void r300_emit_fragment_shader(struct r300_context* r300,
struct r300_fragment_shader* fs)
{
- CS_LOCALS(r300);
int i;
+ CS_LOCALS(r300);
BEGIN_CS(22);
- OUT_CS_REG(R300_US_CONFIG, MAX2(fs->indirections - 1, 0));
+ 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, 0x0);
+ 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, R300_RGBA_OUT);
+ 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),
void r500_emit_fragment_shader(struct r300_context* r300,
struct r500_fragment_shader* fs)
{
- CS_LOCALS(r300);
int i;
+ struct r300_constant_buffer* constants =
+ &r300->shader_constants[PIPE_SHADER_FRAGMENT];
+ CS_LOCALS(r300);
- BEGIN_CS(9 + (fs->instruction_count * 6));
+ 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));
+ R500_US_CODE_END_ADDR(fs->instruction_count));
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);
+ 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].inst4);
OUT_CS(fs->instructions[i].inst5);
}
- END_CS;
-}
-/* Translate pipe_format into US_OUT_FMT. Note that formats are stored from
- * C3 to C0. */
-uint32_t translate_out_fmt(enum pipe_format format)
-{
- switch (format) {
- case PIPE_FORMAT_A8R8G8B8_UNORM:
- return R300_US_OUT_FMT_C4_8 |
- R300_C0_SEL_B | R300_C1_SEL_G |
- R300_C2_SEL_R | R300_C3_SEL_A;
- default:
- return R300_US_OUT_FMT_UNUSED;
- }
- return 0;
+ 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]);
+ }
+ }
+
+ END_CS;
}
-/* XXX add pitch, stride */
+/* XXX add pitch, stride, clean up */
void r300_emit_fb_state(struct r300_context* r300,
struct pipe_framebuffer_state* fb)
{
- CS_LOCALS(r300);
- struct r300_texture* tex;
int i;
+ struct r300_texture* tex;
+ CS_LOCALS(r300);
- BEGIN_CS((5 * fb->nr_cbufs) + (fb->zsbuf ? 5 : 0) + 4);
+ BEGIN_CS((6 * fb->nr_cbufs) + (fb->zsbuf ? 6 : 0) + 4);
for (i = 0; i < fb->nr_cbufs; i++) {
tex = (struct r300_texture*)fb->cbufs[i]->texture;
OUT_CS_REG_SEQ(R300_RB3D_COLOROFFSET0 + (4 * i), 1);
OUT_CS_RELOC(tex->buffer, 0, 0, RADEON_GEM_DOMAIN_VRAM, 0);
OUT_CS_REG(R300_US_OUT_FMT_0 + (4 * i),
- translate_out_fmt(fb->cbufs[i]->format));
+ r300_translate_out_fmt(fb->cbufs[i]->format));
}
if (fb->zsbuf) {
void r300_emit_rs_state(struct r300_context* r300, struct r300_rs_state* rs)
{
- struct r300_screen* r300screen = r300_screen(r300->context.screen);
CS_LOCALS(r300);
- BEGIN_CS(13);
+ BEGIN_CS(20);
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);
+ OUT_CS(rs->point_minmax);
+ OUT_CS(rs->line_control);
OUT_CS_REG_SEQ(R300_SU_POLY_OFFSET_FRONT_SCALE, 6);
OUT_CS(rs->depth_scale_front);
OUT_CS(rs->depth_offset_front);
OUT_CS(rs->cull_mode);
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);
END_CS;
}
void r300_emit_rs_block_state(struct r300_context* r300,
struct r300_rs_block* rs)
{
+ int i;
struct r300_screen* r300screen = r300_screen(r300->context.screen);
CS_LOCALS(r300);
- int i;
BEGIN_CS(21);
if (r300screen->caps->is_r500) {
}
for (i = 0; i < 8; i++) {
OUT_CS(rs->ip[i]);
+ debug_printf("ip %d: 0x%08x\n", i, rs->ip[i]);
}
OUT_CS_REG_SEQ(R300_RS_COUNT, 2);
}
for (i = 0; i < 8; i++) {
OUT_CS(rs->inst[i]);
+ debug_printf("inst %d: 0x%08x\n", i, rs->inst[i]);
}
+ debug_printf("count: 0x%08x inst_count: 0x%08x\n", rs->count,
+ rs->inst_count);
+
+ END_CS;
+}
+
+void r300_emit_sampler(struct r300_context* r300,
+ struct r300_sampler_state* sampler, unsigned offset)
+{
+ CS_LOCALS(r300);
+
+ BEGIN_CS(6);
+ OUT_CS_REG(R300_TX_FILTER0_0 + (offset * 4), sampler->filter0);
+ OUT_CS_REG(R300_TX_FILTER1_0 + (offset * 4), sampler->filter1);
+ OUT_CS_REG(R300_TX_BORDER_COLOR_0 + (offset * 4), sampler->border_color);
END_CS;
}
END_CS;
}
-void r300_emit_vertex_format_state(struct r300_context* r300)
+void r300_emit_texture(struct r300_context* r300,
+ struct r300_texture* tex, unsigned offset)
{
CS_LOCALS(r300);
+
+ BEGIN_CS(10);
+ OUT_CS_REG(R300_TX_FORMAT0_0 + (offset * 4), tex->state.format0);
+ OUT_CS_REG(R300_TX_FORMAT1_0 + (offset * 4), tex->state.format1);
+ OUT_CS_REG(R300_TX_FORMAT2_0 + (offset * 4), tex->state.format2);
+ OUT_CS_REG_SEQ(R300_TX_OFFSET_0 + (offset * 4), 1);
+ OUT_CS_RELOC(tex->buffer, 0,
+ RADEON_GEM_DOMAIN_GTT | RADEON_GEM_DOMAIN_VRAM, 0, 0);
+ END_CS;
+}
+
+void r300_emit_vertex_format_state(struct r300_context* r300)
+{
int i;
+ CS_LOCALS(r300);
+
+ BEGIN_CS(26);
+ OUT_CS_REG(R300_VAP_VTX_SIZE, r300->vertex_info.vinfo.size);
- BEGIN_CS(6);
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_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]);
- END_CS;
+ for (i = 0; i < 4; i++) {
+ debug_printf("hwfmt%d: 0x%08x\n", i,
+ r300->vertex_info.vinfo.hwfmt[i]);
+ }
- BEGIN_CS(18);
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_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]);
}
END_CS;
}
+void r300_emit_vertex_shader(struct r300_context* r300,
+ struct r300_vertex_shader* vs)
+{
+ int i;
+ struct r300_screen* r300screen = r300_screen(r300->context.screen);
+ struct r300_constant_buffer* constants =
+ &r300->shader_constants[PIPE_SHADER_VERTEX];
+ CS_LOCALS(r300);
+
+ if (!r300screen->caps->has_tcl) {
+ debug_printf("r300: Implementation error: emit_vertex_shader called,"
+ " but has_tcl is FALSE!\n");
+ return;
+ }
+
+ BEGIN_CS(13 + (vs->instruction_count * 4) + (constants->count * 4));
+
+ OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_0, R300_PVS_FIRST_INST(0) |
+ R300_PVS_LAST_INST(vs->instruction_count - 1));
+ OUT_CS_REG(R300_VAP_PVS_CODE_CNTL_1, vs->instruction_count - 1);
+
+ /* XXX */
+ OUT_CS_REG(R300_VAP_PVS_CONST_CNTL, 0x0);
+
+ 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_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_viewport_state(struct r300_context* r300,
+ struct r300_viewport_state* viewport)
+{
+ CS_LOCALS(r300);
+
+ BEGIN_CS(9);
+ OUT_CS_REG_SEQ(R300_SE_VPORT_XSCALE, 6);
+ OUT_CS_32F(viewport->xscale);
+ OUT_CS_32F(viewport->xoffset);
+ OUT_CS_32F(viewport->yscale);
+ OUT_CS_32F(viewport->yoffset);
+ OUT_CS_32F(viewport->zscale);
+ OUT_CS_32F(viewport->zoffset);
+
+ OUT_CS_REG(R300_VAP_VTE_CNTL, viewport->vte_control);
+ END_CS;
+}
+
+void r300_flush_textures(struct r300_context* r300)
+{
+ CS_LOCALS(r300);
+
+ BEGIN_CS(4);
+ OUT_CS_REG(R300_TX_INVALTAGS, 0);
+ OUT_CS_REG(R300_TX_ENABLE, (1 << r300->texture_count) - 1);
+ END_CS;
+}
+
/* Emit all dirty state. */
void r300_emit_dirty_state(struct r300_context* r300)
{
struct r300_screen* r300screen = r300_screen(r300->context.screen);
- CS_LOCALS(r300);
+ int i;
+ int dirty_tex = 0;
if (!(r300->dirty_state) && !(r300->dirty_hw)) {
return;
r300->dirty_state &= ~R300_NEW_FRAGMENT_SHADER;
}
+ if (r300->dirty_state & R300_NEW_FRAMEBUFFERS) {
+ r300_emit_fb_state(r300, &r300->framebuffer_state);
+ r300->dirty_state &= ~R300_NEW_FRAMEBUFFERS;
+ }
+
if (r300->dirty_state & R300_NEW_RASTERIZER) {
r300_emit_rs_state(r300, r300->rs_state);
r300->dirty_state &= ~R300_NEW_RASTERIZER;
}
+ if (r300->dirty_state & R300_NEW_RS_BLOCK) {
+ r300_emit_rs_block_state(r300, r300->rs_block);
+ r300->dirty_state &= ~R300_NEW_RS_BLOCK;
+ }
+
+ if (r300->dirty_state & R300_ANY_NEW_SAMPLERS) {
+ for (i = 0; i < r300->sampler_count; i++) {
+ if (r300->dirty_state & (R300_NEW_SAMPLER << i)) {
+ r300_emit_sampler(r300, r300->sampler_states[i], i);
+ r300->dirty_state &= ~(R300_NEW_SAMPLER << i);
+ dirty_tex++;
+ }
+ }
+ }
+
if (r300->dirty_state & R300_NEW_SCISSOR) {
r300_emit_scissor_state(r300, r300->scissor_state);
r300->dirty_state &= ~R300_NEW_SCISSOR;
}
+ if (r300->dirty_state & R300_ANY_NEW_TEXTURES) {
+ for (i = 0; i < r300->texture_count; i++) {
+ if (r300->dirty_state & (R300_NEW_TEXTURE << i)) {
+ r300_emit_texture(r300, r300->textures[i], i);
+ r300->dirty_state &= ~(R300_NEW_TEXTURE << i);
+ dirty_tex++;
+ }
+ }
+ }
+
+ if (r300->dirty_state & R300_NEW_VIEWPORT) {
+ r300_emit_viewport_state(r300, r300->viewport_state);
+ r300->dirty_state &= ~R300_NEW_VIEWPORT;
+ }
+
+ if (dirty_tex) {
+ r300_flush_textures(r300);
+ }
+
if (r300->dirty_state & R300_NEW_VERTEX_FORMAT) {
r300_emit_vertex_format_state(r300);
r300->dirty_state &= ~R300_NEW_VERTEX_FORMAT;