struct r600_context *rctx = (struct r600_context *)ctx;
struct si_pm4_state *pm4;
unsigned num_sgprs, num_user_sgprs;
- unsigned nparams, i;
+ unsigned nparams, i, vgpr_comp_cnt;
uint64_t va;
si_pm4_delete_state(rctx, vs, shader->pm4);
* takes care of adding a dummy export.
*/
for (nparams = 0, i = 0 ; i < shader->shader.noutput; i++) {
- if (shader->shader.output[i].name != TGSI_SEMANTIC_POSITION)
+ switch (shader->shader.output[i].name) {
+ case TGSI_SEMANTIC_POSITION:
+ case TGSI_SEMANTIC_PSIZE:
+ break;
+ default:
nparams++;
+ }
}
if (nparams < 1)
nparams = 1;
si_pm4_set_reg(pm4, R_02870C_SPI_SHADER_POS_FORMAT,
S_02870C_POS0_EXPORT_FORMAT(V_02870C_SPI_SHADER_4COMP) |
- S_02870C_POS1_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
- S_02870C_POS2_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE) |
- S_02870C_POS3_EXPORT_FORMAT(V_02870C_SPI_SHADER_NONE));
+ S_02870C_POS1_EXPORT_FORMAT(shader->shader.vs_out_misc_write ?
+ V_02870C_SPI_SHADER_4COMP :
+ V_02870C_SPI_SHADER_NONE) |
+ S_02870C_POS2_EXPORT_FORMAT((shader->shader.clip_dist_write & 0x0F) ?
+ V_02870C_SPI_SHADER_4COMP :
+ V_02870C_SPI_SHADER_NONE) |
+ S_02870C_POS3_EXPORT_FORMAT((shader->shader.clip_dist_write & 0xF0) ?
+ V_02870C_SPI_SHADER_4COMP :
+ V_02870C_SPI_SHADER_NONE));
va = r600_resource_va(ctx->screen, (void *)shader->bo);
si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
num_sgprs += 2;
assert(num_sgprs <= 104);
+ vgpr_comp_cnt = shader->shader.uses_instanceid ? 3 : 0;
+
si_pm4_set_reg(pm4, R_00B128_SPI_SHADER_PGM_RSRC1_VS,
S_00B128_VGPRS((shader->num_vgprs - 1) / 4) |
- S_00B128_SGPRS((num_sgprs - 1) / 8));
+ S_00B128_SGPRS((num_sgprs - 1) / 8) |
+ S_00B128_VGPR_COMP_CNT(vgpr_comp_cnt));
si_pm4_set_reg(pm4, R_00B12C_SPI_SHADER_PGM_RSRC2_VS,
S_00B12C_USER_SGPR(num_user_sgprs));
unsigned num_sgprs, num_user_sgprs;
boolean have_linear = FALSE, have_centroid = FALSE, have_perspective = FALSE;
unsigned fragcoord_interp_mode = 0;
- unsigned spi_baryc_cntl, spi_ps_input_ena;
+ unsigned spi_baryc_cntl, spi_ps_input_ena, spi_shader_z_format;
uint64_t va;
si_pm4_delete_state(rctx, ps, shader->pm4);
db_shader_control = S_02880C_Z_ORDER(V_02880C_EARLY_Z_THEN_LATE_Z);
for (i = 0; i < shader->shader.ninput; i++) {
- if (shader->shader.input[i].name == TGSI_SEMANTIC_POSITION) {
+ switch (shader->shader.input[i].name) {
+ case TGSI_SEMANTIC_POSITION:
if (shader->shader.input[i].centroid) {
/* fragcoord_interp_mode will be written to
* SPI_BARYC_CNTL.POS_FLOAT_LOCATION
*/
fragcoord_interp_mode = 1;
}
+ /* Fall through */
+ case TGSI_SEMANTIC_FACE:
continue;
}
- /* XXX: Flat shading hangs the GPU */
- if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
- (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
- rctx->queued.named.rasterizer->flatshade))
- have_linear = TRUE;
if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_LINEAR)
have_linear = TRUE;
if (shader->shader.input[i].interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
if (shader->shader.output[i].name == TGSI_SEMANTIC_POSITION)
db_shader_control |= S_02880C_Z_EXPORT_ENABLE(1);
if (shader->shader.output[i].name == TGSI_SEMANTIC_STENCIL)
- db_shader_control |= 0; // XXX OP_VAL or TEST_VAL?
+ db_shader_control |= S_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(1);
}
- if (shader->shader.uses_kill)
+ if (shader->shader.uses_kill || shader->key.ps.alpha_func != PIPE_FUNC_ALWAYS)
db_shader_control |= S_02880C_KILL_ENABLE(1);
exports_ps = 0;
si_pm4_set_reg(pm4, R_0286D0_SPI_PS_INPUT_ADDR, spi_ps_input_ena);
si_pm4_set_reg(pm4, R_0286D8_SPI_PS_IN_CONTROL, spi_ps_in_control);
- /* XXX: Depends on Z buffer format? */
- si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, 0);
+ if (G_02880C_STENCIL_TEST_VAL_EXPORT_ENABLE(db_shader_control))
+ spi_shader_z_format = V_028710_SPI_SHADER_32_GR;
+ else if (G_02880C_Z_EXPORT_ENABLE(db_shader_control))
+ spi_shader_z_format = V_028710_SPI_SHADER_32_R;
+ else
+ spi_shader_z_format = 0;
+ si_pm4_set_reg(pm4, R_028710_SPI_SHADER_Z_FORMAT, spi_shader_z_format);
+ si_pm4_set_reg(pm4, R_028714_SPI_SHADER_COL_FORMAT,
+ shader->spi_shader_col_format);
+ si_pm4_set_reg(pm4, R_02823C_CB_SHADER_MASK, shader->cb_shader_mask);
va = r600_resource_va(ctx->screen, (void *)shader->bo);
si_pm4_add_bo(pm4, shader->bo, RADEON_USAGE_READ);
const struct pipe_draw_info *info)
{
struct si_pm4_state *pm4 = CALLOC_STRUCT(si_pm4_state);
+ struct si_shader *vs = &rctx->vs_shader->current->shader;
unsigned prim = si_conv_pipe_prim(info->mode);
unsigned ls_mask = 0;
return false;
}
- si_pm4_set_reg(pm4, R_008958_VGT_PRIMITIVE_TYPE, prim);
+ if (rctx->chip_class >= CIK)
+ si_pm4_set_reg(pm4, R_030908_VGT_PRIMITIVE_TYPE, prim);
+ else
+ si_pm4_set_reg(pm4, R_008958_VGT_PRIMITIVE_TYPE, prim);
si_pm4_set_reg(pm4, R_028400_VGT_MAX_VTX_INDX, ~0);
si_pm4_set_reg(pm4, R_028404_VGT_MIN_VTX_INDX, 0);
si_pm4_set_reg(pm4, R_028408_VGT_INDX_OFFSET,
info->indexed ? info->index_bias : info->start);
si_pm4_set_reg(pm4, R_02840C_VGT_MULTI_PRIM_IB_RESET_INDX, info->restart_index);
si_pm4_set_reg(pm4, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN, info->primitive_restart);
-#if 0
- si_pm4_set_reg(pm4, R_03CFF0_SQ_VTX_BASE_VTX_LOC, 0);
- si_pm4_set_reg(pm4, R_03CFF4_SQ_VTX_START_INST_LOC, info->start_instance);
-#endif
+ si_pm4_set_reg(pm4, R_00B130_SPI_SHADER_USER_DATA_VS_0 + SI_SGPR_START_INSTANCE * 4,
+ info->start_instance);
if (prim == V_008958_DI_PT_LINELIST)
ls_mask = 1;
si_pm4_set_reg(pm4, R_028814_PA_SU_SC_MODE_CNTL, rctx->pa_su_sc_mode_cntl);
}
si_pm4_set_reg(pm4, R_02881C_PA_CL_VS_OUT_CNTL,
- prim == PIPE_PRIM_POINTS ? rctx->pa_cl_vs_out_cntl : 0
- /*| (rctx->rasterizer->clip_plane_enable &
- rctx->vs_shader->shader.clip_dist_write)*/);
- si_pm4_set_reg(pm4, R_028810_PA_CL_CLIP_CNTL, rctx->pa_cl_clip_cntl
- /*| (rctx->vs_shader->shader.clip_dist_write ||
- rctx->vs_shader->shader.vs_prohibit_ucps ?
- 0 : rctx->rasterizer->clip_plane_enable & 0x3F)*/);
+ S_02881C_USE_VTX_POINT_SIZE(vs->vs_out_point_size) |
+ S_02881C_VS_OUT_CCDIST0_VEC_ENA((vs->clip_dist_write & 0x0F) != 0) |
+ S_02881C_VS_OUT_CCDIST1_VEC_ENA((vs->clip_dist_write & 0xF0) != 0) |
+ S_02881C_VS_OUT_MISC_VEC_ENA(vs->vs_out_misc_write) |
+ (rctx->queued.named.rasterizer->clip_plane_enable &
+ vs->clip_dist_write));
+ si_pm4_set_reg(pm4, R_028810_PA_CL_CLIP_CNTL,
+ rctx->queued.named.rasterizer->pa_cl_clip_cntl |
+ (vs->clip_dist_write ? 0 :
+ rctx->queued.named.rasterizer->clip_plane_enable & 0x3F));
si_pm4_set_state(rctx, draw_info, pm4);
return true;
}
-static void si_update_alpha_ref(struct r600_context *rctx)
-{
-#if 0
- unsigned alpha_ref;
- struct r600_pipe_state rstate;
-
- alpha_ref = rctx->alpha_ref;
- rstate.nregs = 0;
- if (rctx->export_16bpc)
- alpha_ref &= ~0x1FFF;
- si_pm4_set_reg(&rstate, R_028438_SX_ALPHA_REF, alpha_ref);
-
- si_pm4_set_state(rctx, TODO, pm4);
- rctx->alpha_ref_dirty = false;
-#endif
-}
-
static void si_update_spi_map(struct r600_context *rctx)
{
struct si_shader *ps = &rctx->ps_shader->current->shader;
unsigned i, j, tmp;
for (i = 0; i < ps->ninput; i++) {
+ unsigned name = ps->input[i].name;
+ unsigned param_offset = ps->input[i].param_offset;
+
+ if (name == TGSI_SEMANTIC_POSITION)
+ /* Read from preloaded VGPRs, not parameters */
+ continue;
+
+bcolor:
tmp = 0;
-#if 0
- /* XXX: Flat shading hangs the GPU */
- if (ps->input[i].name == TGSI_SEMANTIC_POSITION ||
- ps->input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
+ if (ps->input[i].interpolate == TGSI_INTERPOLATE_CONSTANT ||
(ps->input[i].interpolate == TGSI_INTERPOLATE_COLOR &&
- rctx->rasterizer && rctx->rasterizer->flatshade)) {
+ rctx->ps_shader->current->key.ps.flatshade)) {
tmp |= S_028644_FLAT_SHADE(1);
}
-#endif
- if (ps->input[i].name == TGSI_SEMANTIC_GENERIC &&
+ if (name == TGSI_SEMANTIC_GENERIC &&
rctx->sprite_coord_enable & (1 << ps->input[i].sid)) {
tmp |= S_028644_PT_SPRITE_TEX(1);
}
for (j = 0; j < vs->noutput; j++) {
- if (ps->input[i].name == vs->output[j].name &&
+ if (name == vs->output[j].name &&
ps->input[i].sid == vs->output[j].sid) {
tmp |= S_028644_OFFSET(vs->output[j].param_offset);
break;
}
si_pm4_set_reg(pm4,
- R_028644_SPI_PS_INPUT_CNTL_0 + ps->input[i].param_offset * 4,
+ R_028644_SPI_PS_INPUT_CNTL_0 + param_offset * 4,
tmp);
+
+ if (name == TGSI_SEMANTIC_COLOR &&
+ rctx->ps_shader->current->key.ps.color_two_side) {
+ name = TGSI_SEMANTIC_BCOLOR;
+ param_offset++;
+ goto bcolor;
+ }
}
si_pm4_set_state(rctx, spi, pm4);
static void si_update_derived_state(struct r600_context *rctx)
{
struct pipe_context * ctx = (struct pipe_context*)rctx;
- unsigned ps_dirty = 0;
+ unsigned vs_dirty = 0, ps_dirty = 0;
if (!rctx->blitter->running) {
- if (rctx->have_depth_fb || rctx->have_depth_texture)
- si_flush_depth_textures(rctx);
+ /* Flush depth textures which need to be flushed. */
+ if (rctx->vs_samplers.depth_texture_mask) {
+ si_flush_depth_textures(rctx, &rctx->vs_samplers);
+ }
+ if (rctx->ps_samplers.depth_texture_mask) {
+ si_flush_depth_textures(rctx, &rctx->ps_samplers);
+ }
}
- si_shader_select(ctx, rctx->ps_shader, &ps_dirty);
-
- if (rctx->alpha_ref_dirty) {
- si_update_alpha_ref(rctx);
- }
+ si_shader_select(ctx, rctx->vs_shader, &vs_dirty);
if (!rctx->vs_shader->current->pm4) {
si_pipe_shader_vs(ctx, rctx->vs_shader->current);
+ vs_dirty = 0;
}
+ if (vs_dirty) {
+ si_pm4_bind_state(rctx, vs, rctx->vs_shader->current->pm4);
+ }
+
+
+ si_shader_select(ctx, rctx->ps_shader, &ps_dirty);
+
if (!rctx->ps_shader->current->pm4) {
si_pipe_shader_ps(ctx, rctx->ps_shader->current);
ps_dirty = 0;
}
if (si_pm4_state_changed(rctx, ps) || si_pm4_state_changed(rctx, vs)) {
+ /* XXX: Emitting the PS state even when only the VS changed
+ * fixes random failures with piglit glsl-max-varyings.
+ * Not sure why...
+ */
+ rctx->emitted.named.ps = NULL;
si_update_spi_map(rctx);
}
}
+static void si_constant_buffer_update(struct r600_context *rctx)
+{
+ struct pipe_context *ctx = &rctx->context;
+ struct si_pm4_state *pm4;
+ unsigned shader, i;
+ uint64_t va;
+
+ if (!rctx->constbuf_state[PIPE_SHADER_VERTEX].dirty_mask &&
+ !rctx->constbuf_state[PIPE_SHADER_FRAGMENT].dirty_mask)
+ return;
+
+ for (shader = PIPE_SHADER_VERTEX ; shader <= PIPE_SHADER_FRAGMENT; shader++) {
+ struct r600_constbuf_state *state = &rctx->constbuf_state[shader];
+
+ pm4 = CALLOC_STRUCT(si_pm4_state);
+ if (!pm4)
+ continue;
+
+ si_pm4_inval_shader_cache(pm4);
+ si_pm4_sh_data_begin(pm4);
+
+ for (i = 0; i < 2; i++) {
+ if (state->enabled_mask & (1 << i)) {
+ struct pipe_constant_buffer *cb = &state->cb[i];
+ struct si_resource *rbuffer = si_resource(cb->buffer);
+
+ va = r600_resource_va(ctx->screen, (void*)rbuffer);
+ va += cb->buffer_offset;
+
+ si_pm4_add_bo(pm4, rbuffer, RADEON_USAGE_READ);
+
+ /* Fill in a T# buffer resource description */
+ si_pm4_sh_data_add(pm4, va);
+ si_pm4_sh_data_add(pm4, (S_008F04_BASE_ADDRESS_HI(va >> 32) |
+ S_008F04_STRIDE(0)));
+ si_pm4_sh_data_add(pm4, cb->buffer_size);
+ si_pm4_sh_data_add(pm4, S_008F0C_DST_SEL_X(V_008F0C_SQ_SEL_X) |
+ S_008F0C_DST_SEL_Y(V_008F0C_SQ_SEL_Y) |
+ S_008F0C_DST_SEL_Z(V_008F0C_SQ_SEL_Z) |
+ S_008F0C_DST_SEL_W(V_008F0C_SQ_SEL_W) |
+ S_008F0C_NUM_FORMAT(V_008F0C_BUF_NUM_FORMAT_FLOAT) |
+ S_008F0C_DATA_FORMAT(V_008F0C_BUF_DATA_FORMAT_32));
+ } else {
+ /* Fill in an empty T# buffer resource description */
+ si_pm4_sh_data_add(pm4, 0);
+ si_pm4_sh_data_add(pm4, 0);
+ si_pm4_sh_data_add(pm4, 0);
+ si_pm4_sh_data_add(pm4, 0);
+ }
+ }
+
+ switch (shader) {
+ case PIPE_SHADER_VERTEX:
+ si_pm4_sh_data_end(pm4, R_00B130_SPI_SHADER_USER_DATA_VS_0, SI_SGPR_CONST);
+ si_pm4_set_state(rctx, vs_const, pm4);
+ break;
+
+ case PIPE_SHADER_FRAGMENT:
+ si_pm4_sh_data_end(pm4, R_00B030_SPI_SHADER_USER_DATA_PS_0, SI_SGPR_CONST);
+ si_pm4_set_state(rctx, ps_const, pm4);
+ break;
+
+ default:
+ R600_ERR("unsupported %d\n", shader);
+ FREE(pm4);
+ return;
+ }
+
+ state->dirty_mask = 0;
+ }
+}
+
static void si_vertex_buffer_update(struct r600_context *rctx)
{
struct pipe_context *ctx = &rctx->context;
unsigned i, count;
uint64_t va;
- si_pm4_inval_vertex_cache(pm4);
+ si_pm4_inval_texture_cache(pm4);
/* bind vertex buffer once */
count = rctx->vertex_elements->count;
si_pm4_sh_data_add(pm4, va & 0xFFFFFFFF);
si_pm4_sh_data_add(pm4, (S_008F04_BASE_ADDRESS_HI(va >> 32) |
S_008F04_STRIDE(vb->stride)));
- si_pm4_sh_data_add(pm4, (vb->buffer->width0 - vb->buffer_offset) /
- MAX2(vb->stride, 1));
+ if (vb->stride)
+ /* Round up by rounding down and adding 1 */
+ si_pm4_sh_data_add(pm4,
+ (vb->buffer->width0 - offset -
+ util_format_get_blocksize(ve->src_format)) /
+ vb->stride + 1);
+ else
+ si_pm4_sh_data_add(pm4, vb->buffer->width0 - offset);
si_pm4_sh_data_add(pm4, rctx->vertex_elements->rsrc_word3[i]);
if (!bound[ve->vertex_buffer_index]) {
struct pipe_index_buffer ib = {};
uint32_t cp_coher_cntl;
- if ((!info->count && (info->indexed || !info->count_from_stream_output)) ||
- (info->indexed && !rctx->index_buffer.buffer)) {
+ if (!info->count && (info->indexed || !info->count_from_stream_output))
return;
- }
if (!rctx->ps_shader || !rctx->vs_shader)
return;
si_update_derived_state(rctx);
+ si_constant_buffer_update(rctx);
si_vertex_buffer_update(rctx);
if (info->indexed) {
/* Initialize the index buffer struct. */
pipe_resource_reference(&ib.buffer, rctx->index_buffer.buffer);
+ ib.user_buffer = rctx->index_buffer.user_buffer;
ib.index_size = rctx->index_buffer.index_size;
ib.offset = rctx->index_buffer.offset + info->start * ib.index_size;
/* Translate or upload, if needed. */
r600_translate_index_buffer(rctx, &ib, info->count);
- if (ib.user_buffer) {
+ if (ib.user_buffer && !ib.buffer) {
r600_upload_index_buffer(rctx, &ib, info->count);
}
si_pm4_emit_dirty(rctx);
rctx->pm4_dirty_cdwords = 0;
+#if R600_TRACE_CS
+ if (rctx->screen->trace_bo) {
+ r600_trace_emit(rctx);
+ }
+#endif
+
#if 0
/* Enable stream out if needed. */
if (rctx->streamout_start) {
}
#endif
-
rctx->flags |= R600_CONTEXT_DST_CACHES_DIRTY;
- if (rctx->framebuffer.zsbuf)
- {
- struct pipe_resource *tex = rctx->framebuffer.zsbuf->texture;
- ((struct r600_resource_texture *)tex)->dirty_db = TRUE;
+ /* Set the depth buffer as dirty. */
+ if (rctx->framebuffer.zsbuf) {
+ struct pipe_surface *surf = rctx->framebuffer.zsbuf;
+ struct r600_resource_texture *rtex = (struct r600_resource_texture *)surf->texture;
+
+ rtex->dirty_db_mask |= 1 << surf->u.tex.level;
}
pipe_resource_reference(&ib.buffer, NULL);