From 2240c0d33365189f975b84b06792e2a5ecb8b13a Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Thu, 14 May 2009 02:07:49 +0200
Subject: r300: software fallbacking handling rewrite

Until now falling back to software rasterizer worked only for TCL enabled cards.
For non TCL cards we used to plug our rendering functions in r300InitSwtcl, and we had never restored original functions for software rasterizer.
---
 src/mesa/drivers/dri/r300/r300_render.h | 49 +++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)
 create mode 100644 src/mesa/drivers/dri/r300/r300_render.h

(limited to 'src/mesa/drivers/dri/r300/r300_render.h')

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..fbc9581e06
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -0,0 +1,49 @@
+/*
+ * 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_INVALID_BUFFERS   (1 << 31)
+#define R300_RASTER_FALLBACK_MASK        0xffff0000
+
+extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
+
+#endif
-- 
cgit v1.2.3


From 6f1a86ca1eb4a44a738d0ad99861d948c9749de6 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Wed, 13 May 2009 23:09:39 +0200
Subject: r300: move forward declarations to where they belong

---
 src/mesa/drivers/dri/r300/r300_context.c | 5 +----
 src/mesa/drivers/dri/r300/r300_render.h  | 4 ++++
 2 files changed, 5 insertions(+), 4 deletions(-)

(limited to 'src/mesa/drivers/dri/r300/r300_render.h')

diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 5b22a11bca..be8d480c1c 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"
 
@@ -150,10 +151,6 @@ 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
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
index fbc9581e06..940d2566e2 100644
--- a/src/mesa/drivers/dri/r300/r300_render.h
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -44,6 +44,10 @@
 #define R300_FALLBACK_INVALID_BUFFERS   (1 << 31)
 #define R300_RASTER_FALLBACK_MASK        0xffff0000
 
+extern const struct tnl_pipeline_stage _r300_render_stage;
+
+extern const struct tnl_pipeline_stage _r300_tcl_stage;
+
 extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
 
 #endif
-- 
cgit v1.2.3


From e98082997c8cfe3bf9c1c1cdc40c23ee897d1007 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 30 May 2009 13:28:47 +0200
Subject: r300: prepare for different vertex data type support

---
 src/mesa/drivers/dri/r300/r300_context.h |  32 +++++---
 src/mesa/drivers/dri/r300/r300_render.h  |   6 ++
 src/mesa/drivers/dri/r300/r300_state.c   |  58 ++++++++++++++
 src/mesa/drivers/dri/r300/r300_state.h   |   1 +
 src/mesa/drivers/dri/r300/r300_swtcl.c   | 133 +++++++------------------------
 src/mesa/drivers/dri/r300/r300_swtcl.h   |   6 --
 6 files changed, 116 insertions(+), 120 deletions(-)

(limited to 'src/mesa/drivers/dri/r300/r300_render.h')

diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 2ea064ed45..7694fe4862 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -584,16 +584,6 @@ struct r300_swtcl_info {
     */
    GLuint specoffset;
 
-   struct vertex_attribute{
-       GLuint attr;
-       GLubyte format;
-       GLubyte dst_loc;
-       GLuint swizzle;
-       GLubyte write_mask;
-   } vert_attrs[VERT_ATTRIB_MAX];
-
-   GLubyte vertex_attr_count;
-
    int sw_tcl_inputs[VERT_ATTRIB_MAX];
 };
 
@@ -605,6 +595,27 @@ struct r300_vtable {
 	void (* SetupPixelShader)(GLcontext *ctx);
 };
 
+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;
+};
 
 /**
  * \brief R300 context structure.
@@ -632,6 +643,7 @@ struct r300_context {
 	} options;
 	
 	struct r300_swtcl_info swtcl;
+	struct r300_vertex_buffer vbuf;
 	GLboolean vap_flush_needed;
 
 	uint32_t fallback;
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
index 940d2566e2..002cd613dc 100644
--- a/src/mesa/drivers/dri/r300/r300_render.h
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -44,6 +44,12 @@
 #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
+
 extern const struct tnl_pipeline_stage _r300_render_stage;
 
 extern const struct tnl_pipeline_stage _r300_tcl_stage;
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index bbe993429b..a04326fca9 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -2216,6 +2216,64 @@ static void r500SetupPixelShader(GLcontext *ctx)
 	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);
+	rmesa->hw.vof.cmd[R300_VOF_CNTL_1] = r300VAPOutputCntl1(ctx, OutputsWritten);
+}
+
 void r300UpdateShaderStates(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
diff --git a/src/mesa/drivers/dri/r300/r300_state.h b/src/mesa/drivers/dri/r300/r300_state.h
index cac639d7c6..2328289420 100644
--- a/src/mesa/drivers/dri/r300/r300_state.h
+++ b/src/mesa/drivers/dri/r300/r300_state.h
@@ -58,5 +58,6 @@ void r300UpdateShaderStates (r300ContextPtr rmesa);
 void r300InitState (r300ContextPtr r300);
 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 68e24dec8f..fd6312ae26 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.c
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.c
@@ -38,6 +38,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_swtcl.h"
 #include "r300_emit.h"
 #include "r300_tex.h"
+#include "r300_render.h"
 
 #define EMIT_ATTR( ATTR, STYLE )					\
 do {									\
@@ -54,107 +55,28 @@ 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)
-{
-	r300ContextPtr rmesa = R300_CONTEXT( ctx );
-	struct vertex_attribute *attrs = rmesa->swtcl.vert_attrs;
-	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, 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;
-}
-
-
 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;
+	struct vertex_attribute *attrs = rmesa->vbuf.attribs;
 
 	rmesa->swtcl.coloroffset = rmesa->swtcl.specoffset = 0;
 	rmesa->radeon.swtcl.vertex_attr_count = 0;
@@ -166,7 +88,7 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		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);
+		ADD_ATTR(VERT_ATTRIB_POS, R300_DATA_TYPE_FLOAT_4, SWTCL_OVM_POS, SWIZZLE_XYZW, MASK_XYZW, 0);
 		rmesa->swtcl.coloroffset = 4;
 	}
 
@@ -175,10 +97,10 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		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
 	}
 
@@ -188,10 +110,10 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		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;
 	}
@@ -201,20 +123,21 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		OutputsWritten |= 1 << VERT_RESULT_BFC0;
 #if MESA_LITTLE_ENDIAN
 		EMIT_ATTR( _TNL_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA );
-		ADD_ATTR(VERT_ATTRIB_GENERIC0, EMIT_4UB_4F_RGBA, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW);
+		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, EMIT_4UB_4F_ABGR, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW);
+		ADD_ATTR(VERT_ATTRIB_GENERIC0, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR2, SWIZZLE_XYZW, MASK_XYZW, 1);
 #endif
 		if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 )) {
+			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, EMIT_4UB_4F_RGBA, SWTCL_OVM_COLOR3, swiz, MASK_XYZW);
+			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, EMIT_4UB_4F_ABGR, SWTCL_OVM_COLOR3, swiz, MASK_XYZW);
+			ADD_ATTR(VERT_ATTRIB_GENERIC1, R300_DATA_TYPE_BYTE, SWTCL_OVM_COLOR3, swiz, MASK_XYZW, 1);
 #endif
 		}
 	}
@@ -224,7 +147,7 @@ static void r300SetVertexFormat( GLcontext *ctx )
 		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);
 	}
 
 	/**
@@ -233,24 +156,28 @@ static void r300SetVertexFormat( GLcontext *ctx )
 	 */
 	if (RENDERINPUTS_TEST_RANGE(tnl->render_inputs_bitset, _TNL_FIRST_TEX, _TNL_LAST_TEX )) {
 		int i;
-		GLuint swiz, format;
+		GLuint swiz, format, hw_format;
 		for (i = 0; i < ctx->Const.MaxTextureUnits; i++) {
 			if (RENDERINPUTS_TEST(tnl->render_inputs_bitset, _TNL_ATTRIB_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;
+						hw_format = R300_DATA_TYPE_FLOAT_2;
 						swiz = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ZERO, SWIZZLE_ONE);
 						break;
 					case 3:
 						format = EMIT_3F;
+						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;
 						break;
 					default:
@@ -259,8 +186,7 @@ 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_XYZW);
-				vap_out_fmt_1 |= 4 << (i * 3);
+				ADD_ATTR(VERT_ATTRIB_TEX0 + i, hw_format, SWTCL_OVM_TEX(i), swiz, MASK_XYZW, 0);
 				++first_free_tex;
 			}
 		}
@@ -276,8 +202,7 @@ 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;
 	}
 
@@ -291,12 +216,12 @@ 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;
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
 	rmesa->radeon.swtcl.vertex_size =
 		_tnl_install_attrs( ctx,
diff --git a/src/mesa/drivers/dri/r300/r300_swtcl.h b/src/mesa/drivers/dri/r300/r300_swtcl.h
index 14826f0817..cebc895c47 100644
--- a/src/mesa/drivers/dri/r300/r300_swtcl.h
+++ b/src/mesa/drivers/dri/r300/r300_swtcl.h
@@ -39,12 +39,6 @@ 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
  */
-- 
cgit v1.2.3


From fd80128e15c6393fa85088577ab2ca320c6cb8ae Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 30 May 2009 20:38:29 +0200
Subject: r300: add hw accelerated support for different vertex data formats

---
 src/mesa/drivers/dri/r300/Makefile       |   1 +
 src/mesa/drivers/dri/r300/r300_context.c |   9 +-
 src/mesa/drivers/dri/r300/r300_context.h |  10 +
 src/mesa/drivers/dri/r300/r300_draw.c    | 435 +++++++++++++++++++++++++++++++
 src/mesa/drivers/dri/r300/r300_emit.c    |   5 +
 src/mesa/drivers/dri/r300/r300_render.c  |  77 ++----
 src/mesa/drivers/dri/r300/r300_render.h  |   5 +-
 7 files changed, 485 insertions(+), 57 deletions(-)
 create mode 100644 src/mesa/drivers/dri/r300/r300_draw.c

(limited to 'src/mesa/drivers/dri/r300/r300_render.h')

diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 62715e3b50..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 \
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index dbd5ce589e..b7911f23cc 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -153,10 +153,6 @@ const struct dri_extension gl_20_extension[] = {
 
 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,
@@ -436,8 +432,11 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	_tnl_allow_pixel_fog(ctx, GL_FALSE);
 	_tnl_allow_vertex_fog(ctx, GL_TRUE);
 
-	if (!r300->options.hw_tcl_enabled)
+	if (r300->options.hw_tcl_enabled) {
+		r300InitDraw(ctx);
+	} else {
 		r300InitSwtcl(ctx);
+	}
 
 	radeon_fbo_init(&r300->radeon);
 	radeonInitSpanFuncs( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index 7694fe4862..11cfb55a9b 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -617,6 +617,13 @@ struct r300_vertex_buffer {
 	GLubyte num_attribs;
 };
 
+struct r300_index_buffer {
+	GLvoid *ptr;
+	GLboolean is_32bit;
+	GLboolean free_needed;
+	GLuint count;
+};
+
 /**
  * \brief R300 context structure.
  */
@@ -644,6 +651,7 @@ struct r300_context {
 	
 	struct r300_swtcl_info swtcl;
 	struct r300_vertex_buffer vbuf;
+	struct r300_index_buffer ind_buf;
 	GLboolean vap_flush_needed;
 
 	uint32_t fallback;
@@ -666,6 +674,8 @@ extern int r300VertexProgUpdateParams(GLcontext * ctx,
 
 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..ba74878721
--- /dev/null
+++ b/src/mesa/drivers/dri/r300/r300_draw.c
@@ -0,0 +1,435 @@
+/**************************************************************************
+ *
+ * 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 TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#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(GLuint) * mesa_ind_buf->count);
+		int i;
+
+		for (i = 0; i < mesa_ind_buf->count; ++i) {
+			out[i] = (GLuint) in[i];
+		}
+
+		ind_buf->ptr = out;
+		ind_buf->free_needed = GL_TRUE;
+		ind_buf->is_32bit = GL_TRUE;
+	} else if (mesa_ind_buf->type == GL_UNSIGNED_SHORT) {
+		ind_buf->ptr = src_ptr;
+		ind_buf->free_needed = GL_FALSE;
+		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 += input->StrideB;			\
+		}						\
+	} 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 += input->StrideB;			\
+		}						\
+	}						\
+} 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;
+
+	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;
+
+	if (input->Type == GL_DOUBLE || ((getTypeSize(input->Type) * input->Size) % 4 > 0)) {
+		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", input->StrideB, 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;
+		}
+
+		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 = input->StrideB;
+		r300_attr.dwords = getTypeSize(type) * input->Size / 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;
+			if (input->Size == 2)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+			else if (input->Size == 4)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+			else
+				assert(0);
+			break;
+		case GL_BYTE:
+			assert(input->Size == 4);
+			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;
+			if (input->Size == 2)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_2;
+			else if (input->Size == 4)
+				r300_attr.data_type = R300_DATA_TYPE_SHORT_4;
+			else
+				assert(0);
+			break;
+		case GL_UNSIGNED_BYTE:
+			assert(input->Size == 4);
+			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 1e79a76b47..d6c29ea388 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.c
+++ b/src/mesa/drivers/dri/r300/r300_emit.c
@@ -337,6 +337,11 @@ void r300EmitArrays(GLcontext * ctx)
 	    r300VAPOutputCntl1(ctx, OutputsWritten);
 
 	rmesa->radeon.tcl.aos_count = nr;
+
+	/* Fill index buffer info */
+	rmesa->ind_buf.ptr = vb->Elts;
+	rmesa->ind_buf.is_32bit = GL_TRUE;
+	rmesa->ind_buf.free_needed = GL_FALSE;
 }
 
 void r300EmitCacheFlush(r300ContextPtr rmesa)
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 92310a0264..4982bc63b9 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -172,16 +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;
+	GLbyte el_size;
+
+	el_size = rmesa->ind_buf.is_32bit ? 4 : 2;
 
 	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, n_elts * el_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, n_elts * el_size);
 	radeon_bo_unmap(rmesa->radeon.tcl.elt_dma_bo);
 }
 
@@ -191,12 +194,20 @@ static void r300FireEB(r300ContextPtr rmesa, int vertex_count, int type)
 
     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);
@@ -206,13 +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);
@@ -337,12 +348,10 @@ static void r300FireAOS(r300ContextPtr rmesa, int vertex_count, int 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);
 	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);
@@ -356,7 +365,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	 */
 	rcommonEnsureCmdBufSpace(&rmesa->radeon, 64, __FUNCTION__);
 
-	if (vb->Elts) {
+	if (rmesa->ind_buf.ptr) {
 		if (num_verts > 65535) {
 			/* not implemented yet */
 			WARN_ONCE("Too many elts\n");
@@ -373,7 +382,7 @@ 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);
 		r300FireEB(rmesa, num_verts, type);
 	} else {
@@ -383,8 +392,7 @@ static void r300RunRenderPrimitive(r300ContextPtr rmesa, GLcontext * ctx,
 	COMMIT_BATCH();
 }
 
-static void 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;
@@ -406,7 +414,7 @@ static void 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);
@@ -436,6 +444,8 @@ static const char *getFallbackString(uint32_t bit)
 			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:
@@ -504,30 +514,6 @@ static GLboolean r300RunNonTCLRender(GLcontext * ctx,
 	return GL_FALSE;
 }
 
-static GLboolean r300RunTCLRender(GLcontext * ctx,
-				  struct tnl_pipeline_stage *stage)
-{
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-
-	if (RADEON_DEBUG & DEBUG_PRIMS)
-		fprintf(stderr, "%s\n", __FUNCTION__);
-
-	if (rmesa->options.hw_tcl_enabled == GL_FALSE)
-		return GL_TRUE;
-
-	/* Call it here so we can fallback early */
-	r300UpdateShaders(rmesa);
-
-	r300SwitchFallback(ctx, R300_FALLBACK_INVALID_BUFFERS, !r300ValidateBuffers(ctx));
-
-	if (rmesa->fallback)
-		return GL_TRUE;
-
-	r300RunRender(ctx, stage);
-
-	return GL_FALSE;
-}
-
 const struct tnl_pipeline_stage _r300_render_stage = {
 	"r300 Hardware Rasterization",
 	NULL,
@@ -536,12 +522,3 @@ 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
index 002cd613dc..192d738f5f 100644
--- a/src/mesa/drivers/dri/r300/r300_render.h
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -41,6 +41,7 @@
 #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
 
@@ -52,8 +53,8 @@
 
 extern const struct tnl_pipeline_stage _r300_render_stage;
 
-extern const struct tnl_pipeline_stage _r300_tcl_stage;
-
 extern void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode);
 
+extern void r300RunRenderPrimitive(GLcontext * ctx, int start, int end, int prim);
+
 #endif
-- 
cgit v1.2.3


From 72c78b99926aab6af38248b5aa6b0fb1f9d42005 Mon Sep 17 00:00:00 2001
From: Maciej Cencora <m.cencora@gmail.com>
Date: Sat, 30 May 2009 13:50:50 +0200
Subject: r300: rewrite vertex setup for software T&L path using functions from
 software TCL path

---
 src/mesa/drivers/dri/r300/r300_emit.c   | 230 ++++----------------------------
 src/mesa/drivers/dri/r300/r300_emit.h   |   5 +-
 src/mesa/drivers/dri/r300/r300_render.c |  19 +++
 src/mesa/drivers/dri/r300/r300_render.h |   9 ++
 4 files changed, 54 insertions(+), 209 deletions(-)

(limited to 'src/mesa/drivers/dri/r300/r300_render.h')

diff --git a/src/mesa/drivers/dri/r300/r300_emit.c b/src/mesa/drivers/dri/r300/r300_emit.c
index d6c29ea388..cf31596d9b 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)
 {
@@ -194,154 +129,39 @@ GLuint r300VAPOutputCntl1(GLcontext * ctx, GLuint OutputsWritten)
 	return ret;
 }
 
-/* Emit vertex data to GART memory
- * Route inputs to the vertex processor
- */
-void 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 = rmesa->selected_vp;
-
-	if (rmesa->options.hw_tcl_enabled) {
-		inputs = prog->inputs;
-		InputsRead = prog->key.InputsRead;
-		OutputsWritten = prog->key.OutputsWritten;
-	} else {
-		inputs = rmesa->swtcl.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;
-		}
-
-		if (RENDERINPUTS_TEST(render_inputs_bitset, _TNL_ATTRIB_COLOR1)) {
-			InputsRead |= 1 << VERT_ATTRIB_COLOR1;
-			OutputsWritten |= 1 << VERT_RESULT_COL1;
-		}
-
-		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);
-			}
-		}
-
-		for (i = 0, nr = 0; i < VERT_ATTRIB_MAX; i++) {
-			if (InputsRead & (1 << i)) {
-				inputs[i] = nr++;
-			} else {
-				inputs[i] = -1;
-			}
-		}
+	r300ChooseSwtclVertexFormat(ctx, &InputsRead, &OutputsWritten);
 
-		/* 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->render_inputs_bitset, render_inputs_bitset);
-	}
+	r300SwitchFallback(ctx, R300_FALLBACK_AOS_LIMIT, vbuf->num_attribs > R300_MAX_AOS_ARRAYS);
+	if (r300->fallback & R300_RASTER_FALLBACK_MASK)
+		return GL_FALSE;
 
-	assert(InputsRead);
-	assert(OutputsWritten);
+	{
+		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)) {
-			tab[nr++] = i;
+		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);
 		}
-	}
-
-	assert(nr <= R300_MAX_AOS_ARRAYS);
 
-	for (i = 0; i < nr; i++) {
-		int ci;
+		r300->radeon.tcl.aos_count = vbuf->num_attribs;
 
-		swizzle[i][0] = SWIZZLE_ZERO;
-		swizzle[i][1] = SWIZZLE_ZERO;
-		swizzle[i][2] = SWIZZLE_ZERO;
-		swizzle[i][3] = SWIZZLE_ONE;
-
-		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);
+		/* 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;
 	}
 
-	/* 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;
+	r300SetupVAP(ctx, InputsRead, OutputsWritten);
 
-	/* Fill index buffer info */
-	rmesa->ind_buf.ptr = vb->Elts;
-	rmesa->ind_buf.is_32bit = GL_TRUE;
-	rmesa->ind_buf.free_needed = GL_FALSE;
+	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 e6485e9bd7..d88914ce95 100644
--- a/src/mesa/drivers/dri/r300/r300_emit.h
+++ b/src/mesa/drivers/dri/r300/r300_emit.h
@@ -216,16 +216,13 @@ void static INLINE cp_wait(radeonContextPtr radeon, unsigned char flags)
 	}
 }
 
-extern void 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);
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 4982bc63b9..adda92467a 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -466,6 +466,16 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 			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)) {
@@ -476,6 +486,15 @@ void r300SwitchFallback(GLcontext *ctx, uint32_t bit, GLboolean mode)
 		}
 	} else {
 		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;
+			}
+		}
+
 		/* update only if we have disabled all raster fallbacks */
 		if ((old_fallback & R300_RASTER_FALLBACK_MASK) == bit) {
 			_swrast_flush( ctx );
diff --git a/src/mesa/drivers/dri/r300/r300_render.h b/src/mesa/drivers/dri/r300/r300_render.h
index 192d738f5f..ec785474a6 100644
--- a/src/mesa/drivers/dri/r300/r300_render.h
+++ b/src/mesa/drivers/dri/r300/r300_render.h
@@ -51,6 +51,15 @@
 #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);
-- 
cgit v1.2.3