = viaChooseTexFormat(ctx, internalFormat, 
+					    format, type);
+
+   assert(texImage->TexFormat);
+
+   if (dims == 1) {
+      texImage->FetchTexelc = texImage->TexFormat->FetchTexel1D;
+      texImage->FetchTexelf = texImage->TexFormat->FetchTexel1Df;
+   }
+   else {
+      texImage->FetchTexelc = texImage->TexFormat->FetchTexel2D;
+      texImage->FetchTexelf = texImage->TexFormat->FetchTexel2Df;
+   }
+   texelBytes = texImage->TexFormat->TexelBytes;
+
+   if (texelBytes == 0) {
+      /* compressed format */
+      texImage->IsCompressed = GL_TRUE;
+      texImage->CompressedSize =
+         ctx->Driver.CompressedTextureSize(ctx, texImage->Width,
+                                           texImage->Height, texImage->Depth,
+                                           texImage->TexFormat->MesaFormat);
+   }
+
+   /* Minimum pitch of 32 bytes */
+   if (postConvWidth * texelBytes < 32) {
+      postConvWidth = 32 / texelBytes;
+      texImage->RowStride = postConvWidth;
+   }
+
+   assert(texImage->RowStride == postConvWidth);
+   viaImage->pitchLog2 = logbase2(postConvWidth * texelBytes);
+
+   /* allocate memory */
+   if (texImage->IsCompressed)
+      sizeInBytes = texImage->CompressedSize;
+   else
+      sizeInBytes = postConvWidth * postConvHeight * texelBytes;
+
+
+   /* Attempt to allocate texture memory directly, otherwise use main
+    * memory and this texture will always be a fallback.   FIXME!
+    *
+    * TODO: make room in agp if this fails.
+    * TODO: use fb ram for textures as well.
+    */
+   
+      
+   switch (viaObj->memType) {
+   case VIA_MEM_UNKNOWN:
+      heaps[0] = VIA_MEM_AGP;
+      heaps[1] = VIA_MEM_VIDEO;
+      heaps[2] = VIA_MEM_SYSTEM;
+      nheaps = 3;
+      break;
+   case VIA_MEM_AGP:
+   case VIA_MEM_VIDEO:
+      heaps[0] = viaObj->memType;
+      heaps[1] = VIA_MEM_SYSTEM;
+      nheaps = 2;
+      break;
+   case VIA_MEM_MIXED:
+   case VIA_MEM_SYSTEM:
+   default:
+      heaps[0] = VIA_MEM_SYSTEM;
+      nheaps = 1;
+      break;
+   }
+	
+   for (i = 0; i < nheaps && !viaImage->texMem; i++) {
+      if (VIA_DEBUG & DEBUG_TEXTURE) 
+	 fprintf(stderr, "try %s (obj %s)\n", get_memtype_name(heaps[i]),
+		 get_memtype_name(viaObj->memType));
+      viaImage->texMem = via_alloc_texture(vmesa, sizeInBytes, heaps[i]);
+   }
+
+   if (!viaImage->texMem) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+      return;
+   }
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "upload %d bytes to %s\n", sizeInBytes, 
+	      get_memtype_name(viaImage->texMem->memType));
+
+   viaImage->texMem->image = viaImage;
+   texImage->Data = viaImage->texMem->bufAddr;
+
+   if (viaObj->memType == VIA_MEM_UNKNOWN)
+      viaObj->memType = viaImage->texMem->memType;
+   else if (viaObj->memType != viaImage->texMem->memType)
+      viaObj->memType = VIA_MEM_MIXED;
+
+   if (VIA_DEBUG & DEBUG_TEXTURE)
+      fprintf(stderr, "%s, obj %s, image : %s\n",
+	      __FUNCTION__,	      
+	      get_memtype_name(viaObj->memType),
+	      get_memtype_name(viaImage->texMem->memType));
+
+   vmesa->clearTexCache = 1;
+
+   pixels = _mesa_validate_pbo_teximage(ctx, dims, width, height, 1, 
+					format, type,
+					pixels, packing, "glTexImage");
+   if (!pixels) {
+      /* Note: we check for a NULL image pointer here, _after_ we allocated
+       * memory for the texture.  That's what the GL spec calls for.
+       */
+      return;
+   }
+   else {
+      GLint dstRowStride;
+      GLboolean success;
+      if (texImage->IsCompressed) {
+         dstRowStride = _mesa_compressed_row_stride(texImage->TexFormat->MesaFormat, width);
+      }
+      else {
+         dstRowStride = postConvWidth * texImage->TexFormat->TexelBytes;
+      }
+      ASSERT(texImage->TexFormat->StoreImage);
+      success = texImage->TexFormat->StoreImage(ctx, dims,
+                                                texImage->_BaseFormat,
+                                                texImage->TexFormat,
+                                                texImage->Data,
+                                                0, 0, 0,  /* dstX/Y/Zoffset */
+                                                dstRowStride,
+                                                texImage->ImageOffsets,
+                                                width, height, 1,
+                                                format, type, pixels, packing);
+      if (!success) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+      }
+   }
+
+   /* GL_SGIS_generate_mipmap */
+   if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
+      _mesa_generate_mipmap(ctx, target,
+                            &ctx->Texture.Unit[ctx->Texture.CurrentUnit],
+                            texObj);
+   }
+
+   _mesa_unmap_teximage_pbo(ctx, packing);
+}
+
+static void viaTexImage2D(GLcontext *ctx, 
+			  GLenum target, GLint level,
+			  GLint internalFormat,
+			  GLint width, GLint height, GLint border,
+			  GLenum format, GLenum type, const void *pixels,
+			  const struct gl_pixelstore_attrib *packing,
+			  struct gl_texture_object *texObj,
+			  struct gl_texture_image *texImage)
+{
+   viaTexImage( ctx, 2, target, level, 
+		internalFormat, width, height, border,
+		format, type, pixels,
+		packing, texObj, texImage );
+}
+
+static void viaTexSubImage2D(GLcontext *ctx,
+                             GLenum target,
+                             GLint level,
+                             GLint xoffset, GLint yoffset,
+                             GLsizei width, GLsizei height,
+                             GLenum format, GLenum type,
+                             const GLvoid *pixels,
+                             const struct gl_pixelstore_attrib *packing,
+                             struct gl_texture_object *texObj,
+                             struct gl_texture_image *texImage)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+  
+   viaWaitIdle(vmesa, GL_TRUE);
+   vmesa->clearTexCache = 1;
+
+   _mesa_store_texsubimage2d(ctx, target, level, xoffset, yoffset, width,
+			     height, format, type, pixels, packing, texObj,
+			     texImage);
+}
+
+static void viaTexImage1D(GLcontext *ctx, 
+			  GLenum target, GLint level,
+			  GLint internalFormat,
+			  GLint width, GLint border,
+			  GLenum format, GLenum type, const void *pixels,
+			  const struct gl_pixelstore_attrib *packing,
+			  struct gl_texture_object *texObj,
+			  struct gl_texture_image *texImage)
+{
+   viaTexImage( ctx, 1, target, level, 
+		internalFormat, width, 1, border,
+		format, type, pixels,
+		packing, texObj, texImage );
+}
+
+static void viaTexSubImage1D(GLcontext *ctx,
+                             GLenum target,
+                             GLint level,
+                             GLint xoffset,
+                             GLsizei width,
+                             GLenum format, GLenum type,
+                             const GLvoid *pixels,
+                             const struct gl_pixelstore_attrib *packing,
+                             struct gl_texture_object *texObj,
+                             struct gl_texture_image *texImage)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+   viaWaitIdle(vmesa, GL_TRUE); 
+   vmesa->clearTexCache = 1;
+
+   _mesa_store_texsubimage1d(ctx, target, level, xoffset, width,
+			     format, type, pixels, packing, texObj,
+			     texImage);
+}
+
+
+
+static GLboolean viaIsTextureResident(GLcontext *ctx,
+                                      struct gl_texture_object *texObj)
+{
+   struct via_texture_object *viaObj = 
+      (struct via_texture_object *)texObj;
+
+   return (viaObj->memType == VIA_MEM_AGP ||
+	   viaObj->memType == VIA_MEM_VIDEO);
+}
+
+
+
+static struct gl_texture_image *viaNewTextureImage( GLcontext *ctx )
+{
+   (void) ctx;
+   return (struct gl_texture_image *)CALLOC_STRUCT(via_texture_image);
+}
+
+
+static struct gl_texture_object *viaNewTextureObject( GLcontext *ctx, 
+						      GLuint name, 
+						      GLenum target )
+{
+   struct via_texture_object *obj = CALLOC_STRUCT(via_texture_object);
+
+   _mesa_initialize_texture_object(&obj->obj, name, target);
+   (void) ctx;
+
+   obj->memType = VIA_MEM_UNKNOWN;
+
+   return &obj->obj;
+}
+
+
+static void viaFreeTextureImageData( GLcontext *ctx, 
+				     struct gl_texture_image *texImage )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   struct via_texture_image *image = (struct via_texture_image *)texImage;
+
+   if (image->texMem) {
+      via_free_texture(vmesa, image->texMem);
+      image->texMem = NULL;
+   }
+   
+   texImage->Data = NULL;
+}
+
+
+
+
+void viaInitTextureFuncs(struct dd_function_table * functions)
+{
+   functions->ChooseTextureFormat = viaChooseTexFormat;
+   functions->TexImage1D = viaTexImage1D;
+   functions->TexImage2D = viaTexImage2D;
+   functions->TexSubImage1D = viaTexSubImage1D;
+   functions->TexSubImage2D = viaTexSubImage2D;
+
+   functions->NewTextureObject = viaNewTextureObject;
+   functions->NewTextureImage = viaNewTextureImage;
+   functions->DeleteTexture = _mesa_delete_texture_object;
+   functions->FreeTexImageData = viaFreeTextureImageData;
+
+#if 0 && defined( USE_SSE_ASM )
+   /*
+    * XXX this code is disabled for now because the via_sse_memcpy()
+    * routine causes segfaults with flightgear.
+    * See Mesa3d-dev mail list messages from 7/15/2005 for details.
+    * Note that this function is currently disabled in via_tris.c too.
+    */
+   if (getenv("VIA_NO_SSE"))
+      functions->TextureMemCpy = _mesa_memcpy;
+   else
+      functions->TextureMemCpy = via_sse_memcpy;
+#else
+   functions->TextureMemCpy = _mesa_memcpy;
+#endif
+
+   functions->UpdateTexturePalette = 0;
+   functions->IsTextureResident = viaIsTextureResident;
+}
+
+
diff --git a/src/via_tex.h b/src/via_tex.h
new file mode 100644
index 0000000..73cfa91
--- /dev/null
+++ b/src/via_tex.h
@@ -0,0 +1,41 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+
+#ifndef _VIATEX_H
+#define _VIATEX_H
+
+#include "mtypes.h"
+
+struct via_context;
+
+GLboolean viaUpdateTextureState(GLcontext *ctx);
+void viaInitTextureFuncs(struct dd_function_table * functions);
+GLboolean viaSwapOutWork( struct via_context *vmesa );
+
+#if defined( USE_SSE_ASM )
+void via_sse_memcpy( void *to, const void *from, size_t sz );
+#endif /* defined( USE_SSE_ASM ) */
+
+#endif
diff --git a/src/via_texcombine.c b/src/via_texcombine.c
new file mode 100644
index 0000000..d604457
--- /dev/null
+++ b/src/via_texcombine.c
@@ -0,0 +1,424 @@
+/*
+ * (C) Copyright IBM Corporation 2004
+ * 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
+ * IBM 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.
+ */
+
+/**
+ * \file via_texcombine.c
+ * Calculate texture combine hardware state.
+ *
+ * \author Ian Romanick <idr@us.ibm.com>
+ */
+
+#include <stdio.h>
+
+#include "glheader.h"
+#include "context.h"
+#include "macros.h"
+#include "colormac.h"
+#include "enums.h"
+
+#include "via_context.h"
+#include "via_state.h"
+#include "via_tex.h"
+#include "via_3d_reg.h"
+
+
+#define VIA_USE_ALPHA (HC_XTC_Adif - HC_XTC_Dif)
+
+#define INPUT_A_SHIFT     14
+#define INPUT_B_SHIFT     7
+#define INPUT_C_SHIFT     0
+#define INPUT_CBias_SHIFT 14
+
+#define CONST_ONE         (HC_XTC_0 | HC_XTC_InvTOPC)
+
+static const unsigned color_operand_modifier[4] = {
+   0,
+   HC_XTC_InvTOPC,
+   VIA_USE_ALPHA,
+   VIA_USE_ALPHA | HC_XTC_InvTOPC,
+};
+
+static const unsigned alpha_operand_modifier[2] = {
+   0, HC_XTA_InvTOPA
+};
+
+static const unsigned bias_alpha_operand_modifier[2] = {
+   0, HC_HTXnTBLAbias_Inv
+};
+
+
+static const unsigned c_shift_table[3] = {
+   HC_HTXnTBLCshift_No, HC_HTXnTBLCshift_1, HC_HTXnTBLCshift_2
+};
+
+static const unsigned  a_shift_table[3] = {
+   HC_HTXnTBLAshift_No, HC_HTXnTBLAshift_1, HC_HTXnTBLAshift_2
+};
+
+
+/**
+ * Calculate the hardware state for the specified texture combine mode
+ *
+ * \bug
+ * All forms of DOT3 bumpmapping are completely untested, and are most
+ * likely wrong.  KW: Looks like it will never be quite right as the
+ * hardware seems to experience overflow in color calculation at the
+ * 4x shift levels, which need to be programed for DOT3.  Maybe newer
+ * hardware fixes these issues.
+ *
+ * \bug 
+ * KW: needs attention to the case where texunit 1 is enabled but
+ * texunit 0 is not.
+ */
+GLboolean
+viaTexCombineState( struct via_context *vmesa,
+		    const struct gl_tex_env_combine_state * combine,
+		    unsigned unit )
+{
+   unsigned color_arg[3];
+   unsigned alpha_arg[3];
+   unsigned bias_alpha_arg[3];
+   unsigned color = HC_HTXnTBLCsat_MASK;
+   unsigned alpha = HC_HTXnTBLAsat_MASK;
+   unsigned bias = 0;
+   unsigned op = 0;
+   unsigned a_shift = combine->ScaleShiftA;
+   unsigned c_shift = combine->ScaleShiftRGB;
+   unsigned i;
+   unsigned constant_color[3];
+   unsigned ordered_constant_color[4];
+   unsigned constant_alpha[3];
+   unsigned bias_alpha = 0;
+   unsigned abc_alpha = 0;
+   const struct gl_texture_unit * texUnit = 
+      &vmesa->glCtx->Texture.Unit[unit];
+   unsigned env_color[4];
+
+   /* It seems that the color clamping can be overwhelmed at the 4x
+    * scale settings, necessitating this fallback:
+    */
+   if (c_shift == 2 || a_shift == 2) {
+      return GL_FALSE;
+   }
+
+   CLAMPED_FLOAT_TO_UBYTE(env_color[0], texUnit->EnvColor[0]);
+   CLAMPED_FLOAT_TO_UBYTE(env_color[1], texUnit->EnvColor[1]);
+   CLAMPED_FLOAT_TO_UBYTE(env_color[2], texUnit->EnvColor[2]);
+   CLAMPED_FLOAT_TO_UBYTE(env_color[3], texUnit->EnvColor[3]);
+
+   (void) memset( constant_color, 0, sizeof( constant_color ) );
+   (void) memset( ordered_constant_color, 0, sizeof( ordered_constant_color ) );
+   (void) memset( constant_alpha, 0, sizeof( constant_alpha ) );
+
+   for ( i = 0 ; i < combine->_NumArgsRGB ; i++ ) {
+      const GLint op = combine->OperandRGB[i] - GL_SRC_COLOR;
+
+      switch ( combine->SourceRGB[i] ) {
+      case GL_TEXTURE:
+	 color_arg[i] = HC_XTC_Tex;
+	 color_arg[i] += color_operand_modifier[op];
+	 break;
+      case GL_CONSTANT:
+	 color_arg[i] = HC_XTC_HTXnTBLRC;
+
+	 switch( op ) {
+	 case 0:		/* GL_SRC_COLOR */
+	    constant_color[i] = ((env_color[0] << 16) | 
+				 (env_color[1] << 8) | 
+				 env_color[2]);
+	    break;
+	 case 1:		/* GL_ONE_MINUS_SRC_COLOR */
+	    constant_color[i] = ~((env_color[0] << 16) | 
+				  (env_color[1] << 8) | 
+				  env_color[2]) & 0x00ffffff;
+	    break;
+	 case 2:		/* GL_SRC_ALPHA */
+	    constant_color[i] = ((env_color[3] << 16) | 
+				 (env_color[3] << 8) | 
+				 env_color[3]);
+	    break;
+	 case 3:		/* GL_ONE_MINUS_SRC_ALPHA */
+	    constant_color[i] = ~((env_color[3] << 16) | 
+				  (env_color[3] << 8) | 
+				  env_color[3]) & 0x00ffffff;
+	    break;
+	 }
+	 break;
+      case GL_PRIMARY_COLOR:
+	 color_arg[i] = HC_XTC_Dif;
+	 color_arg[i] += color_operand_modifier[op];
+	 break;
+      case GL_PREVIOUS:
+	 color_arg[i] = (unit == 0) ? HC_XTC_Dif : HC_XTC_Cur;
+	 color_arg[i] += color_operand_modifier[op];
+	 break;
+      }
+   }
+	
+   
+   /* On the Unichrome, all combine operations take on some form of:
+    *
+    *     (xA * (xB op xC) + xBias) << xShift
+    * 
+    * 'op' can be selected as add, subtract, min, max, or mask.  The min, max
+    * and mask modes are currently unused.  With the exception of DOT3, all
+    * standard GL_COMBINE modes can be implemented simply by selecting the
+    * correct inputs for A, B, C, and Bias and the correct operation for op.
+    *
+    * NOTE: xBias (when read from the constant registers) is signed,
+    * and scaled to fit -255..255 in 8 bits, ie 0x1 == 2.
+    */
+
+   switch( combine->ModeRGB ) {
+   /* Ca = 1.0, Cb = arg0, Cc = 0, Cbias = 0
+    */
+   case GL_REPLACE:
+      color |= ((CONST_ONE << INPUT_A_SHIFT) |
+		(color_arg[0] << INPUT_B_SHIFT));
+		
+      ordered_constant_color[1] = constant_color[0];
+      break;
+      
+   /* Ca = arg[0], Cb = arg[1], Cc = 0, Cbias = 0
+    */
+   case GL_MODULATE:
+      color |= ((color_arg[0] << INPUT_A_SHIFT) | 
+		(color_arg[1] << INPUT_B_SHIFT));
+
+      ordered_constant_color[0] = constant_color[0];
+      ordered_constant_color[1] = constant_color[1];
+      break;
+
+   /* Ca = 1.0, Cb = arg[0], Cc = arg[1], Cbias = 0
+    */
+   case GL_ADD:
+   case GL_SUBTRACT:
+      if ( combine->ModeRGB == GL_SUBTRACT ) {
+	 op |= HC_HTXnTBLCop_Sub;
+      }
+
+      color |= ((CONST_ONE << INPUT_A_SHIFT) |
+		(color_arg[0] << INPUT_B_SHIFT) |
+		(color_arg[1] << INPUT_C_SHIFT));
+
+      ordered_constant_color[1] = constant_color[0];
+      ordered_constant_color[2] = constant_color[1];
+      break;
+
+   /* Ca = 1.0, Cb = arg[0], Cc = arg[1], Cbias = -0.5
+    */
+   case GL_ADD_SIGNED:
+      color |= ((CONST_ONE << INPUT_A_SHIFT) |
+		(color_arg[0] << INPUT_B_SHIFT) | 
+		(color_arg[1] << INPUT_C_SHIFT));
+
+      bias |= HC_HTXnTBLCbias_HTXnTBLRC;
+
+      ordered_constant_color[1] = constant_color[0];
+      ordered_constant_color[2] = constant_color[1];
+      ordered_constant_color[3] = 0x00bfbfbf; /* -.5 */
+      break;
+
+   /* Ca = arg[2], Cb = arg[0], Cc = arg[1], Cbias = arg[1]
+    */
+   case GL_INTERPOLATE:
+      op |= HC_HTXnTBLCop_Sub;
+
+      color |= ((color_arg[2] << INPUT_A_SHIFT) |
+		(color_arg[0] << INPUT_B_SHIFT) |
+		(color_arg[1] << INPUT_C_SHIFT));
+
+      bias |= (color_arg[1] << INPUT_CBias_SHIFT);
+
+      ordered_constant_color[0] = constant_color[2];
+      ordered_constant_color[1] = constant_color[0];
+      ordered_constant_color[2] = constant_color[1];
+      ordered_constant_color[3] = (constant_color[1] >> 1) & 0x7f7f7f;
+      break;
+
+#if 0
+   /* At this point this code is completely untested.  It appears that the
+    * Unichrome has the same limitation as the Radeon R100.  The only
+    * supported post-scale when doing DOT3 bumpmapping is 1x.
+    */
+   case GL_DOT3_RGB_EXT:
+   case GL_DOT3_RGBA_EXT:
+   case GL_DOT3_RGB:
+   case GL_DOT3_RGBA:
+      c_shift = 2;
+      a_shift = 2;
+      color |= ((color_arg[0] << INPUT_A_SHIFT) |
+		(color_arg[1] << INPUT_B_SHIFT));
+      op |= HC_HTXnTBLDOT4;
+      break;
+#endif
+
+   default:
+      assert(0);
+      break;
+   }
+
+
+
+
+   /* The alpha blend stage has the annoying quirk of not having a
+    * hard-wired 0 input, like the color stage.  As a result, we have
+    * to program the constant register with 0 and use that as our
+    * 0 input.
+    *
+    *     (xA * (xB op xC) + xBias) << xShift
+    *
+    */
+
+   for ( i = 0 ; i < combine->_NumArgsA ; i++ ) {
+      const GLint op = combine->OperandA[i] - GL_SRC_ALPHA;
+
+      switch ( combine->SourceA[i] ) {
+      case GL_TEXTURE:
+	 alpha_arg[i] = HC_XTA_Atex;
+	 alpha_arg[i] += alpha_operand_modifier[op];
+	 bias_alpha_arg[i] = HC_HTXnTBLAbias_Atex;
+	 bias_alpha_arg[i] += bias_alpha_operand_modifier[op];
+	 break;
+      case GL_CONSTANT:
+	 alpha_arg[i] = HC_XTA_HTXnTBLRA;
+	 bias_alpha_arg[i] = HC_HTXnTBLAbias_HTXnTBLRAbias;
+	 constant_alpha[i] = (op == 0) ? env_color[3] : (~env_color[3] & 0xff);
+	 break;
+      case GL_PRIMARY_COLOR:
+	 alpha_arg[i] = HC_XTA_Adif;
+	 alpha_arg[i] += alpha_operand_modifier[op];
+	 bias_alpha_arg[i] = HC_HTXnTBLAbias_Adif;
+	 bias_alpha_arg[i] += bias_alpha_operand_modifier[op];
+	 break;
+      case GL_PREVIOUS:
+	 alpha_arg[i] = (unit == 0) ? HC_XTA_Adif : HC_XTA_Acur;
+	 alpha_arg[i] += alpha_operand_modifier[op];
+	 bias_alpha_arg[i] = (unit == 0 ? 
+			      HC_HTXnTBLAbias_Adif : 
+			      HC_HTXnTBLAbias_Acur);
+	 bias_alpha_arg[i] += bias_alpha_operand_modifier[op];
+	 break;
+      }
+   }
+
+   switch( combine->ModeA ) {
+   /* Aa = 0, Ab = 0, Ac = 0, Abias = arg0
+    */
+   case GL_REPLACE:
+      alpha |= ((HC_XTA_HTXnTBLRA << INPUT_A_SHIFT) |
+		(HC_XTA_HTXnTBLRA << INPUT_B_SHIFT) |
+		(HC_XTA_HTXnTBLRA << INPUT_C_SHIFT));
+      abc_alpha = 0;
+
+      bias |= bias_alpha_arg[0];
+      bias_alpha = constant_alpha[0] >> 1;
+      break;
+      
+   /* Aa = arg[0], Ab = arg[1], Ac = 0, Abias = 0
+    */
+   case GL_MODULATE:
+      alpha |= ((alpha_arg[1] << INPUT_A_SHIFT) | 
+		(alpha_arg[0] << INPUT_B_SHIFT) | 
+		(HC_XTA_HTXnTBLRA << INPUT_C_SHIFT));
+
+      abc_alpha = ((constant_alpha[1] << HC_HTXnTBLRAa_SHIFT) |
+		   (constant_alpha[0] << HC_HTXnTBLRAb_SHIFT) |
+		   (0 << HC_HTXnTBLRAc_SHIFT));
+
+      bias |= HC_HTXnTBLAbias_HTXnTBLRAbias;
+      bias_alpha = 0;
+      break;
+
+   /* Aa = 1.0, Ab = arg[0], Ac = arg[1], Abias = 0
+    */
+   case GL_ADD:
+   case GL_SUBTRACT:
+      if ( combine->ModeA == GL_SUBTRACT ) {
+	 op |= HC_HTXnTBLAop_Sub;
+      }
+
+      alpha |= ((HC_XTA_HTXnTBLRA << INPUT_A_SHIFT) |
+		(alpha_arg[0] << INPUT_B_SHIFT) |
+		(alpha_arg[1] << INPUT_C_SHIFT));
+
+      abc_alpha = ((0xff << HC_HTXnTBLRAa_SHIFT) |
+		   (constant_alpha[0] << HC_HTXnTBLRAb_SHIFT) |
+		   (constant_alpha[1] << HC_HTXnTBLRAc_SHIFT));
+
+      bias |= HC_HTXnTBLAbias_HTXnTBLRAbias;
+      bias_alpha = 0;
+      break;
+
+   /* Aa = 1.0, Ab = arg[0], Ac = arg[1], Abias = -0.5
+    */
+   case GL_ADD_SIGNED:
+      alpha |= ((HC_XTA_HTXnTBLRA << INPUT_A_SHIFT) |
+		(alpha_arg[0] << INPUT_B_SHIFT) | 
+		(alpha_arg[1] << INPUT_C_SHIFT));
+      abc_alpha = ((0xff << HC_HTXnTBLRAa_SHIFT) |
+		   (constant_alpha[0] << HC_HTXnTBLRAb_SHIFT) |
+		   (constant_alpha[1] << HC_HTXnTBLRAc_SHIFT));
+
+      bias |= HC_HTXnTBLAbias_HTXnTBLRAbias;
+      bias_alpha = 0xbf;
+      break;
+
+   /* Aa = arg[2], Ab = arg[0], Ac = arg[1], Abias = arg[1]
+    */
+   case GL_INTERPOLATE:
+      op |= HC_HTXnTBLAop_Sub;
+
+      alpha |= ((alpha_arg[2] << INPUT_A_SHIFT) |
+		(alpha_arg[0] << INPUT_B_SHIFT) |
+		(alpha_arg[1] << INPUT_C_SHIFT));
+      abc_alpha = ((constant_alpha[2] << HC_HTXnTBLRAa_SHIFT) |
+		   (constant_alpha[0] << HC_HTXnTBLRAb_SHIFT) |
+		   (constant_alpha[1] << HC_HTXnTBLRAc_SHIFT));
+
+      bias |= bias_alpha_arg[1];
+      bias_alpha = constant_alpha[1] >> 1;
+      break;
+   }
+   
+
+   op |= c_shift_table[ c_shift ] | a_shift_table[ a_shift ];
+
+
+   vmesa->regHTXnTBLMPfog[unit] = HC_HTXnTBLMPfog_Fog;
+
+   vmesa->regHTXnTBLCsat[unit] = color;
+   vmesa->regHTXnTBLAsat[unit] = alpha;
+   vmesa->regHTXnTBLCop[unit] = op | bias;
+   vmesa->regHTXnTBLRAa[unit] = abc_alpha;
+   vmesa->regHTXnTBLRFog[unit] = bias_alpha;
+
+   vmesa->regHTXnTBLRCa[unit] = ordered_constant_color[0];
+   vmesa->regHTXnTBLRCb[unit] = ordered_constant_color[1];
+   vmesa->regHTXnTBLRCc[unit] = ordered_constant_color[2];
+   vmesa->regHTXnTBLRCbias[unit] = ordered_constant_color[3];
+
+   return GL_TRUE;
+}
+
diff --git a/src/via_tris.c b/src/via_tris.c
new file mode 100644
index 0000000..b97dacd
--- /dev/null
+++ b/src/via_tris.c
@@ -0,0 +1,1194 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#include <stdio.h>
+#include <math.h>
+
+#include "glheader.h"
+#include "context.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "colormac.h"
+#include "enums.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/t_context.h"
+#include "tnl/t_pipeline.h"
+
+#include "via_context.h"
+#include "via_tris.h"
+#include "via_state.h"
+#include "via_span.h"
+#include "via_ioctl.h"
+#include "via_3d_reg.h"
+#include "via_tex.h"
+
+/***********************************************************************
+ *                    Emit primitives as inline vertices               *
+ ***********************************************************************/
+#define LINE_FALLBACK (0)
+#define POINT_FALLBACK (0)
+#define TRI_FALLBACK (0)
+#define ANY_FALLBACK_FLAGS (POINT_FALLBACK|LINE_FALLBACK|TRI_FALLBACK)
+#define ANY_RASTER_FLAGS (DD_TRI_LIGHT_TWOSIDE|DD_TRI_OFFSET|DD_TRI_UNFILLED)
+
+
+#if 0
+#define COPY_DWORDS(vb, vertsize, v) 		\
+do {						\
+   via_sse_memcpy(vb, v, vertsize * 4);		\
+   vb += vertsize;				\
+} while (0)
+#else
+#if defined( USE_X86_ASM )
+#define COPY_DWORDS(vb, vertsize, v)					\
+    do {								\
+        int j;								\
+        int __tmp;							\
+        __asm__ __volatile__("rep ; movsl"				\
+                              : "=%c" (j), "=D" (vb), "=S" (__tmp)	\
+                              : "0" (vertsize),				\
+                                "D" ((long)vb),				\
+                                "S" ((long)v));				\
+    } while (0)
+#else
+#define COPY_DWORDS(vb, vertsize, v)		\
+    do {					\
+        int j;					\
+        for (j = 0; j < vertsize; j++)		\
+            vb[j] = ((GLuint *)v)[j];		\
+        vb += vertsize;				\
+    } while (0)
+#endif
+#endif
+
+static void via_draw_triangle(struct via_context *vmesa,
+			      viaVertexPtr v0,
+			      viaVertexPtr v1,
+			      viaVertexPtr v2)
+{
+   GLuint vertsize = vmesa->vertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 3 * 4 * vertsize);
+
+   COPY_DWORDS(vb, vertsize, v0);
+   COPY_DWORDS(vb, vertsize, v1);
+   COPY_DWORDS(vb, vertsize, v2);
+}
+
+
+static void via_draw_quad(struct via_context *vmesa,
+			  viaVertexPtr v0,
+			  viaVertexPtr v1,
+			  viaVertexPtr v2,
+			  viaVertexPtr v3)
+{
+   GLuint vertsize = vmesa->vertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 6 * 4 * vertsize);
+
+   COPY_DWORDS(vb, vertsize, v0);
+   COPY_DWORDS(vb, vertsize, v1);
+   COPY_DWORDS(vb, vertsize, v3);
+   COPY_DWORDS(vb, vertsize, v1);
+   COPY_DWORDS(vb, vertsize, v2);
+   COPY_DWORDS(vb, vertsize, v3);
+}
+
+static void via_draw_line(struct via_context *vmesa,
+			  viaVertexPtr v0,
+			  viaVertexPtr v1)
+{
+   GLuint vertsize = vmesa->vertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 2 * 4 * vertsize);
+   COPY_DWORDS(vb, vertsize, v0);
+   COPY_DWORDS(vb, vertsize, v1);
+}
+
+
+static void via_draw_point(struct via_context *vmesa,
+			   viaVertexPtr v0)
+{
+   GLuint vertsize = vmesa->vertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 4 * vertsize);
+   COPY_DWORDS(vb, vertsize, v0);
+}
+
+
+/* Fallback drawing functions for the ptex hack.
+ */
+#define PTEX_VERTEX( tmp, vertex_size, v)	\
+do {							\
+   GLuint j;						\
+   GLfloat rhw = 1.0 / v->f[vertex_size];		\
+   for ( j = 0 ; j < vertex_size ; j++ )		\
+      tmp.f[j] = v->f[j];				\
+   tmp.f[3] *= v->f[vertex_size];			\
+   tmp.f[vertex_size-2] *= rhw;				\
+   tmp.f[vertex_size-1] *= rhw;				\
+} while (0)
+
+static void via_ptex_tri (struct via_context *vmesa,
+			  viaVertexPtr v0,
+			  viaVertexPtr v1,
+			  viaVertexPtr v2)
+{
+   GLuint vertsize = vmesa->hwVertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 3*4*vertsize);
+   viaVertex tmp;
+
+   PTEX_VERTEX(tmp, vertsize, v0); COPY_DWORDS(vb, vertsize, &tmp);
+   PTEX_VERTEX(tmp, vertsize, v1); COPY_DWORDS(vb, vertsize, &tmp);
+   PTEX_VERTEX(tmp, vertsize, v2); COPY_DWORDS(vb, vertsize, &tmp);
+}
+
+static void via_ptex_line (struct via_context *vmesa,
+			   viaVertexPtr v0,
+			   viaVertexPtr v1)
+{
+   GLuint vertsize = vmesa->hwVertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 2*4*vertsize);
+   viaVertex tmp;
+
+   PTEX_VERTEX(tmp, vertsize, v0); COPY_DWORDS(vb, vertsize, &tmp);
+   PTEX_VERTEX(tmp, vertsize, v1); COPY_DWORDS(vb, vertsize, &tmp);
+}
+
+static void via_ptex_point (struct via_context *vmesa,
+			    viaVertexPtr v0)
+{
+   GLuint vertsize = vmesa->hwVertexSize;
+   GLuint *vb = viaExtendPrimitive(vmesa, 1*4*vertsize);
+   viaVertex tmp;
+
+   PTEX_VERTEX(tmp, vertsize, v0); COPY_DWORDS(vb, vertsize, &tmp);
+}
+
+
+
+
+
+/***********************************************************************
+ *          Macros for via_dd_tritmp.h to draw basic primitives        *
+ ***********************************************************************/
+
+#define TRI(a, b, c)                                \
+    do {                                            \
+        if (DO_FALLBACK)                            \
+            vmesa->drawTri(vmesa, a, b, c);         \
+        else                                        \
+            via_draw_triangle(vmesa, a, b, c);      \
+    } while (0)
+
+#define QUAD(a, b, c, d)                            \
+    do {                                            \
+        if (DO_FALLBACK) {                          \
+            vmesa->drawTri(vmesa, a, b, d);         \
+            vmesa->drawTri(vmesa, b, c, d);         \
+        }                                           \
+        else                                        \
+            via_draw_quad(vmesa, a, b, c, d);       \
+    } while (0)
+
+#define LINE(v0, v1)                                \
+    do {                                            \
+        if (DO_FALLBACK)                            \
+            vmesa->drawLine(vmesa, v0, v1);         \
+        else                                        \
+            via_draw_line(vmesa, v0, v1);           \
+    } while (0)
+
+#define POINT(v0)                                    \
+    do {                                             \
+        if (DO_FALLBACK)                             \
+            vmesa->drawPoint(vmesa, v0);             \
+        else                                         \
+            via_draw_point(vmesa, v0);               \
+    } while (0)
+
+
+/***********************************************************************
+ *              Build render functions from dd templates               *
+ ***********************************************************************/
+
+#define VIA_OFFSET_BIT         0x01
+#define VIA_TWOSIDE_BIT        0x02
+#define VIA_UNFILLED_BIT       0x04
+#define VIA_FALLBACK_BIT       0x08
+#define VIA_MAX_TRIFUNC        0x10
+
+
+static struct {
+    tnl_points_func          points;
+    tnl_line_func            line;
+    tnl_triangle_func        triangle;
+    tnl_quad_func            quad;
+} rast_tab[VIA_MAX_TRIFUNC + 1];
+
+
+#define DO_FALLBACK (IND & VIA_FALLBACK_BIT)
+#define DO_OFFSET   (IND & VIA_OFFSET_BIT)
+#define DO_UNFILLED (IND & VIA_UNFILLED_BIT)
+#define DO_TWOSIDE  (IND & VIA_TWOSIDE_BIT)
+#define DO_FLAT      0
+#define DO_TRI       1
+#define DO_QUAD      1
+#define DO_LINE      1
+#define DO_POINTS    1
+#define DO_FULL_QUAD 1
+
+#define HAVE_RGBA         1
+#define HAVE_SPEC         1
+#define HAVE_BACK_COLORS  0
+#define HAVE_HW_FLATSHADE 1
+#define VERTEX            viaVertex
+#define TAB               rast_tab
+
+/* Only used to pull back colors into vertices (ie, we know color is
+ * floating point).
+ */
+#define VIA_COLOR(dst, src)                     \
+    do {                                        \
+        dst[0] = src[2];                        \
+        dst[1] = src[1];                        \
+        dst[2] = src[0];                        \
+        dst[3] = src[3];                        \
+    } while (0)
+
+#define VIA_SPEC(dst, src)                      \
+    do {                                        \
+        dst[0] = src[2];                        \
+        dst[1] = src[1];                        \
+        dst[2] = src[0];                        \
+    } while (0)
+
+
+#define DEPTH_SCALE vmesa->polygon_offset_scale
+#define UNFILLED_TRI unfilled_tri
+#define UNFILLED_QUAD unfilled_quad
+#define VERT_X(_v) _v->v.x
+#define VERT_Y(_v) _v->v.y
+#define VERT_Z(_v) _v->v.z
+#define AREA_IS_CCW(a) (a > 0)
+#define GET_VERTEX(e) (vmesa->verts + (e * vmesa->vertexSize * sizeof(int)))
+
+#define VERT_SET_RGBA( v, c )  					\
+do {								\
+   via_color_t *color = (via_color_t *)&((v)->ui[coloroffset]);	\
+   UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]);		\
+   UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]);		\
+   UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]);		\
+   UNCLAMPED_FLOAT_TO_UBYTE(color->alpha, (c)[3]);		\
+} while (0)
+
+#define VERT_COPY_RGBA( v0, v1 ) v0->ui[coloroffset] = v1->ui[coloroffset]
+
+#define VERT_SET_SPEC( v, c )					\
+do {								\
+   if (specoffset) {						\
+     via_color_t *color = (via_color_t *)&((v)->ui[specoffset]);	\
+     UNCLAMPED_FLOAT_TO_UBYTE(color->red, (c)[0]);		\
+     UNCLAMPED_FLOAT_TO_UBYTE(color->green, (c)[1]);		\
+     UNCLAMPED_FLOAT_TO_UBYTE(color->blue, (c)[2]);		\
+   }								\
+} while (0)
+#define VERT_COPY_SPEC( v0, v1 )			\
+do {							\
+   if (specoffset) {					\
+      v0->ub4[specoffset][0] = v1->ub4[specoffset][0];	\
+      v0->ub4[specoffset][1] = v1->ub4[specoffset][1];	\
+      v0->ub4[specoffset][2] = v1->ub4[specoffset][2];	\
+   }							\
+} while (0)
+
+
+#define VERT_SAVE_RGBA( idx )    color[idx] = v[idx]->ui[coloroffset]
+#define VERT_RESTORE_RGBA( idx ) v[idx]->ui[coloroffset] = color[idx]
+#define VERT_SAVE_SPEC( idx )    if (specoffset) spec[idx] = v[idx]->ui[specoffset]
+#define VERT_RESTORE_SPEC( idx ) if (specoffset) v[idx]->ui[specoffset] = spec[idx]
+
+
+#define LOCAL_VARS(n)                                                   \
+    struct via_context *vmesa = VIA_CONTEXT(ctx);                             \
+    GLuint color[n], spec[n];                                           \
+    GLuint coloroffset = vmesa->coloroffset;              \
+    GLuint specoffset = vmesa->specoffset;                       \
+    (void)color; (void)spec; (void)coloroffset; (void)specoffset;
+
+
+/***********************************************************************
+ *                Helpers for rendering unfilled primitives            *
+ ***********************************************************************/
+
+static const GLenum hwPrim[GL_POLYGON + 2] = {
+    GL_POINTS,
+    GL_LINES,
+    GL_LINES,
+    GL_LINES,
+    GL_TRIANGLES,
+    GL_TRIANGLES,
+    GL_TRIANGLES,
+    GL_TRIANGLES,
+    GL_TRIANGLES,
+    GL_TRIANGLES,
+    GL_POLYGON+1
+};
+
+
+#define RASTERIZE(x) viaRasterPrimitive( ctx, x, hwPrim[x] )
+#define RENDER_PRIMITIVE vmesa->renderPrimitive
+#define TAG(x) x
+#define IND VIA_FALLBACK_BIT
+#include "tnl_dd/t_dd_unfilled.h"
+#undef IND
+#undef RASTERIZE
+
+/***********************************************************************
+ *                      Generate GL render functions                   *
+ ***********************************************************************/
+#define RASTERIZE(x)
+
+#define IND (0)
+#define TAG(x) x
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_OFFSET_BIT)
+#define TAG(x) x##_offset
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT)
+#define TAG(x) x##_twoside
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_OFFSET_BIT)
+#define TAG(x) x##_twoside_offset
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_UNFILLED_BIT)
+#define TAG(x) x##_unfilled
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_OFFSET_BIT|VIA_UNFILLED_BIT)
+#define TAG(x) x##_offset_unfilled
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_UNFILLED_BIT)
+#define TAG(x) x##_twoside_unfilled
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_OFFSET_BIT|VIA_UNFILLED_BIT)
+#define TAG(x) x##_twoside_offset_unfilled
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_FALLBACK_BIT)
+#define TAG(x) x##_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_OFFSET_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_offset_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_twoside_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_OFFSET_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_twoside_offset_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_UNFILLED_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_unfilled_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_OFFSET_BIT|VIA_UNFILLED_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_offset_unfilled_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_UNFILLED_BIT|VIA_FALLBACK_BIT)
+#define TAG(x) x##_twoside_unfilled_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+#define IND (VIA_TWOSIDE_BIT|VIA_OFFSET_BIT|VIA_UNFILLED_BIT| \
+             VIA_FALLBACK_BIT)
+#define TAG(x) x##_twoside_offset_unfilled_fallback
+#include "tnl_dd/t_dd_tritmp.h"
+
+
+/* Catchall case for flat, separate specular triangles (via has flat
+ * diffuse shading, but always does specular color with gouraud).
+ */
+#undef  DO_FALLBACK
+#undef  DO_OFFSET
+#undef  DO_UNFILLED
+#undef  DO_TWOSIDE
+#undef  DO_FLAT
+#define DO_FALLBACK (0)
+#define DO_OFFSET   (ctx->_TriangleCaps & DD_TRI_OFFSET)
+#define DO_UNFILLED (ctx->_TriangleCaps & DD_TRI_UNFILLED)
+#define DO_TWOSIDE  (ctx->_TriangleCaps & DD_TRI_LIGHT_TWOSIDE)
+#define DO_FLAT     1
+#define TAG(x) x##_flat_specular
+#define IND VIA_MAX_TRIFUNC
+#include "tnl_dd/t_dd_tritmp.h"
+
+
+static void init_rast_tab(void)
+{
+    init();
+    init_offset();
+    init_twoside();
+    init_twoside_offset();
+    init_unfilled();
+    init_offset_unfilled();
+    init_twoside_unfilled();
+    init_twoside_offset_unfilled();
+    init_fallback();
+    init_offset_fallback();
+    init_twoside_fallback();
+    init_twoside_offset_fallback();
+    init_unfilled_fallback();
+    init_offset_unfilled_fallback();
+    init_twoside_unfilled_fallback();
+    init_twoside_offset_unfilled_fallback();
+
+    init_flat_specular();	/* special! */
+}
+
+
+/***********************************************************************
+ *                    Rasterization fallback helpers                   *
+ ***********************************************************************/
+
+
+/* This code is hit only when a mix of accelerated and unaccelerated
+ * primitives are being drawn, and only for the unaccelerated
+ * primitives.
+ */
+static void
+via_fallback_tri(struct via_context *vmesa,
+                 viaVertex *v0,
+                 viaVertex *v1,
+                 viaVertex *v2)
+{    
+    GLcontext *ctx = vmesa->glCtx;
+    SWvertex v[3];
+    _swsetup_Translate(ctx, v0, &v[0]);
+    _swsetup_Translate(ctx, v1, &v[1]);
+    _swsetup_Translate(ctx, v2, &v[2]);
+    viaSpanRenderStart( ctx );
+    _swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
+    viaSpanRenderFinish( ctx );
+}
+
+
+static void
+via_fallback_line(struct via_context *vmesa,
+                  viaVertex *v0,
+                  viaVertex *v1)
+{
+    GLcontext *ctx = vmesa->glCtx;
+    SWvertex v[2];
+    _swsetup_Translate(ctx, v0, &v[0]);
+    _swsetup_Translate(ctx, v1, &v[1]);
+    viaSpanRenderStart( ctx );
+    _swrast_Line(ctx, &v[0], &v[1]);
+    viaSpanRenderFinish( ctx );
+}
+
+
+static void
+via_fallback_point(struct via_context *vmesa,
+                   viaVertex *v0)
+{
+    GLcontext *ctx = vmesa->glCtx;
+    SWvertex v[1];
+    _swsetup_Translate(ctx, v0, &v[0]);
+    viaSpanRenderStart( ctx );
+    _swrast_Point(ctx, &v[0]);
+    viaSpanRenderFinish( ctx );
+}
+
+static void viaResetLineStipple( GLcontext *ctx )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   vmesa->regCmdB |= HC_HLPrst_MASK;
+}
+
+/**********************************************************************/
+/*               Render unclipped begin/end objects                   */
+/**********************************************************************/
+#define IND 0
+#define V(x) (viaVertex *)(vertptr + ((x) * vertsize * sizeof(int)))
+#define RENDER_POINTS(start, count)   \
+    for (; start < count; start++) POINT(V(ELT(start)));
+#define RENDER_LINE(v0, v1)         LINE(V(v0), V(v1))
+#define RENDER_TRI( v0, v1, v2)     TRI( V(v0), V(v1), V(v2))
+#define RENDER_QUAD(v0, v1, v2, v3) QUAD(V(v0), V(v1), V(v2), V(v3))
+#define INIT(x) viaRasterPrimitive(ctx, x, hwPrim[x])
+#undef LOCAL_VARS
+#define LOCAL_VARS                                              \
+    struct via_context *vmesa = VIA_CONTEXT(ctx);                     \
+    GLubyte *vertptr = (GLubyte *)vmesa->verts;                 \
+    const GLuint vertsize = vmesa->vertexSize;          \
+    const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;       \
+   const GLboolean stipple = ctx->Line.StippleFlag;		\
+   (void) elt; (void) stipple;
+#define RESET_STIPPLE	if ( stipple ) viaResetLineStipple( ctx );
+#define RESET_OCCLUSION
+#define PRESERVE_VB_DEFS
+#define ELT(x) x
+#define TAG(x) via_##x##_verts
+#include "tnl/t_vb_rendertmp.h"
+#undef ELT
+#undef TAG
+#define TAG(x) via_##x##_elts
+#define ELT(x) elt[x]
+#include "tnl/t_vb_rendertmp.h"
+#undef ELT
+#undef TAG
+#undef NEED_EDGEFLAG_SETUP
+#undef EDGEFLAG_GET
+#undef EDGEFLAG_SET
+#undef RESET_OCCLUSION
+
+
+/**********************************************************************/
+/*                   Render clipped primitives                        */
+/**********************************************************************/
+
+
+
+static void viaRenderClippedPoly(GLcontext *ctx, const GLuint *elts,
+                                 GLuint n)
+{
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+    GLuint prim = VIA_CONTEXT(ctx)->renderPrimitive;
+
+    /* Render the new vertices as an unclipped polygon.
+     */
+    {
+        GLuint *tmp = VB->Elts;
+        VB->Elts = (GLuint *)elts;
+        tnl->Driver.Render.PrimTabElts[GL_POLYGON](ctx, 0, n,
+                                                   PRIM_BEGIN|PRIM_END);
+        VB->Elts = tmp;
+    }
+
+    /* Restore the render primitive
+     */
+    if (prim != GL_POLYGON &&
+	prim != GL_POLYGON + 1)
+       tnl->Driver.Render.PrimitiveNotify( ctx, prim );
+}
+
+static void viaRenderClippedLine(GLcontext *ctx, GLuint ii, GLuint jj)
+{
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    tnl->Driver.Render.Line(ctx, ii, jj);
+}
+
+static void viaFastRenderClippedPoly(GLcontext *ctx, const GLuint *elts,
+                                     GLuint n)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    GLuint vertsize = vmesa->vertexSize;
+    GLuint *vb = viaExtendPrimitive(vmesa, (n - 2) * 3 * 4 * vertsize);
+    GLubyte *vertptr = (GLubyte *)vmesa->verts;
+    const GLuint *start = (const GLuint *)V(elts[0]);
+    int i;
+
+    for (i = 2; i < n; i++) {
+	COPY_DWORDS(vb, vertsize, V(elts[i - 1]));
+        COPY_DWORDS(vb, vertsize, V(elts[i]));
+	COPY_DWORDS(vb, vertsize, start);	
+    }
+}
+
+
+/**********************************************************************/
+/*                    Choose render functions                         */
+/**********************************************************************/
+
+
+#define _VIA_NEW_VERTEX (_NEW_TEXTURE |                         \
+                         _DD_NEW_SEPARATE_SPECULAR |            \
+                         _DD_NEW_TRI_UNFILLED |                 \
+                         _DD_NEW_TRI_LIGHT_TWOSIDE |            \
+                         _NEW_FOG)
+
+#define _VIA_NEW_RENDERSTATE (_DD_NEW_LINE_STIPPLE |            \
+                              _DD_NEW_TRI_UNFILLED |            \
+                              _DD_NEW_TRI_LIGHT_TWOSIDE |       \
+                              _DD_NEW_TRI_OFFSET |              \
+                              _DD_NEW_TRI_STIPPLE |             \
+                              _NEW_POLYGONSTIPPLE)
+
+
+static void viaChooseRenderState(GLcontext *ctx)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   GLuint flags = ctx->_TriangleCaps;
+   GLuint index = 0;
+
+   if (vmesa->ptexHack) {
+      vmesa->drawPoint = via_ptex_point;
+      vmesa->drawLine = via_ptex_line;
+      vmesa->drawTri = via_ptex_tri;
+      index |= VIA_FALLBACK_BIT;
+   }
+   else {
+      vmesa->drawPoint = via_draw_point;
+      vmesa->drawLine = via_draw_line;
+      vmesa->drawTri = via_draw_triangle;
+   }
+
+   if (flags & (ANY_FALLBACK_FLAGS | ANY_RASTER_FLAGS)) {
+      if (ctx->Light.Enabled && ctx->Light.Model.TwoSide)
+         index |= VIA_TWOSIDE_BIT;
+      if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL)
+         index |= VIA_UNFILLED_BIT;
+      if (flags & DD_TRI_OFFSET)
+         index |= VIA_OFFSET_BIT;
+      if (flags & ANY_FALLBACK_FLAGS)
+         index |= VIA_FALLBACK_BIT;
+
+      /* Hook in fallbacks for specific primitives. */
+      if (flags & POINT_FALLBACK)
+	 vmesa->drawPoint = via_fallback_point;
+      
+      if (flags & LINE_FALLBACK)
+	 vmesa->drawLine = via_fallback_line;
+
+      if (flags & TRI_FALLBACK)
+	 vmesa->drawTri = via_fallback_tri;
+   }
+
+   if ((flags & DD_SEPARATE_SPECULAR) && ctx->Light.ShadeModel == GL_FLAT)
+      index = VIA_MAX_TRIFUNC;	/* flat specular */
+
+   if (vmesa->renderIndex != index) {
+      vmesa->renderIndex = index;
+
+      tnl->Driver.Render.Points = rast_tab[index].points;
+      tnl->Driver.Render.Line = rast_tab[index].line;
+      tnl->Driver.Render.Triangle = rast_tab[index].triangle;
+      tnl->Driver.Render.Quad = rast_tab[index].quad;
+
+      if (index == 0) {
+	 tnl->Driver.Render.PrimTabVerts = via_render_tab_verts;
+	 tnl->Driver.Render.PrimTabElts = via_render_tab_elts;
+	 tnl->Driver.Render.ClippedLine = line; /* from tritmp.h */
+	 tnl->Driver.Render.ClippedPolygon = viaFastRenderClippedPoly;
+      }
+      else {
+	 tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
+	 tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
+	 tnl->Driver.Render.ClippedLine = viaRenderClippedLine;
+	 tnl->Driver.Render.ClippedPolygon = viaRenderClippedPoly;
+      }
+   }
+}
+
+
+#define VIA_EMIT_TEX1	0x01
+#define VIA_EMIT_TEX0	0x02
+#define VIA_EMIT_PTEX0	0x04
+#define VIA_EMIT_RGBA	0x08
+#define VIA_EMIT_SPEC	0x10
+#define VIA_EMIT_FOG	0x20
+#define VIA_EMIT_W	0x40
+
+#define EMIT_ATTR( ATTR, STYLE, INDEX, REGB )				\
+do {									\
+   vmesa->vertex_attrs[vmesa->vertex_attr_count].attrib = (ATTR);	\
+   vmesa->vertex_attrs[vmesa->vertex_attr_count].format = (STYLE);	\
+   vmesa->vertex_attr_count++;						\
+   setupIndex |= (INDEX);						\
+   regCmdB |= (REGB);							\
+} while (0)
+
+#define EMIT_PAD( N )							\
+do {									\
+   vmesa->vertex_attrs[vmesa->vertex_attr_count].attrib = 0;		\
+   vmesa->vertex_attrs[vmesa->vertex_attr_count].format = EMIT_PAD;	\
+   vmesa->vertex_attrs[vmesa->vertex_attr_count].offset = (N);		\
+   vmesa->vertex_attr_count++;						\
+} while (0)
+
+
+
+static void viaChooseVertexState( GLcontext *ctx )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   DECLARE_RENDERINPUTS(index_bitset);
+   GLuint regCmdB = HC_HVPMSK_X | HC_HVPMSK_Y | HC_HVPMSK_Z;
+   GLuint setupIndex = 0;
+
+   RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
+   vmesa->vertex_attr_count = 0;
+ 
+   /* EMIT_ATTR's must be in order as they tell t_vertex.c how to
+    * build up a hardware vertex.
+    */
+   if (RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX ) ||
+       RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
+      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, VIA_EMIT_W, HC_HVPMSK_W );
+      vmesa->coloroffset = 4;
+   }
+   else {
+      EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_3F_VIEWPORT, 0, 0 );
+      vmesa->coloroffset = 3;
+   }
+
+   /* t_context.c always includes a diffuse color */
+   EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_BGRA, VIA_EMIT_RGBA, 
+	      HC_HVPMSK_Cd );
+      
+   vmesa->specoffset = 0;
+   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 ) ||
+       RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG )) {
+      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR1 )) {
+	 vmesa->specoffset = vmesa->coloroffset + 1;
+	 EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_3UB_3F_BGR, VIA_EMIT_SPEC, 
+		    HC_HVPMSK_Cs );
+      }
+      else
+	 EMIT_PAD( 3 );
+
+      if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_FOG ))
+	 EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1UB_1F, VIA_EMIT_FOG, HC_HVPMSK_Cs );
+      else
+	 EMIT_PAD( 1 );
+   }
+
+   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 )) {
+      if (vmesa->ptexHack)
+	 EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_3F_XYW, VIA_EMIT_PTEX0, 
+		    (HC_HVPMSK_S | HC_HVPMSK_T) );
+      else 
+	 EMIT_ATTR( _TNL_ATTRIB_TEX0, EMIT_2F, VIA_EMIT_TEX0, 
+		    (HC_HVPMSK_S | HC_HVPMSK_T) );
+   }
+
+   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 )) {
+      EMIT_ATTR( _TNL_ATTRIB_TEX1, EMIT_2F, VIA_EMIT_TEX1, 
+		 (HC_HVPMSK_S | HC_HVPMSK_T) );
+   }
+
+   if (setupIndex != vmesa->setupIndex) {
+      vmesa->vertexSize = _tnl_install_attrs( ctx, 
+					       vmesa->vertex_attrs, 
+					       vmesa->vertex_attr_count,
+					       vmesa->ViewportMatrix.m, 0 );
+      vmesa->vertexSize >>= 2;
+      vmesa->setupIndex = setupIndex;
+      vmesa->regCmdB &= ~HC_HVPMSK_MASK;
+      vmesa->regCmdB |= regCmdB;
+
+      if (vmesa->ptexHack) 
+	 vmesa->hwVertexSize = vmesa->vertexSize - 1;
+      else
+	 vmesa->hwVertexSize = vmesa->vertexSize;
+   }
+}
+
+
+
+
+/* Check if projective texture coordinates are used and if we can fake
+ * them. Fallback to swrast if we can't. Returns GL_TRUE if projective
+ * texture coordinates must be faked, GL_FALSE otherwise.
+ */
+static GLboolean viaCheckPTexHack( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   DECLARE_RENDERINPUTS(index_bitset);
+   GLboolean fallback = GL_FALSE;
+   GLboolean ptexHack = GL_FALSE;
+
+   RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
+
+   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX0 ) && VB->TexCoordPtr[0]->size == 4) {
+      if (!RENDERINPUTS_TEST_RANGE( index_bitset, _TNL_ATTRIB_TEX1, _TNL_LAST_TEX ))
+	 ptexHack = GL_TRUE; 
+      else
+	 fallback = GL_TRUE;
+   }
+   if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_TEX1 ) && VB->TexCoordPtr[1]->size == 4)
+      fallback = GL_TRUE;
+
+   FALLBACK(VIA_CONTEXT(ctx), VIA_FALLBACK_PROJ_TEXTURE, fallback);
+   return ptexHack;
+}
+
+
+
+
+/**********************************************************************/
+/*                 High level hooks for t_vb_render.c                 */
+/**********************************************************************/
+
+
+static void viaRenderStart(GLcontext *ctx)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   {
+      GLboolean ptexHack = viaCheckPTexHack( ctx );
+      if (ptexHack != vmesa->ptexHack) {
+	 vmesa->ptexHack = ptexHack;
+	 vmesa->newRenderState |= _VIA_NEW_RENDERSTATE;
+      }
+   }
+
+   if (vmesa->newState) {
+      vmesa->newRenderState |= vmesa->newState;
+      viaValidateState( ctx );
+   }
+
+   if (vmesa->Fallback) {
+      tnl->Driver.Render.Start(ctx);
+      return;
+   }
+
+   if (vmesa->newRenderState) {
+      viaChooseVertexState(ctx);
+      viaChooseRenderState(ctx);
+      vmesa->newRenderState = 0;
+   }
+
+   /* Important:
+    */
+   VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
+}
+
+static void viaRenderFinish(GLcontext *ctx)
+{
+   VIA_FINISH_PRIM(VIA_CONTEXT(ctx));
+}
+
+
+/* System to flush dma and emit state changes based on the rasterized
+ * primitive.
+ */
+void viaRasterPrimitive(GLcontext *ctx,
+			GLenum glprim,
+			GLenum hwprim)
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+   GLuint regCmdB;
+   RING_VARS;
+
+   if (VIA_DEBUG & DEBUG_PRIMS) 
+      fprintf(stderr, "%s: %s/%s/%s\n", 
+	      __FUNCTION__, _mesa_lookup_enum_by_nr(glprim),
+	      _mesa_lookup_enum_by_nr(hwprim),
+	      _mesa_lookup_enum_by_nr(ctx->Light.ShadeModel));
+
+   assert (!vmesa->newState);
+
+   vmesa->renderPrimitive = glprim;
+
+   if (hwprim != vmesa->hwPrimitive ||
+       ctx->Light.ShadeModel != vmesa->hwShadeModel) {
+
+      VIA_FINISH_PRIM(vmesa);
+
+      /* Ensure no wrapping inside this function  */    
+      viaCheckDma( vmesa, 1024 );	
+
+      if (vmesa->newEmitState) {
+	 viaEmitState(vmesa);
+      }
+       
+      vmesa->regCmdA_End = HC_ACMD_HCmdA;
+
+      if (ctx->Light.ShadeModel == GL_SMOOTH) {
+	 vmesa->regCmdA_End |= HC_HShading_Gouraud;
+      }
+      
+      vmesa->hwShadeModel = ctx->Light.ShadeModel;
+      regCmdB = vmesa->regCmdB;
+
+      switch (hwprim) {
+      case GL_POINTS:
+	 vmesa->regCmdA_End |= HC_HPMType_Point | HC_HVCycle_Full;
+	 vmesa->regCmdA_End |= HC_HShading_Gouraud; /* always Gouraud 
+						       shade points?!? */
+	 break;
+      case GL_LINES:
+	 vmesa->regCmdA_End |= HC_HPMType_Line | HC_HVCycle_Full;
+         regCmdB |= HC_HLPrst_MASK;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatB; 
+	 break;
+      case GL_LINE_LOOP:
+      case GL_LINE_STRIP:
+	 vmesa->regCmdA_End |= HC_HPMType_Line | HC_HVCycle_AFP |
+	    HC_HVCycle_AB | HC_HVCycle_NewB;
+	 regCmdB |= HC_HVCycle_AB | HC_HVCycle_NewB | HC_HLPrst_MASK;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatB; 
+	 break;
+      case GL_TRIANGLES:
+	 vmesa->regCmdA_End |= HC_HPMType_Tri | HC_HVCycle_Full;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatC; 
+	 break;
+      case GL_TRIANGLE_STRIP:
+	 vmesa->regCmdA_End |= HC_HPMType_Tri | HC_HVCycle_AFP |
+	    HC_HVCycle_AC | HC_HVCycle_BB | HC_HVCycle_NewC;
+	 regCmdB |= HC_HVCycle_AA | HC_HVCycle_BC | HC_HVCycle_NewC;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatC; 
+	 break;
+      case GL_TRIANGLE_FAN:
+	 vmesa->regCmdA_End |= HC_HPMType_Tri | HC_HVCycle_AFP |
+	    HC_HVCycle_AA | HC_HVCycle_BC | HC_HVCycle_NewC;
+	 regCmdB |= HC_HVCycle_AA | HC_HVCycle_BC | HC_HVCycle_NewC;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatC; 
+	 break;
+      case GL_QUADS:
+	 abort();
+	 return;
+      case GL_QUAD_STRIP:
+	 abort();
+	 return;
+      case GL_POLYGON:
+	 vmesa->regCmdA_End |= HC_HPMType_Tri | HC_HVCycle_AFP |
+	    HC_HVCycle_AA | HC_HVCycle_BC | HC_HVCycle_NewC;
+	 regCmdB |= HC_HVCycle_AA | HC_HVCycle_BC | HC_HVCycle_NewC;
+	 if (ctx->Light.ShadeModel == GL_FLAT)
+            vmesa->regCmdA_End |= HC_HShading_FlatC; 
+	 break;                          
+      default:
+	 abort();
+	 return;
+      }
+    
+/*     assert((vmesa->dmaLow & 0x4) == 0); */
+
+      if (vmesa->dmaCliprectAddr == ~0) {
+	 if (VIA_DEBUG & DEBUG_DMA) 
+	    fprintf(stderr, "reserve cliprect space at %x\n", vmesa->dmaLow);
+	 vmesa->dmaCliprectAddr = vmesa->dmaLow;
+	 BEGIN_RING(8);
+	 OUT_RING( HC_HEADER2 );    
+	 OUT_RING( (HC_ParaType_NotTex << 16) );
+	 OUT_RING( 0xCCCCCCCC );
+	 OUT_RING( 0xCCCCCCCC );
+	 OUT_RING( 0xCCCCCCCC );
+	 OUT_RING( 0xCCCCCCCC );
+	 OUT_RING( 0xCCCCCCCC );
+	 OUT_RING( 0xCCCCCCCC );
+	 ADVANCE_RING();
+      }
+
+      assert(vmesa->dmaLastPrim == 0);
+
+      BEGIN_RING(8);
+      OUT_RING( HC_HEADER2 );    
+      OUT_RING( (HC_ParaType_NotTex << 16) );
+      OUT_RING( 0xCCCCCCCC );
+      OUT_RING( 0xDDDDDDDD );
+
+      OUT_RING( HC_HEADER2 );    
+      OUT_RING( (HC_ParaType_CmdVdata << 16) );
+      OUT_RING( regCmdB );
+      OUT_RING( vmesa->regCmdA_End );
+      ADVANCE_RING();
+
+      vmesa->hwPrimitive = hwprim;        
+      vmesa->dmaLastPrim = vmesa->dmaLow;
+   }
+   else {
+      assert(!vmesa->newEmitState);
+   }
+}
+
+/* Callback for mesa:
+ */
+static void viaRenderPrimitive( GLcontext *ctx, GLuint prim )
+{
+   viaRasterPrimitive( ctx, prim, hwPrim[prim] );
+}
+
+
+void viaFinishPrimitive(struct via_context *vmesa)
+{
+   if (VIA_DEBUG & (DEBUG_DMA|DEBUG_PRIMS)) 
+      fprintf(stderr, "%s\n", __FUNCTION__);
+
+   if (!vmesa->dmaLastPrim || vmesa->dmaCliprectAddr == ~0) {
+      assert(0);
+   }
+   else if (vmesa->dmaLow != vmesa->dmaLastPrim) {
+      GLuint cmdA = (vmesa->regCmdA_End | HC_HPLEND_MASK | 
+		     HC_HPMValidN_MASK | HC_HE3Fire_MASK); 
+      RING_VARS;
+
+      vmesa->dmaLastPrim = 0;
+
+      /* KW: modified 0x1 to 0x4 below:
+       */
+      if ((vmesa->dmaLow & 0x4) || !vmesa->useAgp) {
+	 BEGIN_RING_NOCHECK( 1 );
+	 OUT_RING( cmdA );
+	 ADVANCE_RING();
+      }   
+      else {      
+	 BEGIN_RING_NOCHECK( 2 );
+	 OUT_RING( cmdA );
+	 OUT_RING( cmdA );
+	 ADVANCE_RING();
+      }   
+
+      if (vmesa->dmaLow > VIA_DMA_HIGHWATER)
+	 viaFlushDma( vmesa );
+   }
+   else {
+      if (VIA_DEBUG & (DEBUG_DMA|DEBUG_PRIMS)) 
+	 fprintf(stderr, "remove empty primitive\n");
+
+      /* Remove the primitive header:
+       */
+      vmesa->dmaLastPrim = 0;
+      vmesa->dmaLow -= 8 * sizeof(GLuint);
+
+      /* Maybe remove the cliprect as well:
+       */
+      if (vmesa->dmaCliprectAddr == vmesa->dmaLow - 8 * sizeof(GLuint)) {
+	 vmesa->dmaLow -= 8 * sizeof(GLuint);
+	 vmesa->dmaCliprectAddr = ~0;
+      }
+   }
+
+   vmesa->renderPrimitive = GL_POLYGON + 1;
+   vmesa->hwPrimitive = GL_POLYGON + 1;
+   vmesa->dmaLastPrim = 0;
+}
+
+
+/**********************************************************************/
+/*           Transition to/from hardware rasterization.               */
+/**********************************************************************/
+
+
+void viaFallback(struct via_context *vmesa, GLuint bit, GLboolean mode)
+{
+    GLcontext *ctx = vmesa->glCtx;
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    GLuint oldfallback = vmesa->Fallback;
+    
+    if (mode) {
+        vmesa->Fallback |= bit;
+        if (oldfallback == 0) {
+	    VIA_FLUSH_DMA(vmesa);
+
+ 	    if (VIA_DEBUG & DEBUG_FALLBACKS) 
+	       fprintf(stderr, "ENTER FALLBACK %x\n", bit);
+
+            _swsetup_Wakeup(ctx);
+            vmesa->renderIndex = ~0;
+        }
+    }
+    else {
+        vmesa->Fallback &= ~bit;
+        if (oldfallback == bit) {
+	    _swrast_flush( ctx );
+
+ 	    if (VIA_DEBUG & DEBUG_FALLBACKS) 
+	       fprintf(stderr, "LEAVE FALLBACK %x\n", bit);
+
+	    tnl->Driver.Render.Start = viaRenderStart;
+            tnl->Driver.Render.PrimitiveNotify = viaRenderPrimitive;
+            tnl->Driver.Render.Finish = viaRenderFinish;
+
+	    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+	    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+	    tnl->Driver.Render.Interp = _tnl_interp;
+    	    tnl->Driver.Render.ResetLineStipple = viaResetLineStipple;
+
+	    _tnl_invalidate_vertex_state( ctx, ~0 );
+	    _tnl_invalidate_vertices( ctx, ~0 );
+	    _tnl_install_attrs( ctx, 
+				vmesa->vertex_attrs, 
+				vmesa->vertex_attr_count,
+				vmesa->ViewportMatrix.m, 0 ); 
+
+            vmesa->newState |= (_VIA_NEW_RENDERSTATE|_VIA_NEW_VERTEX);
+        }
+    }    
+}
+
+static void viaRunPipeline( GLcontext *ctx )
+{
+   struct via_context *vmesa = VIA_CONTEXT(ctx);
+
+   if (vmesa->newState) {
+      vmesa->newRenderState |= vmesa->newState;
+      viaValidateState( ctx );
+   }
+
+   _tnl_run_pipeline( ctx );
+}
+
+
+/**********************************************************************/
+/*                            Initialization.                         */
+/**********************************************************************/
+
+
+void viaInitTriFuncs(GLcontext *ctx)
+{
+    struct via_context *vmesa = VIA_CONTEXT(ctx);
+    TNLcontext *tnl = TNL_CONTEXT(ctx);
+    static int firsttime = 1;
+
+    if (firsttime) {
+        init_rast_tab();
+        firsttime = 0;
+    }
+
+    tnl->Driver.RunPipeline = viaRunPipeline;
+    tnl->Driver.Render.Start = viaRenderStart;
+    tnl->Driver.Render.Finish = viaRenderFinish;
+    tnl->Driver.Render.PrimitiveNotify = viaRenderPrimitive;
+    tnl->Driver.Render.ResetLineStipple = viaResetLineStipple;
+    tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+    tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+    tnl->Driver.Render.Interp = _tnl_interp;
+
+    _tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
+			(6 + 2*ctx->Const.MaxTextureUnits) * sizeof(GLfloat) );
+   
+    vmesa->verts = (GLubyte *)tnl->clipspace.vertex_buf;
+
+}
diff --git a/src/via_tris.h b/src/via_tris.h
new file mode 100644
index 0000000..b4f0880
--- /dev/null
+++ b/src/via_tris.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 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
+ * 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
+ * VIA, S3 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.
+ */
+
+#ifndef _VIATRIS_H
+#define _VIATRIS_H
+
+#include "mtypes.h"
+
+extern void viaPrintRenderState(const char *msg, GLuint state);
+extern void viaInitTriFuncs(GLcontext *ctx);
+extern void viaRasterPrimitive(GLcontext *ctx, GLenum rPrim, GLuint hwPrim);
+extern void viaRasterPrimitiveFinish(GLcontext *ctx);
+
+#endif
author	Luc Verhaegen <libv@skynet.be>	2010-03-11 14:18:41 +0100
committer	Luc Verhaegen <libv@skynet.be>	2010-03-11 14:18:41 +0100
commit	bf6a158b95d810d06344b4ca9395e4ebbca8bda9 (patch)
tree	36a709a98723329d45cf43154e0fb93278a07cbd
parent	5dee9b7b19c1aa3a13618b08bc24f00677b5364b (diff)