X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fr300%2Fr300_emit.c;h=b44c7bdbb3d05abc299095d958a09f308b40d690;hb=ccc888c39ee8a7c460dca5b1b659d28dbbc4c689;hp=ac510ffc2ed5ed3f9b8bc766cc6afeb197406044;hpb=ca83d5a8db510756eb95423a52b19ff52a2d6dc1;p=mesa.git

diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index ac510ffc2ed..b44c7bdbb3d 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -22,16 +22,27 @@
 
 /* r300_emit: Functions for emitting state. */
 
+#include "util/u_math.h"
+
+#include "r300_context.h"
+#include "r300_cs.h"
 #include "r300_emit.h"
+#include "r300_fs.h"
+#include "r300_screen.h"
+#include "r300_state_derived.h"
+#include "r300_state_inlines.h"
+#include "r300_texture.h"
+#include "r300_vs.h"
 
 void r300_emit_blend_state(struct r300_context* r300,
                            struct r300_blend_state* blend)
 {
     CS_LOCALS(r300);
-    BEGIN_CS(7);
-    OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 2);
+    BEGIN_CS(8);
+    OUT_CS_REG_SEQ(R300_RB3D_CBLEND, 3);
     OUT_CS(blend->blend_control);
     OUT_CS(blend->alpha_blend_control);
+    OUT_CS(blend->color_channel_mask);
     OUT_CS_REG(R300_RB3D_ROPCNTL, blend->rop);
     OUT_CS_REG(R300_RB3D_DITHER_CTL, blend->dither);
     END_CS;
@@ -92,147 +103,387 @@ void r300_emit_dsa_state(struct r300_context* r300,
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
-    BEGIN_CS(r300screen->caps->is_r500 ? 8 : 8);
+    BEGIN_CS(r300screen->caps->is_r500 ? 10 : 8);
     OUT_CS_REG(R300_FG_ALPHA_FUNC, dsa->alpha_function);
-    /* XXX figure out the r300 counterpart for this */
-    if (r300screen->caps->is_r500) {
-        /* OUT_CS_REG(R500_FG_ALPHA_VALUE, dsa->alpha_reference); */
-    }
+
+    /* not needed since we use the 8bit alpha ref */
+    /*if (r300screen->caps->is_r500) {
+        OUT_CS_REG(R500_FG_ALPHA_VALUE, dsa->alpha_reference);
+    }*/
+
     OUT_CS_REG_SEQ(R300_ZB_CNTL, 3);
     OUT_CS(dsa->z_buffer_control);
     OUT_CS(dsa->z_stencil_control);
     OUT_CS(dsa->stencil_ref_mask);
-    OUT_CS_REG(R300_ZB_ZTOP, dsa->z_buffer_top);
+    OUT_CS_REG(R300_ZB_ZTOP, r300->ztop_state.z_buffer_top);
+
+    /* XXX it seems r3xx doesn't support STENCILREFMASK_BF */
     if (r300screen->caps->is_r500) {
-        /* OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf); */
+        OUT_CS_REG(R500_ZB_STENCILREFMASK_BF, dsa->stencil_ref_bf);
     }
     END_CS;
 }
 
-void r300_emit_fragment_shader(struct r300_context* r300,
-                               struct r3xx_fragment_shader* fs)
+static const float * get_shader_constant(
+    struct r300_context * r300,
+    struct rc_constant * constant,
+    struct r300_constant_buffer * externals)
+{
+    static const float zero[4] = { 0.0, 0.0, 0.0, 0.0 };
+    switch(constant->Type) {
+        case RC_CONSTANT_EXTERNAL:
+            return externals->constants[constant->u.External];
+
+        case RC_CONSTANT_IMMEDIATE:
+            return constant->u.Immediate;
+
+        default:
+            debug_printf("r300: Implementation error: Unhandled constant type %i\n",
+                constant->Type);
+            return zero;
+    }
+}
+
+/* Convert a normal single-precision float into the 7.16 format
+ * used by the R300 fragment shader.
+ */
+static uint32_t pack_float24(float f)
+{
+    union {
+        float fl;
+        uint32_t u;
+    } u;
+    float mantissa;
+    int exponent;
+    uint32_t float24 = 0;
+
+    if (f == 0.0)
+        return 0;
+
+    u.fl = f;
+
+    mantissa = frexpf(f, &exponent);
+
+    /* Handle -ve */
+    if (mantissa < 0) {
+        float24 |= (1 << 23);
+        mantissa = mantissa * -1.0;
+    }
+    /* Handle exponent, bias of 63 */
+    exponent += 62;
+    float24 |= (exponent << 16);
+    /* Kill 7 LSB of mantissa */
+    float24 |= (u.u & 0x7FFFFF) >> 7;
+
+    return float24;
+}
+
+void r300_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code)
 {
+    struct r300_fragment_program_code * code = &generic_code->code.r300;
     int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS(22);
+    BEGIN_CS(15 +
+             code->alu.length * 4 +
+             (code->tex.length ? (1 + code->tex.length) : 0));
+
+    OUT_CS_REG(R300_US_CONFIG, code->config);
+    OUT_CS_REG(R300_US_PIXSIZE, code->pixsize);
+    OUT_CS_REG(R300_US_CODE_OFFSET, code->code_offset);
+
+    OUT_CS_REG_SEQ(R300_US_CODE_ADDR_0, 4);
+    for(i = 0; i < 4; ++i)
+        OUT_CS(code->code_addr[i]);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_RGB_INST_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].rgb_inst);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_RGB_ADDR_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].rgb_addr);
 
-    OUT_CS_REG(R300_US_CONFIG, fs->indirections);
-    OUT_CS_REG(R300_US_PIXSIZE, fs->shader.stack_size);
-    /* XXX figure out exactly how big the sizes are on this reg */
-    OUT_CS_REG(R300_US_CODE_OFFSET, 0x40);
-    /* XXX figure these ones out a bit better kthnx */
-    OUT_CS_REG(R300_US_CODE_ADDR_0, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_1, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_2, 0x0);
-    OUT_CS_REG(R300_US_CODE_ADDR_3, 0x40 | R300_RGBA_OUT);
-
-    for (i = 0; i < fs->alu_instruction_count; i++) {
-        OUT_CS_REG(R300_US_ALU_RGB_INST_0 + (4 * i),
-            fs->instructions[i].alu_rgb_inst);
-        OUT_CS_REG(R300_US_ALU_RGB_ADDR_0 + (4 * i),
-            fs->instructions[i].alu_rgb_addr);
-        OUT_CS_REG(R300_US_ALU_ALPHA_INST_0 + (4 * i),
-            fs->instructions[i].alu_alpha_inst);
-        OUT_CS_REG(R300_US_ALU_ALPHA_ADDR_0 + (4 * i),
-            fs->instructions[i].alu_alpha_addr);
+    OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_INST_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].alpha_inst);
+
+    OUT_CS_REG_SEQ(R300_US_ALU_ALPHA_ADDR_0, code->alu.length);
+    for (i = 0; i < code->alu.length; i++)
+        OUT_CS(code->alu.inst[i].alpha_addr);
+
+    if (code->tex.length) {
+        OUT_CS_REG_SEQ(R300_US_TEX_INST_0, code->tex.length);
+        for(i = 0; i < code->tex.length; ++i)
+            OUT_CS(code->tex.inst[i]);
     }
 
     END_CS;
 }
 
-void r500_emit_fragment_shader(struct r300_context* r300,
-                               struct r5xx_fragment_shader* fs)
+void r300_emit_fs_constant_buffer(struct r300_context* r300,
+                                  struct rc_constant_list* constants)
 {
     int i;
-    struct r300_constant_buffer* constants =
-        &r300->shader_constants[PIPE_SHADER_FRAGMENT];
     CS_LOCALS(r300);
 
-    BEGIN_CS(9 + (fs->instruction_count * 6) + (constants->count ? 3 : 0) +
-            (constants->count * 4));
-    OUT_CS_REG(R500_US_CONFIG, R500_ZERO_TIMES_ANYTHING_EQUALS_ZERO);
-    OUT_CS_REG(R500_US_PIXSIZE, fs->shader.stack_size);
-    OUT_CS_REG(R500_US_CODE_ADDR, R500_US_CODE_START_ADDR(0) |
-            R500_US_CODE_END_ADDR(fs->instruction_count));
+    if (constants->Count == 0)
+        return;
+
+    BEGIN_CS(constants->Count * 4 + 1);
+    OUT_CS_REG_SEQ(R300_PFS_PARAM_0_X, constants->Count * 4);
+    for(i = 0; i < constants->Count; ++i) {
+        const float * data = get_shader_constant(r300,
+                                                 &constants->Constants[i],
+                                                 &r300->shader_constants[PIPE_SHADER_FRAGMENT]);
+        OUT_CS(pack_float24(data[0]));
+        OUT_CS(pack_float24(data[1]));
+        OUT_CS(pack_float24(data[2]));
+        OUT_CS(pack_float24(data[3]));
+    }
+    END_CS;
+}
+
+void r500_emit_fragment_program_code(struct r300_context* r300,
+                                     struct rX00_fragment_program_code* generic_code)
+{
+    struct r500_fragment_program_code * code = &generic_code->code.r500;
+    int i;
+    CS_LOCALS(r300);
+
+    BEGIN_CS(13 +
+             ((code->inst_end + 1) * 6));
+    OUT_CS_REG(R500_US_CONFIG, 0);
+    OUT_CS_REG(R500_US_PIXSIZE, code->max_temp_idx);
+    OUT_CS_REG(R500_US_CODE_RANGE,
+               R500_US_CODE_RANGE_ADDR(0) | R500_US_CODE_RANGE_SIZE(code->inst_end));
+    OUT_CS_REG(R500_US_CODE_OFFSET, 0);
+    OUT_CS_REG(R500_US_CODE_ADDR,
+               R500_US_CODE_START_ADDR(0) | R500_US_CODE_END_ADDR(code->inst_end));
 
     OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_INSTR);
-    OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, fs->instruction_count * 6);
-    for (i = 0; i < fs->instruction_count; i++) {
-        OUT_CS(fs->instructions[i].inst0);
-        OUT_CS(fs->instructions[i].inst1);
-        OUT_CS(fs->instructions[i].inst2);
-        OUT_CS(fs->instructions[i].inst3);
-        OUT_CS(fs->instructions[i].inst4);
-        OUT_CS(fs->instructions[i].inst5);
-    }
-
-    if (constants->count) {
-        OUT_CS_REG(R500_GA_US_VECTOR_INDEX,
-                R500_GA_US_VECTOR_INDEX_TYPE_CONST);
-        OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->count * 4);
-        for (i = 0; i < constants->count; i++) {
-            OUT_CS_32F(constants->constants[i][0]);
-            OUT_CS_32F(constants->constants[i][1]);
-            OUT_CS_32F(constants->constants[i][2]);
-            OUT_CS_32F(constants->constants[i][3]);
-        }
+    OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, (code->inst_end + 1) * 6);
+    for (i = 0; i <= code->inst_end; i++) {
+        OUT_CS(code->inst[i].inst0);
+        OUT_CS(code->inst[i].inst1);
+        OUT_CS(code->inst[i].inst2);
+        OUT_CS(code->inst[i].inst3);
+        OUT_CS(code->inst[i].inst4);
+        OUT_CS(code->inst[i].inst5);
     }
 
     END_CS;
 }
 
+void r500_emit_fs_constant_buffer(struct r300_context* r300,
+                                  struct rc_constant_list* constants)
+{
+    int i;
+    CS_LOCALS(r300);
+
+    if (constants->Count == 0)
+        return;
+
+    BEGIN_CS(constants->Count * 4 + 3);
+    OUT_CS_REG(R500_GA_US_VECTOR_INDEX, R500_GA_US_VECTOR_INDEX_TYPE_CONST);
+    OUT_CS_ONE_REG(R500_GA_US_VECTOR_DATA, constants->Count * 4);
+    for (i = 0; i < constants->Count; i++) {
+        const float * data = get_shader_constant(r300,
+                                                 &constants->Constants[i],
+                                                 &r300->shader_constants[PIPE_SHADER_FRAGMENT]);
+        OUT_CS_32F(data[0]);
+        OUT_CS_32F(data[1]);
+        OUT_CS_32F(data[2]);
+        OUT_CS_32F(data[3]);
+    }
+    END_CS;
+}
+
 void r300_emit_fb_state(struct r300_context* r300,
                         struct pipe_framebuffer_state* fb)
 {
     struct r300_texture* tex;
-    unsigned pixpitch;
+    struct pipe_surface* surf;
     int i;
     CS_LOCALS(r300);
 
-    BEGIN_CS((8 * fb->nr_cbufs) + (fb->zsbuf ? 8 : 0) + 4);
+    BEGIN_CS((10 * fb->nr_cbufs) + (fb->zsbuf ? 10 : 0) + 4);
+    OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
+        R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
+        R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
+    OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
+        R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
+        R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
+
     for (i = 0; i < fb->nr_cbufs; i++) {
-        tex = (struct r300_texture*)fb->cbufs[i]->texture;
+        surf = fb->cbufs[i];
+        tex = (struct r300_texture*)surf->texture;
         assert(tex && tex->buffer && "cbuf is marked, but NULL!");
-        pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
-        OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+        OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
-        OUT_CS_REG(R300_RB3D_COLORPITCH0 + (4 * i), pixpitch |
-            r300_translate_colorformat(tex->tex.format));
+        OUT_CS_REG_SEQ(R300_RB3D_COLORPITCH0 + (4 * i), 1);
+        OUT_CS_RELOC(tex->buffer, tex->pitch[surf->level] |
+                     r300_translate_colorformat(tex->tex.format), 0,
+                     RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i),
-            r300_translate_out_fmt(fb->cbufs[i]->format));
+            r300_translate_out_fmt(surf->format));
     }
 
     if (fb->zsbuf) {
-        tex = (struct r300_texture*)fb->zsbuf->texture;
+        surf = fb->zsbuf;
+        tex = (struct r300_texture*)surf->texture;
         assert(tex && tex->buffer && "zsbuf is marked, but NULL!");
-        pixpitch = tex->stride / tex->tex.block.size;
 
         OUT_CS_REG_SEQ(R300_ZB_DEPTHOFFSET, 1);
-        OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
+        OUT_CS_RELOC(tex->buffer, surf->offset, 0, RADEON_GEM_DOMAIN_VRAM, 0);
 
         OUT_CS_REG(R300_ZB_FORMAT, r300_translate_zsformat(tex->tex.format));
 
-        OUT_CS_REG(R300_ZB_DEPTHPITCH, pixpitch);
+        OUT_CS_REG_SEQ(R300_ZB_DEPTHPITCH, 1);
+        OUT_CS_RELOC(tex->buffer, tex->pitch[surf->level], 0,
+                     RADEON_GEM_DOMAIN_VRAM, 0);
     }
 
-    OUT_CS_REG(R300_RB3D_DSTCACHE_CTLSTAT,
-        R300_RB3D_DSTCACHE_CTLSTAT_DC_FREE_FREE_3D_TAGS |
-        R300_RB3D_DSTCACHE_CTLSTAT_DC_FLUSH_FLUSH_DIRTY_3D);
-    OUT_CS_REG(R300_ZB_ZCACHE_CTLSTAT,
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FLUSH_FLUSH_AND_FREE |
-        R300_ZB_ZCACHE_CTLSTAT_ZC_FREE_FREE);
     END_CS;
 }
 
+static void r300_emit_query_start(struct r300_context *r300)
+{
+    struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps;
+    struct r300_query *query = r300->query_current;
+    CS_LOCALS(r300);
+
+    if (!query)
+	return;
+
+    /* XXX This will almost certainly not return good results
+     * for overlapping queries. */
+    BEGIN_CS(4);
+    if (caps->family == CHIP_FAMILY_RV530) {
+        OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
+    } else {
+        OUT_CS_REG(R300_SU_REG_DEST, R300_RASTER_PIPE_SELECT_ALL);
+    }
+    OUT_CS_REG(R300_ZB_ZPASS_DATA, 0);
+    END_CS;
+    query->begin_emitted = TRUE;
+}
+
+
+static void r300_emit_query_finish(struct r300_context *r300,
+                                   struct r300_query *query)
+{
+    struct r300_capabilities* caps = r300_screen(r300->context.screen)->caps;
+    CS_LOCALS(r300);
+
+    assert(caps->num_frag_pipes);
+
+    BEGIN_CS(6 * caps->num_frag_pipes + 2);
+    /* I'm not so sure I like this switch, but it's hard to be elegant
+     * when there's so many special cases...
+     *
+     * So here's the basic idea. For each pipe, enable writes to it only,
+     * then put out the relocation for ZPASS_ADDR, taking into account a
+     * 4-byte offset for each pipe. RV380 and older are special; they have
+     * only two pipes, and the second pipe's enable is on bit 3, not bit 1,
+     * so there's a chipset cap for that. */
+    switch (caps->num_frag_pipes) {
+        case 4:
+            /* pipe 3 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 3);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 3),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 3:
+            /* pipe 2 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 2);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 2),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 2:
+            /* pipe 1 only */
+            /* As mentioned above, accomodate RV380 and older. */
+            OUT_CS_REG(R300_SU_REG_DEST,
+                    1 << (caps->high_second_pipe ? 3 : 1));
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 1),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+        case 1:
+            /* pipe 0 only */
+            OUT_CS_REG(R300_SU_REG_DEST, 1 << 0);
+            OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+            OUT_CS_RELOC(r300->oqbo, query->offset + (sizeof(uint32_t) * 0),
+                    0, RADEON_GEM_DOMAIN_GTT, 0);
+            break;
+        default:
+            debug_printf("r300: Implementation error: Chipset reports %d"
+                    " pixel pipes!\n", caps->num_frag_pipes);
+            assert(0);
+    }
+
+    /* And, finally, reset it to normal... */
+    OUT_CS_REG(R300_SU_REG_DEST, 0xF);
+    END_CS;
+}
+
+static void rv530_emit_query_single(struct r300_context *r300,
+                                    struct r300_query *query)
+{
+    CS_LOCALS(r300);
+
+    BEGIN_CS(8);
+    OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
+    OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+    OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
+    END_CS;
+}
+
+static void rv530_emit_query_double(struct r300_context *r300,
+                                    struct r300_query *query)
+{
+    CS_LOCALS(r300);
+
+    BEGIN_CS(14);
+    OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_0);
+    OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+    OUT_CS_RELOC(r300->oqbo, query->offset, 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_1);
+    OUT_CS_REG_SEQ(R300_ZB_ZPASS_ADDR, 1);
+    OUT_CS_RELOC(r300->oqbo, query->offset + sizeof(uint32_t), 0, RADEON_GEM_DOMAIN_GTT, 0);
+    OUT_CS_REG(RV530_FG_ZBREG_DEST, RV530_FG_ZBREG_DEST_PIPE_SELECT_ALL);
+    END_CS;
+}
+
+void r300_emit_query_end(struct r300_context* r300)
+{
+    struct r300_capabilities *caps = r300_screen(r300->context.screen)->caps;
+    struct r300_query *query = r300->query_current;
+
+    if (!query)
+	return;
+
+    if (query->begin_emitted == FALSE)
+        return;
+
+    if (caps->family == CHIP_FAMILY_RV530) {
+        if (caps->num_z_pipes == 2)
+            rv530_emit_query_double(r300, query);
+        else
+            rv530_emit_query_single(r300, query);
+    } else 
+        r300_emit_query_finish(r300, query);
+}
+
 void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
 {
     CS_LOCALS(r300);
 
-    BEGIN_CS(20);
+    BEGIN_CS(22);
     OUT_CS_REG(R300_VAP_CNTL_STATUS, rs->vap_control_status);
     OUT_CS_REG(R300_GA_POINT_SIZE, rs->point_size);
     OUT_CS_REG_SEQ(R300_GA_POINT_MINMAX, 2);
@@ -248,6 +499,7 @@ void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
     OUT_CS_REG(R300_GA_LINE_STIPPLE_CONFIG, rs->line_stipple_config);
     OUT_CS_REG(R300_GA_LINE_STIPPLE_VALUE, rs->line_stipple_value);
     OUT_CS_REG(R300_GA_COLOR_CONTROL, rs->color_control);
+    OUT_CS_REG(R300_GA_POLY_MODE, rs->polygon_mode);
     END_CS;
 }
 
@@ -258,6 +510,8 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
     CS_LOCALS(r300);
 
+    DBG(r300, DBG_DRAW, "r300: RS emit:\n");
+
     BEGIN_CS(21);
     if (r300screen->caps->is_r500) {
         OUT_CS_REG_SEQ(R500_RS_IP_0, 8);
@@ -266,7 +520,7 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->ip[i]);
-        /* debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]); */
+        DBG(r300, DBG_DRAW, "    : ip %d: 0x%08x\n", i, rs->ip[i]);
     }
 
     OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
@@ -280,11 +534,11 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     }
     for (i = 0; i < 8; i++) {
         OUT_CS(rs->inst[i]);
-        /* debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]); */
+        DBG(r300, DBG_DRAW, "    : inst %d: 0x%08x\n", i, rs->inst[i]);
     }
 
-    /* debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
-     *        rs->inst_count); */
+    DBG(r300, DBG_DRAW, "    : count: 0x%08x inst_count: 0x%08x\n",
+        rs->count, rs->inst_count);
 
     END_CS;
 }
@@ -306,10 +560,18 @@ void r300_emit_texture(struct r300_context* r300,
                        struct r300_texture* tex,
                        unsigned offset)
 {
+    uint32_t filter0 = sampler->filter0;
     CS_LOCALS(r300);
 
+    /* to emulate 1D textures through 2D ones correctly */
+    if (tex->tex.target == PIPE_TEXTURE_1D) {
+        filter0 &= ~R300_TX_WRAP_T_MASK;
+        filter0 |= R300_TX_WRAP_T(R300_TX_CLAMP_TO_EDGE);
+    }
+
     BEGIN_CS(16);
-    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
+    OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), filter0 |
+        (offset << 28));
     OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
     OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
 
@@ -322,13 +584,58 @@ void r300_emit_texture(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_vertex_buffer(struct r300_context* r300)
+/* XXX I can't read this and that's not good */
+void r300_emit_aos(struct r300_context* r300, unsigned offset)
+{
+    struct pipe_vertex_buffer *vbuf = r300->vertex_buffer;
+    struct pipe_vertex_element *velem = r300->vertex_element;
+    CS_LOCALS(r300);
+    int i;
+    unsigned aos_count = r300->vertex_element_count;
+
+    unsigned packet_size = (aos_count * 3 + 1) / 2;
+    BEGIN_CS(2 + packet_size + aos_count * 2);
+    OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, packet_size);
+    OUT_CS(aos_count);
+    for (i = 0; i < aos_count - 1; i += 2) {
+        int buf_num1 = velem[i].vertex_buffer_index;
+        int buf_num2 = velem[i+1].vertex_buffer_index;
+        assert(vbuf[buf_num1].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
+        assert(vbuf[buf_num2].stride % 4 == 0 && pf_get_size(velem[i+1].src_format) % 4 == 0);
+        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num1].stride << 6) |
+               (pf_get_size(velem[i+1].src_format) << 14) | (vbuf[buf_num2].stride << 22));
+        OUT_CS(vbuf[buf_num1].buffer_offset + velem[i].src_offset +
+               offset * vbuf[buf_num1].stride);
+        OUT_CS(vbuf[buf_num2].buffer_offset + velem[i+1].src_offset +
+               offset * vbuf[buf_num2].stride);
+    }
+    if (aos_count & 1) {
+        int buf_num = velem[i].vertex_buffer_index;
+        assert(vbuf[buf_num].stride % 4 == 0 && pf_get_size(velem[i].src_format) % 4 == 0);
+        OUT_CS((pf_get_size(velem[i].src_format) >> 2) | (vbuf[buf_num].stride << 6));
+        OUT_CS(vbuf[buf_num].buffer_offset + velem[i].src_offset +
+               offset * vbuf[buf_num].stride);
+    }
+
+    /* XXX bare CS reloc */
+    for (i = 0; i < aos_count; i++) {
+        cs_winsys->write_cs_reloc(cs_winsys,
+                                  vbuf[velem[i].vertex_buffer_index].buffer,
+                                  RADEON_GEM_DOMAIN_GTT,
+                                  0,
+                                  0);
+        cs_count -= 2;
+    }
+    END_CS;
+}
+#if 0
+void r300_emit_draw_packet(struct r300_context* r300)
 {
     CS_LOCALS(r300);
 
-    debug_printf("r300: Preparing vertex buffer %p for render, "
+    DBG(r300, DBG_DRAW, "r300: Preparing vertex buffer %p for render, "
             "vertex size %d\n", r300->vbo,
-            r300->vertex_info.vinfo.size);
+            r300->vertex_info->vinfo.size);
     /* Set the pointer to our vertex buffer. The emitted values are this:
      * PACKET3 [3D_LOAD_VBPNTR]
      * COUNT   [1]
@@ -339,54 +646,56 @@ void r300_emit_vertex_buffer(struct r300_context* r300)
     BEGIN_CS(7);
     OUT_CS_PKT3(R300_PACKET3_3D_LOAD_VBPNTR, 3);
     OUT_CS(1);
-    OUT_CS(r300->vertex_info.vinfo.size |
-            (r300->vertex_info.vinfo.size << 8));
+    OUT_CS(r300->vertex_info->vinfo.size |
+            (r300->vertex_info->vinfo.size << 8));
     OUT_CS(r300->vbo_offset);
     OUT_CS_RELOC(r300->vbo, 0, RADEON_GEM_DOMAIN_GTT, 0, 0);
     END_CS;
 }
+#endif
 
 void r300_emit_vertex_format_state(struct r300_context* r300)
 {
     int i;
     CS_LOCALS(r300);
 
+    DBG(r300, DBG_DRAW, "r300: VAP/PSC emit:\n");
+
     BEGIN_CS(26);
-    OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info.vinfo.size);
+    OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info->vinfo.size);
 
     OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2);
-    OUT_CS(r300->vertex_info.vinfo.hwfmt[0]);
-    OUT_CS(r300->vertex_info.vinfo.hwfmt[1]);
+    OUT_CS(r300->vertex_info->vinfo.hwfmt[0]);
+    OUT_CS(r300->vertex_info->vinfo.hwfmt[1]);
     OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
-    OUT_CS(r300->vertex_info.vinfo.hwfmt[2]);
-    OUT_CS(r300->vertex_info.vinfo.hwfmt[3]);
-    /* for (i = 0; i < 4; i++) {
-     *    debug_printf("hwfmt%d: 0x%08x\n", i,
-     *            r300->vertex_info.vinfo.hwfmt[i]);
-     * } */
+    OUT_CS(r300->vertex_info->vinfo.hwfmt[2]);
+    OUT_CS(r300->vertex_info->vinfo.hwfmt[3]);
+    for (i = 0; i < 4; i++) {
+       DBG(r300, DBG_DRAW, "    : hwfmt%d: 0x%08x\n", i,
+               r300->vertex_info->vinfo.hwfmt[i]);
+    }
 
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_0, 8);
     for (i = 0; i < 8; i++) {
-        OUT_CS(r300->vertex_info.vap_prog_stream_cntl[i]);
-        /* debug_printf("prog_stream_cntl%d: 0x%08x\n", i,
-         *        r300->vertex_info.vap_prog_stream_cntl[i]); */
+        OUT_CS(r300->vertex_info->vap_prog_stream_cntl[i]);
+        DBG(r300, DBG_DRAW, "    : prog_stream_cntl%d: 0x%08x\n", i,
+               r300->vertex_info->vap_prog_stream_cntl[i]);
     }
     OUT_CS_REG_SEQ(R300_VAP_PROG_STREAM_CNTL_EXT_0, 8);
     for (i = 0; i < 8; i++) {
-        OUT_CS(r300->vertex_info.vap_prog_stream_cntl_ext[i]);
-        /* debug_printf("prog_stream_cntl_ext%d: 0x%08x\n", i,
-         *        r300->vertex_info.vap_prog_stream_cntl_ext[i]); */
+        OUT_CS(r300->vertex_info->vap_prog_stream_cntl_ext[i]);
+        DBG(r300, DBG_DRAW, "    : prog_stream_cntl_ext%d: 0x%08x\n", i,
+               r300->vertex_info->vap_prog_stream_cntl_ext[i]);
     }
     END_CS;
 }
 
-void r300_emit_vertex_shader(struct r300_context* r300,
-                             struct r300_vertex_shader* vs)
+void r300_emit_vertex_program_code(struct r300_context* r300,
+                                   struct r300_vertex_program_code* code)
 {
     int i;
     struct r300_screen* r300screen = r300_screen(r300->context.screen);
-    struct r300_constant_buffer* constants =
-        &r300->shader_constants[PIPE_SHADER_VERTEX];
+    unsigned instruction_count = code->length / 4;
     CS_LOCALS(r300);
 
     if (!r300screen->caps->has_tcl) {
@@ -395,12 +704,7 @@ void r300_emit_vertex_shader(struct r300_context* r300,
         return;
     }
 
-    if (constants->count) {
-        BEGIN_CS(14 + (vs->instruction_count * 4) + (constants->count * 4));
-    } else {
-        BEGIN_CS(11 + (vs->instruction_count * 4));
-    }
-
+    BEGIN_CS(9 + code->length);
     /* R300_VAP_PVS_CODE_CNTL_0
      * R300_VAP_PVS_CONST_CNTL
      * R300_VAP_PVS_CODE_CNTL_1
@@ -408,38 +712,59 @@ void r300_emit_vertex_shader(struct r300_context* r300,
      * XXX these could be optimized to select better values... */
     OUT_CS_REG_SEQ(R300_VAP_PVS_CODE_CNTL_0, 3);
     OUT_CS(R300_PVS_FIRST_INST(0) |
-            R300_PVS_XYZW_VALID_INST(vs->instruction_count - 1) |
-            R300_PVS_LAST_INST(vs->instruction_count - 1));
-    OUT_CS(R300_PVS_MAX_CONST_ADDR(constants->count - 1));
-    OUT_CS(vs->instruction_count - 1);
+            R300_PVS_XYZW_VALID_INST(instruction_count - 1) |
+            R300_PVS_LAST_INST(instruction_count - 1));
+    OUT_CS(R300_PVS_MAX_CONST_ADDR(code->constants.Count - 1));
+    OUT_CS(instruction_count - 1);
 
     OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG, 0);
-    OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, vs->instruction_count * 4);
-    for (i = 0; i < vs->instruction_count; i++) {
-        OUT_CS(vs->instructions[i].inst0);
-        OUT_CS(vs->instructions[i].inst1);
-        OUT_CS(vs->instructions[i].inst2);
-        OUT_CS(vs->instructions[i].inst3);
-    }
-
-    if (constants->count) {
-        OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
-                (r300screen->caps->is_r500 ?
-                 R500_PVS_CONST_START : R300_PVS_CONST_START));
-        OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->count * 4);
-        for (i = 0; i < constants->count; i++) {
-            OUT_CS_32F(constants->constants[i][0]);
-            OUT_CS_32F(constants->constants[i][1]);
-            OUT_CS_32F(constants->constants[i][2]);
-            OUT_CS_32F(constants->constants[i][3]);
-        }
-    }
+    OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, code->length);
+    for (i = 0; i < code->length; i++)
+        OUT_CS(code->body.d[i]);
 
     OUT_CS_REG(R300_VAP_CNTL, R300_PVS_NUM_SLOTS(10) |
             R300_PVS_NUM_CNTLRS(5) |
             R300_PVS_NUM_FPUS(r300screen->caps->num_vert_fpus) |
             R300_PVS_VF_MAX_VTX_NUM(12));
-    OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
+    END_CS;
+}
+
+void r300_emit_vertex_shader(struct r300_context* r300,
+                             struct r300_vertex_shader* vs)
+{
+    r300_emit_vertex_program_code(r300, &vs->code);
+}
+
+void r300_emit_vs_constant_buffer(struct r300_context* r300,
+                                  struct rc_constant_list* constants)
+{
+    int i;
+    struct r300_screen* r300screen = r300_screen(r300->context.screen);
+    CS_LOCALS(r300);
+
+    if (!r300screen->caps->has_tcl) {
+        debug_printf("r300: Implementation error: emit_vertex_shader called,"
+        " but has_tcl is FALSE!\n");
+        return;
+    }
+
+    if (constants->Count == 0)
+        return;
+
+    BEGIN_CS(constants->Count * 4 + 3);
+    OUT_CS_REG(R300_VAP_PVS_VECTOR_INDX_REG,
+               (r300screen->caps->is_r500 ?
+               R500_PVS_CONST_START : R300_PVS_CONST_START));
+    OUT_CS_ONE_REG(R300_VAP_PVS_UPLOAD_DATA, constants->Count * 4);
+    for (i = 0; i < constants->Count; i++) {
+        const float * data = get_shader_constant(r300,
+                                                 &constants->Constants[i],
+                                                 &r300->shader_constants[PIPE_SHADER_VERTEX]);
+        OUT_CS_32F(data[0]);
+        OUT_CS_32F(data[1]);
+        OUT_CS_32F(data[2]);
+        OUT_CS_32F(data[3]);
+    }
     END_CS;
 }
 
@@ -475,6 +800,15 @@ void r300_flush_textures(struct r300_context* r300)
     END_CS;
 }
 
+static void r300_flush_pvs(struct r300_context* r300)
+{
+    CS_LOCALS(r300);
+
+    BEGIN_CS(2);
+    OUT_CS_REG(R300_VAP_PVS_STATE_FLUSH_REG, 0x0);
+    END_CS;
+}
+
 /* Emit all dirty state. */
 void r300_emit_dirty_state(struct r300_context* r300)
 {
@@ -487,7 +821,8 @@ void r300_emit_dirty_state(struct r300_context* r300)
         return;
     }
 
-    r300_update_derived_state(r300);
+    /* Clean out BOs. */
+    r300->winsys->reset_bos(r300->winsys);
 
     /* XXX check size */
 validate:
@@ -514,13 +849,20 @@ validate:
     /* ...textures... */
     for (i = 0; i < r300->texture_count; i++) {
         tex = r300->textures[i];
-        assert(tex && tex->buffer && "texture is marked, but NULL!");
+        if (!tex)
+            continue;
         if (!r300->winsys->add_buffer(r300->winsys, tex->buffer,
                     RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0)) {
             r300->context.flush(&r300->context, 0, NULL);
             goto validate;
         }
     }
+    /* ...occlusion query buffer... */
+    if (!r300->winsys->add_buffer(r300->winsys, r300->oqbo,
+                0, RADEON_GEM_DOMAIN_GTT)) {
+        r300->context.flush(&r300->context, 0, NULL);
+        goto validate;
+    }
     /* ...and vertex buffer. */
     if (r300->vbo) {
         if (!r300->winsys->add_buffer(r300->winsys, r300->vbo,
@@ -529,7 +871,7 @@ validate:
             goto validate;
         }
     } else {
-        debug_printf("No VBO while emitting dirty state!\n");
+        /* debug_printf("No VBO while emitting dirty state!\n"); */
     }
     if (!r300->winsys->validate(r300->winsys)) {
         r300->context.flush(&r300->context, 0, NULL);
@@ -542,6 +884,11 @@ validate:
         goto validate;
     }
 
+    if (r300->dirty_state & R300_NEW_QUERY) {
+        r300_emit_query_start(r300);
+        r300->dirty_state &= ~R300_NEW_QUERY;
+    }
+
     if (r300->dirty_state & R300_NEW_BLEND) {
         r300_emit_blend_state(r300, r300->blend_state);
         r300->dirty_state &= ~R300_NEW_BLEND;
@@ -564,15 +911,22 @@ validate:
 
     if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER) {
         if (r300screen->caps->is_r500) {
-            r500_emit_fragment_shader(r300,
-                (struct r5xx_fragment_shader*)r300->fs);
+            r500_emit_fragment_program_code(r300, &r300->fs->code);
         } else {
-            r300_emit_fragment_shader(r300,
-                (struct r3xx_fragment_shader*)r300->fs);
+            r300_emit_fragment_program_code(r300, &r300->fs->code);
         }
         r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER;
     }
 
+    if (r300->dirty_state & R300_NEW_FRAGMENT_SHADER_CONSTANTS) {
+        if (r300screen->caps->is_r500) {
+            r500_emit_fs_constant_buffer(r300, &r300->fs->code.constants);
+        } else {
+            r300_emit_fs_constant_buffer(r300, &r300->fs->code.constants);
+        }
+        r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER_CONSTANTS;
+    }
+
     if (r300->dirty_state & R300_NEW_FRAMEBUFFERS) {
         r300_emit_fb_state(r300, &r300->framebuffer_state);
         r300->dirty_state &= ~R300_NEW_FRAMEBUFFERS;
@@ -597,12 +951,13 @@ validate:
     if (r300->dirty_state &
             (R300_ANY_NEW_SAMPLERS | R300_ANY_NEW_TEXTURES)) {
         for (i = 0; i < MIN2(r300->sampler_count, r300->texture_count); i++) {
-            if (r300->dirty_state &
-                    ((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) {
-                r300_emit_texture(r300,
-                        r300->sampler_states[i],
-                        r300->textures[i],
-                        i);
+  	    if (r300->dirty_state &
+		((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i))) {
+		if (r300->textures[i]) 
+		    r300_emit_texture(r300,
+				      r300->sampler_states[i],
+				      r300->textures[i],
+				      i);
                 r300->dirty_state &=
                     ~((R300_NEW_SAMPLER << i) | (R300_NEW_TEXTURE << i));
                 dirty_tex++;
@@ -625,17 +980,26 @@ validate:
         r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT;
     }
 
+    if (r300->dirty_state & (R300_NEW_VERTEX_SHADER | R300_NEW_VERTEX_SHADER_CONSTANTS)) {
+        r300_flush_pvs(r300);
+    }
+
     if (r300->dirty_state & R300_NEW_VERTEX_SHADER) {
         r300_emit_vertex_shader(r300, r300->vs);
         r300->dirty_state &= ~R300_NEW_VERTEX_SHADER;
     }
 
+    if (r300->dirty_state & R300_NEW_VERTEX_SHADER_CONSTANTS) {
+        r300_emit_vs_constant_buffer(r300, &r300->vs->code.constants);
+        r300->dirty_state &= ~R300_NEW_VERTEX_SHADER_CONSTANTS;
+    }
+
     /* XXX
     assert(r300->dirty_state == 0);
     */
 
     /* Finally, emit the VBO. */
-    r300_emit_vertex_buffer(r300);
+    /* r300_emit_vertex_buffer(r300); */
 
     r300->dirty_hw++;
 }