Merge master and fix conflicts

32 files changed, 2183 insertions, 2136 deletions

diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 0dff9a1273..bdb09624be 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -37,6 +37,7 @@ RADEON_COMMON_SOURCES = \
 DRIVER_SOURCES = \
 		 radeon_screen.c \
 		 r300_context.c \
+		 r300_draw.c \
 		 r300_ioctl.c \
 		 r300_cmdbuf.c \
 		 r300_state.c \
@@ -48,6 +49,7 @@ DRIVER_SOURCES = \
 		 radeon_program_pair.c \
 		 radeon_nqssadce.c \
 		 r300_vertprog.c \
+		 r300_fragprog_common.c \
 		 r300_fragprog.c \
 		 r300_fragprog_swizzle.c \
 		 r300_fragprog_emit.c \
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.c b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
index c575c9ac49..0261a5b1d8 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.c
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.c
@@ -55,9 +55,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_state.h"
 #include "radeon_reg.h"
 
-#define R300_VAP_PVS_UPLOAD_ADDRESS 0x2200
-#   define RADEON_ONE_REG_WR        (1 << 15)
-
 /** # of dwords reserved for additional instructions that may need to be written
  * during flushing.
  */
@@ -71,7 +68,6 @@ static unsigned packet0_count(r300ContextPtr r300, uint32_t *pkt)
         drm_r300_cmd_header_t *t = (drm_r300_cmd_header_t*)pkt;
         return t->packet0.count;
     }
-    return 0;
 }
 
 #define vpu_count(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
@@ -83,7 +79,7 @@ void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom)
 	BATCH_LOCALS(&r300->radeon);
 	drm_r300_cmd_header_t cmd;
 	uint32_t addr, ndw, i;
-	
+
 	if (!r300->radeon.radeonScreen->kernel_mm) {
 		uint32_t dwords;
 		dwords = (*atom->check) (ctx, atom);
@@ -92,7 +88,7 @@ void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom)
 		END_BATCH();
 		return;
 	}
-	
+
 	cmd.u = atom->cmd[0];
 	addr = (cmd.vpu.adrhi << 8) | cmd.vpu.adrlo;
 	ndw = cmd.vpu.count * 4;
@@ -111,7 +107,7 @@ void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom)
 		} else {
 			BEGIN_BATCH_NO_AUTOSTATE(5 + ndw);
 		}
-		OUT_BATCH_REGVAL(R300_VAP_PVS_UPLOAD_ADDRESS, addr);
+		OUT_BATCH_REGVAL(R300_VAP_PVS_VECTOR_INDX_REG, addr);
 		OUT_BATCH(CP_PACKET0(R300_VAP_PVS_UPLOAD_DATA, ndw-1) | RADEON_ONE_REG_WR);
 		for (i = 0; i < ndw; i++) {
 			OUT_BATCH(atom->cmd[i+1]);
@@ -175,7 +171,7 @@ static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom)
 
 		for(i = 0; i < numtmus; ++i) {
 		    radeonTexObj *t = r300->hw.textures[i];
-		
+
 		    if (!t)
 			notexture = 1;
 		}
@@ -183,30 +179,75 @@ static void emit_tex_offsets(GLcontext *ctx, struct radeon_state_atom * atom)
 		if (r300->radeon.radeonScreen->kernel_mm && notexture) {
 			return;
 		}
-		BEGIN_BATCH_NO_AUTOSTATE(4 * numtmus);
 		for(i = 0; i < numtmus; ++i) {
 		    radeonTexObj *t = r300->hw.textures[i];
-		    OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
 		    if (t && !t->image_override) {
+                BEGIN_BATCH_NO_AUTOSTATE(4);
+                OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
 			    OUT_BATCH_RELOC(t->tile_bits, t->mt->bo, 0,
 					    RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+                END_BATCH();
 		    } else if (!t) {
-			    OUT_BATCH(r300->radeon.radeonScreen->texOffset[0]);
+                /* Texture unit hasn't a texture bound nothings to do */
 		    } else { /* override cases */
 			    if (t->bo) {
+                    BEGIN_BATCH_NO_AUTOSTATE(4);
+                    OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
 				    OUT_BATCH_RELOC(t->tile_bits, t->bo, 0,
 						    RADEON_GEM_DOMAIN_GTT|RADEON_GEM_DOMAIN_VRAM, 0, 0);
+                    END_BATCH();
 			    } else if (!r300->radeon.radeonScreen->kernel_mm) {
+                    BEGIN_BATCH_NO_AUTOSTATE(2);
+                    OUT_BATCH_REGSEQ(R300_TX_OFFSET_0 + (i * 4), 1);
 				    OUT_BATCH(t->override_offset);
-			    }
-			    else
-			    	OUT_BATCH(r300->radeon.radeonScreen->texOffset[0]);
+                    END_BATCH();
+			    } else {
+                    /* Texture unit hasn't a texture bound nothings to do */
+                }
 		    }
 		}
-		END_BATCH();
 	}
 }
 
+void r300_emit_scissor(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    unsigned x1, y1, x2, y2;
+	struct radeon_renderbuffer *rrb;
+
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        return;
+    }
+	rrb = radeon_get_colorbuffer(&r300->radeon);
+	if (!rrb || !rrb->bo) {
+		fprintf(stderr, "no rrb\n");
+		return;
+	}
+    if (r300->radeon.state.scissor.enabled) {
+        x1 = r300->radeon.state.scissor.rect.x1;
+        y1 = r300->radeon.state.scissor.rect.y1;
+        x2 = r300->radeon.state.scissor.rect.x2 - 1;
+        y2 = r300->radeon.state.scissor.rect.y2 - 1;
+    } else {
+        x1 = 0;
+        y1 = 0;
+        x2 = rrb->width - 1;
+        y2 = rrb->height - 1;
+    }
+    if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
+        x1 += R300_SCISSORS_OFFSET;
+        y1 += R300_SCISSORS_OFFSET;
+        x2 += R300_SCISSORS_OFFSET;
+        y2 += R300_SCISSORS_OFFSET;
+    }
+    BEGIN_BATCH_NO_AUTOSTATE(3);
+    OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
+    OUT_BATCH((x1 << R300_SCISSORS_X_SHIFT)|(y1 << R300_SCISSORS_Y_SHIFT));
+    OUT_BATCH((x2 << R300_SCISSORS_X_SHIFT)|(y2 << R300_SCISSORS_Y_SHIFT));
+    END_BATCH();
+}
+
 static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
@@ -215,6 +256,7 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	uint32_t cbpitch;
 	uint32_t offset = r300->radeon.state.color.draw_offset;
 	uint32_t dw = 6;
+    int i;
 
 	rrb = radeon_get_colorbuffer(&r300->radeon);
 	if (!rrb || !rrb->bo) {
@@ -247,16 +289,39 @@ static void emit_cb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
             BEGIN_BATCH_NO_AUTOSTATE(3);
             OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
             OUT_BATCH(0);
-            OUT_BATCH((rrb->width << R300_SCISSORS_X_SHIFT) |
-                    (rrb->height << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->width - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->height - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((0 << R300_CLIPRECT_X_SHIFT) | (0 << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((rrb->width - 1) << R300_CLIPRECT_X_SHIFT) | ((rrb->height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
             END_BATCH();
         } else {
             BEGIN_BATCH_NO_AUTOSTATE(3);
             OUT_BATCH_REGSEQ(R300_SC_SCISSORS_TL, 2);
             OUT_BATCH((R300_SCISSORS_OFFSET << R300_SCISSORS_X_SHIFT) |
                     (R300_SCISSORS_OFFSET << R300_SCISSORS_Y_SHIFT));
-            OUT_BATCH(((rrb->width + R300_SCISSORS_OFFSET) << R300_SCISSORS_X_SHIFT) |
-                    ((rrb->height + R300_SCISSORS_OFFSET) << R300_SCISSORS_Y_SHIFT));
+            OUT_BATCH(((rrb->width + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_X_SHIFT) |
+                    ((rrb->height + R300_SCISSORS_OFFSET - 1) << R300_SCISSORS_Y_SHIFT));
+            END_BATCH();
+            BEGIN_BATCH_NO_AUTOSTATE(16);
+            for (i = 0; i < 4; i++) {
+                OUT_BATCH_REGSEQ(R300_SC_CLIPRECT_TL_0 + (i * 8), 2);
+                OUT_BATCH((R300_SCISSORS_OFFSET << R300_CLIPRECT_X_SHIFT) | (R300_SCISSORS_OFFSET << R300_CLIPRECT_Y_SHIFT));
+                OUT_BATCH(((R300_SCISSORS_OFFSET + rrb->width - 1) << R300_CLIPRECT_X_SHIFT) |
+                          ((R300_SCISSORS_OFFSET + rrb->height - 1) << R300_CLIPRECT_Y_SHIFT));
+            }
+            OUT_BATCH_REGSEQ(R300_SC_CLIP_RULE, 1);
+            OUT_BATCH(0xAAAA);
+            OUT_BATCH_REGSEQ(R300_SC_SCREENDOOR, 1);
+            OUT_BATCH(0xffffff);
             END_BATCH();
         }
     }
@@ -280,7 +345,7 @@ static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	if (rrb->bo->flags & RADEON_BO_FLAGS_MICRO_TILE){
 		zbpitch |= R300_DEPTHMICROTILE_TILED;
 	}
-	
+
 	BEGIN_BATCH_NO_AUTOSTATE(6);
 	OUT_BATCH_REGSEQ(R300_ZB_DEPTHOFFSET, 1);
 	OUT_BATCH_RELOC(0, rrb->bo, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
@@ -288,6 +353,46 @@ static void emit_zb_offset(GLcontext *ctx, struct radeon_state_atom * atom)
 	END_BATCH();
 }
 
+static void emit_gb_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        BEGIN_BATCH_NO_AUTOSTATE(4);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        OUT_BATCH(atom->cmd[2]);
+        OUT_BATCH(atom->cmd[3]);
+        END_BATCH();
+    }
+}
+
+static void emit_threshold_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+    if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+        BEGIN_BATCH_NO_AUTOSTATE(3);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        OUT_BATCH(atom->cmd[2]);
+        END_BATCH();
+    }
+}
+
+static void emit_shade_misc(GLcontext *ctx, struct radeon_state_atom * atom)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&r300->radeon);
+
+    if (!r300->radeon.radeonScreen->driScreen->dri2.enabled) {
+        BEGIN_BATCH_NO_AUTOSTATE(2);
+        OUT_BATCH(atom->cmd[0]);
+        OUT_BATCH(atom->cmd[1]);
+        END_BATCH();
+    }
+}
+
 static void emit_zstencil_format(GLcontext *ctx, struct radeon_state_atom * atom)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
@@ -372,12 +477,10 @@ int check_r500fp_const(GLcontext *ctx, struct radeon_state_atom *atom)
 void r300InitCmdBuf(r300ContextPtr r300)
 {
 	int mtu;
-	int has_tcl = 1;
+	int has_tcl;
 	int is_r500 = 0;
-	int i;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
 		is_r500 = 1;
@@ -444,7 +547,10 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	ALLOC_STATE(gb_enable, always, 2, 0);
 	r300->hw.gb_enable.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_ENABLE, 1);
 	ALLOC_STATE(gb_misc, always, R300_GB_MISC_CMDSIZE, 0);
-	r300->hw.gb_misc.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_MSPOS0, 5);
+	r300->hw.gb_misc.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GB_MSPOS0, 3);
+	r300->hw.gb_misc.emit = emit_gb_misc;
+	ALLOC_STATE(gb_misc2, always, R300_GB_MISC2_CMDSIZE, 0);
+    r300->hw.gb_misc2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x401C, 2);
 	ALLOC_STATE(txe, always, R300_TXE_CMDSIZE, 0);
 	r300->hw.txe.cmd[R300_TXE_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_TX_ENABLE, 1);
 	ALLOC_STATE(ga_point_s0, always, 5, 0);
@@ -459,8 +565,11 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	r300->hw.lcntl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_CNTL, 1);
 	ALLOC_STATE(ga_line_stipple, always, 4, 0);
 	r300->hw.ga_line_stipple.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_LINE_STIPPLE_VALUE, 3);
-	ALLOC_STATE(shade, always, 5, 0);
-	r300->hw.shade.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_ENHANCE, 4);
+	ALLOC_STATE(shade, always, 2, 0);
+	r300->hw.shade.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_ENHANCE, 1);
+	r300->hw.shade.emit = emit_shade_misc;
+	ALLOC_STATE(shade2, always, 4, 0);
+	r300->hw.shade2.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x4278, 3);
 	ALLOC_STATE(polygon_mode, always, 4, 0);
 	r300->hw.polygon_mode.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_GA_POLY_MODE, 3);
 	ALLOC_STATE(fogp, always, 3, 0);
@@ -479,19 +588,12 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	ALLOC_STATE(rc, always, R300_RC_CMDSIZE, 0);
 	r300->hw.rc.cmd[R300_RC_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_COUNT, 2);
 	if (is_r500) {
-		ALLOC_STATE(ri, always, R500_RI_CMDSIZE, 0);
+		ALLOC_STATE(ri, variable, R500_RI_CMDSIZE, 0);
 		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, 16);
-		for (i = 0; i < 8; i++) {
-			r300->hw.ri.cmd[R300_RI_CMD_0 + i +1] =
-			  (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
-                          (R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
-                          (R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT);
-		}
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
 		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, 1);
 	} else {
-		ALLOC_STATE(ri, always, R300_RI_CMDSIZE, 0);
+		ALLOC_STATE(ri, variable, R300_RI_CMDSIZE, 0);
 		r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, 8);
 		ALLOC_STATE(rr, variable, R300_RR_CMDSIZE, 0);
 		r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, 1);
@@ -567,8 +669,9 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	r300->hw.rb3d_dither_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_DITHER_CTL, 9);
 	ALLOC_STATE(rb3d_aaresolve_ctl, always, 2, 0);
 	r300->hw.rb3d_aaresolve_ctl.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_RB3D_AARESOLVE_CTL, 1);
-	ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
+    ALLOC_STATE(rb3d_discard_src_pixel_lte_threshold, always, 3, 0);
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R500_RB3D_DISCARD_SRC_PIXEL_LTE_THRESHOLD, 2);
+	r300->hw.rb3d_discard_src_pixel_lte_threshold.emit = emit_threshold_misc;
 	ALLOC_STATE(zs, always, R300_ZS_CMDSIZE, 0);
 	r300->hw.zs.cmd[R300_ZS_CMD_0] =
 	    cmdpacket0(r300->radeon.radeonScreen, R300_ZB_CNTL, 3);
@@ -582,8 +685,8 @@ void r300InitCmdBuf(r300ContextPtr r300)
 	r300->hw.zb.emit = emit_zb_offset;
 	ALLOC_STATE(zb_depthclearvalue, always, 2, 0);
 	r300->hw.zb_depthclearvalue.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_DEPTHCLEARVALUE, 1);
-	ALLOC_STATE(unk4F30, always, 3, 0);
-	r300->hw.unk4F30.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, 0x4F30, 2);
+	ALLOC_STATE(zb_zmask, always, 3, 0);
+	r300->hw.zb_zmask.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_ZMASK_OFFSET, 2);
 	ALLOC_STATE(zb_hiz_offset, always, 2, 0);
 	r300->hw.zb_hiz_offset.cmd[0] = cmdpacket0(r300->radeon.radeonScreen, R300_ZB_HIZ_OFFSET, 1);
 	ALLOC_STATE(zb_hiz_pitch, always, 2, 0);
diff --git a/src/mesa/drivers/dri/r300/r300_cmdbuf.h b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
index 3786813de3..53bcc0eeb4 100644
--- a/src/mesa/drivers/dri/r300/r300_cmdbuf.h
+++ b/src/mesa/drivers/dri/r300/r300_cmdbuf.h
@@ -39,6 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_context.h"
 
 extern void r300InitCmdBuf(r300ContextPtr r300);
+void r300_emit_scissor(GLcontext *ctx);
 
 void emit_vpu(GLcontext *ctx, struct radeon_state_atom * atom);
 int check_vpu(GLcontext *ctx, struct radeon_state_atom *atom);
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 5f279d6629..394521a051 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -64,6 +64,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_ioctl.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
+#include "r300_render.h"
 #include "r300_swtcl.h"
 #include "radeon_bocs_wrapper.h"
 
@@ -72,10 +73,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "utils.h"
 #include "xmlpool.h"		/* for symbolic values of enum-type options */
 
-/* hw_tcl_on derives from future_hw_tcl_on when its safe to change it. */
-int future_hw_tcl_on = 1;
-int hw_tcl_on = 1;
-
 #define need_GL_VERSION_2_0
 #define need_GL_ARB_point_parameters
 #define need_GL_ARB_vertex_program
@@ -127,6 +124,7 @@ const struct dri_extension card_extensions[] = {
   {"GL_EXT_texture_lod_bias",		NULL},
   {"GL_EXT_texture_mirror_clamp",	NULL},
   {"GL_EXT_texture_rectangle",		NULL},
+  {"GL_EXT_vertex_array_bgra",		NULL},
   {"GL_ATI_separate_stencil",		GL_ATI_separate_stencil_functions},
   {"GL_ATI_texture_env_combine3",	NULL},
   {"GL_ATI_texture_mirror_once",	NULL},
@@ -154,16 +152,8 @@ const struct dri_extension gl_20_extension[] = {
   {"GL_VERSION_2_0",			GL_VERSION_2_0_functions },
 };
 
-
-extern struct tnl_pipeline_stage _r300_render_stage;
-extern const struct tnl_pipeline_stage _r300_tcl_stage;
-
 static const struct tnl_pipeline_stage *r300_pipeline[] = {
 
-	/* Try and go straight to t&l
-	 */
-	&_r300_tcl_stage,
-
 	/* Catch any t&l fallbacks
 	 */
 	&_tnl_vertex_transform_stage,
@@ -172,6 +162,7 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	&_tnl_fog_coordinate_stage,
 	&_tnl_texgen_stage,
 	&_tnl_texture_transform_stage,
+	&_tnl_point_attenuation_stage,
 	&_tnl_vertex_program_stage,
 
 	/* Try again to go to tcl?
@@ -191,17 +182,6 @@ static const struct tnl_pipeline_stage *r300_pipeline[] = {
 	0,
 };
 
-static void r300RunPipeline(GLcontext * ctx)
-{
-    _mesa_lock_context_textures(ctx);
-
-    if (ctx->NewState)
-        _mesa_update_state_locked(ctx);
-    
-    _tnl_run_pipeline(ctx);
-    _mesa_unlock_context_textures(ctx);
-}
-
 static void r300_get_lock(radeonContextPtr rmesa)
 {
 	drm_radeon_sarea_t *sarea = rmesa->sarea;
@@ -211,7 +191,7 @@ static void r300_get_lock(radeonContextPtr rmesa)
 		if (!rmesa->radeonScreen->kernel_mm)
 			radeon_bo_legacy_texture_age(rmesa->radeonScreen->bom);
 	}
-}		  
+}
 
 static void r300_vtbl_emit_cs_header(struct radeon_cs *cs, radeonContextPtr rmesa)
 {
@@ -246,9 +226,9 @@ static void r300_vtbl_pre_emit_atoms(radeonContextPtr radeon)
 {
 	r300ContextPtr r300 = (r300ContextPtr)radeon;
 	BATCH_LOCALS(radeon);
-	
+
 	r300->vap_flush_needed = GL_TRUE;
-	
+
 	cp_wait(radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
 	BEGIN_BATCH_NO_AUTOSTATE(2);
 	OUT_BATCH_REGVAL(R300_TX_INVALTAGS, R300_TX_FLUSH);
@@ -275,6 +255,112 @@ static void r300_init_vtbl(radeonContextPtr radeon)
 	radeon->vtbl.fallback = r300_fallback;
 }
 
+static void r300InitConstValues(GLcontext *ctx, radeonScreenPtr screen)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	ctx->Const.MaxTextureImageUnits =
+	    driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
+	ctx->Const.MaxTextureCoordUnits =
+	    driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
+	ctx->Const.MaxTextureUnits = MIN2(ctx->Const.MaxTextureImageUnits,
+		 ctx->Const.MaxTextureCoordUnits);
+
+	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
+	ctx->Const.MaxTextureLodBias = 16.0;
+
+	if (screen->chip_family >= CHIP_FAMILY_RV515)
+		ctx->Const.MaxTextureLevels = 13;
+	else
+		ctx->Const.MaxTextureLevels = 12;
+
+	ctx->Const.MinPointSize = 1.0;
+	ctx->Const.MinPointSizeAA = 1.0;
+	ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
+	ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
+
+	ctx->Const.MinLineWidth = 1.0;
+	ctx->Const.MinLineWidthAA = 1.0;
+	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
+	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
+
+	ctx->Const.MaxDrawBuffers = 1;
+
+	/* currently bogus data */
+	if (r300->options.hw_tcl_enabled) {
+		ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
+		ctx->Const.VertexProgram.MaxNativeInstructions =
+		  VSF_MAX_FRAGMENT_LENGTH / 4;
+		ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
+		ctx->Const.VertexProgram.MaxTemps = 32;
+		ctx->Const.VertexProgram.MaxNativeTemps =
+		  /*VSF_MAX_FRAGMENT_TEMPS */ 32;
+		ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
+		ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
+	}
+
+	if (screen->chip_family >= CHIP_FAMILY_RV515) {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R500_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R500_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R500_PFS_MAX_INST;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	} else {
+		ctx->Const.FragmentProgram.MaxNativeTemps = R300_PFS_NUM_TEMP_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
+		ctx->Const.FragmentProgram.MaxNativeParameters = R300_PFS_NUM_CONST_REGS;
+		ctx->Const.FragmentProgram.MaxNativeAluInstructions = R300_PFS_MAX_ALU_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexInstructions = R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeInstructions = R300_PFS_MAX_ALU_INST + R300_PFS_MAX_TEX_INST;
+		ctx->Const.FragmentProgram.MaxNativeTexIndirections = R300_PFS_MAX_TEX_INDIRECT;
+		ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;
+	}
+}
+
+static void r300ParseOptions(r300ContextPtr r300, radeonScreenPtr screen)
+{
+	struct r300_options options = { 0 };
+
+	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
+			    screen->driScreen->myNum, "r300");
+
+	r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache, "def_max_anisotropy");
+
+	options.stencil_two_side_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side");
+	options.s3tc_force_enabled = driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable");
+	options.s3tc_force_disabled = driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc");
+
+	if (!(screen->chip_flags & RADEON_CHIPSET_TCL) || driQueryOptioni(&r300->radeon.optionCache, "tcl_mode") == DRI_CONF_TCL_SW)
+		options.hw_tcl_enabled = 0;
+	else
+		options.hw_tcl_enabled = 1;
+
+	options.conformance_mode = !driQueryOptionb(&r300->radeon.optionCache, "disable_lowimpact_fallback");
+
+	r300->options = options;
+}
+
+static void r300InitGLExtensions(GLcontext *ctx)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+	driInitExtensions(ctx, card_extensions, GL_TRUE);
+	if (r300->radeon.radeonScreen->kernel_mm)
+		driInitExtensions(ctx, mm_extensions, GL_FALSE);
+
+	if (r300->options.stencil_two_side_disabled)
+		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
+
+	if (ctx->Mesa_DXTn && !r300->options.s3tc_force_enabled) {
+		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+		_mesa_enable_extension(ctx, "GL_S3_s3tc");
+	} else if (r300->options.s3tc_force_disabled) {
+		_mesa_disable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+	}
+}
 
 /* Create the device specific rendering context.
  */
@@ -287,33 +373,19 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	struct dd_function_table functions;
 	r300ContextPtr r300;
 	GLcontext *ctx;
-	int tcl_mode;
 
 	assert(glVisual);
 	assert(driContextPriv);
 	assert(screen);
 
-	/* Allocate the R300 context */
 	r300 = (r300ContextPtr) CALLOC(sizeof(*r300));
 	if (!r300)
 		return GL_FALSE;
 
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-		hw_tcl_on = future_hw_tcl_on = 0;
+	r300ParseOptions(r300, screen);
 
 	r300_init_vtbl(&r300->radeon);
-	/* Parse configuration files.
-	 * Do this here so that initialMaxAnisotropy is set before we create
-	 * the default textures.
-	 */
-	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
-			    screen->driScreen->myNum, "r300");
-	r300->radeon.initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
-						     "def_max_anisotropy");
 
-	/* Init default driver functions then plug in our R300-specific functions
-	 * (the texture functions are especially important)
-	 */
 	_mesa_init_driver_functions(&functions);
 	r300InitIoctlFuncs(&functions);
 	r300InitStateFuncs(&functions);
@@ -327,46 +399,15 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 		return GL_FALSE;
 	}
 
-	/* Init r300 context data */
-	/* Set the maximum texture size small enough that we can guarentee that
-	 * all texture units can bind a maximal texture and have them both in
-	 * texturable memory at once.
-	 */
-
 	ctx = r300->radeon.glCtx;
 
-	ctx->Const.MaxTextureImageUnits =
-	    driQueryOptioni(&r300->radeon.optionCache, "texture_image_units");
-	ctx->Const.MaxTextureCoordUnits =
-	    driQueryOptioni(&r300->radeon.optionCache, "texture_coord_units");
-	ctx->Const.MaxTextureUnits =
-	    MIN2(ctx->Const.MaxTextureImageUnits,
-		 ctx->Const.MaxTextureCoordUnits);
-	ctx->Const.MaxTextureMaxAnisotropy = 16.0;
-	ctx->Const.MaxTextureLodBias = 16.0;
-
-	if (screen->chip_family >= CHIP_FAMILY_RV515)
-	    ctx->Const.MaxTextureLevels = 13;
-	else
-	    ctx->Const.MaxTextureLevels = 12;
-
-	ctx->Const.MinPointSize = 1.0;
-	ctx->Const.MinPointSizeAA = 1.0;
-	ctx->Const.MaxPointSize = R300_POINTSIZE_MAX;
-	ctx->Const.MaxPointSizeAA = R300_POINTSIZE_MAX;
+	r300->fallback = 0;
+	if (r300->options.hw_tcl_enabled)
+		ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
 
-	ctx->Const.MinLineWidth = 1.0;
-	ctx->Const.MinLineWidthAA = 1.0;
-	ctx->Const.MaxLineWidth = R300_LINESIZE_MAX;
-	ctx->Const.MaxLineWidthAA = R300_LINESIZE_MAX;
-
-	/* Needs further modifications */
-#if 0
-	ctx->Const.MaxArrayLockSize =
-	    ( /*512 */ RADEON_BUFFER_SIZE * 16 * 1024) / (4 * 4);
-#endif
+	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
-	ctx->Const.MaxDrawBuffers = 1;
+	r300InitConstValues(ctx, screen);
 
 	/* Initialize the software rasterizer and helper modules.
 	 */
@@ -375,16 +416,12 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_CreateContext(ctx);
 	_swsetup_CreateContext(ctx);
 	_swsetup_Wakeup(ctx);
-	_ae_create_context(ctx);
 
 	/* Install the customized pipeline:
 	 */
 	_tnl_destroy_pipeline(ctx);
 	_tnl_install_pipeline(ctx, r300_pipeline);
-
-	/* Try and keep materials and vertices separate:
-	 */
-/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
+	TNL_CONTEXT(ctx)->Driver.RunPipeline = _tnl_run_pipeline;
 
 	/* Configure swrast and TNL to match hardware characteristics:
 	 */
@@ -393,80 +430,25 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_allow_pixel_fog(ctx, GL_FALSE);
 	_tnl_allow_vertex_fog(ctx, GL_TRUE);
 
-	/* currently bogus data */
-	if (screen->chip_flags & RADEON_CHIPSET_TCL) {
-	        ctx->Const.VertexProgram.MaxInstructions = VSF_MAX_FRAGMENT_LENGTH / 4;
-		ctx->Const.VertexProgram.MaxNativeInstructions =
-		  VSF_MAX_FRAGMENT_LENGTH / 4;
-		ctx->Const.VertexProgram.MaxNativeAttribs = 16;	/* r420 */
-		ctx->Const.VertexProgram.MaxTemps = 32;
-		ctx->Const.VertexProgram.MaxNativeTemps =
-		  /*VSF_MAX_FRAGMENT_TEMPS */ 32;
-		ctx->Const.VertexProgram.MaxNativeParameters = 256;	/* r420 */
-		ctx->Const.VertexProgram.MaxNativeAddressRegs = 1;
-	}
-
-	ctx->Const.FragmentProgram.MaxNativeTemps = PFS_NUM_TEMP_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAttribs = 11;	/* copy i915... */
-	ctx->Const.FragmentProgram.MaxNativeParameters = PFS_NUM_CONST_REGS;
-	ctx->Const.FragmentProgram.MaxNativeAluInstructions = PFS_MAX_ALU_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexInstructions = PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeInstructions =
-	    PFS_MAX_ALU_INST + PFS_MAX_TEX_INST;
-	ctx->Const.FragmentProgram.MaxNativeTexIndirections =
-	    PFS_MAX_TEX_INDIRECT;
-	ctx->Const.FragmentProgram.MaxNativeAddressRegs = 0;	/* and these are?? */
-	ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
-	ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
-
-	driInitExtensions(ctx, card_extensions, GL_TRUE);
-	if (r300->radeon.radeonScreen->kernel_mm)
-	  driInitExtensions(ctx, mm_extensions, GL_FALSE);
-
-	if (driQueryOptionb
-	    (&r300->radeon.optionCache, "disable_stencil_two_side"))
-		_mesa_disable_extension(ctx, "GL_EXT_stencil_two_side");
-
-	if (r300->radeon.glCtx->Mesa_DXTn
-	    && !driQueryOptionb(&r300->radeon.optionCache, "disable_s3tc")) {
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
-		_mesa_enable_extension(ctx, "GL_S3_s3tc");
-	} else
-	    if (driQueryOptionb(&r300->radeon.optionCache, "force_s3tc_enable"))
-	{
-		_mesa_enable_extension(ctx, "GL_EXT_texture_compression_s3tc");
+	if (r300->options.hw_tcl_enabled) {
+		r300InitDraw(ctx);
+	} else {
+		r300InitSwtcl(ctx);
 	}
 
-	r300->disable_lowimpact_fallback =
-	    driQueryOptionb(&r300->radeon.optionCache,
-			    "disable_lowimpact_fallback");
 	radeon_fbo_init(&r300->radeon);
-   	radeonInitSpanFuncs( ctx );
+	radeonInitSpanFuncs( ctx );
 	r300InitCmdBuf(r300);
 	r300InitState(r300);
-	if (!(screen->chip_flags & RADEON_CHIPSET_TCL))
-	        r300InitSwtcl(ctx);
-
-	TNL_CONTEXT(ctx)->Driver.RunPipeline = r300RunPipeline;
+	r300InitShaderFunctions(r300);
 
-	tcl_mode = driQueryOptioni(&r300->radeon.optionCache, "tcl_mode");
-	if (driQueryOptionb(&r300->radeon.optionCache, "no_rast")) {
-		fprintf(stderr, "disabling 3D acceleration\n");
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_DISABLE, 1);
-#endif
-	}
-	if (tcl_mode == DRI_CONF_TCL_SW ||
-	    !(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL)) {
-		if (r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-			r300->radeon.radeonScreen->chip_flags &=
-			    ~RADEON_CHIPSET_TCL;
-			fprintf(stderr, "Disabling HW TCL support\n");
-		}
-		TCL_FALLBACK(r300->radeon.glCtx,
-			     RADEON_TCL_FALLBACK_TCL_DISABLE, 1);
+	if (screen->chip_family == CHIP_FAMILY_RS600 ||	screen->chip_family == CHIP_FAMILY_RS690 ||
+		screen->chip_family == CHIP_FAMILY_RS740) {
+		r300->radeon.texture_row_align = 64;
 	}
 
+	r300InitGLExtensions(ctx);
+
 	return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 602f86ba66..026c33c67c 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -37,24 +37,19 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #ifndef __R300_CONTEXT_H__
 #define __R300_CONTEXT_H__
 
-#include "tnl/t_vertex.h"
 #include "drm.h"
 #include "radeon_drm.h"
 #include "dri_util.h"
-#include "texmem.h"
 #include "radeon_common.h"
 
-#include "main/macros.h"
 #include "main/mtypes.h"
-#include "main/colormac.h"
+#include "shader/prog_instruction.h"
 
 struct r300_context;
 typedef struct r300_context r300ContextRec;
 typedef struct r300_context *r300ContextPtr;
 
 
-#include "main/mm.h"
-
 /* From http://gcc. gnu.org/onlinedocs/gcc-3.2.3/gcc/Variadic-Macros.html .
    I suppose we could inline this and use macro to fetch out __LINE__ and stuff in case we run into trouble
    with other compilers ... GLUE!
@@ -72,8 +67,6 @@ typedef struct r300_context *r300ContextPtr;
 	}
 
 #include "r300_vertprog.h"
-#include "r500_fragprog.h"
-
 
 
 /* The blit width for texture uploads
@@ -81,9 +74,6 @@ typedef struct r300_context *r300ContextPtr;
 #define R300_BLIT_WIDTH_BYTES 1024
 #define R300_MAX_TEXTURE_UNITS 8
 
-struct r300_texture_state {
-	int tc_count;		/* number of incoming texture coordinates from VAP */
-};
 
 
 #define R300_VPT_CMD_0		0
@@ -126,9 +116,11 @@ struct r300_texture_state {
 #define R300_GB_MISC_MSPOS_0		1
 #define R300_GB_MISC_MSPOS_1		2
 #define R300_GB_MISC_TILE_CONFIG	3
-#define R300_GB_MISC_SELECT		4
-#define R300_GB_MISC_AA_CONFIG		5
-#define R300_GB_MISC_CMDSIZE		6
+#define R300_GB_MISC_CMDSIZE		4
+#define R300_GB_MISC2_CMD_0		    0
+#define R300_GB_MISC2_SELECT		1
+#define R300_GB_MISC2_AA_CONFIG		2
+#define R300_GB_MISC2_CMDSIZE		3
 
 #define R300_TXE_CMD_0		0
 #define R300_TXE_ENABLE		1
@@ -303,7 +295,7 @@ struct r300_texture_state {
 struct r300_hw_state {
 	struct radeon_state_atom vpt;	/* viewport (1D98) */
 	struct radeon_state_atom vap_cntl;
-        struct radeon_state_atom vap_index_offset; /* 0x208c r5xx only */
+	struct radeon_state_atom vap_index_offset; /* 0x208c r5xx only */
 	struct radeon_state_atom vof;	/* VAP output format register 0x2090 */
 	struct radeon_state_atom vte;	/* (20B0) */
 	struct radeon_state_atom vap_vf_max_vtx_indx;	/* Maximum Vertex Indx Clamp (2134) */
@@ -317,6 +309,7 @@ struct r300_hw_state {
 	struct radeon_state_atom pvs;	/* pvs_cntl (22D0) */
 	struct radeon_state_atom gb_enable;	/* (4008) */
 	struct radeon_state_atom gb_misc;	/* Multisampling position shifts ? (4010) */
+	struct radeon_state_atom gb_misc2;	/* Multisampling position shifts ? (4010) */
 	struct radeon_state_atom ga_point_s0;	/* S Texture Coordinate of Vertex 0 for Point texture stuffing (LLC) (4200) */
 	struct radeon_state_atom ga_triangle_stipple;	/* (4214) */
 	struct radeon_state_atom ps;	/* pointsize (421C) */
@@ -324,6 +317,7 @@ struct r300_hw_state {
 	struct radeon_state_atom lcntl;	/* line control */
 	struct radeon_state_atom ga_line_stipple;	/* (4260) */
 	struct radeon_state_atom shade;
+	struct radeon_state_atom shade2;
 	struct radeon_state_atom polygon_mode;
 	struct radeon_state_atom fogp;	/* fog parameters (4294) */
 	struct radeon_state_atom ga_soft_reset;	/* (429C) */
@@ -361,7 +355,7 @@ struct r300_hw_state {
 	struct radeon_state_atom zstencil_format;
 	struct radeon_state_atom zb;	/* z buffer (4F20) */
 	struct radeon_state_atom zb_depthclearvalue;	/* (4F28) */
-	struct radeon_state_atom unk4F30;	/* (4F30) */
+	struct radeon_state_atom zb_zmask;	/* (4F30) */
 	struct radeon_state_atom zb_hiz_offset;	/* (4F44) */
 	struct radeon_state_atom zb_hiz_pitch;	/* (4F54) */
 
@@ -405,73 +399,55 @@ struct r300_hw_state {
 #define STATE_R300_WINDOW_DIMENSION (STATE_INTERNAL_DRIVER+0)
 #define STATE_R300_TEXRECT_FACTOR (STATE_INTERNAL_DRIVER+1)
 
-struct r300_vertex_shader_fragment {
-	int length;
-	union {
-		GLuint d[VSF_MAX_FRAGMENT_LENGTH];
-		float f[VSF_MAX_FRAGMENT_LENGTH];
-		GLuint i[VSF_MAX_FRAGMENT_LENGTH];
-	} body;
-};
-
-struct r300_vertex_shader_state {
-	struct r300_vertex_shader_fragment program;
-};
-
-extern int hw_tcl_on;
-
 #define COLOR_IS_RGBA
 #define TAG(x) r300##x
 #include "tnl_dd/t_dd_vertex.h"
 #undef TAG
 
-//#define CURRENT_VERTEX_SHADER(ctx) (ctx->VertexProgram._Current)
-#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->selected_vp)
-
-/* Should but doesnt work */
-//#define CURRENT_VERTEX_SHADER(ctx) (R300_CONTEXT(ctx)->curr_vp)
-
-/* r300_vertex_shader_state and r300_vertex_program should probably be merged together someday.
- * Keeping them them seperate for now should ensure fixed pipeline keeps functioning properly.
- */
-
-struct r300_vertex_program_key {
-	GLuint InputsRead;
-	GLuint OutputsWritten;
-	GLuint OutputsAdded;
-};
-
 struct r300_vertex_program {
 	struct r300_vertex_program *next;
-	struct r300_vertex_program_key key;
-	int translated;
 
-	struct r300_vertex_shader_fragment program;
+	struct r300_vertex_program_key {
+		GLuint InputsRead;
+		GLuint OutputsWritten;
+		GLuint OutputsAdded;
+	} key;
+	
+	struct r300_vertex_shader_hw_code {
+		int length;
+		union {
+			GLuint d[VSF_MAX_FRAGMENT_LENGTH];
+			float f[VSF_MAX_FRAGMENT_LENGTH];
+		} body;
+	} hw_code;
+
+	GLboolean translated;
+	GLboolean error;
 
 	int pos_end;
 	int num_temporaries;	/* Number of temp vars used by program */
 	int wpos_idx;
 	int inputs[VERT_ATTRIB_MAX];
 	int outputs[VERT_RESULT_MAX];
-	int native;
-	int ref_count;
-	int use_ref_count;
 };
 
 struct r300_vertex_program_cont {
 	struct gl_vertex_program mesa_program;	/* Must be first */
-	struct r300_vertex_shader_fragment params;
 	struct r300_vertex_program *progs;
 };
 
-#define PFS_MAX_ALU_INST	64
-#define PFS_MAX_TEX_INST	64
-#define PFS_MAX_TEX_INDIRECT 4
-#define PFS_NUM_TEMP_REGS	32
-#define PFS_NUM_CONST_REGS	16
+#define R300_PFS_MAX_ALU_INST	64
+#define R300_PFS_MAX_TEX_INST	32
+#define R300_PFS_MAX_TEX_INDIRECT 4
+#define R300_PFS_NUM_TEMP_REGS	32
+#define R300_PFS_NUM_CONST_REGS	32
 
-struct r300_pfs_compile_state;
+#define R500_PFS_MAX_INST 512
+#define R500_PFS_NUM_TEMP_REGS 128
+#define R500_PFS_NUM_CONST_REGS 256
 
+struct r300_pfs_compile_state;
+struct r500_pfs_compile_state;
 
 /**
  * Stores state that influences the compilation of a fragment program.
@@ -514,7 +490,7 @@ struct r300_fragment_program_node {
 struct r300_fragment_program_code {
 	struct {
 		int length; /**< total # of texture instructions used */
-		GLuint inst[PFS_MAX_TEX_INST];
+		GLuint inst[R300_PFS_MAX_TEX_INST];
 	} tex;
 
 	struct {
@@ -524,7 +500,7 @@ struct r300_fragment_program_code {
 			GLuint inst1;
 			GLuint inst2;
 			GLuint inst3;
-		} inst[PFS_MAX_ALU_INST];
+		} inst[R300_PFS_MAX_ALU_INST];
 	} alu;
 
 	struct r300_fragment_program_node node[4];
@@ -535,53 +511,12 @@ struct r300_fragment_program_code {
 	 * Remember which program register a given hardware constant
 	 * belongs to.
 	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
+	struct prog_src_register constant[R300_PFS_NUM_CONST_REGS];
 	int const_nr;
 
 	int max_temp_idx;
 };
 
-/**
- * Store everything about a fragment program that is needed
- * to render with that program.
- */
-struct r300_fragment_program {
-	struct gl_fragment_program mesa_program;
-
-	GLboolean translated;
-	GLboolean error;
-
-	struct r300_fragment_program_external_state state;
-	struct r300_fragment_program_code code;
-
-	GLboolean WritesDepth;
-	GLuint optimization;
-};
-
-struct r500_pfs_compile_state;
-
-struct r500_fragment_program_external_state {
-	struct {
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is:
-		 *  0 - GL_LUMINANCE
-		 *  1 - GL_INTENSITY
-		 *  2 - GL_ALPHA
-		 * depending on the depth texture mode.
-		 */
-		GLuint depth_texture_mode : 2;
-
-		/**
-		 * If the sampler is used as a shadow sampler,
-		 * this field is (texture_compare_func - GL_NEVER).
-		 * [e.g. if compare function is GL_LEQUAL, this field is 3]
-		 *
-		 * Otherwise, this field is 0.
-		 */
-		GLuint texture_compare_func : 3;
-	} unit[16];
-};
 
 struct r500_fragment_program_code {
 	struct {
@@ -591,7 +526,7 @@ struct r500_fragment_program_code {
 		GLuint inst3;
 		GLuint inst4;
 		GLuint inst5;
-	} inst[512];
+	} inst[R500_PFS_MAX_INST];
 
 	int inst_offset;
 	int inst_end;
@@ -600,47 +535,41 @@ struct r500_fragment_program_code {
 	 * Remember which program register a given hardware constant
 	 * belongs to.
 	 */
-	struct prog_src_register constant[PFS_NUM_CONST_REGS];
+	struct prog_src_register constant[R500_PFS_NUM_CONST_REGS];
 	int const_nr;
 
 	int max_temp_idx;
 };
 
-struct r500_fragment_program {
-	struct gl_fragment_program mesa_program;
+/**
+* Store everything about a fragment program that is needed
+* to render with that program.
+*/
+struct r300_fragment_program {
+	struct gl_fragment_program Base;
 
-	GLcontext *ctx;
 	GLboolean translated;
 	GLboolean error;
 
-	struct r500_fragment_program_external_state state;
-	struct r500_fragment_program_code code;
+	struct r300_fragment_program_external_state state;
+	union rX00_fragment_program_code {
+		struct r300_fragment_program_code r300;
+		struct r500_fragment_program_code r500;
+	} code;
 
 	GLboolean writes_depth;
-
 	GLuint optimization;
 };
 
-#define R300_MAX_AOS_ARRAYS		16
-
-#define REG_COORDS	0
-#define REG_COLOR0	1
-#define REG_TEX0	2
-
-struct r300_state {
-	struct r300_texture_state texture;
-	int sw_tcl_inputs[VERT_ATTRIB_MAX];
-	struct r300_vertex_shader_state vertex_shader;
-
-
-	DECLARE_RENDERINPUTS(render_inputs_bitset);	/* actual render inputs that R300 was configured for.
-							   They are the same as tnl->render_inputs for fixed pipeline */
-
+struct r300_fragment_program_compiler {
+	r300ContextPtr r300;
+	struct r300_fragment_program *fp;
+	union rX00_fragment_program_code *code;
+	struct gl_program *program;
 };
 
-#define R300_FALLBACK_NONE 0
-#define R300_FALLBACK_TCL 1
-#define R300_FALLBACK_RAST 2
+#define R300_MAX_AOS_ARRAYS		16
+
 
 /* r300_swtcl.c
  */
@@ -654,18 +583,44 @@ struct r300_swtcl_info {
     * Offset of the 3UB specular color data within a hardware (swtcl) vertex.
     */
    GLuint specoffset;
+};
 
-   struct vertex_attribute{
-       GLuint attr;
-       GLubyte format;
-       GLubyte dst_loc;
-       GLuint swizzle;
-       GLubyte write_mask;
-   } vert_attrs[VERT_ATTRIB_MAX];
+struct r300_vtable {
+	void (* SetupRSUnit)(GLcontext *ctx);
+	void (* SetupFragmentShaderTextures)(GLcontext *ctx, int *tmu_mappings);
+	GLboolean (* BuildFragmentProgramHwCode)(struct r300_fragment_program_compiler *compiler);
+	void (* FragmentProgramDump)(union rX00_fragment_program_code *code);
+	void (* SetupPixelShader)(GLcontext *ctx);
+};
 
-   GLubyte vertex_attr_count;
+struct r300_vertex_buffer {
+	struct vertex_attribute {
+		/* generic */
+		GLubyte element;
+		GLvoid *data;
+		GLboolean free_needed;
+		GLuint stride;
+		GLuint dwords;
+		GLubyte size; /* number of components */
+
+		/* hw specific */
+		uint32_t data_type:4;
+		uint32_t dst_loc:5;
+		uint32_t _signed:1;
+		uint32_t normalize:1;
+		uint32_t swizzle:12;
+		uint32_t write_mask:4;
+	} attribs[VERT_ATTRIB_MAX];
+
+	GLubyte num_attribs;
 };
 
+struct r300_index_buffer {
+	GLvoid *ptr;
+	GLboolean is_32bit;
+	GLboolean free_needed;
+	GLuint count;
+};
 
 /**
  * \brief R300 context structure.
@@ -673,10 +628,10 @@ struct r300_swtcl_info {
 struct r300_context {
 	struct radeon_context radeon;	/* parent class, must be first */
 
+	struct r300_vtable vtbl;
+
 	struct r300_hw_state hw;
 
-	struct r300_state state;
-	struct gl_vertex_program *curr_vp;
 	struct r300_vertex_program *selected_vp;
 
 	/* Vertex buffers
@@ -684,10 +639,22 @@ struct r300_context {
 	GLvector4f dummy_attrib[_TNL_ATTRIB_MAX];
 	GLvector4f *temp_attrib[_TNL_ATTRIB_MAX];
 
-	GLboolean disable_lowimpact_fallback;
-
+	struct r300_options {
+		uint32_t conformance_mode:1;
+		uint32_t hw_tcl_enabled:1;
+		uint32_t s3tc_force_enabled:1;
+		uint32_t s3tc_force_disabled:1;
+		uint32_t stencil_two_side_disabled:1;
+	} options;
+	
 	struct r300_swtcl_info swtcl;
+	struct r300_vertex_buffer vbuf;
+	struct r300_index_buffer ind_buf;
 	GLboolean vap_flush_needed;
+
+	uint32_t fallback;
+
+	DECLARE_RENDERINPUTS(render_inputs_bitset);
 };
 
 #define R300_CONTEXT(ctx)		((r300ContextPtr)(ctx->DriverCtx))
@@ -703,10 +670,9 @@ extern int r300VertexProgUpdateParams(GLcontext * ctx,
 				      struct r300_vertex_program_cont *vp,
 				      float *dst);
 
-#define RADEON_D_CAPTURE 0
-#define RADEON_D_PLAYBACK 1
-#define RADEON_D_PLAYBACK_RAW 2
-#define RADEON_D_T 3
+extern void r300InitShaderFunctions(r300ContextPtr r300);
+
+extern void r300InitDraw(GLcontext *ctx);
 
 #define r300PackFloat32 radeonPackFloat32
 #define r300PackFloat24 radeonPackFloat24
diff --git a/src/mesa/drivers/dri/r300/r300_draw.c b/src/mesa/drivers/dri/r300/r300_draw.c
new file mode 100644
index 0000000000..cc5650fb7c
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -0,0 +1,475 @@
+/**************************************************************************
+ *
+ * Copyright 2009 Maciej Cencora
+ *
+ * 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 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 "main/glheader.h"
+#include "main/context.h"
+#include "main/state.h"
+#include "main/api_validate.h"
+#include "main/enums.h"
+
+#include "r300_reg.h"
+#include "r300_context.h"
+#include "r300_emit.h"
+#include "r300_render.h"
+#include "r300_state.h"
+#include "r300_tex.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
+#include "vbo/vbo_context.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+
+static void r300FixupIndexBuffer(GLcontext *ctx, const struct _mesa_index_buffer *mesa_ind_buf, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_index_buffer *ind_buf = &r300->ind_buf;
+	GLvoid *src_ptr;
+
+	if (!mesa_ind_buf) {
+		ind_buf->ptr = NULL;
+		return;
+	}
+
+	ind_buf->count = mesa_ind_buf->count;
+	if (mesa_ind_buf->obj->Name && !mesa_ind_buf->obj->Pointer) {
+		bo[*nr_bo] = mesa_ind_buf->obj;
+		(*nr_bo)++;
+		ctx->Driver.MapBuffer(ctx, GL_ELEMENT_ARRAY_BUFFER, GL_READ_ONLY_ARB, mesa_ind_buf->obj);
+		assert(mesa_ind_buf->obj->Pointer != NULL);
+	}
+	src_ptr = ADD_POINTERS(mesa_ind_buf->obj->Pointer, mesa_ind_buf->ptr);
+
+	if (mesa_ind_buf->type == GL_UNSIGNED_BYTE) {
+		GLubyte *in = (GLubyte *)src_ptr;
+		GLuint *out = _mesa_malloc(sizeof(GLushort) * ((mesa_ind_buf->count + 1) & ~1));
+		int i;
+
+		ind_buf->ptr = out;
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+
+		ind_buf->free_needed = GL_TRUE;
+		ind_buf->is_32bit = GL_FALSE;
+	} else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
+#if MESA_BIG_ENDIAN
+		GLushort *in = (GLushort *)src_ptr;
+		GLuint *out = _mesa_malloc(sizeof(GLushort) *
+					   ((mesa_ind_buf->count + 1) & ~1));
+		int i;
+
+		ind_buf->ptr = out;
+
+		for (i = 0; i + 1 < mesa_ind_buf->count; i += 2) {
+			*out++ = in[i] | in[i + 1] << 16;
+		}
+
+		if (i < mesa_ind_buf->count) {
+			*out++ = in[i];
+		}
+
+		ind_buf->free_needed = GL_TRUE;
+#else
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+#endif
+		ind_buf->is_32bit = GL_FALSE;
+	} else {
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+		ind_buf->is_32bit = GL_TRUE;
+	}
+}
+
+static int getTypeSize(GLenum type)
+{
+	switch (type) {
+		case GL_DOUBLE:
+			return sizeof(GLdouble);
+		case GL_FLOAT:
+			return sizeof(GLfloat);
+		case GL_INT:
+			return sizeof(GLint);
+		case GL_UNSIGNED_INT:
+			return sizeof(GLuint);
+		case GL_SHORT:
+			return sizeof(GLshort);
+		case GL_UNSIGNED_SHORT:
+			return sizeof(GLushort);
+		case GL_BYTE:
+			return sizeof(GLbyte);
+		case GL_UNSIGNED_BYTE:
+			return sizeof(GLubyte);
+		default:
+			assert(0);
+			return 0;
+	}
+}
+
+#define CONVERT( TYPE, MACRO ) do {		\
+	GLuint i, j, sz;				\
+	sz = input->Size;				\
+	if (input->Normalized) {			\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = MACRO(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	} else {					\
+		for (i = 0; i < count; i++) {		\
+			const TYPE *in = (TYPE *)src_ptr;		\
+			for (j = 0; j < sz; j++) {		\
+				*dst_ptr++ = (GLfloat)(*in);		\
+				in++;				\
+			}					\
+			src_ptr += stride;			\
+		}						\
+	}						\
+} while (0)
+
+static void r300TranslateAttrib(GLcontext *ctx, GLuint attr, int count, const struct gl_client_array *input, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	struct vertex_attribute r300_attr;
+	const void *src_ptr;
+	GLenum type;
+	GLuint stride;
+
+	if (input->BufferObj->Name) {
+		if (!input->BufferObj->Pointer) {
+			bo[*nr_bo] = input->BufferObj;
+			(*nr_bo)++;
+			ctx->Driver.MapBuffer(ctx, GL_ARRAY_BUFFER, GL_READ_ONLY_ARB, input->BufferObj);
+			assert(input->BufferObj->Pointer != NULL);
+		}
+
+		src_ptr = ADD_POINTERS(input->BufferObj->Pointer, input->Ptr);
+	} else
+		src_ptr = input->Ptr;
+
+	stride = (input->StrideB == 0) ? getTypeSize(input->Type) * input->Size : input->StrideB;
+
+	if (input->Type == GL_DOUBLE || input->Type == GL_UNSIGNED_INT || input->Type == GL_INT ||
+#if MESA_BIG_ENDIAN
+	    getTypeSize(input->Type) != 4 ||
+#endif
+	    stride < 4) {
+		if (RADEON_DEBUG & DEBUG_FALLBACKS) {
+			fprintf(stderr, "%s: Converting vertex attributes, attribute data format %x,", __FUNCTION__, input->Type);
+			fprintf(stderr, "stride %d, components %d\n", stride, input->Size);
+		}
+
+		GLfloat *dst_ptr, *tmp;
+		tmp = dst_ptr = _mesa_malloc(sizeof(GLfloat) * input->Size * count);
+
+		switch (input->Type) {
+			case GL_DOUBLE:
+				CONVERT(GLdouble, (GLfloat));
+				break;
+			case GL_UNSIGNED_INT:
+				CONVERT(GLuint, UINT_TO_FLOAT);
+				break;
+			case GL_INT:
+				CONVERT(GLint, INT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_SHORT:
+				CONVERT(GLushort, USHORT_TO_FLOAT);
+				break;
+			case GL_SHORT:
+				CONVERT(GLshort, SHORT_TO_FLOAT);
+				break;
+			case GL_UNSIGNED_BYTE:
+				assert(input->Format != GL_BGRA);
+				CONVERT(GLubyte, UBYTE_TO_FLOAT);
+				break;
+			case GL_BYTE:
+				CONVERT(GLbyte, BYTE_TO_FLOAT);
+				break;
+			default:
+				assert(0);
+				break;
+		}
+
+		type = GL_FLOAT;
+		r300_attr.free_needed = GL_TRUE;
+		r300_attr.data = tmp;
+		r300_attr.stride = sizeof(GLfloat) * input->Size;
+		r300_attr.dwords = input->Size;
+	} else {
+		type = input->Type;
+		r300_attr.free_needed = GL_FALSE;
+		r300_attr.data = (GLvoid *)src_ptr;
+		r300_attr.stride = stride;
+		r300_attr.dwords = (getTypeSize(type) * input->Size  + 3)/ 4;
+	}
+
+	r300_attr.size = input->Size;
+	r300_attr.element = attr;
+	r300_attr.dst_loc = vbuf->num_attribs;
+
+	switch (type) {
+		case GL_FLOAT:
+			switch (input->Size) {
+				case 1: r300_attr.data_type = R300_DATA_TYPE_FLOAT_1; break;
+				case 2: r300_attr.data_type = R300_DATA_TYPE_FLOAT_2; break;
+				case 3: r300_attr.data_type = R300_DATA_TYPE_FLOAT_3; break;
+				case 4: r300_attr.data_type = R300_DATA_TYPE_FLOAT_4; break;
+			}
+			r300_attr._signed = 0;
+			r300_attr.normalize = 0;
+			break;
+		case GL_SHORT:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_BYTE:
+			r300_attr._signed = 1;
+			r300_attr.normalize = input->Normalized;
+			r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+		case GL_UNSIGNED_SHORT:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			switch (input->Size) {
+				case 1:
+				case 2:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+					break;
+				case 3:
+				case 4:
+					r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+					break;
+			}
+			break;
+		case GL_UNSIGNED_BYTE:
+			r300_attr._signed = 0;
+			r300_attr.normalize = input->Normalized;
+			if (input->Format == GL_BGRA)
+				r300_attr.data_type = R300_DATA_TYPE_D3DCOLOR;
+			else
+				r300_attr.data_type = R300_DATA_TYPE_BYTE;
+			break;
+
+		default:
+		case GL_DOUBLE:
+		case GL_INT:
+		case GL_UNSIGNED_INT:
+			assert(0);
+			break;
+	}
+
+	switch (input->Size) {
+		case 4:
+			r300_attr.swizzle = SWIZZLE_XYZW;
+			break;
+		case 3:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			break;
+		case 2:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+		case 1:
+			r300_attr.swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+			break;
+	}
+
+	r300_attr.write_mask = MASK_XYZW;
+
+	vbuf->attribs[vbuf->num_attribs] = r300_attr;
+	++vbuf->num_attribs;
+}
+
+static void r300SetVertexFormat(GLcontext *ctx, const struct gl_client_array *arrays[], int count, struct gl_buffer_object **bo, GLuint *nr_bo)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+
+	{
+		int i, tmp;
+
+		tmp = r300->selected_vp->key.InputsRead;
+		i = 0;
+		vbuf->num_attribs = 0;
+		while (tmp) {
+			/* find first enabled bit */
+			while (!(tmp & 1)) {
+				tmp >>= 1;
+				++i;
+			}
+
+			r300TranslateAttrib(ctx, i, count, arrays[i], bo, nr_bo);
+
+			tmp >>= 1;
+			++i;
+		}
+	}
+
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback)
+		return;
+
+	{
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i],
+						vbuf->attribs[i].data, vbuf->attribs[i].dwords,
+						vbuf->attribs[i].stride, count);
+		}
+
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
+	}
+}
+
+static void r300FreeData(GLcontext *ctx, struct gl_buffer_object **bo, GLuint nr_bo)
+{
+	{
+		struct r300_vertex_buffer *vbuf = &R300_CONTEXT(ctx)->vbuf;
+		int i;
+
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			if (vbuf->attribs[i].free_needed)
+				_mesa_free(vbuf->attribs[i].data);
+		}
+	}
+
+	{
+		struct r300_index_buffer *ind_buf = &R300_CONTEXT(ctx)->ind_buf;
+		if (ind_buf->free_needed)
+			_mesa_free(ind_buf->ptr);
+	}
+
+	{
+		int i;
+
+		for (i = 0; i < nr_bo; ++i) {
+			ctx->Driver.UnmapBuffer(ctx, 0, bo[i]);
+		}
+	}
+}
+
+static GLboolean r300TryDrawPrims(GLcontext *ctx,
+					 const struct gl_client_array *arrays[],
+					 const struct _mesa_prim *prim,
+					 GLuint nr_prims,
+					 const struct _mesa_index_buffer *ib,
+					 GLuint min_index,
+					 GLuint max_index )
+{
+	struct r300_context *r300 = R300_CONTEXT(ctx);
+	struct gl_buffer_object *bo[VERT_ATTRIB_MAX+1];
+	GLuint i, nr_bo = 0;
+
+	if (ctx->NewState)
+		_mesa_update_state( ctx );
+
+	if (r300->options.hw_tcl_enabled)
+		_tnl_UpdateFixedFunctionProgram(ctx);
+
+	r300UpdateShaders(r300);
+
+	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
+
+	r300FixupIndexBuffer(ctx, ib, bo, &nr_bo);
+
+	r300SetVertexFormat(ctx, arrays, max_index + 1, bo, &nr_bo);
+
+	if (r300->fallback)
+		return GL_FALSE;
+
+	r300SetupVAP(ctx, r300->selected_vp->key.InputsRead, r300->selected_vp->key.OutputsWritten);
+
+	r300UpdateShaderStates(r300);
+
+	r300EmitCacheFlush(r300);
+	radeonEmitState(&r300->radeon);
+
+	for (i = 0; i < nr_prims; ++i) {
+		r300RunRenderPrimitive(ctx, prim[i].start, prim[i].start + prim[i].count, prim[i].mode);
+	}
+
+	r300EmitCacheFlush(r300);
+
+	radeonReleaseArrays(ctx, ~0);
+
+	r300FreeData(ctx, bo, nr_bo);
+
+	return GL_TRUE;
+}
+
+/* TODO: rebase if number of indices in any of primitives is > 8192 for 32bit indices or 16384 for 16bit indices */
+
+static void r300DrawPrims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+	GLboolean retval;
+
+	if (min_index) {
+		vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, min_index, max_index, r300DrawPrims );
+		return;
+	}
+
+	/* Make an attempt at drawing */
+	retval = r300TryDrawPrims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+
+	/* If failed run tnl pipeline - it should take care of fallbacks */
+	if (!retval)
+		_tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
+
+void r300InitDraw(GLcontext *ctx)
+{
+	struct vbo_context *vbo = vbo_context(ctx);
+
+	vbo->draw_prims = r300DrawPrims;
+}
diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index 4fd6ba9b91..c3817721dc 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -31,6 +31,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
  * \file
  *
  * \author Keith Whitwell <keith@tungstengraphics.com>
+ * \author Maciej Cencora <m.cencora@gmail.com>
  */
 
 #include "main/glheader.h"
@@ -49,74 +50,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_state.h"
 #include "r300_emit.h"
 #include "r300_ioctl.h"
-
-
-#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
-    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
-    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
-    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
-    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
-    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
-#error Cannot change these!
-#endif
-
-#define DEBUG_ALL DEBUG_VERTS
-
-#define DW_SIZE(x) ((inputs[tab[(x)]] << R300_DST_VEC_LOC_SHIFT) |	\
-		    (attribptr[tab[(x)]]->size - 1) << R300_DATA_TYPE_0_SHIFT)
-
-GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr)
-{
-	GLuint i, dw;
-
-	/* type, inputs, stop bit, size */
-	for (i = 0; i < nr; i += 2) {
-		/* make sure input is valid, would lockup the gpu */
-		assert(inputs[tab[i]] != -1);
-		dw = (R300_SIGNED | DW_SIZE(i));
-		if (i + 1 == nr) {
-			dw |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
-		} else {
-			assert(inputs[tab[i + 1]] != -1);
-			dw |= (R300_SIGNED |
-			       DW_SIZE(i + 1)) << R300_DATA_TYPE_1_SHIFT;
-			if (i + 2 == nr) {
-				dw |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
-			}
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
-
-static GLuint r300VAPInputRoute1Swizzle(int swizzle[4])
-{
-	return (swizzle[0] << R300_SWIZZLE_SELECT_X_SHIFT) |
-	    (swizzle[1] << R300_SWIZZLE_SELECT_Y_SHIFT) |
-	    (swizzle[2] << R300_SWIZZLE_SELECT_Z_SHIFT) |
-	    (swizzle[3] << R300_SWIZZLE_SELECT_W_SHIFT);
-}
-
-GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr)
-{
-	GLuint i, dw;
-
-	for (i = 0; i < nr; i += 2) {
-		dw = (r300VAPInputRoute1Swizzle(swizzle[i]) |
-		      ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-			R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE0_SHIFT;
-		if (i + 1 < nr) {
-			dw |= (r300VAPInputRoute1Swizzle(swizzle[i + 1]) |
-			       ((R300_WRITE_ENA_X | R300_WRITE_ENA_Y |
-				 R300_WRITE_ENA_Z | R300_WRITE_ENA_W) << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
-		}
-		dst[i >> 1] = dw;
-	}
-
-	return (nr + 1) >> 1;
-}
+#include "r300_render.h"
+#include "r300_swtcl.h"
 
 GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 {
@@ -127,7 +62,6 @@ GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead)
 
 GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	GLuint i, vic_1 = 0;
 
 	if (InputsRead & (1 << VERT_ATTRIB_POS))
@@ -139,213 +73,99 @@ GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead)
 	if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
 		vic_1 |= R300_INPUT_CNTL_COLOR;
 
-	rmesa->state.texture.tc_count = 0;
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
 		if (InputsRead & (1 << (VERT_ATTRIB_TEX0 + i))) {
-			rmesa->state.texture.tc_count++;
 			vic_1 |= R300_INPUT_CNTL_TC0 << i;
 		}
 
 	return vic_1;
 }
 
-GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
 {
 	GLuint ret = 0;
 
-	if (OutputsWritten & (1 << VERT_RESULT_HPOS))
+	if (vp_writes & (1 << VERT_RESULT_HPOS))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL0))
+	if (vp_writes & (1 << VERT_RESULT_COL0) && fp_reads & FRAG_BIT_COL0)
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_COL1))
+	if (vp_writes & (1 << VERT_RESULT_COL1) && fp_reads & FRAG_BIT_COL1)
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_BFC0)
-	    || OutputsWritten & (1 << VERT_RESULT_BFC1))
-		ret |=
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT |
-		    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
+	/* Two sided lighting works only if all 4 colors are written */
+	if (vp_writes & (1 << VERT_RESULT_BFC0) || vp_writes & (1 << VERT_RESULT_BFC1))
+		ret |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_1_PRESENT |
+			   R300_VAP_OUTPUT_VTX_FMT_0__COLOR_2_PRESENT | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_3_PRESENT;
 
-	if (OutputsWritten & (1 << VERT_RESULT_PSIZ))
+	if (vp_writes & (1 << VERT_RESULT_PSIZ))
 		ret |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
 
 	return ret;
 }
 
-GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
+GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads)
 {
 	GLuint i, ret = 0, first_free_texcoord = 0;
 
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-		if (OutputsWritten & (1 << (VERT_RESULT_TEX0 + i))) {
-			ret |= (4 << (3 * i));
+		if (vp_writes & (1 << (VERT_RESULT_TEX0 + i)) && fp_reads & FRAG_BIT_TEX(i)) {
+			ret |= (4 << (3 * first_free_texcoord));
 			++first_free_texcoord;
 		}
 	}
 
-	if (OutputsWritten & (1 << VERT_RESULT_FOGC)) {
-		if (first_free_texcoord > 8) {
-			fprintf(stderr, "\tout of free texcoords to write fog coord\n");
-			_mesa_exit(-1);
-		}
-		ret |= 1 << (3 * first_free_texcoord);
+	if (fp_reads & FRAG_BIT_WPOS) {
+		ret |= (4 << (3 * first_free_texcoord));
+		++first_free_texcoord;
+	}
+
+	if (vp_writes & (1 << VERT_RESULT_FOGC) && fp_reads & FRAG_BIT_FOGC) {
+		ret |= 4 << (3 * first_free_texcoord);
+	}
+
+	if (first_free_texcoord > 8) {
+		fprintf(stderr, "\tout of free texcoords\n");
+		_mesa_exit(-1);
 	}
 
 	return ret;
 }
 
-/* Emit vertex data to GART memory
- * Route inputs to the vertex processor
- * This function should never return R300_FALLBACK_TCL when using software tcl.
- */
-int r300EmitArrays(GLcontext * ctx)
+GLboolean r300EmitArrays(GLcontext * ctx)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
-	GLuint nr;
-	GLuint count = vb->Count;
-	GLuint i;
-	GLuint InputsRead = 0, OutputsWritten = 0;
-	int *inputs = NULL;
-	int vir_inputs[VERT_ATTRIB_MAX];
-	GLint tab[VERT_ATTRIB_MAX];
-	int swizzle[VERT_ATTRIB_MAX][4];
-	struct r300_vertex_program *prog =
-	    (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-
-	if (hw_tcl_on) {
-		inputs = prog->inputs;
-		InputsRead = prog->key.InputsRead;
-		OutputsWritten = prog->key.OutputsWritten;
-	} else {
-		inputs = rmesa->state.sw_tcl_inputs;
-
-		DECLARE_RENDERINPUTS(render_inputs_bitset);
-		RENDERINPUTS_COPY(render_inputs_bitset, tnl->render_inputs_bitset);
-
-		vb->AttribPtr[VERT_ATTRIB_POS] = vb->ClipPtr;
-
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS));
-		assert(RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_NORMAL) == 0);
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_POS)) {
-			InputsRead |= 1 << VERT_ATTRIB_POS;
-			OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		}
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_vertex_buffer *vbuf = &r300->vbuf;
+	GLuint InputsRead, OutputsWritten;
 
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR0)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR0;
-			OutputsWritten |= 1 << VERT_RESULT_COL0;
-		}
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
 
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR1)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR1;
-			OutputsWritten |= 1 << VERT_RESULT_COL1;
-		}
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback & R300_RASTER_FALLBACK_MASK)
+		return GL_FALSE;
 
-		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_TEX(i))) {
-				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
-				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
-			}
-		}
+	{
+		struct vertex_buffer *mesa_vb = &TNL_CONTEXT(ctx)->vb;
+		GLuint attr, i;
 
-		for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-			if (InputsRead & (1 << i)) {
-				inputs[i] = nr++;
-			} else {
-				inputs[i] = -1;
-			}
+		for (i = 0; i < vbuf->num_attribs; i++) {
+			attr = vbuf->attribs[i].element;
+			rcommon_emit_vector(ctx, &r300->radeon.tcl.aos[i], mesa_vb->AttribPtr[attr]->data,
+					mesa_vb->AttribPtr[attr]->size, mesa_vb->AttribPtr[attr]->stride, mesa_vb->Count);
 		}
 
-		/* Fixed, apply to vir0 only */
-		memcpy(vir_inputs, inputs, VERT_ATTRIB_MAX * sizeof(int));
-		inputs = vir_inputs;
-		if (InputsRead & VERT_ATTRIB_POS)
-			inputs[VERT_ATTRIB_POS] = 0;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR0))
-			inputs[VERT_ATTRIB_COLOR0] = 2;
-		if (InputsRead & (1 << VERT_ATTRIB_COLOR1))
-			inputs[VERT_ATTRIB_COLOR1] = 3;
-		for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX7; i++)
-			if (InputsRead & (1 << i))
-				inputs[i] = 6 + (i - VERT_ATTRIB_TEX0);
-
-		RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, render_inputs_bitset);
-	}
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
 
-	assert(InputsRead);
-	assert(OutputsWritten);
-
-	for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-		if (InputsRead & (1 << i)) {
-			tab[nr++] = i;
-		}
+		/* Fill index buffer info */
+		r300->ind_buf.ptr = mesa_vb->Elts;
+		r300->ind_buf.is_32bit = GL_TRUE;
+		r300->ind_buf.free_needed = GL_FALSE;
 	}
 
-	if (nr > R300_MAX_AOS_ARRAYS) {
-		return R300_FALLBACK_TCL;
-	}
-
-	for (i = 0; i < nr; i++) {
-		int ci;
-
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
-		for (ci = 0; ci < vb->AttribPtr[tab[i]]->size; ci++) {
-			swizzle[i][ci] = ci;
-		}
-		rcommon_emit_vector(ctx, &rmesa->radeon.tcl.aos[i],
-				    vb->AttribPtr[tab[i]]->data,
-				    vb->AttribPtr[tab[i]]->size,
-				    vb->AttribPtr[tab[i]]->stride, count);
-	}
-
-	/* Setup INPUT_ROUTE. */
-	if (rmesa->radeon.radeonScreen->kernel_mm) {
-		R300_STATECHANGE(rmesa, vir[0]);
-		rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
-		rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
-		rmesa->hw.vir[0].cmd[0] |=
-			(r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-					    vb->AttribPtr, inputs, tab, nr) & 0x3FFF) << 16;
-		R300_STATECHANGE(rmesa, vir[1]);
-		rmesa->hw.vir[1].cmd[0] |=
-			(r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-					    nr) & 0x3FFF) << 16;
-	} else {
-		R300_STATECHANGE(rmesa, vir[0]);
-		((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count =
-			r300VAPInputRoute0(&rmesa->hw.vir[0].cmd[R300_VIR_CNTL_0],
-					   vb->AttribPtr, inputs, tab, nr);
-		R300_STATECHANGE(rmesa, vir[1]);
-		((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count =
-			r300VAPInputRoute1(&rmesa->hw.vir[1].cmd[R300_VIR_CNTL_0], swizzle,
-					   nr);
-	}
-	
-	/* Setup INPUT_CNTL. */
-	R300_STATECHANGE(rmesa, vic);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-
-	/* Setup OUTPUT_VTX_FMT. */
-	R300_STATECHANGE(rmesa, vof);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] =
-	    r300VAPOutputCntl0(ctx, OutputsWritten);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] =
-	    r300VAPOutputCntl1(ctx, OutputsWritten);
-
-	rmesa->radeon.tcl.aos_count = nr;
-
-	return R300_FALLBACK_NONE;
+	return GL_TRUE;
 }
 
 void r300EmitCacheFlush(r300ContextPtr rmesa)
diff --git a/src/mesa/drivers/dri/r300/r300_emit.h b/src/mesa/drivers/dri/r300/r300_emit.h
index 80c22d5e9a..2fb8b82d3a 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -216,19 +216,16 @@ void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags)
 	}
 }
 
-extern int r300EmitArrays(GLcontext * ctx);
+extern GLboolean r300EmitArrays(GLcontext * ctx);
 
 extern int r300PrimitiveType(r300ContextPtr rmesa, int prim);
 extern int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim);
 
 extern void r300EmitCacheFlush(r300ContextPtr rmesa);
 
-extern GLuint r300VAPInputRoute0(uint32_t * dst, GLvector4f ** attribptr,
-				 int *inputs, GLint * tab, GLuint nr);
-extern GLuint r300VAPInputRoute1(uint32_t * dst, int swizzle[][4], GLuint nr);
 extern GLuint r300VAPInputCntl0(GLcontext * ctx, GLuint InputsRead);
 extern GLuint r300VAPInputCntl1(GLcontext * ctx, GLuint InputsRead);
-extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint OutputsWritten);
-extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten);
+extern GLuint r300VAPOutputCntl0(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads);
+extern GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint vp_writes, GLuint fp_reads);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
index 32182bb667..55c1cfe631 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.c
@@ -25,32 +25,12 @@
  *
  */
 
-/**
- * \file
- *
- * Fragment program compiler. Perform transformations on the intermediate
- * representation until the program is in a form where we can translate
- * it more or less directly into machine-readable form.
- *
- * \author Ben Skeggs <darktama@iinet.net.au>
- * \author Jerome Glisse <j.glisse@gmail.com>
- */
+#include "r300_fragprog.h"
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
-#include "shader/prog_instruction.h"
 #include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
 
 #include "r300_context.h"
-#include "r300_fragprog.h"
 #include "r300_fragprog_swizzle.h"
-#include "r300_state.h"
-
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
 
 static void reset_srcreg(struct prog_src_register* reg)
 {
@@ -81,7 +61,7 @@ static struct prog_src_register shadow_ambient(struct gl_program *program, int t
  * \todo If/when r5xx uses the radeon_program architecture, this can probably
  * be reused.
  */
-static GLboolean transform_TEX(
+GLboolean r300_transform_TEX(
 	struct radeon_transform_context *t,
 	struct prog_instruction* orig_inst, void* data)
 {
@@ -160,6 +140,8 @@ static GLboolean transform_TEX(
 			inst.DstReg.Index = tempreg;
 			inst.DstReg.WriteMask = WRITEMASK_XYZW;
 			destredirect = GL_TRUE;
+		} else if (inst.SaturateMode) {
+			destredirect = GL_TRUE;
 		}
 	}
 
@@ -175,7 +157,7 @@ static GLboolean transform_TEX(
 		inst.SrcReg[0].File = PROGRAM_TEMPORARY;
 		inst.SrcReg[0].Index = tmpreg;
 	}
-	
+
 	tgt = radeonAppendInstructions(t->Program, 1);
 	_mesa_copy_instructions(tgt, &inst, 1);
 
@@ -214,9 +196,9 @@ static GLboolean transform_TEX(
 		 *   r  < tex  <=>      -tex+r < 0
 		 *   r >= tex  <=> not (-tex+r < 0 */
 		if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
-			tgt[1].SrcReg[2].NegateBase = tgt[0].SrcReg[2].NegateBase ^ NEGATE_XYZW;
+			tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW;
 		else
-			tgt[1].SrcReg[0].NegateBase = tgt[0].SrcReg[0].NegateBase ^ NEGATE_XYZW;
+			tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW;
 
 		tgt[2].Opcode = OPCODE_CMP;
 		tgt[2].DstReg = orig_inst->DstReg;
@@ -239,6 +221,7 @@ static GLboolean transform_TEX(
 
 		tgt->Opcode = OPCODE_MOV;
 		tgt->DstReg = orig_inst->DstReg;
+		tgt->SaturateMode = inst.SaturateMode;
 		tgt->SrcReg[0].File = PROGRAM_TEMPORARY;
 		tgt->SrcReg[0].Index = inst.DstReg.Index;
 	}
@@ -246,241 +229,10 @@ static GLboolean transform_TEX(
 	return GL_TRUE;
 }
 
-
-static void update_params(r300ContextPtr r300, struct r300_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-
-void r300TranslateFragmentShader(r300ContextPtr r300,
-				 struct r300_fragment_program *fp)
-{
-	struct r300_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r300_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformTrigSimple, 0 }
-		};
-		radeonLocalTransform(
-			r300->radeon.glCtx,
-			compiler.program,
-			3, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Fragment Program: After native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &r300FPIsNativeSwizzle,
-			.BuildSwizzle = &r300FPBuildSwizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		if (!r300FragmentProgramEmit(&compiler))
-			fp->error = GL_TRUE;
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, NULL);
-
-		if (!fp->error)
-			fp->translated = GL_TRUE;
-		if (fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
-			r300FragmentProgramDump(fp, &fp->code);
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-	}
-
-	update_params(r300, fp);
-}
-
 /* just some random things... */
-void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code)
+void r300FragmentProgramDump(union rX00_fragment_program_code *c)
 {
+	struct r300_fragment_program_code *code = &c->r300;
 	int n, i, j;
 	static int pc = 0;
 
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.h b/src/mesa/drivers/dri/r300/r300_fragprog.h
index 94fb554fb3..5ce6f33cee 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.h
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.h
@@ -33,9 +33,6 @@
 #ifndef __R300_FRAGPROG_H_
 #define __R300_FRAGPROG_H_
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
 #include "shader/program.h"
 #include "shader/prog_instruction.h"
 
@@ -105,28 +102,10 @@
 
 #endif
 
-struct r300_fragment_program;
+extern GLboolean r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
 
-extern void r300TranslateFragmentShader(r300ContextPtr r300,
-					struct r300_fragment_program *fp);
+extern void r300FragmentProgramDump(union rX00_fragment_program_code *c);
 
-
-/**
- * Used internally by the r300 fragment program code to store compile-time
- * only data.
- */
-struct r300_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r300_fragment_program *fp;
-	struct r300_fragment_program_code *code;
-	struct gl_program *program;
-};
-
-extern GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler);
-
-
-extern void r300FragmentProgramDump(
-	struct r300_fragment_program *fp,
-	struct r300_fragment_program_code *code);
+extern GLboolean r300_transform_TEX(struct radeon_transform_context *t, struct prog_instruction* orig_inst, void* data);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.c b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
new file mode 100644
index 0000000000..abc8757ba1
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.c
@@ -0,0 +1,291 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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, sublicense, 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 NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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
+ *
+ * Fragment program compiler. Perform transformations on the intermediate
+ * representation until the program is in a form where we can translate
+ * it more or less directly into machine-readable form.
+ *
+ * \author Ben Skeggs <darktama@iinet.net.au>
+ * \author Jerome Glisse <j.glisse@gmail.com>
+ */
+
+#include "r300_fragprog_common.h"
+
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "r300_state.h"
+#include "r300_fragprog.h"
+#include "r300_fragprog_swizzle.h"
+#include "r500_fragprog.h"
+
+#include "radeon_program.h"
+#include "radeon_program_alu.h"
+
+static void update_params(GLcontext *ctx, struct gl_fragment_program *fp)
+{
+	/* Ask Mesa nicely to fill in ParameterValues for us */
+	if (fp->Base.Parameters)
+		_mesa_load_state_parameters(ctx, fp->Base.Parameters);
+}
+
+static void nqssadce_init(struct nqssadce_state* s)
+{
+	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
+	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
+}
+
+/**
+ * Transform the program to support fragment.position.
+ *
+ * Introduce a small fragment at the start of the program that will be
+ * the only code that directly reads the FRAG_ATTRIB_WPOS input.
+ * All other code pieces that reference that input will be rewritten
+ * to read from a newly allocated temporary.
+ *
+ */
+static void insert_WPOS_trailer(struct r300_fragment_program_compiler *compiler)
+{
+	GLuint InputsRead = compiler->fp->Base.Base.InputsRead;
+
+	if (!(InputsRead & FRAG_BIT_WPOS))
+		return;
+
+	static gl_state_index tokens[STATE_LENGTH] = {
+		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
+	};
+	struct prog_instruction *fpi;
+	GLuint window_index;
+	int i = 0;
+	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
+
+	_mesa_insert_instructions(compiler->program, 0, 3);
+	fpi = compiler->program->Instructions;
+
+	/* perspective divide */
+	fpi[i].Opcode = OPCODE_RCP;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_W;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
+	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
+	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
+	i++;
+
+	fpi[i].Opcode = OPCODE_MUL;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
+	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
+	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
+	fpi[i].SrcReg[1].Index = tempregi;
+	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
+	i++;
+
+	/* viewport transformation */
+	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
+
+	fpi[i].Opcode = OPCODE_MAD;
+
+	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
+	fpi[i].DstReg.Index = tempregi;
+	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
+	fpi[i].DstReg.CondMask = COND_TR;
+
+	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
+	fpi[i].SrcReg[0].Index = tempregi;
+	fpi[i].SrcReg[0].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
+	fpi[i].SrcReg[1].Index = window_index;
+	fpi[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+
+	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
+	fpi[i].SrcReg[2].Index = window_index;
+	fpi[i].SrcReg[2].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
+	i++;
+
+	for (; i < compiler->program->NumInstructions; ++i) {
+		int reg;
+		for (reg = 0; reg < 3; reg++) {
+			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
+			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
+				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
+				fpi[i].SrcReg[reg].Index = tempregi;
+			}
+		}
+	}
+}
+
+static GLuint build_dtm(GLuint depthmode)
+{
+	switch(depthmode) {
+	default:
+	case GL_LUMINANCE: return 0;
+	case GL_INTENSITY: return 1;
+	case GL_ALPHA: return 2;
+	}
+}
+
+static GLuint build_func(GLuint comparefunc)
+{
+	return comparefunc - GL_NEVER;
+}
+
+/**
+ * Collect all external state that is relevant for compiling the given
+ * fragment program.
+ */
+static void build_state(
+	r300ContextPtr r300,
+	struct r300_fragment_program *fp,
+	struct r300_fragment_program_external_state *state)
+{
+	int unit;
+
+	_mesa_bzero(state, sizeof(*state));
+
+	for(unit = 0; unit < 16; ++unit) {
+		if (fp->Base.Base.ShadowSamplers & (1 << unit)) {
+			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
+
+			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
+			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
+		}
+	}
+}
+
+void r300TranslateFragmentShader(GLcontext *ctx, struct gl_fragment_program *fp)
+{
+	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)fp;
+	struct r300_fragment_program_external_state state;
+
+	build_state(r300, r300_fp, &state);
+	if (_mesa_memcmp(&r300_fp->state, &state, sizeof(state))) {
+		/* TODO: cache compiled programs */
+		r300_fp->translated = GL_FALSE;
+		_mesa_memcpy(&r300_fp->state, &state, sizeof(state));
+	}
+
+	if (!r300_fp->translated) {
+		struct r300_fragment_program_compiler compiler;
+
+		compiler.r300 = r300;
+		compiler.fp = r300_fp;
+		compiler.code = &r300_fp->code;
+		compiler.program = _mesa_clone_program(ctx, &fp->Base);
+
+		if (RADEON_DEBUG & DEBUG_PIXEL) {
+			fflush(stdout);
+			_mesa_printf("Fragment Program: Initial program:\n");
+			_mesa_print_program(compiler.program);
+			fflush(stdout);
+		}
+
+		insert_WPOS_trailer(&compiler);
+
+		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+			struct radeon_program_transformation transformations[] = {
+				{ &r500_transform_TEX, &compiler },
+				{ &radeonTransformALU, 0 },
+				{ &radeonTransformDeriv, 0 },
+				{ &radeonTransformTrigScale, 0 }
+			};
+			radeonLocalTransform(ctx, compiler.program, 4, transformations);
+		} else {
+			struct radeon_program_transformation transformations[] = {
+				{ &r300_transform_TEX, &compiler },
+				{ &radeonTransformALU, 0 },
+				{ &radeonTransformTrigSimple, 0 }
+			};
+			radeonLocalTransform(ctx, compiler.program, 3, transformations);
+		}
+
+		if (RADEON_DEBUG & DEBUG_PIXEL) {
+			_mesa_printf("Fragment Program: After native rewrite:\n");
+			_mesa_print_program(compiler.program);
+			fflush(stdout);
+		}
+
+		if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+			struct radeon_nqssadce_descr nqssadce = {
+				.Init = &nqssadce_init,
+				.IsNativeSwizzle = &r500FPIsNativeSwizzle,
+				.BuildSwizzle = &r500FPBuildSwizzle,
+				.RewriteDepthOut = GL_TRUE
+			};
+			radeonNqssaDce(ctx, compiler.program, &nqssadce);
+		} else {
+			struct radeon_nqssadce_descr nqssadce = {
+				.Init = &nqssadce_init,
+				.IsNativeSwizzle = &r300FPIsNativeSwizzle,
+				.BuildSwizzle = &r300FPBuildSwizzle,
+				.RewriteDepthOut = GL_TRUE
+			};
+			radeonNqssaDce(ctx, compiler.program, &nqssadce);
+		}
+
+		if (RADEON_DEBUG & DEBUG_PIXEL) {
+			_mesa_printf("Compiler: after NqSSA-DCE:\n");
+			_mesa_print_program(compiler.program);
+			fflush(stdout);
+		}
+
+		if (!r300->vtbl.BuildFragmentProgramHwCode(&compiler))
+			r300_fp->error = GL_TRUE;
+
+		/* Subtle: Rescue any parameters that have been added during transformations */
+		_mesa_free_parameter_list(fp->Base.Parameters);
+		fp->Base.Parameters = compiler.program->Parameters;
+		compiler.program->Parameters = 0;
+
+		_mesa_reference_program(ctx, &compiler.program, NULL);
+
+		r300_fp->translated = GL_TRUE;
+
+		r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+
+		if (r300_fp->error || (RADEON_DEBUG & DEBUG_PIXEL))
+			r300->vtbl.FragmentProgramDump(&r300_fp->code);
+	}
+
+	update_params(ctx, fp);
+}
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_common.h b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
new file mode 100644
index 0000000000..85ea86fecb
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_common.h
@@ -0,0 +1,35 @@
+/*
+ * Copyright (C) 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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, sublicense, 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 NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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 __R300_FRAGPROG_COMMON_H_
+#define __R300_FRAGPROG_COMMON_H_
+
+#include "main/mtypes.h"
+
+extern void r300TranslateFragmentShader(GLcontext *ctx, struct gl_fragment_program *fp);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c b/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
index 9f0b7e3534..b75656e7ee 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_emit.c
@@ -47,7 +47,7 @@
 
 #define PROG_CODE \
 	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
-	struct r300_fragment_program_code *code = c->code
+	struct r300_fragment_program_code *code = &c->code->r300
 
 #define error(fmt, args...) do {			\
 		fprintf(stderr, "%s::%s(): " fmt "\n",	\
@@ -66,7 +66,7 @@ static GLboolean emit_const(void* data, GLuint file, GLuint index, GLuint *hwind
 	}
 
 	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
+		if (*hwindex >= R300_PFS_NUM_CONST_REGS) {
 			error("Out of hw constants!\n");
 			return GL_FALSE;
 		}
@@ -138,7 +138,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->alu.length >= PFS_MAX_ALU_INST) {
+	if (code->alu.length >= R300_PFS_MAX_ALU_INST) {
 		error("Too many ALU instructions");
 		return GL_FALSE;
 	}
@@ -201,7 +201,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
 	if (inst->Alpha.DepthWriteMask) {
 		code->alu.inst[ip].inst3 |= R300_ALU_DSTA_DEPTH;
 		code->node[code->cur_node].flags |= R300_W_OUT;
-		c->fp->WritesDepth = GL_TRUE;
+		c->fp->writes_depth = GL_TRUE;
 	}
 
 	return GL_TRUE;
@@ -213,7 +213,7 @@ static GLboolean emit_alu(void* data, struct radeon_pair_instruction* inst)
  */
 static GLboolean finish_node(struct r300_fragment_program_compiler *c)
 {
-	struct r300_fragment_program_code *code = c->code;
+	struct r300_fragment_program_code *code = &c->code->r300;
 	struct r300_fragment_program_node *node = &code->node[code->cur_node];
 
 	if (node->alu_end < 0) {
@@ -275,7 +275,7 @@ static GLboolean emit_tex(void* data, struct prog_instruction* inst)
 {
 	PROG_CODE;
 
-	if (code->tex.length >= PFS_MAX_TEX_INST) {
+	if (code->tex.length >= R300_PFS_MAX_TEX_INST) {
 		error("Too many TEX instructions");
 		return GL_FALSE;
 	}
@@ -318,16 +318,16 @@ static const struct radeon_pair_handler pair_handler = {
 	.EmitPaired = &emit_alu,
 	.EmitTex = &emit_tex,
 	.BeginTexBlock = &begin_tex,
-	.MaxHwTemps = PFS_NUM_TEMP_REGS
+	.MaxHwTemps = R300_PFS_NUM_TEMP_REGS
 };
 
 /**
  * Final compilation step: Turn the intermediate radeon_program into
  * machine-readable instructions.
  */
-GLboolean r300FragmentProgramEmit(struct r300_fragment_program_compiler *compiler)
+GLboolean r300BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r300_fragment_program_code *code = compiler->code;
+	struct r300_fragment_program_code *code = &compiler->code->r300;
 
 	_mesa_bzero(code, sizeof(struct r300_fragment_program_code));
 	code->node[0].alu_end = -1;
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c b/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c
index a86d2bd471..fc9d855bce 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog_swizzle.c
@@ -92,7 +92,7 @@ static const struct swizzle_data* lookup_native_swizzle(GLuint swizzle)
 GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 {
 	if (reg.Abs)
-		reg.NegateBase = 0;
+		reg.Negate = NEGATE_NONE;
 
 	if (opcode == OPCODE_KIL ||
 	    opcode == OPCODE_TEX ||
@@ -100,7 +100,7 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 	    opcode == OPCODE_TXP) {
 		int j;
 
-		if (reg.Abs || reg.NegateBase != (15*reg.NegateAbs))
+		if (reg.Abs || reg.Negate)
 			return GL_FALSE;
 
 		for(j = 0; j < 4; ++j) {
@@ -121,7 +121,7 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 		if (GET_SWZ(reg.Swizzle, j) != SWIZZLE_NIL)
 			relevant |= 1 << j;
 
-	if ((reg.NegateBase & relevant) && (reg.NegateBase & relevant) != relevant)
+	if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant))
 		return GL_FALSE;
 
 	if (!lookup_native_swizzle(reg.Swizzle))
@@ -137,13 +137,12 @@ GLboolean r300FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src)
 {
 	if (src.Abs)
-		src.NegateBase = 0;
+		src.Negate = NEGATE_NONE;
 
 	while(dst.WriteMask) {
 		const struct swizzle_data *best_swizzle = 0;
 		GLuint best_matchcount = 0;
 		GLuint best_matchmask = 0;
-		GLboolean rgbnegate;
 		int i, comp;
 
 		for(i = 0; i < num_native_swizzles; ++i) {
@@ -157,6 +156,11 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst,
 				if (swz == SWIZZLE_NIL)
 					continue;
 				if (swz == GET_SWZ(sd->hash, comp)) {
+					/* check if the negate bit of current component
+					 * is the same for already matched components */
+					if (matchmask && (!!(src.Negate & matchmask) != !!(src.Negate & (1 << comp))))
+						continue;
+
 					matchcount++;
 					matchmask |= 1 << comp;
 				}
@@ -170,13 +174,6 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst,
 			}
 		}
 
-		if ((src.NegateBase & best_matchmask) != 0) {
-			best_matchmask &= src.NegateBase;
-			rgbnegate = !src.NegateAbs;
-		} else {
-			rgbnegate = src.NegateAbs;
-		}
-
 		struct prog_instruction *inst;
 
 		_mesa_insert_instructions(s->Program, s->IP, 1);
@@ -185,6 +182,7 @@ void r300FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst,
 		inst->DstReg = dst;
 		inst->DstReg.WriteMask &= (best_matchmask | WRITEMASK_W);
 		inst->SrcReg[0] = src;
+		inst->SrcReg[0].Negate = (best_matchmask & src.Negate) ? NEGATE_XYZW : NEGATE_NONE;
 		/* Note: We rely on NqSSA/DCE to set unused swizzle components to NIL */
 
 		dst.WriteMask &= ~inst->DstReg.WriteMask;
diff --git a/src/mesa/drivers/dri/r300/r300_ioctl.c b/src/mesa/drivers/dri/r300/r300_ioctl.c
index a7f5121da7..104079b4db 100644
--- a/src/mesa/drivers/dri/r300/r300_ioctl.c
+++ b/src/mesa/drivers/dri/r300/r300_ioctl.c
@@ -79,7 +79,7 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags,
 {
 	BATCH_LOCALS(&r300->radeon);
 	GLcontext *ctx = r300->radeon.glCtx;
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	GLuint cbpitch = 0;
 	r300ContextPtr rmesa = r300;
 
@@ -200,7 +200,7 @@ static void r300ClearBuffer(r300ContextPtr r300, int flags,
 		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[2]);
 		OUT_BATCH_FLOAT32(ctx->Color.ClearColor[3]);
 	}
-	
+
 	r300EmitCacheFlush(rmesa);
 	cp_wait(&r300->radeon, R300_WAIT_3D | R300_WAIT_3D_CLEAN);
 
@@ -213,14 +213,13 @@ static void r300EmitClearState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	BATCH_LOCALS(&r300->radeon);
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	int i;
-	int has_tcl = 1;
+	int has_tcl;
 	int is_r500 = 0;
 	GLuint vap_cntl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
 		is_r500 = 1;
@@ -448,7 +447,7 @@ static void r300EmitClearState(GLcontext * ctx)
 			R500_ALU_RGBA_G_SWIZ_0 |
 			R500_ALU_RGBA_B_SWIZ_0 |
 			R500_ALU_RGBA_A_SWIZ_0;
-		
+
 		r500fp.cmd[7] = 0;
 		emit_r500fp(ctx, &r500fp);
 	}
@@ -542,9 +541,9 @@ static void r300EmitClearState(GLcontext * ctx)
 }
 
 static void r300KernelClear(GLcontext *ctx, GLuint flags)
-{	  	
+{
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	struct radeon_framebuffer *rfb = dPriv->driverPrivate;
 	struct radeon_renderbuffer *rrb;
 	struct radeon_renderbuffer *rrbd;
@@ -566,7 +565,7 @@ static void r300KernelClear(GLcontext *ctx, GLuint flags)
 		r300ClearBuffer(r300, CLEARBUFFER_COLOR, rrb, NULL);
 		bits = 0;
 	}
-		
+
 	if (flags & BUFFER_BIT_FRONT_LEFT) {
 		rrb = radeon_get_renderbuffer(&rfb->base, BUFFER_FRONT_LEFT);
 		r300ClearBuffer(r300, bits | CLEARBUFFER_COLOR, rrb, rrbd);
@@ -591,7 +590,7 @@ static void r300KernelClear(GLcontext *ctx, GLuint flags)
 static void r300Clear(GLcontext * ctx, GLbitfield mask)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&r300->radeon);
 	const GLuint colorMask = *((GLuint *) & ctx->Color.ColorMask);
 	GLbitfield swrast_mask = 0, tri_mask = 0;
 	int i;
diff --git a/src/mesa/drivers/dri/r300/r300_reg.h b/src/mesa/drivers/dri/r300/r300_reg.h
index ed552d09bb..c22616b95f 100644
--- a/src/mesa/drivers/dri/r300/r300_reg.h
+++ b/src/mesa/drivers/dri/r300/r300_reg.h
@@ -2432,6 +2432,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 /* Z Buffer Clear Value */
 #define R300_ZB_DEPTHCLEARVALUE                  0x4f28
 
+#define R300_ZB_ZMASK_OFFSET                     0x4f30
+#define R300_ZB_ZMASK_PITCH                      0x4f34
+#define R300_ZB_ZMASK_WRINDEX                    0x4f38
+#define R300_ZB_ZMASK_DWORD                      0x4f3c
+#define R300_ZB_ZMASK_RDINDEX                    0x4f40
+
 /* Hierarchical Z Memory Offset */
 #define R300_ZB_HIZ_OFFSET                       0x4f44
 
@@ -3172,6 +3178,9 @@ enum {
 #   define R300_W_SRC_RAS				(1 << 2)
 
 
+/* Packet0 field ordering to write all values to the same reg */
+#define RADEON_ONE_REG_WR        (1 << 15)
+
 /* Draw a primitive from vertex data in arrays loaded via 3D_LOAD_VBPNTR.
  * Two parameter dwords:
  * 0. VAP_VTX_FMT: The first parameter is not written to hardware
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 924305dd12..1356305a21 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -50,6 +50,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * no bugs...
  */
 
+#include "r300_render.h"
+
 #include "main/glheader.h"
 #include "main/state.h"
 #include "main/imports.h"
@@ -72,8 +74,8 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_reg.h"
 #include "r300_tex.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
-extern int future_hw_tcl_on;
+#include "r300_fragprog_common.h"
+#include "r300_swtcl.h"
 
 /**
  * \brief Convert a OpenGL primitive type into a R300 primitive type.
@@ -170,16 +172,19 @@ int r300NumVerts(r300ContextPtr rmesa, int num_verts, int prim)
 	return num_verts - verts_off;
 }
 
-static void r300EmitElts(GLcontext * ctx, void *elts, unsigned long n_elts)
+static void r300EmitElts(GLcontext * ctx, unsigned long n_elts)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	void *out;
+	GLuint size;
+
+	size = ((rmesa->ind_buf.is_32bit ? 4 : 2) * n_elts + 3) & ~3;
 
 	radeonAllocDmaRegion(&rmesa->radeon, &rmesa->radeon.tcl.elt_dma_bo,
-			     &rmesa->radeon.tcl.elt_dma_offset, n_elts * 4, 4);
+			     &rmesa->radeon.tcl.elt_dma_offset, size, 4);
 	radeon_bo_map(rmesa->radeon.tcl.elt_dma_bo, 1);
 	out = rmesa->radeon.tcl.elt_dma_bo->ptr + rmesa->radeon.tcl.elt_dma_offset;
-	memcpy(out, elts, n_elts * 4);
+	memcpy(out, rmesa->ind_buf.ptr, size);
 	radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 }
 
@@ -187,14 +192,23 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 {
 	BATCH_LOCALS(&rmesa->radeon);
 
+    r300_emit_scissor(rmesa->radeon.glCtx);
 	if (vertex_count > 0) {
+		int size;
+
 		BEGIN_BATCH(10);
 		OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_INDX_2, 0);
-		OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
-			  ((vertex_count + 0) << 16) |
-			  type |
+		if (rmesa->ind_buf.is_32bit) {
+			size = vertex_count;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			  ((vertex_count + 0) << 16) | type |
 			  R300_VAP_VF_CNTL__INDEX_SIZE_32bit);
-		
+		} else {
+			size = (vertex_count + 1) >> 1;
+			OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_INDICES |
+			   ((vertex_count + 0) << 16) | type);
+		}
+
 		if (!rmesa->radeon.radeonScreen->kernel_mm) {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
 			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
@@ -203,13 +217,13 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 					rmesa->radeon.tcl.elt_dma_bo,
 					rmesa->radeon.tcl.elt_dma_offset,
 					RADEON_GEM_DOMAIN_GTT, 0, 0);
-			OUT_BATCH(vertex_count);
+			OUT_BATCH(size);
 		} else {
 			OUT_BATCH_PACKET3(R300_PACKET3_INDX_BUFFER, 2);
 			OUT_BATCH(R300_INDX_BUFFER_ONE_REG_WR | (0 << R300_INDX_BUFFER_SKIP_SHIFT) |
 	    			 (R300_VAP_PORT_IDX0 >> 2));
 			OUT_BATCH(rmesa->radeon.tcl.elt_dma_offset);
-			OUT_BATCH(vertex_count);
+			OUT_BATCH(size);
 			radeon_cs_write_reloc(rmesa->radeon.cmdbuf.cs,
 					      rmesa->radeon.tcl.elt_dma_bo,
 					      RADEON_GEM_DOMAIN_GTT, 0, 0);
@@ -224,12 +238,12 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 	uint32_t voffset;
 	int sz = 1 + (nr >> 1) * 3 + (nr & 1) * 2;
 	int i;
-	
+
 	if (RADEON_DEBUG & DEBUG_VERTS)
 		fprintf(stderr, "%s: nr=%d, ofs=0x%08x\n", __FUNCTION__, nr,
 			offset);
 
-    
+
 	if (!rmesa->radeon.radeonScreen->kernel_mm) {
 		BEGIN_BATCH(sz+2+(nr * 2));
 		OUT_BATCH_PACKET3(R300_PACKET3_3D_LOAD_VBPNTR, sz - 1);
@@ -240,7 +254,7 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 				  (rmesa->radeon.tcl.aos[i].stride << 8) |
 				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
 				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
-			
+
 			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
 				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
 			OUT_BATCH_RELOC(voffset,
@@ -256,7 +270,7 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 					RADEON_GEM_DOMAIN_GTT,
 					0, 0);
 		}
-		
+
 		if (nr & 1) {
 			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
 				  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
@@ -280,7 +294,7 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 				  (rmesa->radeon.tcl.aos[i].stride << 8) |
 				  (rmesa->radeon.tcl.aos[i + 1].components << 16) |
 				  (rmesa->radeon.tcl.aos[i + 1].stride << 24));
-			
+
 			voffset =  rmesa->radeon.tcl.aos[i + 0].offset +
 				offset * 4 * rmesa->radeon.tcl.aos[i + 0].stride;
 			OUT_BATCH(voffset);
@@ -288,7 +302,7 @@ static void r300EmitAOS(r300ContextPtr rmesa, GLuint nr, GLuint offset)
 				offset * 4 * rmesa->radeon.tcl.aos[i + 1].stride;
 			OUT_BATCH(voffset);
 		}
-		
+
 		if (nr & 1) {
 			OUT_BATCH((rmesa->radeon.tcl.aos[nr - 1].components << 0) |
 			  (rmesa->radeon.tcl.aos[nr - 1].stride << 8));
@@ -327,18 +341,18 @@ static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int type)
 {
 	BATCH_LOCALS(&rmesa->radeon);
 
+    r300_emit_scissor(rmesa->radeon.glCtx);
 	BEGIN_BATCH(3);
 	OUT_BATCH_PACKET3(R300_PACKET3_3D_DRAW_VBUF_2, 0);
 	OUT_BATCH(R300_VAP_VF_CNTL__PRIM_WALK_VERTEX_LIST | (vertex_count << 16) | type);
 	END_BATCH();
 }
 
-static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
-				   int start, int end, int prim)
+void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim)
 {
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	BATCH_LOCALS(&rmesa->radeon);
 	int type, num_verts;
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *vb = &tnl->vb;
 
 	type = r300PrimitiveType(rmesa, prim);
 	num_verts = r300NumVerts(rmesa, end - start, prim);
@@ -350,9 +364,9 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	 * This is supposed to ensure that we can get all rendering
 	 * commands into a single command buffer.
 	 */
-	rcommonEnsureCmdBufSpace(&rmesa->radeon, 64, __FUNCTION__);
+	rcommonEnsureCmdBufSpace(&rmesa->radeon, 128, __FUNCTION__);
 
-	if (vb->Elts) {
+	if (rmesa->ind_buf.ptr) {
 		if (num_verts > 65535) {
 			/* not implemented yet */
 			WARN_ONCE("Too many elts\n");
@@ -369,8 +383,14 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 		 * allocating the index array might actually evict the vertex
 		 * arrays. *sigh*
 		 */
-		r300EmitElts(ctx, vb->Elts, num_verts);
+		r300EmitElts(ctx, num_verts);
 		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start);
+		if (rmesa->radeon.radeonScreen->kernel_mm) {
+			BEGIN_BATCH_NO_AUTOSTATE(2);
+			OUT_BATCH_REGSEQ(R300_VAP_VF_MAX_VTX_INDX, 1);
+			OUT_BATCH(rmesa->radeon.tcl.aos[0].count);
+			END_BATCH();
+		}
 		r300FireEB(rmesa, num_verts, type);
 	} else {
 		r300EmitAOS(rmesa, rmesa->radeon.tcl.aos_count, start);
@@ -379,8 +399,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	COMMIT_BATCH();
 }
 
-static GLboolean r300RunRender(GLcontext * ctx,
-			       struct tnl_pipeline_stage *stage)
+static void r300RunRender(GLcontext * ctx, struct tnl_pipeline_stage *stage)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	int i;
@@ -391,8 +410,7 @@ static GLboolean r300RunRender(GLcontext * ctx,
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
 	r300UpdateShaders(rmesa);
-	if (r300EmitArrays(ctx))
-		return GL_TRUE;
+	r300EmitArrays(ctx);
 
 	r300UpdateShaderStates(rmesa);
 
@@ -403,77 +421,104 @@ static GLboolean r300RunRender(GLcontext * ctx,
 		GLuint prim = _tnl_translate_prim(&vb->Primitive[i]);
 		GLuint start = vb->Primitive[i].start;
 		GLuint end = vb->Primitive[i].start + vb->Primitive[i].count;
-		r300RunRenderPrimitive(rmesa, ctx, start, end, prim);
+		r300RunRenderPrimitive(ctx, start, end, prim);
 	}
 
 	r300EmitCacheFlush(rmesa);
 
 	radeonReleaseArrays(ctx, ~0);
+}
 
-	return GL_FALSE;
+
+static const char *getFallbackString(uint32_t bit)
+{
+	switch (bit) {
+		case R300_FALLBACK_VERTEX_PROGRAM :
+			return "vertex program";
+		case R300_FALLBACK_LINE_SMOOTH:
+			return "smooth lines";
+		case R300_FALLBACK_POINT_SMOOTH:
+			return "smooth points";
+		case R300_FALLBACK_POLYGON_SMOOTH:
+			return "smooth polygons";
+		case R300_FALLBACK_LINE_STIPPLE:
+			return "line stipple";
+		case R300_FALLBACK_POLYGON_STIPPLE:
+			return "polygon stipple";
+		case R300_FALLBACK_STENCIL_TWOSIDE:
+			return "two-sided stencil";
+		case R300_FALLBACK_RENDER_MODE:
+			return "render mode != GL_RENDER";
+		case R300_FALLBACK_FRAGMENT_PROGRAM:
+			return "fragment program";
+		case R300_FALLBACK_AOS_LIMIT:
+			return "aos limit";
+		case R300_FALLBACK_INVALID_BUFFERS:
+			return "invalid buffers";
+		default:
+			return "unknown";
+	}
 }
 
-#define FALLBACK_IF(expr)						\
-	do {								\
-		if (expr) {						\
-			if (1 || RADEON_DEBUG & DEBUG_FALLBACKS)	\
-				WARN_ONCE("Software fallback:%s\n",	\
-					  #expr);			\
-			return R300_FALLBACK_RAST;			\
-		}							\
-	} while(0)
-
-static int r300Fallback(GLcontext * ctx)
+void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 {
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
-	const unsigned back = ctx->Stencil._BackFace;
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	uint32_t old_fallback = rmesa->fallback;
+	static uint32_t fallback_warn = 0;
 	
-	FALLBACK_IF(r300->radeon.Fallback);
-	/* Do we need to use new-style shaders?
-	 * Also is there a better way to do this? */
-	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-		struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r500TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
+	if (mode) {
+		if ((fallback_warn & bit) == 0) {
+			_mesa_fprintf(stderr, "WARNING! Falling back to software for %s\n", getFallbackString(bit));
+			fallback_warn |= bit;
+		}
+		rmesa->fallback |= bit;
+
+		/* update only if we change from no tcl fallbacks to some tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if (((old_fallback & R300_TCL_FALLBACK_MASK) == 0) &&
+				((bit & R300_TCL_FALLBACK_MASK) > 0)) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] |= R300_VAP_TCL_BYPASS;
 			}
 		}
+
+		/* update only if we change from no raster fallbacks to some raster fallbacks */
+		if (((old_fallback & R300_RASTER_FALLBACK_MASK) == 0) &&
+			((bit & R300_RASTER_FALLBACK_MASK) > 0)) {
+			
+			radeon_firevertices(&rmesa->radeon);
+			rmesa->radeon.swtcl.RenderIndex = ~0;
+			_swsetup_Wakeup( ctx );
+		}
 	} else {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-		if (fp) {
-			if (!fp->translated) {
-				r300TranslateFragmentShader(r300, fp);
-				FALLBACK_IF(!fp->translated);
+		rmesa->fallback &= ~bit;
+
+		/* update only if we have disabled all tcl fallbacks */
+		if (rmesa->options.hw_tcl_enabled) {
+			if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+				R300_STATECHANGE(rmesa, vap_cntl_status);
+				rmesa->hw.vap_cntl_status.cmd[1] &= ~R300_VAP_TCL_BYPASS;
 			}
 		}
-	}
 
-	FALLBACK_IF(ctx->RenderMode != GL_RENDER);
-
-	/* If GL_EXT_stencil_two_side is disabled, this fallback check can
-	 * be removed.
-	 */
-	FALLBACK_IF(ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
-		    || ctx->Stencil.ValueMask[0] !=
-		    ctx->Stencil.ValueMask[back]
-		    || ctx->Stencil.WriteMask[0] !=
-		    ctx->Stencil.WriteMask[back]);
-
-	if (ctx->Extensions.NV_point_sprite || ctx->Extensions.ARB_point_sprite)
-		FALLBACK_IF(ctx->Point.PointSprite);
-
-	if (!r300->disable_lowimpact_fallback) {
-		FALLBACK_IF(ctx->Polygon.StippleFlag);
-		FALLBACK_IF(ctx->Multisample._Enabled);
-		FALLBACK_IF(ctx->Line.StippleFlag);
-		FALLBACK_IF(ctx->Line.SmoothFlag);
-		FALLBACK_IF(ctx->Point.SmoothFlag);
+		/* update only if we have disabled all raster fallbacks */
+		if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
+			_swrast_flush( ctx );
+			
+			tnl->Driver.Render.Start = r300RenderStart;
+			tnl->Driver.Render.Finish = r300RenderFinish;
+			tnl->Driver.Render.PrimitiveNotify = r300RenderPrimitive;
+			tnl->Driver.Render.ResetLineStipple = r300ResetLineStipple;
+			tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+			tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+			tnl->Driver.Render.Interp = _tnl_interp;
+			
+			_tnl_invalidate_vertex_state( ctx, ~0 );
+			_tnl_invalidate_vertices( ctx, ~0 );
+		}
 	}
-
-	return R300_FALLBACK_NONE;
+	
 }
 
 static GLboolean r300RunNonTCLRender(GLcontext * ctx,
@@ -484,49 +529,15 @@ static GLboolean r300RunNonTCLRender(GLcontext * ctx,
 	if (RADEON_DEBUG & DEBUG_PRIMS)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 
-	if (r300Fallback(ctx) >= R300_FALLBACK_RAST)
+	if (rmesa->fallback & R300_RASTER_FALLBACK_MASK)
 		return GL_TRUE;
 
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
- 	        return GL_TRUE;
-
-	if (!r300ValidateBuffers(ctx))
-	    return GL_TRUE;
-	
-	return r300RunRender(ctx, stage);
-}
-
-static GLboolean r300RunTCLRender(GLcontext * ctx,
-				  struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	struct r300_vertex_program *vp;
-
-	hw_tcl_on = future_hw_tcl_on;
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (hw_tcl_on == GL_FALSE)
-		return GL_TRUE;
-
-	if (r300Fallback(ctx) >= R300_FALLBACK_TCL) {
-		hw_tcl_on = GL_FALSE;
+	if (rmesa->options.hw_tcl_enabled == GL_FALSE)
 		return GL_TRUE;
-	}
 
-	if (!r300ValidateBuffers(ctx))
-	    return GL_TRUE;
-	
-	r300UpdateShaders(rmesa);
-
-	vp = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	if (vp->native == GL_FALSE) {
-		hw_tcl_on = GL_FALSE;
-		return GL_TRUE;
-	}
+	r300RunRender(ctx, stage);
 
-	return r300RunRender(ctx, stage);
+	return GL_FALSE;
 }
 
 const struct tnl_pipeline_stage _r300_render_stage = {
@@ -537,12 +548,3 @@ const struct tnl_pipeline_stage _r300_render_stage = {
 	NULL,
 	r300RunNonTCLRender
 };
-
-const struct tnl_pipeline_stage _r300_tcl_stage = {
-	"r300 Hardware Transform, Clipping and Lighting",
-	NULL,
-	NULL,
-	NULL,
-	NULL,
-	r300RunTCLRender
-};
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
new file mode 100644
index 0000000000..ec785474a6
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -0,0 +1,69 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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, sublicense, 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 NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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 __R300_RENDER_H__
+#define __R300_RENDER_H__
+
+#include "main/mtypes.h"
+
+#define R300_FALLBACK_VERTEX_PROGRAM    (1 << 0)
+#define R300_TCL_FALLBACK_MASK           0x0000ffff
+
+#define R300_FALLBACK_LINE_SMOOTH       (1 << 16)
+#define R300_FALLBACK_POINT_SMOOTH      (1 << 17)
+#define R300_FALLBACK_POLYGON_SMOOTH    (1 << 18)
+#define R300_FALLBACK_LINE_STIPPLE      (1 << 19)
+#define R300_FALLBACK_POLYGON_STIPPLE   (1 << 20)
+#define R300_FALLBACK_STENCIL_TWOSIDE   (1 << 21)
+#define R300_FALLBACK_RENDER_MODE       (1 << 22)
+#define R300_FALLBACK_FRAGMENT_PROGRAM  (1 << 23)
+#define R300_FALLBACK_AOS_LIMIT         (1 << 30)
+#define R300_FALLBACK_INVALID_BUFFERS   (1 << 31)
+#define R300_RASTER_FALLBACK_MASK        0xffff0000
+
+#define MASK_XYZW (R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W)
+#define MASK_X R300_WRITE_ENA_X
+#define MASK_Y R300_WRITE_ENA_Y
+#define MASK_Z R300_WRITE_ENA_Z
+#define MASK_W R300_WRITE_ENA_W
+
+#if SWIZZLE_X != R300_INPUT_ROUTE_SELECT_X || \
+    SWIZZLE_Y != R300_INPUT_ROUTE_SELECT_Y || \
+    SWIZZLE_Z != R300_INPUT_ROUTE_SELECT_Z || \
+    SWIZZLE_W != R300_INPUT_ROUTE_SELECT_W || \
+    SWIZZLE_ZERO != R300_INPUT_ROUTE_SELECT_ZERO || \
+    SWIZZLE_ONE != R300_INPUT_ROUTE_SELECT_ONE
+#error Cannot change these!
+#endif
+
+extern const struct tnl_pipeline_stage _r300_render_stage;
+
+extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
+
+extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim);
+
+#endif
diff --git a/src/mesa/drivers/dri/r300/r300_shader.c b/src/mesa/drivers/dri/r300/r300_shader.c
index f30fd986e0..0133b83796 100644
--- a/src/mesa/drivers/dri/r300/r300_shader.c
+++ b/src/mesa/drivers/dri/r300/r300_shader.c
@@ -1,18 +1,42 @@
+/*
+ * Copyright 2009 Maciej Cencora <m.cencora@gmail.com>
+ *
+ * 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, sublicense, 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 NONINFRINGEMENT.
+ * IN NO EVENT SHALL THE COPYRIGHT OWNER(S) 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 "main/glheader.h"
 
 #include "shader/program.h"
 #include "tnl/tnl.h"
 #include "r300_context.h"
-#include "r300_fragprog.h"
+#include "r300_fragprog_common.h"
 
 static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 					 GLuint id)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program_cont *vp;
-	struct r300_fragment_program *r300_fp;
-	struct r500_fragment_program *r500_fp;
+	struct r300_fragment_program *fp;
 
 	switch (target) {
 	case GL_VERTEX_STATE_PROGRAM_NV:
@@ -20,28 +44,12 @@ static struct gl_program *r300NewProgram(GLcontext * ctx, GLenum target,
 		vp = CALLOC_STRUCT(r300_vertex_program_cont);
 		return _mesa_init_vertex_program(ctx, &vp->mesa_program,
 						 target, id);
-	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			r500_fp->ctx = ctx;
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
 
 	case GL_FRAGMENT_PROGRAM_NV:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
-			r500_fp = CALLOC_STRUCT(r500_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r500_fp->mesa_program,
-							   target, id);
-		} else {
-			r300_fp = CALLOC_STRUCT(r300_fragment_program);
-			return _mesa_init_fragment_program(ctx, &r300_fp->mesa_program,
-							   target, id);
-		}
+	case GL_FRAGMENT_PROGRAM_ARB:
+		fp = CALLOC_STRUCT(r300_fragment_program);
+		return _mesa_init_fragment_program(ctx, &fp->Base, target, id);
+
 	default:
 		_mesa_problem(ctx, "Bad target in r300NewProgram");
 	}
@@ -57,20 +65,15 @@ static void r300DeleteProgram(GLcontext * ctx, struct gl_program *prog)
 static void
 r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct r300_vertex_program_cont *vp = (void *)prog;
 	struct r300_fragment_program *r300_fp = (struct r300_fragment_program *)prog;
-	struct r500_fragment_program *r500_fp = (struct r500_fragment_program *)prog;
 
 	switch (target) {
 	case GL_VERTEX_PROGRAM_ARB:
 		vp->progs = NULL;
 		break;
 	case GL_FRAGMENT_PROGRAM_ARB:
-		if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-			r500_fp->translated = GL_FALSE;
-		else
-			r300_fp->translated = GL_FALSE;
+		r300_fp->translated = GL_FALSE;
 		break;
 	}
 
@@ -81,7 +84,14 @@ r300ProgramStringNotify(GLcontext * ctx, GLenum target, struct gl_program *prog)
 static GLboolean
 r300IsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
 {
-	return GL_TRUE;
+	if (target == GL_FRAGMENT_PROGRAM_ARB) {
+		struct r300_fragment_program *fp = (struct r300_fragment_program *)prog;
+		if (!fp->translated)
+			r300TranslateFragmentShader(ctx, &fp->Base);
+
+		return !fp->error;
+	} else
+		return GL_TRUE;
 }
 
 void r300InitShaderFuncs(struct dd_function_table *functions)
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 46c3df7099..c0eda977db 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -53,20 +53,22 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "shader/prog_statevars.h"
 #include "vbo/vbo.h"
 #include "tnl/tnl.h"
+#include "tnl/t_vp_build.h"
 
 #include "r300_context.h"
 #include "r300_ioctl.h"
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_emit.h"
-#include "r300_fragprog.h"
 #include "r300_tex.h"
+#include "r300_fragprog_common.h"
+#include "r300_fragprog.h"
+#include "r500_fragprog.h"
+#include "r300_render.h"
+#include "r300_vertprog.h"
 
 #include "drirenderbuffer.h"
 
-extern int future_hw_tcl_on;
-extern void _tnl_UpdateFixedFunctionProgram(GLcontext * ctx);
-
 static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4])
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
@@ -365,7 +367,7 @@ static void r300ClipPlane( GLcontext *ctx, GLenum plane, const GLfloat *eq )
 	GLint *ip;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!rmesa->options.hw_tcl_enabled)
 			return;
 
 	p = (GLint) plane - (GLint) GL_CLIP_PLANE0;
@@ -384,7 +386,7 @@ static void r300SetClipPlaneState(GLcontext * ctx, GLenum cap, GLboolean state)
 	GLuint p;
 
 	/* no VAP UCP on non-TCL chipsets */
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (!r300->options.hw_tcl_enabled)
 		return;
 
 	p = cap - GL_CLIP_PLANE0;
@@ -450,28 +452,22 @@ static void r300SetPolygonOffsetState(GLcontext * ctx, GLboolean state)
 
 static GLboolean current_fragment_program_writes_depth(GLcontext* ctx)
 {
-	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 
-	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		struct r300_fragment_program *fp = (struct r300_fragment_program *)
-			(char *)ctx->FragmentProgram._Current;
-		return (fp && fp->WritesDepth);
-	} else {
-		struct r500_fragment_program* fp =
-			(struct r500_fragment_program*)(char*)
-			ctx->FragmentProgram._Current;
-		return (fp && fp->writes_depth);
-	}
+	return (fp && fp->writes_depth);
 }
 
 static void r300SetEarlyZState(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	GLuint topZ = R300_ZTOP_ENABLE;
+	GLuint w_fmt, fgdepthsrc;
 
 	if (ctx->Color.AlphaEnabled && ctx->Color.AlphaFunc != GL_ALWAYS)
 		topZ = R300_ZTOP_DISABLE;
-	if (current_fragment_program_writes_depth(ctx))
+	else if (current_fragment_program_writes_depth(ctx))
+		topZ = R300_ZTOP_DISABLE;
+	else if (ctx->FragmentProgram._Current && ctx->FragmentProgram._Current->UsesKill)
 		topZ = R300_ZTOP_DISABLE;
 
 	if (topZ != r300->hw.zstencil_format.cmd[2]) {
@@ -482,6 +478,26 @@ static void r300SetEarlyZState(GLcontext * ctx)
 		R300_STATECHANGE(r300, zstencil_format);
 		r300->hw.zstencil_format.cmd[2] = topZ;
 	}
+
+	/* w_fmt value is set to get best performance
+	* see p.130 R5xx 3D acceleration guide v1.3 */
+	if (current_fragment_program_writes_depth(ctx)) {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
+		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
+	} else {
+		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
+		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
+	}
+
+	if (w_fmt != r300->hw.us_out_fmt.cmd[5]) {
+		R300_STATECHANGE(r300, us_out_fmt);
+		r300->hw.us_out_fmt.cmd[5] = w_fmt;
+	}
+
+	if (fgdepthsrc != r300->hw.fg_depth_src.cmd[1]) {
+		R300_STATECHANGE(r300, fg_depth_src);
+		r300->hw.fg_depth_src.cmd[1] = fgdepthsrc;
+	}
 }
 
 static void r300SetAlphaState(GLcontext * ctx)
@@ -532,8 +548,6 @@ static void r300SetAlphaState(GLcontext * ctx)
 	R300_STATECHANGE(r300, at);
 	r300->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc;
 	r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
-
-	r300SetEarlyZState(ctx);
 }
 
 static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref)
@@ -581,14 +595,28 @@ static void r300SetDepthState(GLcontext * ctx)
 		r300->hw.zs.cmd[R300_ZS_CNTL_1] |=
 		    translate_func(ctx->Depth.Func) << R300_Z_FUNC_SHIFT;
 	}
+}
 
-	r300SetEarlyZState(ctx);
+static void r300CatchStencilFallback(GLcontext *ctx)
+{
+	const unsigned back = ctx->Stencil._BackFace;
+
+	if (ctx->Stencil._Enabled && (ctx->Stencil.Ref[0] != ctx->Stencil.Ref[back]
+		|| ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[back]
+		|| ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[back])) {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_TRUE);
+	} else {
+		r300SwitchFallback(ctx, R300_FALLBACK_STENCIL_TWOSIDE, GL_FALSE);
+	}
 }
 
 static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	GLboolean hw_stencil = GL_FALSE;
+
+	r300CatchStencilFallback(ctx);
+
 	if (ctx->DrawBuffer) {
 		struct radeon_renderbuffer *rrbStencil
 			= radeon_get_renderbuffer(ctx->DrawBuffer, BUFFER_STENCIL);
@@ -604,10 +632,6 @@ static void r300SetStencilState(GLcontext * ctx, GLboolean state)
 			r300->hw.zs.cmd[R300_ZS_CNTL_0] &=
 			    ~R300_STENCIL_ENABLE;
 		}
-	} else {
-#if R200_MERGED
-		FALLBACK(&r300->radeon, RADEON_FALLBACK_STENCIL, state);
-#endif
 	}
 }
 
@@ -740,7 +764,12 @@ static void r300ColorMask(GLcontext * ctx,
 static void r300PointSize(GLcontext * ctx, GLfloat size)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        /* same size limits for AA, non-AA points */
+
+	/* We need to clamp to user defined range here, because
+	 * the HW clamping happens only for per vertex point size. */
+	size = CLAMP(size, ctx->Point.MinSize, ctx->Point.MaxSize);
+
+	/* same size limits for AA, non-AA points */
 	size = CLAMP(size, ctx->Const.MinPointSize, ctx->Const.MaxPointSize);
 
 	R300_STATECHANGE(r300, ps);
@@ -833,29 +862,33 @@ static void r300ShadeModel(GLcontext * ctx, GLenum mode)
 
 	R300_STATECHANGE(rmesa, shade);
 	rmesa->hw.shade.cmd[1] = 0x00000002;
+	R300_STATECHANGE(rmesa, shade2);
 	switch (mode) {
 	case GL_FLAT:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_FLAT;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_FLAT;
 		break;
 	case GL_SMOOTH:
-		rmesa->hw.shade.cmd[2] = R300_RE_SHADE_MODEL_SMOOTH;
+		rmesa->hw.shade2.cmd[1] = R300_RE_SHADE_MODEL_SMOOTH;
 		break;
 	default:
 		return;
 	}
-	rmesa->hw.shade.cmd[3] = 0x00000000;
-	rmesa->hw.shade.cmd[4] = 0x00000000;
+	rmesa->hw.shade2.cmd[2] = 0x00000000;
+	rmesa->hw.shade2.cmd[3] = 0x00000000;
 }
 
 static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 				    GLenum func, GLint ref, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint refmask =
-	    ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
-	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
-	const unsigned back = ctx->Stencil._BackFace;
+	GLuint refmask;
 	GLuint flag;
+	const unsigned back = ctx->Stencil._BackFace;
+
+	r300CatchStencilFallback(ctx);
+
+	refmask = ((ctx->Stencil.Ref[0] & 0xff) << R300_STENCILREF_SHIFT)
+	     | ((ctx->Stencil.ValueMask[0] & 0xff) << R300_STENCILMASK_SHIFT);
 
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_0] |= R300_STENCIL_FRONT_BACK;
@@ -883,6 +916,8 @@ static void r300StencilMaskSeparate(GLcontext * ctx, GLenum face, GLuint mask)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &=
 	    ~(R300_STENCILREF_MASK <<
@@ -899,6 +934,8 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	const unsigned back = ctx->Stencil._BackFace;
 
+	r300CatchStencilFallback(ctx);
+
 	R300_STATECHANGE(rmesa, zs);
 	/* It is easier to mask what's left.. */
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &=
@@ -927,16 +964,10 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face,
  * Window position and viewport transformation
  */
 
-/*
- * To correctly position primitives:
- */
-#define SUBPIXEL_X 0.125
-#define SUBPIXEL_Y 0.125
-
 static void r300UpdateWindow(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
 	GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
@@ -953,9 +984,9 @@ static void r300UpdateWindow(GLcontext * ctx)
 	}
 
 	GLfloat sx = v[MAT_SX];
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
+	GLfloat tx = v[MAT_TX] + xoffset;
 	GLfloat sy = v[MAT_SY] * y_scale;
-	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias + SUBPIXEL_Y;
+	GLfloat ty = (v[MAT_TY] * y_scale) + y_bias;
 	GLfloat sz = v[MAT_SZ] * depthScale;
 	GLfloat tz = v[MAT_TZ] * depthScale;
 
@@ -989,13 +1020,13 @@ static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
 void r300UpdateViewportOffset(GLcontext * ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	__DRIdrawablePrivate *dPriv = ((radeonContextPtr) rmesa)->dri.drawable;
+	__DRIdrawablePrivate *dPriv = radeon_get_drawable(&rmesa->radeon);
 	GLfloat xoffset = (GLfloat) dPriv->x;
 	GLfloat yoffset = (GLfloat) dPriv->y + dPriv->h;
 	const GLfloat *v = ctx->Viewport._WindowMap.m;
 
-	GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
-	GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
+	GLfloat tx = v[MAT_TX] + xoffset;
+	GLfloat ty = (-v[MAT_TY]) + yoffset;
 
 	if (rmesa->hw.vpt.cmd[R300_VPT_XOFFSET] != r300PackFloat32(tx) ||
 	    rmesa->hw.vpt.cmd[R300_VPT_YOFFSET] != r300PackFloat32(ty)) {
@@ -1021,12 +1052,14 @@ r300FetchStateParameter(GLcontext * ctx,
 	switch (state[0]) {
 	case STATE_INTERNAL:
 		switch (state[1]) {
-		case STATE_R300_WINDOW_DIMENSION:
-			value[0] = r300->radeon.dri.drawable->w * 0.5f;	/* width*0.5 */
-			value[1] = r300->radeon.dri.drawable->h * 0.5f;	/* height*0.5 */
-			value[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
-			value[3] = 1.0F;	/* not used */
-			break;
+		case STATE_R300_WINDOW_DIMENSION: {
+				__DRIdrawablePrivate * drawable = radeon_get_drawable(&r300->radeon);
+				value[0] = drawable->w * 0.5f;	/* width*0.5 */
+				value[1] = drawable->h * 0.5f;	/* height*0.5 */
+				value[2] = 0.5F;	/* for moving range [-1 1] -> [0 1] */
+				value[3] = 1.0F;	/* not used */
+				break;
+			}
 
 		case STATE_R300_TEXRECT_FACTOR:{
 				struct gl_texture_object *t =
@@ -1066,14 +1099,14 @@ void r300UpdateStateParameters(GLcontext * ctx, GLuint new_state)
 	struct gl_program_parameter_list *paramList;
 	GLuint i;
 
-	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM)))
+	if (!(new_state & (_NEW_BUFFERS | _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS)))
 		return;
 
 	fp = (struct r300_fragment_program *)ctx->FragmentProgram._Current;
 	if (!fp)
 		return;
 
-	paramList = fp->mesa_program.Base.Parameters;
+	paramList = fp->Base.Base.Parameters;
 
 	if (!paramList)
 		return;
@@ -1192,9 +1225,8 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
 	int i;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r300_fragment_program_code *code = &fp->code;
+	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
+	struct r300_fragment_program_code *code = &fp->code.r300;
 
 	R300_STATECHANGE(r300, fpt);
 
@@ -1235,9 +1267,8 @@ static void r300SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 static void r500SetupFragmentShaderTextures(GLcontext *ctx, int *tmu_mappings)
 {
 	int i;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
-	struct r500_fragment_program_code *code = &fp->code;
+	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
+	struct r500_fragment_program_code *code = &fp->code.r500;
 
 	/* find all the texture instructions and relocate the texture units */
 	for (i = 0; i < code->inst_end + 1; i++) {
@@ -1388,11 +1419,8 @@ static void r300SetupTextures(GLcontext * ctx)
 	r300->hw.tex.border_color.cmd[R300_TEX_CMD_0] =
 	    cmdpacket0(r300->radeon.radeonScreen, R300_TX_BORDER_COLOR_0, last_hw_tmu + 1);
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
 	if (r300->radeon.radeonScreen->chip_family < CHIP_FAMILY_RV515) {
-		if (fp->mesa_program.UsesKill && last_hw_tmu < 0) {
+		if (fp->Base.UsesKill && last_hw_tmu < 0) {
 			// The KILL operation requires the first texture unit
 			// to be enabled.
 			r300->hw.txe.cmd[R300_TXE_ENABLE] |= 1;
@@ -1400,9 +1428,8 @@ static void r300SetupTextures(GLcontext * ctx)
 			r300->hw.tex.filter.cmd[R300_TEX_CMD_0] =
 				cmdpacket0(r300->radeon.radeonScreen, R300_TX_FILTER0_0, 1);
 		}
-		r300SetupFragmentShaderTextures(ctx, tmu_mappings);
-	} else
-		r500SetupFragmentShaderTextures(ctx, tmu_mappings);
+	}
+	r300->vtbl.SetupFragmentShaderTextures(ctx, tmu_mappings);
 
 	if (RADEON_DEBUG & DEBUG_STATE)
 		fprintf(stderr, "TX_ENABLE: %08x  last_hw_tmu=%d\n",
@@ -1421,26 +1448,21 @@ union r300_outputs_written {
 static void r300SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->key.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1459,15 +1481,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1479,15 +1493,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R300_RS_COL_PTR(col_ip) | R300_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R300_RS_INST_COL_ID(col_ip) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1497,6 +1503,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1506,26 +1513,10 @@ static void r300SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-			count = 4;
-		else
-			count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		switch(count) {
-		case 4: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3); break;
-		case 3: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		default:
-		case 1:
-		case 2: swiz = R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(R300_RS_SEL_K0) | R300_RS_SEL_Q(R300_RS_SEL_K1); break;
-		};
-
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz | R300_RS_TEX_PTR(rs_tex_count);
+		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_S(0) | R300_RS_SEL_T(1) | R300_RS_SEL_R(2) | R300_RS_SEL_Q(3) | R300_RS_TEX_PTR(rs_tex_count);
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
 		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
+		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
 	}
@@ -1545,7 +1536,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= R300_RS_SEL_Q(R300_RS_SEL_K1) | R300_RS_TEX_PTR(rs_tex_count);
 			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R300_RS_INST_TEX_ID(tex_ip) | R300_RS_INST_TEX_CN_WRITE | R300_RS_INST_TEX_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_FOGC;
-			rs_tex_count += 1;
+			rs_tex_count += 4;
 			++tex_ip;
 			++fp_reg;
 		} else {
@@ -1555,7 +1546,8 @@ static void r300SetupRSUnit(GLcontext * ctx)
 
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_CN_WRITE | R300_RS_INST_COL_ADDR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R300_RS_INST_COL_ID(0) | R300_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R300_RS_COL_PTR(0) | R300_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
@@ -1564,6 +1556,7 @@ static void r300SetupRSUnit(GLcontext * ctx)
 	r300->hw.rc.cmd[2] |= high_rr - 1;
 
 	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R300_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
@@ -1572,26 +1565,21 @@ static void r300SetupRSUnit(GLcontext * ctx)
 static void r500SetupRSUnit(GLcontext * ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
-        TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
 	union r300_outputs_written OutputsWritten;
 	GLuint InputsRead;
 	int fp_reg, high_rr;
 	int col_ip, tex_ip;
 	int rs_tex_count = 0;
-	int i, count, col_fmt;
+	int i, col_fmt, hw_tcl_on;
+
+	hw_tcl_on = r300->options.hw_tcl_enabled;
 
 	if (hw_tcl_on)
-		OutputsWritten.vp_outputs = CURRENT_VERTEX_SHADER(ctx)->key.OutputsWritten;
+		OutputsWritten.vp_outputs = r300->selected_vp->key.OutputsWritten;
 	else
-		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->state.render_inputs_bitset);
+		RENDERINPUTS_COPY(OutputsWritten.index_bitset, r300->render_inputs_bitset);
 
-	if (ctx->FragmentProgram._Current)
-		InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
-	else {
-		fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
-		return;		/* This should only ever happen once.. */
-	}
+	InputsRead = ctx->FragmentProgram._Current->Base.InputsRead;
 
 	R300_STATECHANGE(r300, ri);
 	R300_STATECHANGE(r300, rc);
@@ -1610,15 +1598,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL0, _TNL_ATTRIB_COLOR0)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR0]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL0;
 			++col_ip;
@@ -1630,15 +1610,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 
 	if (InputsRead & FRAG_BIT_COL1) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_COL1, _TNL_ATTRIB_COLOR1)) {
-			count = VB->AttribPtr[_TNL_ATTRIB_COLOR1]->size;
-			if (count == 4)
-			    col_fmt = R300_RS_COL_FMT_RGBA;
-			else if (count == 3)
-			    col_fmt = R300_RS_COL_FMT_RGB1;
-			else
-			    col_fmt = R300_RS_COL_FMT_0001;
-
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(col_fmt);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + col_ip] = R500_RS_COL_PTR(col_ip) | R500_RS_COL_FMT(R300_RS_COL_FMT_RGBA);
 			r300->hw.rr.cmd[R300_RR_INST_0 + col_ip] = R500_RS_INST_COL_ID(col_ip) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_COL1;
 			++col_ip;
@@ -1648,7 +1620,7 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
-
+	/* We always route 4 texcoord components */
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (! ( InputsRead & FRAG_BIT_TEX(i) ) )
 		    continue;
@@ -1658,55 +1630,37 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		    continue;
 		}
 
-		int swiz = 0;
-
-		/* with TCL we always seem to route 4 components */
-		if (hw_tcl_on)
-		  count = 4;
-		else
-		  count = VB->AttribPtr[_TNL_ATTRIB_TEX(i)]->size;
-
-		if (count == 4) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= (rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 3) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= (rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 2) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= (rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else if (count == 1) {
-			swiz |= (rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		} else {
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_S_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT;
-			swiz |= R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT;
-			swiz |= R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT;
-		}
+		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
+			((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
+			((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
+			((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
 
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= swiz;
 		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
 		InputsRead &= ~(FRAG_BIT_TEX0 << i);
-		rs_tex_count += count;
+		rs_tex_count += 4;
+		++tex_ip;
+		++fp_reg;
+	}
+
+	if (InputsRead & FRAG_BIT_WPOS) {
+		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
+			((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
+			((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
+			((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
+
+		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
+		InputsRead &= ~FRAG_BIT_WPOS;
+		rs_tex_count += 4;
 		++tex_ip;
 		++fp_reg;
 	}
 
 	if (InputsRead & FRAG_BIT_FOGC) {
 		if (R300_OUTPUTS_WRITTEN_TEST(OutputsWritten, VERT_RESULT_FOGC, _TNL_ATTRIB_FOG)) {
-			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
+			r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= (rs_tex_count << R500_RS_IP_TEX_PTR_S_SHIFT) |
+				(R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_T_SHIFT) |
+				(R500_RS_IP_PTR_K0 << R500_RS_IP_TEX_PTR_R_SHIFT) |
+				(R500_RS_IP_PTR_K1 << R500_RS_IP_TEX_PTR_Q_SHIFT);
 
 			r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
 			InputsRead &= ~FRAG_BIT_FOGC;
@@ -1718,87 +1672,27 @@ static void r500SetupRSUnit(GLcontext * ctx)
 		}
 	}
 
-	if (InputsRead & FRAG_BIT_WPOS) {
-		r300->hw.ri.cmd[R300_RI_INTERP_0 + tex_ip] |= ((rs_tex_count + 0) << R500_RS_IP_TEX_PTR_S_SHIFT) |
-				((rs_tex_count + 1) << R500_RS_IP_TEX_PTR_T_SHIFT) |
-				((rs_tex_count + 2) << R500_RS_IP_TEX_PTR_R_SHIFT) |
-				((rs_tex_count + 3) << R500_RS_IP_TEX_PTR_Q_SHIFT);
-
-		r300->hw.rr.cmd[R300_RR_INST_0 + tex_ip] |= R500_RS_INST_TEX_ID(tex_ip) | R500_RS_INST_TEX_CN_WRITE | R500_RS_INST_TEX_ADDR(fp_reg);
-		InputsRead &= ~FRAG_BIT_WPOS;
-		rs_tex_count += 4;
-		++tex_ip;
-		++fp_reg;
-	}
-
 	/* Setup default color if no color or tex was set */
 	if (rs_tex_count == 0 && col_ip == 0) {
-		r300->hw.rr.cmd[R300_RR_INST_0] |= R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_CN_WRITE | R500_RS_INST_COL_ADDR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
+		r300->hw.rr.cmd[R300_RR_INST_0] = R500_RS_INST_COL_ID(0) | R500_RS_INST_COL_ADDR(0);
+		r300->hw.ri.cmd[R300_RI_INTERP_0] = R500_RS_COL_PTR(0) | R500_RS_COL_FMT(R300_RS_COL_FMT_0001);
 		++col_ip;
 	}
 
 	high_rr = (col_ip > tex_ip) ? col_ip : tex_ip;
-	r300->hw.rc.cmd[1] |= (rs_tex_count << R300_IT_COUNT_SHIFT)  | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
-	r300->hw.rc.cmd[2] |= 0xC0 | (high_rr - 1);
+	r300->hw.rc.cmd[1] = (rs_tex_count << R300_IT_COUNT_SHIFT) | (col_ip << R300_IC_COUNT_SHIFT) | R300_HIRES_EN;
+	r300->hw.rc.cmd[2] = 0xC0 | (high_rr - 1);
 
 	r300->hw.rr.cmd[R300_RR_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_INST_0, high_rr);
+	r300->hw.ri.cmd[R300_RI_CMD_0] = cmdpacket0(r300->radeon.radeonScreen, R500_RS_IP_0, high_rr);
 
 	if (InputsRead)
 		WARN_ONCE("Don't know how to satisfy InputsRead=0x%08x\n", InputsRead);
 }
 
-
-
-
-#define bump_vpu_count(ptr, new_count)   do{\
-	drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\
-	int _nc=(new_count)/4; \
-	assert(_nc < 256); \
-	if(_nc>_p->vpu.count)_p->vpu.count=_nc;\
-	}while(0)
-
-static INLINE void r300SetupVertexProgramFragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf)
-{
-	int i;
-
-	if (vsf->length == 0)
-		return;
-
-	if (vsf->length & 0x3) {
-		fprintf(stderr, "VERTEX_SHADER_FRAGMENT must have length divisible by 4\n");
-		_mesa_exit(-1);
-	}
-
-	switch ((dest >> 8) & 0xf) {
-	case 0:
-		R300_STATECHANGE(r300, vpi);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpi.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-
-	case 2:
-		R300_STATECHANGE(r300, vpp);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vpp.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	case 4:
-		R300_STATECHANGE(r300, vps);
-		for (i = 0; i < vsf->length; i++)
-			r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (vsf->body.d[i]);
-		bump_vpu_count(r300->hw.vps.cmd, vsf->length + 4 * (dest & 0xff));
-		break;
-	default:
-		fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
-		_mesa_exit(-1);
-	}
-}
-
 #define MIN3(a, b, c)	((a) < (b) ? MIN2(a, c) : MIN2(b, c))
 
-
-static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 			GLuint output_count, GLuint temp_count)
 {
     int vtx_mem_size;
@@ -1822,7 +1716,7 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
     pvs_num_cntrls = MIN2(6, vtx_mem_size/temp_count);
 
     R300_STATECHANGE(rmesa, vap_cntl);
-    if (rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
+    if (rmesa->options.hw_tcl_enabled) {
 	rmesa->hw.vap_cntl.cmd[R300_VAP_CNTL_INSTR] =
 	    (pvs_num_slots << R300_PVS_NUM_SLOTS_SHIFT) |
 	    (pvs_num_cntrls << R300_PVS_NUM_CNTLRS_SHIFT) |
@@ -1852,115 +1746,6 @@ static void r300VapCntl(r300ContextPtr rmesa, GLuint input_count,
 
 }
 
-static void r300SetupDefaultVertexProgram(r300ContextPtr rmesa)
-{
-	struct r300_vertex_shader_state *prog = &(rmesa->state.vertex_shader);
-	GLuint o_reg = 0;
-	GLuint i_reg = 0;
-	int i;
-	int inst_count = 0;
-	int param_count = 0;
-	int program_end = 0;
-
-	for (i = VERT_ATTRIB_POS; i < VERT_ATTRIB_MAX; i++) {
-		if (rmesa->state.sw_tcl_inputs[i] != -1) {
-			prog->program.body.i[program_end + 0] = PVS_OP_DST_OPERAND(VE_MULTIPLY, GL_FALSE, GL_FALSE, o_reg++, VSF_FLAG_ALL, PVS_DST_REG_OUT);
-			prog->program.body.i[program_end + 1] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y, PVS_SRC_SELECT_Z, PVS_SRC_SELECT_W, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 2] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			prog->program.body.i[program_end + 3] = PVS_SRC_OPERAND(rmesa->state.sw_tcl_inputs[i], PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_SELECT_FORCE_1, PVS_SRC_REG_INPUT, VSF_FLAG_NONE);
-			program_end += 4;
-			i_reg++;
-		}
-	}
-
-	prog->program.length = program_end;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START,
-				       &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, i_reg, o_reg, 0);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	    (0 << R300_PVS_FIRST_INST_SHIFT) |
-	    (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	    (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	    (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	    (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	    (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-static int bit_count (int x)
-{
-    x = ((x & 0xaaaaaaaaU) >> 1) + (x & 0x55555555U);
-    x = ((x & 0xccccccccU) >> 2) + (x & 0x33333333U);
-    x = (x >> 16) + (x & 0xffff);
-    x = ((x & 0xf0f0) >> 4) + (x & 0x0f0f);
-    return (x >> 8) + (x & 0x00ff);
-}
-
-static void r300SetupRealVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_vertex_program *prog = (struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx);
-	int inst_count = 0;
-	int param_count = 0;
-
-	/* FIXME: r300SetupVertexProgramFragment */
-	R300_STATECHANGE(rmesa, vpp);
-	param_count =
-	    r300VertexProgUpdateParams(ctx,
-				       (struct r300_vertex_program_cont *)
-				       ctx->VertexProgram._Current,
-				       (float *)&rmesa->hw.vpp.
-				       cmd[R300_VPP_PARAM_0]);
-	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
-	param_count /= 4;
-
-	r300SetupVertexProgramFragment(rmesa, R300_PVS_CODE_START, &(prog->program));
-	inst_count = (prog->program.length / 4) - 1;
-
-	r300VapCntl(rmesa, bit_count(prog->key.InputsRead),
-		    bit_count(prog->key.OutputsWritten), prog->num_temporaries);
-
-	R300_STATECHANGE(rmesa, pvs);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] =
-	  (0 << R300_PVS_FIRST_INST_SHIFT) |
-	  (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
-	  (inst_count << R300_PVS_LAST_INST_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] =
-	  (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) |
-	  (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
-	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] =
-	  (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
-}
-
-
-static void r300SetupVertexProgram(r300ContextPtr rmesa)
-{
-	GLcontext *ctx = rmesa->radeon.glCtx;
-
-	/* Reset state, in case we don't use something */
-	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
-	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
-
-	/* Not sure why this doesnt work...
-	   0x400 area might have something to do with pixel shaders as it appears right after pfs programming.
-	   0x406 is set to { 0.0, 0.0, 1.0, 0.0 } most of the time but should change with smooth points and in other rare cases. */
-	//setup_vertex_shader_fragment(rmesa, 0x406, &unk4);
-	if (hw_tcl_on && ((struct r300_vertex_program *)CURRENT_VERTEX_SHADER(ctx))->translated) {
-		r300SetupRealVertexProgram(rmesa);
-	} else {
-		/* FIXME: This needs to be replaced by vertex shader generation code. */
-		r300SetupDefaultVertexProgram(rmesa);
-	}
-
-}
-
 /**
  * Enable/Disable states.
  *
@@ -1975,14 +1760,6 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 			state ? "GL_TRUE" : "GL_FALSE");
 
 	switch (cap) {
-	case GL_TEXTURE_1D:
-	case GL_TEXTURE_2D:
-	case GL_TEXTURE_3D:
-		/* empty */
-		break;
-	case GL_FOG:
-		/* empty */
-		break;
 	case GL_ALPHA_TEST:
 		r300SetAlphaState(ctx);
 		break;
@@ -2000,14 +1777,31 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 	case GL_CLIP_PLANE5:
 		r300SetClipPlaneState(ctx, cap, state);
 		break;
+	case GL_CULL_FACE:
+		r300UpdateCulling(ctx);
+		break;
 	case GL_DEPTH_TEST:
 		r300SetDepthState(ctx);
 		break;
-	case GL_STENCIL_TEST:
-		r300SetStencilState(ctx, state);
+	case GL_LINE_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_SMOOTH, ctx->Line.SmoothFlag);
 		break;
-	case GL_CULL_FACE:
-		r300UpdateCulling(ctx);
+	case GL_LINE_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_LINE_STIPPLE, ctx->Line.StippleFlag);
+		break;
+	case GL_POINT_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POINT_SMOOTH, ctx->Point.SmoothFlag);
+		break;
+	case GL_POLYGON_SMOOTH:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_SMOOTH, ctx->Polygon.SmoothFlag);
+		break;
+	case GL_POLYGON_STIPPLE:
+		if (rmesa->options.conformance_mode)
+			r300SwitchFallback(ctx, R300_FALLBACK_POLYGON_STIPPLE, ctx->Polygon.StippleFlag);
 		break;
 	case GL_POLYGON_OFFSET_POINT:
 	case GL_POLYGON_OFFSET_LINE:
@@ -2019,6 +1813,9 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 		rmesa->radeon.state.scissor.enabled = state;
 		radeonUpdateScissor( ctx );
 		break;
+	case GL_STENCIL_TEST:
+		r300SetStencilState(ctx, state);
+		break;
 	default:
 		break;
 	}
@@ -2030,10 +1827,9 @@ static void r300Enable(GLcontext * ctx, GLenum cap, GLboolean state)
 static void r300ResetHwState(r300ContextPtr r300)
 {
 	GLcontext *ctx = r300->radeon.glCtx;
-	int has_tcl = 1;
+	int has_tcl;
 
-	if (!(r300->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
-		has_tcl = 0;
+	has_tcl = r300->options.hw_tcl_enabled;
 
 	if (RADEON_DEBUG & DEBUG_STATE)
 		fprintf(stderr, "%s\n", __FUNCTION__);
@@ -2138,8 +1934,8 @@ static void r300ResetHwState(r300ContextPtr r300)
 	}
 
 	/* XXX: Enable anti-aliasing? */
-	r300->hw.gb_misc.cmd[R300_GB_MISC_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
-	r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = 0;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_AA_CONFIG] = GB_AA_CONFIG_AA_DISABLE;
+	r300->hw.gb_misc2.cmd[R300_GB_MISC2_SELECT] = 0;
 
 	r300->hw.ga_point_s0.cmd[1] = r300PackFloat32(0.0);
 	r300->hw.ga_point_s0.cmd[2] = r300PackFloat32(0.0);
@@ -2210,8 +2006,8 @@ static void r300ResetHwState(r300ContextPtr r300)
 
 	r300->hw.rb3d_aaresolve_ctl.cmd[1] = 0;
 
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
-	r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[1] = 0x00000000;
+    r300->hw.rb3d_discard_src_pixel_lte_threshold.cmd[2] = 0xffffffff;
 
 	r300->hw.zb_depthclearvalue.cmd[1] = 0;
 
@@ -2220,8 +2016,8 @@ static void r300ResetHwState(r300ContextPtr r300)
 	r300->hw.zstencil_format.cmd[4] = 0x00000000;
 	r300SetEarlyZState(ctx);
 
-	r300->hw.unk4F30.cmd[1] = 0;
-	r300->hw.unk4F30.cmd[2] = 0;
+	r300->hw.zb_zmask.cmd[1] = 0;
+	r300->hw.zb_zmask.cmd[2] = 0;
 
 	r300->hw.zb_hiz_offset.cmd[1] = 0;
 
@@ -2241,14 +2037,20 @@ static void r300ResetHwState(r300ContextPtr r300)
 void r300UpdateShaders(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
-	struct r300_vertex_program *vp;
+	struct r300_fragment_program *fp;
 	int i;
 
 	ctx = rmesa->radeon.glCtx;
+	fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 
-	if (rmesa->radeon.NewGLState && hw_tcl_on) {
-		rmesa->radeon.NewGLState = 0;
+	/* should only happenen once, just after context is created */
+	/* TODO: shouldn't we fallback to sw here? */
+	if (!fp) {
+		_mesa_fprintf(stderr, "No ctx->FragmentProgram._Current!!\n");
+		return;
+	}
 
+	if (rmesa->radeon.NewGLState && rmesa->options.hw_tcl_enabled) {
 		for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
 			rmesa->temp_attrib[i] =
 			    TNL_CONTEXT(ctx)->vb.AttribPtr[i];
@@ -2264,20 +2066,16 @@ void r300UpdateShaders(r300ContextPtr rmesa)
 		}
 
 		r300SelectVertexShader(rmesa);
-		vp = (struct r300_vertex_program *)
-		    CURRENT_VERTEX_SHADER(ctx);
-		/*if (vp->translated == GL_FALSE)
-		   r300TranslateVertexShader(vp); */
-		if (vp->translated == GL_FALSE) {
-			fprintf(stderr, "Failing back to sw-tcl\n");
-			hw_tcl_on = future_hw_tcl_on = 0;
-			r300ResetHwState(rmesa);
-
-			r300UpdateStateParameters(ctx, _NEW_PROGRAM);
-			return;
-		}
+		r300SwitchFallback(ctx, R300_FALLBACK_VERTEX_PROGRAM, rmesa->selected_vp->error);
 	}
-	r300UpdateStateParameters(ctx, _NEW_PROGRAM);
+
+	if (!fp->translated || rmesa->radeon.NewGLState)
+		r300TranslateFragmentShader(ctx, ctx->FragmentProgram._Current);
+
+	r300SwitchFallback(ctx, R300_FALLBACK_FRAGMENT_PROGRAM, fp->error);
+
+	r300UpdateStateParameters(ctx, _NEW_PROGRAM | _NEW_PROGRAM_CONSTANTS);
+	rmesa->radeon.NewGLState = 0;
 }
 
 static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
@@ -2301,26 +2099,14 @@ static const GLfloat *get_fragmentprogram_constant(GLcontext *ctx,
 }
 
 
-static void r300SetupPixelShader(r300ContextPtr rmesa)
+static void r300SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r300_fragment_program *fp = (struct r300_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 	struct r300_fragment_program_code *code;
 	int i, k;
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
-	r300TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
-
-	r300SetupTextures(ctx);
+	code = &fp->code.r300;
 
 	R300_STATECHANGE(rmesa, fpi[0]);
 	R300_STATECHANGE(rmesa, fpi[1]);
@@ -2363,7 +2149,7 @@ static void r300SetupPixelShader(r300ContextPtr rmesa)
 	rmesa->hw.fpp.cmd[R300_FPP_CMD_0] = cmdpacket0(rmesa->radeon.radeonScreen, R300_PFS_PARAM_0_X, code->const_nr * 4);
 	for (i = 0; i < code->const_nr; i++) {
 		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+			&fp->Base.Base, code->constant[i]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat24(constant[0]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat24(constant[1]);
 		rmesa->hw.fpp.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat24(constant[2]);
@@ -2385,29 +2171,17 @@ static void r300SetupPixelShader(r300ContextPtr rmesa)
 	if(_nc>_p->r500fp.count)_p->r500fp.count=_nc;\
 } while(0)
 
-static void r500SetupPixelShader(r300ContextPtr rmesa)
+static void r500SetupPixelShader(GLcontext *ctx)
 {
-	GLcontext *ctx = rmesa->radeon.glCtx;
-	struct r500_fragment_program *fp = (struct r500_fragment_program *)
-	    (char *)ctx->FragmentProgram._Current;
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	struct r300_fragment_program *fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 	int i;
 	struct r500_fragment_program_code *code;
 
-	if (!fp)		/* should only happenen once, just after context is created */
-		return;
-
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp.cmd)->r500fp.count = 0;
 	((drm_r300_cmd_header_t *) rmesa->hw.r500fp_const.cmd)->r500fp.count = 0;
 
-	r500TranslateFragmentShader(rmesa, fp);
-	if (!fp->translated) {
-		fprintf(stderr, "%s: No valid fragment shader, exiting\n",
-			__FUNCTION__);
-		return;
-	}
-	code = &fp->code;
-
-	r300SetupTextures(ctx);
+	code = &fp->code.r500;
 
 	R300_STATECHANGE(rmesa, fp);
 	rmesa->hw.fp.cmd[R500_FP_PIXSIZE] = code->max_temp_idx;
@@ -2437,57 +2211,96 @@ static void r500SetupPixelShader(r300ContextPtr rmesa)
 	R300_STATECHANGE(rmesa, r500fp_const);
 	for (i = 0; i < code->const_nr; i++) {
 		const GLfloat *constant = get_fragmentprogram_constant(ctx,
-			&fp->mesa_program.Base, code->constant[i]);
+			&fp->Base.Base, code->constant[i]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 0] = r300PackFloat32(constant[0]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 1] = r300PackFloat32(constant[1]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 2] = r300PackFloat32(constant[2]);
 		rmesa->hw.r500fp_const.cmd[R300_FPP_PARAM_0 + 4 * i + 3] = r300PackFloat32(constant[3]);
 	}
 	bump_r500fp_const_count(rmesa->hw.r500fp_const.cmd, code->const_nr * 4);
+}
+
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten)
+{
+	r300ContextPtr rmesa = R300_CONTEXT( ctx );
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
+	int i, j, reg_count;
+	uint32_t *vir0 = &rmesa->hw.vir[0].cmd[1];
+	uint32_t *vir1 = &rmesa->hw.vir[1].cmd[1];
+
+	for (i = 0; i < R300_VIR_CMDSIZE-1; ++i)
+		vir0[i] = vir1[i] = 0;
+
+	for (i = 0, j = 0; i < rmesa->vbuf.num_attribs; ++i) {
+		int tmp;
+
+		tmp = attrs[i].data_type | (attrs[i].dst_loc << R300_DST_VEC_LOC_SHIFT);
+		if (attrs[i]._signed)
+			tmp |= R300_SIGNED;
+		if (attrs[i].normalize)
+			tmp |= R300_NORMALIZE;
+
+		if (i % 2 == 0) {
+			vir0[j] = tmp << R300_DATA_TYPE_0_SHIFT;
+			vir1[j] = attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT);
+		} else {
+			vir0[j] |= tmp << R300_DATA_TYPE_1_SHIFT;
+			vir1[j] |= (attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
+			++j;
+		}
+	}
 
+	reg_count = (rmesa->vbuf.num_attribs + 1) >> 1;
+	if (rmesa->vbuf.num_attribs % 2 != 0) {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
+	} else {
+		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
+	}
+
+	R300_STATECHANGE(rmesa, vir[0]);
+	R300_STATECHANGE(rmesa, vir[1]);
+	R300_STATECHANGE(rmesa, vof);
+	R300_STATECHANGE(rmesa, vic);
+
+	if (rmesa->radeon.radeonScreen->kernel_mm) {
+		rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
+		rmesa->hw.vir[0].cmd[0] |= (reg_count & 0x3FFF) << 16;
+		rmesa->hw.vir[1].cmd[0] |= (reg_count & 0x3FFF) << 16;
+	} else {
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = reg_count;
+		((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = reg_count;
+	}
+
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
+	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten, ctx->FragmentProgram._Current->Base.InputsRead);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten, ctx->FragmentProgram._Current->Base.InputsRead);
 }
 
 void r300UpdateShaderStates(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
 	ctx = rmesa->radeon.glCtx;
+	struct r300_fragment_program *r300_fp;
 
-	r300SetEarlyZState(ctx);
+	r300_fp = (struct r300_fragment_program *) ctx->FragmentProgram._Current;
 
-	/* w_fmt value is set to get best performance
-	 * see p.130 R5xx 3D acceleration guide v1.3 */
-	GLuint w_fmt, fgdepthsrc;
-	if (current_fragment_program_writes_depth(ctx)) {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SHADER;
-		w_fmt = R300_W_FMT_W24 | R300_W_SRC_US;
-	} else {
-		fgdepthsrc = R300_FG_DEPTH_SRC_SCAN;
-		w_fmt = R300_W_FMT_W0 | R300_W_SRC_US;
-	}
+	/* should only happenen once, just after context is created */
+	if (!r300_fp)
+		return;
 
-	if (w_fmt != rmesa->hw.us_out_fmt.cmd[5]) {
-		R300_STATECHANGE(rmesa, us_out_fmt);
-		rmesa->hw.us_out_fmt.cmd[5] = w_fmt;
-	}
+	r300SetEarlyZState(ctx);
 
-	if (fgdepthsrc != rmesa->hw.fg_depth_src.cmd[1]) {
-		R300_STATECHANGE(rmesa, fg_depth_src);
-		rmesa->hw.fg_depth_src.cmd[1] = fgdepthsrc;
-	}
+	r300SetupTextures(ctx);
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupPixelShader(rmesa);
-	else
-		r300SetupPixelShader(rmesa);
+	rmesa->vtbl.SetupPixelShader(ctx);
 
-	if (rmesa->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515)
-		r500SetupRSUnit(ctx);
-	else
-		r300SetupRSUnit(ctx);
+	rmesa->vtbl.SetupRSUnit(ctx);
 
-	if ((rmesa->radeon.radeonScreen->chip_flags & RADEON_CHIPSET_TCL))
+	if (rmesa->options.hw_tcl_enabled) {
 		r300SetupVertexProgram(rmesa);
-
+	}
 }
 
 /**
@@ -2501,9 +2314,8 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 	_swsetup_InvalidateState(ctx, new_state);
 	_vbo_InvalidateState(ctx, new_state);
 	_tnl_InvalidateState(ctx, new_state);
-	_ae_invalidate_state(ctx, new_state);
 
-	if (new_state & (_NEW_BUFFERS | _NEW_COLOR | _NEW_PIXEL)) {
+	if (new_state & _NEW_BUFFERS) {
 		_mesa_update_framebuffer(ctx);
 		/* this updates the DrawBuffer's Width/Height if it's a FBO */
 		_mesa_update_draw_buffer_bounds(ctx);
@@ -2523,34 +2335,12 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
  */
 void r300InitState(r300ContextPtr r300)
 {
-	memset(&(r300->state.texture), 0, sizeof(r300->state.texture));
-
 	r300ResetHwState(r300);
 }
 
 static void r300RenderMode(GLcontext * ctx, GLenum mode)
 {
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	(void)rmesa;
-	(void)mode;
-}
-
-void r300UpdateClipPlanes( GLcontext *ctx )
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	GLuint p;
-
-	for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
-		if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
-			GLint *ip = (GLint *)ctx->Transform._ClipUserPlane[p];
-
-			R300_STATECHANGE( rmesa, vpucp[p] );
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_X] = ip[0];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Y] = ip[1];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_Z] = ip[2];
-			rmesa->hw.vpucp[p].cmd[R300_VPUCP_W] = ip[3];
-		}
-	}
+	r300SwitchFallback(ctx, R300_FALLBACK_RENDER_MODE, ctx->RenderMode != GL_RENDER);
 }
 
 /**
@@ -2598,3 +2388,20 @@ void r300InitStateFuncs(struct dd_function_table *functions)
 	functions->DrawBuffer		= radeonDrawBuffer;
 	functions->ReadBuffer		= radeonReadBuffer;
 }
+
+void r300InitShaderFunctions(r300ContextPtr r300)
+{
+	if (r300->radeon.radeonScreen->chip_family >= CHIP_FAMILY_RV515) {
+		r300->vtbl.SetupRSUnit = r500SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r500SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r500SetupFragmentShaderTextures;
+		r300->vtbl.BuildFragmentProgramHwCode = r500BuildFragmentProgramHwCode;
+		r300->vtbl.FragmentProgramDump = r500FragmentProgramDump;
+	} else {
+		r300->vtbl.SetupRSUnit = r300SetupRSUnit;
+		r300->vtbl.SetupPixelShader = r300SetupPixelShader;
+		r300->vtbl.SetupFragmentShaderTextures = r300SetupFragmentShaderTextures;
+		r300->vtbl.BuildFragmentProgramHwCode = r300BuildFragmentProgramHwCode;
+		r300->vtbl.FragmentProgramDump = r300FragmentProgramDump;
+	}
+}
diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h
index 247a20ee51..2328289420 100644
--- a/src/mesa/drivers/dri/r300/r300_state.h
+++ b/src/mesa/drivers/dri/r300/r300_state.h
@@ -50,16 +50,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 		r300->radeon.hw.is_dirty = GL_TRUE;		\
 	} while(0)
 
-// r300_state.c
-extern int future_hw_tcl_on;
-void _tnl_UpdateFixedFunctionProgram (GLcontext * ctx);
 void r300UpdateViewportOffset (GLcontext * ctx);
 void r300UpdateDrawBuffer (GLcontext * ctx);
 void r300UpdateStateParameters (GLcontext * ctx, GLuint new_state);
 void r300UpdateShaders (r300ContextPtr rmesa);
 void r300UpdateShaderStates (r300ContextPtr rmesa);
 void r300InitState (r300ContextPtr r300);
-void r300UpdateClipPlanes (GLcontext * ctx);
 void r300InitStateFuncs (struct dd_function_table *functions);
+void r300VapCntl(r300ContextPtr rmesa, GLuint input_count, GLuint output_count, GLuint temp_count);
+void r300SetupVAP(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten);
 
 #endif				/* __R300_STATE_H__ */
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.c b/src/mesa/drivers/dri/r300/r300_swtcl.c
index 03c1521de7..ce4179208e 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -34,9 +34,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
+#include "r300_state.h"
 #include "r300_swtcl.h"
 #include "r300_emit.h"
 #include "r300_tex.h"
+#include "r300_render.h"
 
 #define EMIT_ATTR( ATTR, STYLE )					\
 do {									\
@@ -53,199 +55,130 @@ do {									\
    rmesa->radeon.swtcl.vertex_attr_count++;					\
 } while (0)
 
-#define ADD_ATTR(_attr, _format, _dst_loc, _swizzle, _write_mask) \
+#define ADD_ATTR(_attr, _format, _dst_loc, _swizzle, _write_mask, _normalize) \
 do { \
-	attrs[num_attrs].attr = (_attr); \
-	attrs[num_attrs].format = (_format); \
+	attrs[num_attrs].element = (_attr); \
+	attrs[num_attrs].data_type = (_format); \
 	attrs[num_attrs].dst_loc = (_dst_loc); \
 	attrs[num_attrs].swizzle = (_swizzle); \
 	attrs[num_attrs].write_mask = (_write_mask); \
+	attrs[num_attrs]._signed = 0; \
+	attrs[num_attrs].normalize = (_normalize); \
 	++num_attrs; \
 } while (0)
 
-static void r300SwtclVAPSetup(GLcontext *ctx, GLuint InputsRead, GLuint OutputsWritten, GLuint vap_out_fmt_1)
+void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *_InputsRead,  GLuint *_OutputsWritten)
 {
 	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 	TNLcontext *tnl = TNL_CONTEXT(ctx);
 	struct vertex_buffer *VB = &tnl->vb;
-	struct vertex_attribute *attrs = rmesa->swtcl.vert_attrs;
-	int vte = 0;
-	int i, j, reg_count;
-	uint32_t *vir0 = &rmesa->hw.vir[0].cmd[1];
-	uint32_t *vir1 = &rmesa->hw.vir[1].cmd[1];
-
-	for (i = 0; i < R300_VIR_CMDSIZE-1; ++i)
-		vir0[i] = vir1[i] = 0;
-
-	for (i = 0, j = 0; i < rmesa->radeon.swtcl.vertex_attr_count; ++i) {
-		int tmp, data_format;
-		switch (attrs[i].format) {
-			case EMIT_1F:
-				data_format = R300_DATA_TYPE_FLOAT_1;
-				break;
-			case EMIT_2F:
-				data_format = R300_DATA_TYPE_FLOAT_2;
-				break;
-			case EMIT_3F:
-				data_format = R300_DATA_TYPE_FLOAT_3;
-				break;
-			case EMIT_4F:
-				data_format = R300_DATA_TYPE_FLOAT_4;
-				break;
-			case EMIT_4UB_4F_RGBA:
-			case EMIT_4UB_4F_ABGR:
-				data_format = R300_DATA_TYPE_BYTE | R300_NORMALIZE;
-				break;
-			default:
-				fprintf(stderr, "%s: Invalid data format type", __FUNCTION__);
-				_mesa_exit(-1);
-				break;
-		}
-
-		tmp = data_format | (attrs[i].dst_loc << R300_DST_VEC_LOC_SHIFT);
-		if (i % 2 == 0) {
-			vir0[j] = tmp << R300_DATA_TYPE_0_SHIFT;
-			vir1[j] = attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT);
-		} else {
-			vir0[j] |= tmp << R300_DATA_TYPE_1_SHIFT;
-			vir1[j] |= (attrs[i].swizzle | (attrs[i].write_mask << R300_WRITE_ENA_SHIFT)) << R300_SWIZZLE1_SHIFT;
-			++j;
-		}
-	}
-
-	reg_count = (rmesa->radeon.swtcl.vertex_attr_count + 1) >> 1;
-	if (rmesa->radeon.swtcl.vertex_attr_count % 2 != 0) {
-		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_0_SHIFT;
-	} else {
-		vir0[reg_count-1] |= R300_LAST_VEC << R300_DATA_TYPE_1_SHIFT;
-	}
-
-	R300_STATECHANGE(rmesa, vir[0]);
-	R300_STATECHANGE(rmesa, vir[1]);
-	R300_STATECHANGE(rmesa, vof);
-	R300_STATECHANGE(rmesa, vte);
-	R300_STATECHANGE(rmesa, vic);
-
-	if (rmesa->radeon.radeonScreen->kernel_mm) {
-		rmesa->hw.vir[0].cmd[0] &= 0xC000FFFF;
-		rmesa->hw.vir[1].cmd[0] &= 0xC000FFFF;
-		rmesa->hw.vir[0].cmd[0] |= (reg_count & 0x3FFF) << 16;
-		rmesa->hw.vir[1].cmd[0] |= (reg_count & 0x3FFF) << 16;
-	} else {
-		((drm_r300_cmd_header_t *) rmesa->hw.vir[0].cmd)->packet0.count = reg_count;
-		((drm_r300_cmd_header_t *) rmesa->hw.vir[1].cmd)->packet0.count = reg_count;
-	}
-
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_0] = r300VAPInputCntl0(ctx, InputsRead);
-	rmesa->hw.vic.cmd[R300_VIC_CNTL_1] = r300VAPInputCntl1(ctx, InputsRead);
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_0] = r300VAPOutputCntl0(ctx, OutputsWritten);
-	/**
-	  * Can't use r300VAPOutputCntl1 function because it assumes
-	  * that all texture coords have 4 components and that's the case
-	  * for HW TCL path, but not for SW TCL.
-	  */
-	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = vap_out_fmt_1;
-
-	vte = rmesa->hw.vte.cmd[1];
-	vte &= ~(R300_VTX_XY_FMT | R300_VTX_Z_FMT | R300_VTX_W0_FMT);
-	/* Important:
-	 */
-	if ( VB->NdcPtr != NULL ) {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->NdcPtr;
-		vte |= R300_VTX_XY_FMT | R300_VTX_Z_FMT;
-	}
-	else {
-		VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
-		vte |= R300_VTX_W0_FMT;
-	}
-
-	assert( VB->AttribPtr[VERT_ATTRIB_POS] != NULL );
-
-	rmesa->hw.vte.cmd[1] = vte;
-	rmesa->hw.vte.cmd[2] = rmesa->radeon.swtcl.vertex_size;
-}
-
-
-static void r300SetVertexFormat( GLcontext *ctx )
-{
-	r300ContextPtr rmesa = R300_CONTEXT( ctx );
-	TNLcontext *tnl = TNL_CONTEXT(ctx);
-	struct vertex_buffer *VB = &tnl->vb;
-	int first_free_tex = 0, vap_out_fmt_1 = 0;
+	int first_free_tex = 0;
 	GLuint InputsRead = 0;
 	GLuint OutputsWritten = 0;
 	int num_attrs = 0;
-	struct vertex_attribute *attrs = rmesa->swtcl.vert_attrs;
+	GLuint fp_reads = ctx->FragmentProgram._Current->Base.InputsRead;
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
 
 	rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0;
 	rmesa->radeon.swtcl.vertex_attr_count = 0;
 
-	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_POS)) {
-		InputsRead |= 1 << VERT_ATTRIB_POS;
-		OutputsWritten |= 1 << VERT_RESULT_HPOS;
-		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
-		ADD_ATTR(VERT_ATTRIB_POS, EMIT_4F, SWTCL_OVM_POS, SWIZZLE_XYZW, MASK_XYZW);
-		rmesa->swtcl.coloroffset = 4;
-	}
+	/* We always want non Ndc coords format */
+	VB->AttribPtr[VERT_ATTRIB_POS] = VB->ClipPtr;
+
+	/* Always write position vector */
+	InputsRead |= 1 << VERT_ATTRIB_POS;
+	OutputsWritten |= 1 << VERT_RESULT_HPOS;
+	EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
+	ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_POS, SWIZZLE_XYZW, MASK_XYZW, 0);
+	rmesa->swtcl.coloroffset = 4;
 
-	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR0)) {
+	if (fp_reads & FRAG_BIT_COL0) {
 		InputsRead |= 1 << VERT_ATTRIB_COLOR0;
 		OutputsWritten |= 1 << VERT_RESULT_COL0;
 #if MESA_LITTLE_ENDIAN
 		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA );
-		ADD_ATTR(VERT_ATTRIB_COLOR0, EMIT_4UB_4F_RGBA, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW);
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
 #else
 		EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR );
-		ADD_ATTR(VERT_ATTRIB_COLOR0, EMIT_4UB_4F_ABGR, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW);
+		ADD_ATTR(VERT_ATTRIB_COLOR0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR0, SWIZZLE_XYZW, MASK_XYZW, 1);
 #endif
 	}
 
-	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 )) {
+	if (fp_reads & FRAG_BIT_COL1) {
 		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 		InputsRead |= 1 << VERT_ATTRIB_COLOR1;
 		OutputsWritten |= 1 << VERT_RESULT_COL1;
 #if MESA_LITTLE_ENDIAN
 		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_RGBA );
-		ADD_ATTR(VERT_ATTRIB_COLOR1, EMIT_4UB_4F_RGBA, SWTCL_OVM_COLOR1, swiz, MASK_XYZW);
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
 #else
 		EMIT_ATTR( _TNL_ATTRIB_COLOR1, EMIT_4UB_4F_ABGR );
-		ADD_ATTR(VERT_ATTRIB_COLOR1, EMIT_4UB_4F_ABGR, SWTCL_OVM_COLOR1, swiz, MASK_XYZW);
+		ADD_ATTR(VERT_ATTRIB_COLOR1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR1, swiz, MASK_XYZW, 1);
 #endif
 		rmesa->swtcl.specoffset = rmesa->swtcl.coloroffset + 1;
 	}
 
+	if (ctx->Light.Enabled && ctx->Light.Model.TwoSide) {
+		VB->AttribPtr[VERT_ATTRIB_GENERIC0] = VB->ColorPtr[1];
+		OutputsWritten |= 1 << VERT_RESULT_BFC0;
+#if MESA_LITTLE_ENDIAN
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#else
+		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_ABGR );
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
+#endif
+		if (fp_reads & FRAG_BIT_COL1) {
+			VB->AttribPtr[VERT_ATTRIB_GENERIC1] = VB->SecondaryColorPtr[1];
+			GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
+			OutputsWritten |= 1 << VERT_RESULT_BFC1;
+#if MESA_LITTLE_ENDIAN
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_RGBA );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#else
+			EMIT_ATTR( _TNL_ATTRIB_GENERIC1, EMIT_4UB_4F_ABGR );
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
+#endif
+		}
+	}
+
 	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_POINTSIZE )) {
 		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO);
 		InputsRead |= 1 << VERT_ATTRIB_POINT_SIZE;
 		OutputsWritten |= 1 << VERT_RESULT_PSIZ;
 		EMIT_ATTR( _TNL_ATTRIB_POINTSIZE, EMIT_1F );
-		ADD_ATTR(VERT_ATTRIB_POINT_SIZE, EMIT_1F, SWTCL_OVM_POINT_SIZE, swiz, MASK_X);
+		ADD_ATTR(VERT_ATTRIB_POINT_SIZE, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_POINT_SIZE, swiz, MASK_X, 0);
 	}
 
-	if (RENDERINPUTS_TEST_RANGE(tnl->render_inputs_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
-		int i, size;
-		GLuint swiz, mask, format;
+	/**
+	 *  Sending only one texcoord component may lead to lock up,
+	 *  so for all textures always output 4 texcoord components to RS.
+	 */
+	{
+		int i;
+		GLuint swiz, format, hw_format;
 		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
-			if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_TEX(i) )) {
+			if (fp_reads & FRAG_BIT_TEX(i)) {
 				switch (VB->TexCoordPtr[i]->size) {
 					case 1:
+						format = EMIT_1F;
+						hw_format = R300_DATA_TYPE_FLOAT_1;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ONE);
+						break;
 					case 2:
 						format = EMIT_2F;
-						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ZERO);
-						mask = MASK_X | MASK_Y;
-						size = 2;
+						hw_format = R300_DATA_TYPE_FLOAT_2;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
 						break;
 					case 3:
 						format = EMIT_3F;
-						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-						mask = MASK_X | MASK_Y | MASK_Z;
-						size = 3;
+						hw_format = R300_DATA_TYPE_FLOAT_3;
+						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ONE);
 						break;
 					case 4:
 						format = EMIT_4F;
+						hw_format = R300_DATA_TYPE_FLOAT_4;
 						swiz = SWIZZLE_XYZW;
-						mask = MASK_XYZW;
-						size = 4;
 						break;
 					default:
 						continue;
@@ -253,15 +186,14 @@ static void r300SetVertexFormat( GLcontext *ctx )
 				InputsRead |= 1 << (VERT_ATTRIB_TEX0 + i);
 				OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
 				EMIT_ATTR(_TNL_ATTRIB_TEX(i), format);
-				ADD_ATTR(VERT_ATTRIB_TEX0 + i, format, SWTCL_OVM_TEX(i), swiz, mask);
-				vap_out_fmt_1 |= size << (i * 3);
+				ADD_ATTR(VERT_ATTRIB_TEX0 + i, hw_format, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_XYZW, 0);
 				++first_free_tex;
 			}
 		}
 	}
 
 	/* RS can't put fragment position on the pixel stack, so stuff it in texcoord if needed */
-	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_POS) && (ctx->FragmentProgram._Current->Base.InputsRead & FRAG_BIT_WPOS)) {
+	if (fp_reads & FRAG_BIT_WPOS) {
 		if (first_free_tex >= ctx->Const.MaxTextureUnits) {
 			fprintf(stderr, "\tout of free texcoords to write w pos\n");
 			_mesa_exit(-1);
@@ -270,12 +202,11 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		InputsRead |= 1 << (VERT_ATTRIB_TEX0 + first_free_tex);
 		OutputsWritten |= 1 << (VERT_RESULT_TEX0 + first_free_tex);
 		EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F );
-		ADD_ATTR(VERT_ATTRIB_POS, EMIT_4F, SWTCL_OVM_TEX(first_free_tex), SWIZZLE_XYZW, MASK_XYZW);
-		vap_out_fmt_1 |= 4 << (first_free_tex * 3);
+		ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_TEX(first_free_tex), SWIZZLE_XYZW, MASK_XYZW, 0);
 		++first_free_tex;
 	}
 
-	if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_FOG)) {
+	if (fp_reads & FRAG_BIT_FOGC) {
 		if (first_free_tex >= ctx->Const.MaxTextureUnits) {
 			fprintf(stderr, "\tout of free texcoords to write fog coordinate\n");
 			_mesa_exit(-1);
@@ -285,12 +216,24 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		OutputsWritten |= 1 << VERT_RESULT_FOGC;
 		GLuint swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_ZERO, SWIZZLE_ZERO, SWIZZLE_ZERO);
 		EMIT_ATTR( _TNL_ATTRIB_FOG, EMIT_1F );
-		ADD_ATTR(VERT_ATTRIB_FOG, EMIT_1F, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_X);
-		vap_out_fmt_1 |=  1 << (first_free_tex * 3);
+		ADD_ATTR(VERT_ATTRIB_FOG, R300_DATA_TYPE_FLOAT_1, SWTCL_OVM_TEX(first_free_tex), swiz, MASK_XYZW, 0);
 	}
 
 	R300_NEWPRIM(rmesa);
-	r300SwtclVAPSetup(ctx, InputsRead, OutputsWritten, vap_out_fmt_1);
+	rmesa->vbuf.num_attribs = num_attrs;
+	*_InputsRead = InputsRead;
+	*_OutputsWritten = OutputsWritten;
+
+	RENDERINPUTS_COPY(rmesa->render_inputs_bitset, tnl->render_inputs_bitset);
+}
+
+static void r300PrepareVertices(GLcontext *ctx)
+{
+	r300ContextPtr rmesa = R300_CONTEXT(ctx);
+	GLuint InputsRead, OutputsWritten;
+
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
 	rmesa->radeon.swtcl.vertex_size =
 		_tnl_install_attrs( ctx,
@@ -299,8 +242,6 @@ static void r300SetVertexFormat( GLcontext *ctx )
 				    NULL, 0 );
 
 	rmesa->radeon.swtcl.vertex_size /= 4;
-
-	RENDERINPUTS_COPY(rmesa->state.render_inputs_bitset, tnl->render_inputs_bitset);
 }
 
 
@@ -318,7 +259,6 @@ static GLuint reduced_prim[] = {
 };
 
 static void r300RasterPrimitive( GLcontext *ctx, GLuint prim );
-static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
 
 /***********************************************************************
  *                    Emit primitives as inline vertices               *
@@ -366,9 +306,8 @@ static void r300RenderPrimitive( GLcontext *ctx, GLenum prim );
  *              Build render functions from dd templates               *
  ***********************************************************************/
 
-#define R300_TWOSIDE_BIT	0x01
-#define R300_UNFILLED_BIT	0x02
-#define R300_MAX_TRIFUNC	0x04
+#define R300_UNFILLED_BIT	0x01
+#define R300_MAX_TRIFUNC	0x02
 
 static struct {
    tnl_points_func	        points;
@@ -379,9 +318,9 @@ static struct {
 
 #define DO_FALLBACK  0
 #define DO_UNFILLED (IND & R300_UNFILLED_BIT)
-#define DO_TWOSIDE  (IND & R300_TWOSIDE_BIT)
+#define DO_TWOSIDE   0
 #define DO_FLAT      0
-#define DO_OFFSET     0
+#define DO_OFFSET    0
 #define DO_TRI       1
 #define DO_QUAD      1
 #define DO_LINE      1
@@ -469,26 +408,15 @@ do { \
 #define TAG(x) x
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT)
-#define TAG(x) x##_twoside
-#include "tnl_dd/t_dd_tritmp.h"
-
 #define IND (R300_UNFILLED_BIT)
 #define TAG(x) x##_unfilled
 #include "tnl_dd/t_dd_tritmp.h"
 
-#define IND (R300_TWOSIDE_BIT|R300_UNFILLED_BIT)
-#define TAG(x) x##_twoside_unfilled
-#include "tnl_dd/t_dd_tritmp.h"
-
-
 
 static void init_rast_tab( void )
 {
    init();
-   init_twoside();
    init_unfilled();
-   init_twoside_unfilled();
 }
 
 /**********************************************************************/
@@ -540,7 +468,6 @@ static void r300ChooseRenderState( GLcontext *ctx )
 	GLuint index = 0;
 	GLuint flags = ctx->_TriangleCaps;
 
-	if (flags & DD_TRI_LIGHT_TWOSIDE) index |= R300_TWOSIDE_BIT;
 	if (flags & DD_TRI_UNFILLED)      index |= R300_UNFILLED_BIT;
 
 	if (index != rmesa->radeon.swtcl.RenderIndex) {
@@ -565,12 +492,12 @@ static void r300ChooseRenderState( GLcontext *ctx )
 }
 
 
-static void r300RenderStart(GLcontext *ctx)
+void r300RenderStart(GLcontext *ctx)
 {
 	r300ContextPtr rmesa = R300_CONTEXT( ctx );
 
 	r300ChooseRenderState(ctx);
-	r300SetVertexFormat(ctx);
+	r300PrepareVertices(ctx);
 
 	r300ValidateBuffers(ctx);
 
@@ -585,7 +512,7 @@ static void r300RenderStart(GLcontext *ctx)
 	}
 }
 
-static void r300RenderFinish(GLcontext *ctx)
+void r300RenderFinish(GLcontext *ctx)
 {
 }
 
@@ -599,7 +526,7 @@ static void r300RasterPrimitive( GLcontext *ctx, GLuint hwprim )
 	}
 }
 
-static void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
+void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
 {
 
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
@@ -611,7 +538,7 @@ static void r300RenderPrimitive(GLcontext *ctx, GLenum prim)
 	r300RasterPrimitive( ctx, reduced_prim[prim] );
 }
 
-static void r300ResetLineStipple(GLcontext *ctx)
+void r300ResetLineStipple(GLcontext *ctx)
 {
 }
 
@@ -647,7 +574,6 @@ void r300InitSwtcl(GLcontext *ctx)
 	_tnl_invalidate_vertices( ctx, ~0 );
 
 	_tnl_need_projected_coords( ctx, GL_FALSE );
-	r300ChooseRenderState(ctx);
 }
 
 void r300DestroySwtcl(GLcontext *ctx)
@@ -692,6 +618,7 @@ void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset)
 			   rmesa->radeon.hw.max_state_size + (12*sizeof(int)),
 			   __FUNCTION__);
 	radeonEmitState(&rmesa->radeon);
+    r300_emit_scissor(ctx);
 	r300EmitVertexAOS(rmesa,
 			rmesa->radeon.swtcl.vertex_size,
 			rmesa->radeon.dma.current,
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h
index ebc99c9e8a..c271d26546 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.h
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.h
@@ -39,24 +39,27 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "r300_context.h"
 
-#define MASK_XYZW (R300_WRITE_ENA_X | R300_WRITE_ENA_Y | R300_WRITE_ENA_Z | R300_WRITE_ENA_W)
-#define MASK_X R300_WRITE_ENA_X
-#define MASK_Y R300_WRITE_ENA_Y
-#define MASK_Z R300_WRITE_ENA_Z
-#define MASK_W R300_WRITE_ENA_W
-
 /*
  * Here are definitions of OVM locations of vertex attributes for non TCL hw
  */
 #define SWTCL_OVM_POS 0
 #define SWTCL_OVM_COLOR0 2
 #define SWTCL_OVM_COLOR1 3
+#define SWTCL_OVM_COLOR2 4
+#define SWTCL_OVM_COLOR3 5
 #define SWTCL_OVM_TEX(n) ((n) + 6)
 #define SWTCL_OVM_POINT_SIZE 15
 
+extern void r300ChooseSwtclVertexFormat(GLcontext *ctx, GLuint *InputsRead,  GLuint *OutputsWritten);
 
 extern void r300InitSwtcl( GLcontext *ctx );
 extern void r300DestroySwtcl( GLcontext *ctx );
 
+extern void r300RenderStart(GLcontext *ctx);
+extern void r300RenderFinish(GLcontext *ctx);
+extern void r300RenderPrimitive(GLcontext *ctx, GLenum prim);
+extern void r300ResetLineStipple(GLcontext *ctx);
+
 extern void r300_swtcl_flush(GLcontext *ctx, uint32_t current_offset);
+
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_texstate.c b/src/mesa/drivers/dri/r300/r300_texstate.c
index cf4cad73d2..6d6a90aa88 100644
--- a/src/mesa/drivers/dri/r300/r300_texstate.c
+++ b/src/mesa/drivers/dri/r300/r300_texstate.c
@@ -117,6 +117,7 @@ static const struct tx_table {
 	_ASSIGN(INTENSITY_FLOAT16, R300_EASY_TX_FORMAT(X, X, X, X, FL_I16)),
 	_ASSIGN(Z16, R300_EASY_TX_FORMAT(X, X, X, X, X16)),
 	_ASSIGN(Z24_S8, R300_EASY_TX_FORMAT(X, X, X, X, X24_Y8)),
+	_ASSIGN(S8_Z24, R300_EASY_TX_FORMAT(Y, Y, Y, Y, X24_Y8)),
 	_ASSIGN(Z32, R300_EASY_TX_FORMAT(X, X, X, X, X32)),
 	/* *INDENT-ON* */
 };
@@ -431,7 +432,6 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 		radeon_miptree_unreference(rImage->mt);
 		rImage->mt = NULL;
 	}
-	fprintf(stderr,"settexbuf %dx%d@%d %d targ %x format %x\n", rb->width, rb->height, rb->cpp, rb->pitch, target, format);
 	_mesa_init_teximage_fields(radeon->glCtx, target, texImage,
 				   rb->width, rb->height, 1, 0, rb->cpp);
 	texImage->RowStride = rb->pitch / rb->cpp;
@@ -449,7 +449,10 @@ void r300SetTexBuffer2(__DRIcontext *pDRICtx, GLint target, GLint glx_texture_fo
 	pitch_val = rb->pitch;
 	switch (rb->cpp) {
 	case 4:
-		t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
+		if (glx_texture_format == GLX_TEXTURE_FORMAT_RGB_EXT)
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, ONE, W8Z8Y8X8);
+		else
+			t->pp_txformat = R300_EASY_TX_FORMAT(X, Y, Z, W, W8Z8Y8X8);
 		t->pp_txfilter |= tx_table[2].filter;
 		pitch_val /= 4;
 		break;
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.c b/src/mesa/drivers/dri/r300/r300_vertprog.c
index 50806575ce..c41a8fdd62 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.c
@@ -34,10 +34,12 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "shader/program.h"
 #include "shader/prog_instruction.h"
 #include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
 #include "shader/prog_statevars.h"
 #include "tnl/tnl.h"
 
 #include "r300_context.h"
+#include "r300_state.h"
 
 /* TODO: Get rid of t_src_class call */
 #define CMP_SRCS(a, b) ((a.RelAddr != b.RelAddr) || (a.Index != b.Index && \
@@ -64,7 +66,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 		int u_temp_used = (VSF_MAX_FRAGMENT_TEMPS - 1) - u_temp_i; \
 		if((vp->num_temporaries + u_temp_used) > VSF_MAX_FRAGMENT_TEMPS) { \
 			WARN_ONCE("Ran out of temps, num temps %d, us %d\n", vp->num_temporaries, u_temp_used); \
-			vp->native = GL_FALSE; \
+			vp->error = GL_TRUE; \
 		} \
 		u_temp_i=VSF_MAX_FRAGMENT_TEMPS-1; \
 	} while (0)
@@ -214,21 +216,8 @@ static void vp_dump_inputs(struct r300_vertex_program *vp, char *caller)
 static unsigned long t_src_index(struct r300_vertex_program *vp,
 				 struct prog_src_register *src)
 {
-	int i;
-	int max_reg = -1;
-
 	if (src->File == PROGRAM_INPUT) {
-		if (vp->inputs[src->Index] != -1)
-			return vp->inputs[src->Index];
-
-		for (i = 0; i < VERT_ATTRIB_MAX; i++)
-			if (vp->inputs[i] > max_reg)
-				max_reg = vp->inputs[i];
-
-		vp->inputs[src->Index] = max_reg + 1;
-
-		//vp_dump_inputs(vp, __FUNCTION__);
-
+		assert(vp->inputs[src->Index] != -1);
 		return vp->inputs[src->Index];
 	} else {
 		if (src->Index < 0) {
@@ -245,7 +234,7 @@ static unsigned long t_src_index(struct r300_vertex_program *vp,
 static unsigned long t_src(struct r300_vertex_program *vp,
 			   struct prog_src_register *src)
 {
-	/* src->NegateBase uses the NEGATE_ flags from program_instruction.h,
+	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
 	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
 	 */
 	return PVS_SRC_OPERAND(t_src_index(vp, src),
@@ -254,13 +243,13 @@ static unsigned long t_src(struct r300_vertex_program *vp,
 			       t_swizzle(GET_SWZ(src->Swizzle, 2)),
 			       t_swizzle(GET_SWZ(src->Swizzle, 3)),
 			       t_src_class(src->File),
-			       src->NegateBase) | (src->RelAddr << 4);
+			       src->Negate) | (src->RelAddr << 4);
 }
 
 static unsigned long t_src_scalar(struct r300_vertex_program *vp,
 				  struct prog_src_register *src)
 {
-	/* src->NegateBase uses the NEGATE_ flags from program_instruction.h,
+	/* src->Negate uses the NEGATE_ flags from program_instruction.h,
 	 * which equal our VSF_FLAGS_ values, so it's safe to just pass it here.
 	 */
 	return PVS_SRC_OPERAND(t_src_index(vp, src),
@@ -269,8 +258,7 @@ static unsigned long t_src_scalar(struct r300_vertex_program *vp,
 			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
 			       t_swizzle(GET_SWZ(src->Swizzle, 0)),
 			       t_src_class(src->File),
-			       src->
-			       NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+			       src->Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src->RelAddr << 4);
 }
 
@@ -307,7 +295,7 @@ static GLuint *r300TranslateOpcodeABS(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),
 				  t_src_class(src[0].File),
 				  (!src[0].
-				   NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[3] = 0;
 
@@ -369,8 +357,7 @@ static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
 				  SWIZZLE_ZERO,
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[2] =
 	    PVS_SRC_OPERAND(t_src_index(vp, &src[1]),
@@ -378,8 +365,7 @@ static GLuint *r300TranslateOpcodeDP3(struct r300_vertex_program *vp,
 			    t_swizzle(GET_SWZ(src[1].Swizzle, 1)),
 			    t_swizzle(GET_SWZ(src[1].Swizzle, 2)), SWIZZLE_ZERO,
 			    t_src_class(src[1].File),
-			    src[1].
-			    NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+			    src[1].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
 	    (src[1].RelAddr << 4);
 	inst[3] = __CONST(1, SWIZZLE_ZERO);
 
@@ -422,8 +408,7 @@ static GLuint *r300TranslateOpcodeDPH(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 2)),
 				  PVS_SRC_SELECT_FORCE_1,
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_XYZ : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[2] = t_src(vp, &src[1]);
 	inst[3] = __CONST(1, SWIZZLE_ZERO);
@@ -519,7 +504,7 @@ static GLuint *r300TranslateOpcodeFLR(struct r300_vertex_program *vp,
 				  PVS_SRC_SELECT_W, PVS_SRC_REG_TEMPORARY,
 				  /* Not 100% sure about this */
 				  (!src[0].
-				   NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE
+				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE
 				  /*VSF_FLAG_ALL */ );
 	inst[3] = __CONST(0, SWIZZLE_ZERO);
 	(*u_temp_i)--;
@@ -564,8 +549,7 @@ static GLuint *r300TranslateOpcodeLG2(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[2] = __CONST(0, SWIZZLE_ZERO);
 	inst[3] = __CONST(0, SWIZZLE_ZERO);
@@ -592,24 +576,21 @@ static GLuint *r300TranslateOpcodeLIT(struct r300_vertex_program *vp,
 				  PVS_SRC_SELECT_FORCE_0,	// Z
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
 				  PVS_SRC_SELECT_FORCE_0,	// Z
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[3] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
 				  PVS_SRC_SELECT_FORCE_0,	// Z
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 
 	return inst;
@@ -837,7 +818,7 @@ static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
 				  t_src_class(src[1].File),
 				  (!src[1].
-				   NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[1].RelAddr << 4);
 	inst[3] = 0;
 #else
@@ -857,7 +838,7 @@ static GLuint *r300TranslateOpcodeSUB(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),
 				  t_src_class(src[1].File),
 				  (!src[1].
-				   NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[1].RelAddr << 4);
 #endif
 
@@ -905,16 +886,14 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[1]), t_swizzle(GET_SWZ(src[1].Swizzle, 2)),	// Z
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 0)),	// X
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 1)),	// Y
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// W
 				  t_src_class(src[1].File),
-				  src[1].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[1].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[1].RelAddr << 4);
 	inst[3] = __CONST(1, SWIZZLE_ZERO);
 	inst += 4;
@@ -931,15 +910,14 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp,
 				  t_swizzle(GET_SWZ(src[1].Swizzle, 3)),	// W
 				  t_src_class(src[1].File),
 				  (!src[1].
-				   NegateBase) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				   Negate) ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[1].RelAddr << 4);
 	inst[2] = PVS_SRC_OPERAND(t_src_index(vp, &src[0]), t_swizzle(GET_SWZ(src[0].Swizzle, 2)),	// Z
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 0)),	// X
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 1)),	// Y
 				  t_swizzle(GET_SWZ(src[0].Swizzle, 3)),	// W
 				  t_src_class(src[0].File),
-				  src[0].
-				  NegateBase ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
+				  src[0].Negate ? VSF_FLAG_ALL : VSF_FLAG_NONE) |
 	    (src[0].RelAddr << 4);
 	inst[3] =
 	    PVS_SRC_OPERAND(*u_temp_i, PVS_SRC_SELECT_X, PVS_SRC_SELECT_Y,
@@ -954,11 +932,17 @@ static GLuint *r300TranslateOpcodeXPD(struct r300_vertex_program *vp,
 static void t_inputs_outputs(struct r300_vertex_program *vp)
 {
 	int i;
-	int cur_reg = 0;
+	int cur_reg;
 
-	for (i = 0; i < VERT_ATTRIB_MAX; i++)
-		vp->inputs[i] = -1;
+	cur_reg = -1;
+	for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+		if (vp->key.InputsRead & (1 << i))
+			vp->inputs[i] = ++cur_reg;
+		else
+			vp->inputs[i] = -1;
+	}
 
+	cur_reg = 0;
 	for (i = 0; i < VERT_RESULT_MAX; i++)
 		vp->outputs[i] = -1;
 
@@ -972,26 +956,36 @@ static void t_inputs_outputs(struct r300_vertex_program *vp)
 		vp->outputs[VERT_RESULT_PSIZ] = cur_reg++;
 	}
 
+	/* If we're writing back facing colors we need to send
+	 * four colors to make front/back face colors selection work.
+	 * If the vertex program doesn't write all 4 colors, lets
+	 * pretend it does by skipping output index reg so the colors
+	 * get written into appropriate output vectors.
+	 */
 	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL0)) {
 		vp->outputs[VERT_RESULT_COL0] = cur_reg++;
+	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
 	if (vp->key.OutputsWritten & (1 << VERT_RESULT_COL1)) {
-		vp->outputs[VERT_RESULT_COL1] =
-		    vp->outputs[VERT_RESULT_COL0] + 1;
-		cur_reg = vp->outputs[VERT_RESULT_COL1] + 1;
+		vp->outputs[VERT_RESULT_COL1] = cur_reg++;
+	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0) ||
+		vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
 	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
-		vp->outputs[VERT_RESULT_BFC0] =
-		    vp->outputs[VERT_RESULT_COL0] + 2;
-		cur_reg = vp->outputs[VERT_RESULT_BFC0] + 2;
+		vp->outputs[VERT_RESULT_BFC0] = cur_reg++;
+	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
+		cur_reg++;
 	}
 
 	if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC1)) {
-		vp->outputs[VERT_RESULT_BFC1] =
-		    vp->outputs[VERT_RESULT_COL0] + 3;
-		cur_reg = vp->outputs[VERT_RESULT_BFC1] + 1;
+		vp->outputs[VERT_RESULT_BFC1] = cur_reg++;
+	} else if (vp->key.OutputsWritten & (1 << VERT_RESULT_BFC0)) {
+		cur_reg++;
 	}
 
 	for (i = VERT_RESULT_TEX0; i <= VERT_RESULT_TEX7; i++) {
@@ -1018,14 +1012,13 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 	struct prog_src_register src[3];
 
 	vp->pos_end = 0;	/* Not supported yet */
-	vp->program.length = 0;
-	/*vp->num_temporaries=mesa_vp->Base.NumTemporaries; */
+	vp->hw_code.length = 0;
 	vp->translated = GL_TRUE;
-	vp->native = GL_TRUE;
+	vp->error = GL_FALSE;
 
 	t_inputs_outputs(vp);
 
-	for (inst = vp->program.body.i; vpi->Opcode != OPCODE_END;
+	for (inst = vp->hw_code.body.d; vpi->Opcode != OPCODE_END;
 	     vpi++, inst += 4) {
 
 		FREE_TEMPS();
@@ -1187,38 +1180,15 @@ static void r300TranslateVertexShader(struct r300_vertex_program *vp,
 						      &u_temp_i);
 			break;
 		default:
-			assert(0);
+			vp->error = GL_TRUE;
 			break;
 		}
 	}
 
-	/* Some outputs may be artificially added, to match the inputs
-	   of the fragment program. Blank the outputs here. */
-	for (i = 0; i < VERT_RESULT_MAX; i++) {
-		if (vp->key.OutputsAdded & (1 << i)) {
-			inst[0] = PVS_OP_DST_OPERAND(VE_ADD,
-						     GL_FALSE,
-						     GL_FALSE,
-						     vp->outputs[i],
-						     VSF_FLAG_ALL,
-						     PVS_DST_REG_OUT);
-			inst[1] = __CONST(0, SWIZZLE_ZERO);
-			inst[2] = __CONST(0, SWIZZLE_ZERO);
-			inst[3] = __CONST(0, SWIZZLE_ZERO);
-			inst += 4;
-		}
+	vp->hw_code.length = (inst - vp->hw_code.body.d);
+	if (vp->hw_code.length >= VSF_MAX_FRAGMENT_LENGTH) {
+		vp->error = GL_TRUE;
 	}
-
-	vp->program.length = (inst - vp->program.body.i);
-	if (vp->program.length >= VSF_MAX_FRAGMENT_LENGTH) {
-		vp->program.length = 0;
-		vp->native = GL_FALSE;
-	}
-#if 0
-	fprintf(stderr, "hw program:\n");
-	for (i = 0; i < vp->program.length; i++)
-		fprintf(stderr, "%08x\n", vp->program.body.d[i]);
-#endif
 }
 
 /* DP4 version seems to trigger some hw peculiarity */
@@ -1397,6 +1367,49 @@ static struct r300_vertex_program *build_program(struct r300_vertex_program_key
 		pos_as_texcoord(vp, &mesa_vp->Base);
 	}
 
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Vertex program after native rewrite:\n");
+		_mesa_print_program(&mesa_vp->Base);
+		fflush(stdout);
+	}
+
+	/* Some outputs may be artificially added, to match the inputs of the fragment program.
+	 * Issue 16 of vertex program spec says that all vertex attributes that are unwritten by
+	 * vertex program are undefined, so just use MOV [vertex_result], CONST[0]
+	 */
+	{
+		int i, count = 0;
+		for (i = 0; i < VERT_RESULT_MAX; ++i) {
+			if (vp->key.OutputsAdded & (1 << i)) {
+				++count;
+			}
+		}
+
+		if (count > 0) {
+			struct prog_instruction *inst;
+
+			_mesa_insert_instructions(&mesa_vp->Base, mesa_vp->Base.NumInstructions - 1, count);
+			inst = &mesa_vp->Base.Instructions[mesa_vp->Base.NumInstructions - 1 - count];
+
+			for (i = 0; i < VERT_RESULT_MAX; ++i) {
+				if (vp->key.OutputsAdded & (1 << i)) {
+					inst->Opcode = OPCODE_MOV;
+
+					inst->DstReg.File = PROGRAM_OUTPUT;
+					inst->DstReg.Index = i;
+					inst->DstReg.WriteMask = WRITEMASK_XYZW;
+					inst->DstReg.CondMask = COND_TR;
+
+					inst->SrcReg[0].File = PROGRAM_CONSTANT;
+					inst->SrcReg[0].Index = 0;
+					inst->SrcReg[0].Swizzle = SWIZZLE_XYZW;
+
+					++inst;
+				}
+			}
+		}
+	}
+
 	assert(mesa_vp->Base.NumInstructions);
 	vp->num_temporaries = mesa_vp->Base.NumTemporaries;
 	r300TranslateVertexShader(vp, mesa_vp->Base.Instructions);
@@ -1443,7 +1456,12 @@ void r300SelectVertexShader(r300ContextPtr r300)
 		wpos_idx = i;
 	}
 
-	add_outputs(&wanted_key, VERT_RESULT_HPOS);
+	if (vpc->mesa_program.IsPositionInvariant) {
+		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
+		wanted_key.OutputsWritten |= (1 << VERT_RESULT_HPOS);
+	} else {
+		add_outputs(&wanted_key, VERT_RESULT_HPOS);
+	}
 
 	if (InputsRead & FRAG_BIT_COL0) {
 		add_outputs(&wanted_key, VERT_RESULT_COL0);
@@ -1453,27 +1471,103 @@ void r300SelectVertexShader(r300ContextPtr r300)
 		add_outputs(&wanted_key, VERT_RESULT_COL1);
 	}
 
+	if (InputsRead & FRAG_BIT_FOGC) {
+		add_outputs(&wanted_key, VERT_RESULT_FOGC);
+	}
+
 	for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 		if (InputsRead & (FRAG_BIT_TEX0 << i)) {
 			add_outputs(&wanted_key, VERT_RESULT_TEX0 + i);
 		}
 	}
 
-	if (vpc->mesa_program.IsPositionInvariant) {
-		/* we wan't position don't we ? */
-		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
-	}
-
 	for (vp = vpc->progs; vp; vp = vp->next)
 		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key))
 		    == 0) {
 			r300->selected_vp = vp;
 			return;
 		}
-	//_mesa_print_program(&vpc->mesa_program.Base);
+
+	if (RADEON_DEBUG & DEBUG_VERTS) {
+		fprintf(stderr, "Initial vertex program:\n");
+		_mesa_print_program(&vpc->mesa_program.Base);
+		fflush(stdout);
+	}
 
 	vp = build_program(&wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
 	r300->selected_vp = vp;
 }
+
+#define bump_vpu_count(ptr, new_count)   do { \
+		drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr)); \
+		int _nc=(new_count)/4; \
+		assert(_nc < 256); \
+		if(_nc>_p->vpu.count)_p->vpu.count=_nc; \
+	} while(0)
+
+static void r300EmitVertexProgram(r300ContextPtr r300, int dest, struct r300_vertex_shader_hw_code *code)
+{
+	int i;
+
+	assert((code->length > 0) && (code->length % 4 == 0));
+
+	switch ((dest >> 8) & 0xf) {
+		case 0:
+			R300_STATECHANGE(r300, vpi);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpi.cmd[R300_VPI_INSTR_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpi.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 2:
+			R300_STATECHANGE(r300, vpp);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vpp.cmd[R300_VPP_PARAM_0 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vpp.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		case 4:
+			R300_STATECHANGE(r300, vps);
+			for (i = 0; i < code->length; i++)
+				r300->hw.vps.cmd[1 + i + 4 * (dest & 0xff)] = (code->body.d[i]);
+			bump_vpu_count(r300->hw.vps.cmd, code->length + 4 * (dest & 0xff));
+			break;
+		default:
+			fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
+			_mesa_exit(-1);
+	}
+}
+
+void r300SetupVertexProgram(r300ContextPtr rmesa)
+{
+	GLcontext *ctx = rmesa->radeon.glCtx;
+	struct r300_vertex_program *prog = rmesa->selected_vp;
+	int inst_count = 0;
+	int param_count = 0;
+	
+	/* Reset state, in case we don't use something */
+	((drm_r300_cmd_header_t *) rmesa->hw.vpp.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vpi.cmd)->vpu.count = 0;
+	((drm_r300_cmd_header_t *) rmesa->hw.vps.cmd)->vpu.count = 0;
+	
+	R300_STATECHANGE(rmesa, vpp);
+	param_count = r300VertexProgUpdateParams(ctx,
+								(struct r300_vertex_program_cont *)
+								ctx->VertexProgram._Current,
+								(float *)&rmesa->hw.vpp.cmd[R300_VPP_PARAM_0]);
+	bump_vpu_count(rmesa->hw.vpp.cmd, param_count);
+	param_count /= 4;
+
+	r300EmitVertexProgram(rmesa, R300_PVS_CODE_START, &(prog->hw_code));
+	inst_count = (prog->hw_code.length / 4) - 1;
+
+	r300VapCntl(rmesa, _mesa_bitcount(prog->key.InputsRead),
+				 _mesa_bitcount(prog->key.OutputsWritten), prog->num_temporaries);
+
+	R300_STATECHANGE(rmesa, pvs);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_1] = (0 << R300_PVS_FIRST_INST_SHIFT) | (inst_count << R300_PVS_XYZW_VALID_INST_SHIFT) |
+				(inst_count << R300_PVS_LAST_INST_SHIFT);
+
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_2] = (0 << R300_PVS_CONST_BASE_OFFSET_SHIFT) | (param_count << R300_PVS_MAX_CONST_ADDR_SHIFT);
+	rmesa->hw.pvs.cmd[R300_PVS_CNTL_3] = (inst_count << R300_PVS_LAST_VTX_SRC_INST_SHIFT);
+}
diff --git a/src/mesa/drivers/dri/r300/r300_vertprog.h b/src/mesa/drivers/dri/r300/r300_vertprog.h
index 2f35f02bc8..b552e3fb1b 100644
--- a/src/mesa/drivers/dri/r300/r300_vertprog.h
+++ b/src/mesa/drivers/dri/r300/r300_vertprog.h
@@ -32,4 +32,6 @@
 
 #endif
 
+void r300SetupVertexProgram(r300ContextPtr rmesa);
+
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.c b/src/mesa/drivers/dri/r300/r500_fragprog.c
index 07a2a7b17c..4d58cf2162 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r500_fragprog.c
@@ -27,10 +27,6 @@
 
 #include "r500_fragprog.h"
 
-#include "radeon_nqssadce.h"
-#include "radeon_program_alu.h"
-
-
 static void reset_srcreg(struct prog_src_register* reg)
 {
 	_mesa_bzero(reg, sizeof(*reg));
@@ -58,12 +54,12 @@ static struct prog_src_register shadow_ambient(struct gl_program *program, int t
  *  - introduce a temporary register when write masks are needed
  *
  */
-static GLboolean transform_TEX(
+GLboolean r500_transform_TEX(
 	struct radeon_transform_context *t,
 	struct prog_instruction* orig_inst, void* data)
 {
-	struct r500_fragment_program_compiler *compiler =
-		(struct r500_fragment_program_compiler*)data;
+	struct r300_fragment_program_compiler *compiler =
+		(struct r300_fragment_program_compiler*)data;
 	struct prog_instruction inst = *orig_inst;
 	struct prog_instruction* tgt;
 	GLboolean destredirect = GL_FALSE;
@@ -156,9 +152,9 @@ static GLboolean transform_TEX(
 		 *   r  < tex  <=>      -tex+r < 0
 		 *   r >= tex  <=> not (-tex+r < 0 */
 		if (comparefunc == GL_LESS || comparefunc == GL_GEQUAL)
-			tgt[1].SrcReg[2].NegateBase = tgt[0].SrcReg[2].NegateBase ^ NEGATE_XYZW;
+			tgt[1].SrcReg[2].Negate = tgt[0].SrcReg[2].Negate ^ NEGATE_XYZW;
 		else
-			tgt[1].SrcReg[0].NegateBase = tgt[0].SrcReg[0].NegateBase ^ NEGATE_XYZW;
+			tgt[1].SrcReg[0].Negate = tgt[0].SrcReg[0].Negate ^ NEGATE_XYZW;
 
 		tgt[2].Opcode = OPCODE_CMP;
 		tgt[2].DstReg = orig_inst->DstReg;
@@ -188,121 +184,7 @@ static GLboolean transform_TEX(
 	return GL_TRUE;
 }
 
-
-static void update_params(r300ContextPtr r300, struct r500_fragment_program *fp)
-{
-	struct gl_fragment_program *mp = &fp->mesa_program;
-
-	/* Ask Mesa nicely to fill in ParameterValues for us */
-	if (mp->Base.Parameters)
-		_mesa_load_state_parameters(r300->radeon.glCtx, mp->Base.Parameters);
-}
-
-
-/**
- * Transform the program to support fragment.position.
- *
- * Introduce a small fragment at the start of the program that will be
- * the only code that directly reads the FRAG_ATTRIB_WPOS input.
- * All other code pieces that reference that input will be rewritten
- * to read from a newly allocated temporary.
- *
- * \todo if/when r5xx supports the radeon_program architecture, this is a
- * likely candidate for code sharing.
- */
-static void insert_WPOS_trailer(struct r500_fragment_program_compiler *compiler)
-{
-	GLuint InputsRead = compiler->fp->mesa_program.Base.InputsRead;
-
-	if (!(InputsRead & FRAG_BIT_WPOS))
-		return;
-
-	static gl_state_index tokens[STATE_LENGTH] = {
-		STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0
-	};
-	struct prog_instruction *fpi;
-	GLuint window_index;
-	int i = 0;
-	GLuint tempregi = _mesa_find_free_register(compiler->program, PROGRAM_TEMPORARY);
-
-	_mesa_insert_instructions(compiler->program, 0, 3);
-	fpi = compiler->program->Instructions;
-
-	/* perspective divide */
-	fpi[i].Opcode = OPCODE_RCP;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_W;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	fpi[i].Opcode = OPCODE_MUL;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_INPUT;
-	fpi[i].SrcReg[0].Index = FRAG_ATTRIB_WPOS;
-	fpi[i].SrcReg[0].Swizzle = SWIZZLE_XYZW;
-
-	fpi[i].SrcReg[1].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[1].Index = tempregi;
-	fpi[i].SrcReg[1].Swizzle = SWIZZLE_WWWW;
-	i++;
-
-	/* viewport transformation */
-	window_index = _mesa_add_state_reference(compiler->program->Parameters, tokens);
-
-	fpi[i].Opcode = OPCODE_MAD;
-
-	fpi[i].DstReg.File = PROGRAM_TEMPORARY;
-	fpi[i].DstReg.Index = tempregi;
-	fpi[i].DstReg.WriteMask = WRITEMASK_XYZ;
-	fpi[i].DstReg.CondMask = COND_TR;
-
-	fpi[i].SrcReg[0].File = PROGRAM_TEMPORARY;
-	fpi[i].SrcReg[0].Index = tempregi;
-	fpi[i].SrcReg[0].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[1].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[1].Index = window_index;
-	fpi[i].SrcReg[1].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-
-	fpi[i].SrcReg[2].File = PROGRAM_STATE_VAR;
-	fpi[i].SrcReg[2].Index = window_index;
-	fpi[i].SrcReg[2].Swizzle =
-	    MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_Z, SWIZZLE_ZERO);
-	i++;
-
-	for (; i < compiler->program->NumInstructions; ++i) {
-		int reg;
-		for (reg = 0; reg < 3; reg++) {
-			if (fpi[i].SrcReg[reg].File == PROGRAM_INPUT &&
-			    fpi[i].SrcReg[reg].Index == FRAG_ATTRIB_WPOS) {
-				fpi[i].SrcReg[reg].File = PROGRAM_TEMPORARY;
-				fpi[i].SrcReg[reg].Index = tempregi;
-			}
-		}
-	}
-}
-
-
-static void nqssadce_init(struct nqssadce_state* s)
-{
-	s->Outputs[FRAG_RESULT_COLOR].Sourced = WRITEMASK_XYZW;
-	s->Outputs[FRAG_RESULT_DEPTH].Sourced = WRITEMASK_W;
-}
-
-static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
+GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg)
 {
 	GLuint relevant;
 	int i;
@@ -314,33 +196,30 @@ static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
 		if (reg.Abs)
 			return GL_FALSE;
 
-		if (reg.NegateAbs)
-			reg.NegateBase ^= 15;
+		if (opcode == OPCODE_KIL && (reg.Swizzle != SWIZZLE_NOOP || reg.Negate != NEGATE_NONE))
+			return GL_FALSE;
 
-		if (opcode == OPCODE_KIL) {
-			if (reg.Swizzle != SWIZZLE_NOOP)
-				return GL_FALSE;
-		} else {
-			for(i = 0; i < 4; ++i) {
-				GLuint swz = GET_SWZ(reg.Swizzle, i);
-				if (swz == SWIZZLE_NIL) {
-					reg.NegateBase &= ~(1 << i);
-					continue;
-				}
-				if (swz >= 4)
-					return GL_FALSE;
+		if (reg.Negate)
+			reg.Negate ^= NEGATE_XYZW;
+
+		for(i = 0; i < 4; ++i) {
+			GLuint swz = GET_SWZ(reg.Swizzle, i);
+			if (swz == SWIZZLE_NIL) {
+				reg.Negate &= ~(1 << i);
+				continue;
 			}
+			if (swz >= 4)
+				return GL_FALSE;
 		}
 
-		if (reg.NegateBase)
+		if (reg.Negate)
 			return GL_FALSE;
 
 		return GL_TRUE;
 	} else if (opcode == OPCODE_DDX || opcode == OPCODE_DDY) {
 		/* DDX/MDH and DDY/MDV explicitly ignore incoming swizzles;
 		 * if it doesn't fit perfectly into a .xyzw case... */
-		if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs
-				&& !reg.NegateBase && !reg.NegateAbs)
+		if (reg.Swizzle == SWIZZLE_NOOP && !reg.Abs && !reg.Negate)
 			return GL_TRUE;
 
 		return GL_FALSE;
@@ -355,7 +234,7 @@ static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
 			if (swz != SWIZZLE_NIL && swz != SWIZZLE_ZERO)
 				relevant |= 1 << i;
 		}
-		if ((reg.NegateBase & relevant) && ((reg.NegateBase & relevant) != relevant))
+		if ((reg.Negate & relevant) && ((reg.Negate & relevant) != relevant))
 			return GL_FALSE;
 
 		return GL_TRUE;
@@ -368,8 +247,7 @@ static GLboolean is_native_swizzle(GLuint opcode, struct prog_src_register reg)
  * The only thing we *cannot* do in an ALU instruction is per-component
  * negation. Therefore, we split the MOV into two instructions when necessary.
  */
-static void nqssadce_build_swizzle(struct nqssadce_state *s,
-	struct prog_dst_register dst, struct prog_src_register src)
+void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src)
 {
 	struct prog_instruction *inst;
 	GLuint negatebase[2] = { 0, 0 };
@@ -379,7 +257,7 @@ static void nqssadce_build_swizzle(struct nqssadce_state *s,
 		GLuint swz = GET_SWZ(src.Swizzle, i);
 		if (swz == SWIZZLE_NIL)
 			continue;
-		negatebase[GET_BIT(src.NegateBase, i)] |= 1 << i;
+		negatebase[GET_BIT(src.Negate, i)] |= 1 << i;
 	}
 
 	_mesa_insert_instructions(s->Program, s->IP, (negatebase[0] ? 1 : 0) + (negatebase[1] ? 1 : 0));
@@ -393,129 +271,12 @@ static void nqssadce_build_swizzle(struct nqssadce_state *s,
 		inst->DstReg = dst;
 		inst->DstReg.WriteMask = negatebase[i];
 		inst->SrcReg[0] = src;
+		inst->SrcReg[0].Negate = (i == 0) ? NEGATE_NONE : NEGATE_XYZW;
 		inst++;
 		s->IP++;
 	}
 }
 
-static GLuint build_dtm(GLuint depthmode)
-{
-	switch(depthmode) {
-	default:
-	case GL_LUMINANCE: return 0;
-	case GL_INTENSITY: return 1;
-	case GL_ALPHA: return 2;
-	}
-}
-
-static GLuint build_func(GLuint comparefunc)
-{
-	return comparefunc - GL_NEVER;
-}
-
-
-/**
- * Collect all external state that is relevant for compiling the given
- * fragment program.
- */
-static void build_state(
-	r300ContextPtr r300,
-	struct r500_fragment_program *fp,
-	struct r500_fragment_program_external_state *state)
-{
-	int unit;
-
-	_mesa_bzero(state, sizeof(*state));
-
-	for(unit = 0; unit < 16; ++unit) {
-		if (fp->mesa_program.Base.ShadowSamplers & (1 << unit)) {
-			struct gl_texture_object* tex = r300->radeon.glCtx->Texture.Unit[unit]._Current;
-
-			state->unit[unit].depth_texture_mode = build_dtm(tex->DepthMode);
-			state->unit[unit].texture_compare_func = build_func(tex->CompareFunc);
-		}
-	}
-}
-
-static void dump_program(struct r500_fragment_program_code *code);
-
-void r500TranslateFragmentShader(r300ContextPtr r300,
-				 struct r500_fragment_program *fp)
-{
-	struct r500_fragment_program_external_state state;
-
-	build_state(r300, fp, &state);
-	if (_mesa_memcmp(&fp->state, &state, sizeof(state))) {
-		/* TODO: cache compiled programs */
-		fp->translated = GL_FALSE;
-		_mesa_memcpy(&fp->state, &state, sizeof(state));
-	}
-
-	if (!fp->translated) {
-		struct r500_fragment_program_compiler compiler;
-
-		compiler.r300 = r300;
-		compiler.fp = fp;
-		compiler.code = &fp->code;
-		compiler.program = _mesa_clone_program(r300->radeon.glCtx, &fp->mesa_program.Base);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: Initial program:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		insert_WPOS_trailer(&compiler);
-
-		struct radeon_program_transformation transformations[] = {
-			{ &transform_TEX, &compiler },
-			{ &radeonTransformALU, 0 },
-			{ &radeonTransformDeriv, 0 },
-			{ &radeonTransformTrigScale, 0 }
-		};
-		radeonLocalTransform(r300->radeon.glCtx, compiler.program,
-			4, transformations);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after native rewrite:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		struct radeon_nqssadce_descr nqssadce = {
-			.Init = &nqssadce_init,
-			.IsNativeSwizzle = &is_native_swizzle,
-			.BuildSwizzle = &nqssadce_build_swizzle,
-			.RewriteDepthOut = GL_TRUE
-		};
-		radeonNqssaDce(r300->radeon.glCtx, compiler.program, &nqssadce);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			_mesa_printf("Compiler: after NqSSA-DCE:\n");
-			_mesa_print_program(compiler.program);
-		}
-
-		fp->translated = r500FragmentProgramEmit(&compiler);
-
-		/* Subtle: Rescue any parameters that have been added during transformations */
-		_mesa_free_parameter_list(fp->mesa_program.Base.Parameters);
-		fp->mesa_program.Base.Parameters = compiler.program->Parameters;
-		compiler.program->Parameters = 0;
-
-		_mesa_reference_program(r300->radeon.glCtx, &compiler.program, 0);
-
-		r300UpdateStateParameters(r300->radeon.glCtx, _NEW_PROGRAM);
-
-		if (RADEON_DEBUG & DEBUG_PIXEL) {
-			if (fp->translated) {
-				_mesa_printf("Machine-readable code:\n");
-				dump_program(&fp->code);
-			}
-		}
-
-	}
-
-	update_params(r300, fp);
-
-}
 
 static char *toswiz(int swiz_val) {
   switch(swiz_val) {
@@ -614,9 +375,9 @@ static char *to_texop(int val)
   return NULL;
 }
 
-static void dump_program(struct r500_fragment_program_code *code)
+void r500FragmentProgramDump(union rX00_fragment_program_code *c)
 {
-
+  struct r500_fragment_program_code *code = &c->r500;
   fprintf(stderr, "R500 Fragment Program:\n--------\n");
 
   int n;
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog.h b/src/mesa/drivers/dri/r300/r500_fragprog.h
index 1e45538f80..1179bf6607 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog.h
+++ b/src/mesa/drivers/dri/r300/r500_fragprog.h
@@ -33,30 +33,20 @@
 #ifndef __R500_FRAGPROG_H_
 #define __R500_FRAGPROG_H_
 
-#include "main/glheader.h"
-#include "main/macros.h"
-#include "main/enums.h"
 #include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-#include "shader/program.h"
 #include "shader/prog_instruction.h"
 
 #include "r300_context.h"
-#include "r300_state.h"
-#include "radeon_program.h"
+#include "radeon_nqssadce.h"
 
-struct r500_fragment_program;
+extern GLboolean r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler);
 
-extern void r500TranslateFragmentShader(r300ContextPtr r300,
-					struct r500_fragment_program *fp);
+extern void r500FragmentProgramDump(union rX00_fragment_program_code *c);
 
-struct r500_fragment_program_compiler {
-	r300ContextPtr r300;
-	struct r500_fragment_program *fp;
-	struct r500_fragment_program_code *code;
-	struct gl_program *program;
-};
+extern GLboolean r500FPIsNativeSwizzle(GLuint opcode, struct prog_src_register reg);
 
-extern GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler);
+extern void r500FPBuildSwizzle(struct nqssadce_state *s, struct prog_dst_register dst, struct prog_src_register src);
+
+extern GLboolean r500_transform_TEX(struct radeon_transform_context *t, struct prog_instruction* orig_inst, void* data);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c b/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
index 4631235f0d..30f4514897 100644
--- a/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
+++ b/src/mesa/drivers/dri/r300/r500_fragprog_emit.c
@@ -49,8 +49,8 @@
 
 
 #define PROG_CODE \
-	struct r500_fragment_program_compiler *c = (struct r500_fragment_program_compiler*)data; \
-	struct r500_fragment_program_code *code = c->code
+	struct r300_fragment_program_compiler *c = (struct r300_fragment_program_compiler*)data; \
+	struct r500_fragment_program_code *code = &c->code->r500
 
 #define error(fmt, args...) do {			\
 		fprintf(stderr, "%s::%s(): " fmt "\n",	\
@@ -72,7 +72,7 @@ static GLboolean emit_const(void *data, GLuint file, GLuint idx, GLuint *hwindex
 	}
 
 	if (*hwindex >= code->const_nr) {
-		if (*hwindex >= PFS_NUM_CONST_REGS) {
+		if (*hwindex >= R500_PFS_NUM_CONST_REGS) {
 			error("Out of hw constants!\n");
 			return GL_FALSE;
 		}
@@ -299,9 +299,9 @@ static const struct radeon_pair_handler pair_handler = {
 	.MaxHwTemps = 128
 };
 
-GLboolean r500FragmentProgramEmit(struct r500_fragment_program_compiler *compiler)
+GLboolean r500BuildFragmentProgramHwCode(struct r300_fragment_program_compiler *compiler)
 {
-	struct r500_fragment_program_code *code = compiler->code;
+	struct r500_fragment_program_code *code = &compiler->code->r500;
 
 	_mesa_bzero(code, sizeof(*code));
 	code->max_temp_idx = 1;
diff --git a/src/mesa/drivers/dri/r300/radeon_nqssadce.c b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
index a083c3d243..4a2e1cba40 100644
--- a/src/mesa/drivers/dri/r300/radeon_nqssadce.c
+++ b/src/mesa/drivers/dri/r300/radeon_nqssadce.c
@@ -61,12 +61,12 @@ static struct prog_src_register lmul_swizzle(GLuint swizzle, struct prog_src_reg
 	struct prog_src_register tmp = srcreg;
 	int i;
 	tmp.Swizzle = 0;
-	tmp.NegateBase = 0;
+	tmp.Negate = NEGATE_NONE;
 	for(i = 0; i < 4; ++i) {
 		GLuint swz = GET_SWZ(swizzle, i);
 		if (swz < 4) {
 			tmp.Swizzle |= GET_SWZ(srcreg.Swizzle, swz) << (i*3);
-			tmp.NegateBase |= GET_BIT(srcreg.NegateBase, swz) << i;
+			tmp.Negate |= GET_BIT(srcreg.Negate, swz) << i;
 		} else {
 			tmp.Swizzle |= swz << (i*3);
 		}
@@ -103,9 +103,8 @@ static struct prog_instruction* track_used_srcreg(struct nqssadce_state* s,
 		inst->SrcReg[src].File = PROGRAM_TEMPORARY;
 		inst->SrcReg[src].Index = dstreg.Index;
 		inst->SrcReg[src].Swizzle = 0;
-		inst->SrcReg[src].NegateBase = 0;
+		inst->SrcReg[src].Negate = NEGATE_NONE;
 		inst->SrcReg[src].Abs = 0;
-		inst->SrcReg[src].NegateAbs = 0;
 		for(i = 0; i < 4; ++i) {
 			if (GET_BIT(sourced, i))
 				inst->SrcReg[src].Swizzle |= i << (3*i);
diff --git a/src/mesa/drivers/dri/r300/radeon_program_alu.c b/src/mesa/drivers/dri/r300/radeon_program_alu.c
index 1ef71e74dc..8283723bad 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_alu.c
+++ b/src/mesa/drivers/dri/r300/radeon_program_alu.c
@@ -81,18 +81,6 @@ static struct prog_instruction *emit3(struct gl_program* p,
 	return fpi;
 }
 
-static void set_swizzle(struct prog_src_register *SrcReg, int coordinate, int swz)
-{
-	SrcReg->Swizzle &= ~(7 << (3*coordinate));
-	SrcReg->Swizzle |= swz << (3*coordinate);
-}
-
-static void set_negate_base(struct prog_src_register *SrcReg, int coordinate, int negate)
-{
-	SrcReg->NegateBase &= ~(1 << coordinate);
-	SrcReg->NegateBase |= (negate << coordinate);
-}
-
 static struct prog_dst_register dstreg(int file, int index)
 {
 	struct prog_dst_register dst;
@@ -156,15 +144,14 @@ static struct prog_src_register absolute(struct prog_src_register reg)
 {
 	struct prog_src_register newreg = reg;
 	newreg.Abs = 1;
-	newreg.NegateBase = 0;
-	newreg.NegateAbs = 0;
+	newreg.Negate = NEGATE_NONE;
 	return newreg;
 }
 
 static struct prog_src_register negate(struct prog_src_register reg)
 {
 	struct prog_src_register newreg = reg;
-	newreg.NegateAbs = !newreg.NegateAbs;
+	newreg.Negate = newreg.Negate ^ NEGATE_XYZW;
 	return newreg;
 }
 
@@ -189,8 +176,7 @@ static void transform_ABS(struct radeon_transform_context* t,
 {
 	struct prog_src_register src = inst->SrcReg[0];
 	src.Abs = 1;
-	src.NegateBase = 0;
-	src.NegateAbs = 0;
+	src.Negate = NEGATE_NONE;
 	emit1(t->Program, OPCODE_MOV, inst->SaturateMode, inst->DstReg, src);
 }
 
@@ -198,18 +184,9 @@ static void transform_DPH(struct radeon_transform_context* t,
 	struct prog_instruction* inst)
 {
 	struct prog_src_register src0 = inst->SrcReg[0];
-	if (src0.NegateAbs) {
-		if (src0.Abs) {
-			int tempreg = radeonFindFreeTemporary(t);
-			emit1(t->Program, OPCODE_MOV, 0, dstreg(PROGRAM_TEMPORARY, tempreg), src0);
-			src0 = srcreg(src0.File, src0.Index);
-		} else {
-			src0.NegateAbs = 0;
-			src0.NegateBase ^= NEGATE_XYZW;
-		}
-	}
-	set_swizzle(&src0, 3, SWIZZLE_ONE);
-	set_negate_base(&src0, 3, 0);
+	src0.Negate &= ~NEGATE_W;
+	src0.Swizzle &= ~(7 << (3 * 3));
+	src0.Swizzle |= SWIZZLE_ONE << (3 * 3);
 	emit2(t->Program, OPCODE_DP4, inst->SaturateMode, inst->DstReg, src0, inst->SrcReg[1]);
 }
 
@@ -649,7 +626,7 @@ GLboolean radeonTransformDeriv(struct radeon_transform_context* t,
 
 	B.Swizzle = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE,
 						SWIZZLE_ONE, SWIZZLE_ONE);
-	B.NegateBase = NEGATE_XYZW;
+	B.Negate = NEGATE_XYZW;
 
 	emit2(t->Program, inst->Opcode, inst->SaturateMode, inst->DstReg,
 		inst->SrcReg[0], B);
diff --git a/src/mesa/drivers/dri/r300/radeon_program_pair.c b/src/mesa/drivers/dri/r300/radeon_program_pair.c
index 49aa90dd94..906d36e522 100644
--- a/src/mesa/drivers/dri/r300/radeon_program_pair.c
+++ b/src/mesa/drivers/dri/r300/radeon_program_pair.c
@@ -255,8 +255,7 @@ static void final_rewrite(struct pair_state *s, struct prog_instruction *inst)
 		inst->SrcReg[2] = inst->SrcReg[1];
 		inst->SrcReg[1].File = PROGRAM_BUILTIN;
 		inst->SrcReg[1].Swizzle = SWIZZLE_1111;
-		inst->SrcReg[1].NegateBase = 0;
-		inst->SrcReg[1].NegateAbs = 0;
+		inst->SrcReg[1].Negate = NEGATE_NONE;
 		inst->Opcode = OPCODE_MAD;
 		break;
 	case OPCODE_CMP:
@@ -610,6 +609,7 @@ static void emit_all_tex(struct pair_state *s)
 		if (s->Debug) {
 			_mesa_printf("   ");
 			_mesa_print_instruction(inst);
+			fflush(stdout);
 		}
 		s->Error = s->Error || !s->Handler->EmitTex(s->UserData, inst);
 	}
@@ -722,7 +722,6 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_
 		if (pairinst->NeedRGB && !pairinst->IsTranscendent) {
 			GLboolean srcrgb = GL_FALSE;
 			GLboolean srcalpha = GL_FALSE;
-			GLuint negatebase = 0;
 			int j;
 			for(j = 0; j < 3; ++j) {
 				GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, j);
@@ -730,8 +729,6 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_
 					srcrgb = GL_TRUE;
 				else if (swz < 4)
 					srcalpha = GL_TRUE;
-				if (swz != SWIZZLE_NIL && GET_BIT(inst->SrcReg[i].NegateBase, j))
-					negatebase = 1;
 			}
 			source = alloc_pair_source(s, pair, inst->SrcReg[i], srcrgb, srcalpha);
 			if (source < 0)
@@ -739,12 +736,11 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_
 			pair->RGB.Arg[i].Source = source;
 			pair->RGB.Arg[i].Swizzle = inst->SrcReg[i].Swizzle & 0x1ff;
 			pair->RGB.Arg[i].Abs = inst->SrcReg[i].Abs;
-			pair->RGB.Arg[i].Negate = (negatebase & ~pair->RGB.Arg[i].Abs) ^ inst->SrcReg[i].NegateAbs;
+			pair->RGB.Arg[i].Negate = !!(inst->SrcReg[i].Negate & (NEGATE_X | NEGATE_Y | NEGATE_Z));
 		}
 		if (pairinst->NeedAlpha) {
 			GLboolean srcrgb = GL_FALSE;
 			GLboolean srcalpha = GL_FALSE;
-			GLuint negatebase = GET_BIT(inst->SrcReg[i].NegateBase, pairinst->IsTranscendent ? 0 : 3);
 			GLuint swz = GET_SWZ(inst->SrcReg[i].Swizzle, pairinst->IsTranscendent ? 0 : 3);
 			if (swz < 3)
 				srcrgb = GL_TRUE;
@@ -756,7 +752,7 @@ static GLboolean fill_instruction_into_pair(struct pair_state *s, struct radeon_
 			pair->Alpha.Arg[i].Source = source;
 			pair->Alpha.Arg[i].Swizzle = swz;
 			pair->Alpha.Arg[i].Abs = inst->SrcReg[i].Abs;
-			pair->Alpha.Arg[i].Negate = (negatebase & ~pair->RGB.Arg[i].Abs) ^ inst->SrcReg[i].NegateAbs;
+			pair->Alpha.Arg[i].Negate = !!(inst->SrcReg[i].Negate & NEGATE_W);
 		}
 	}