#include "radv_private.h"
#include "radv_radeon_winsys.h"
+#include "radv_shader.h"
#include "radv_cs.h"
#include "sid.h"
+#include "gfx9d.h"
#include "vk_format.h"
+#include "radv_debug.h"
#include "radv_meta.h"
#include "ac_debug.h"
VkCommandBuffer* pCommandBuffer)
{
struct radv_cmd_buffer *cmd_buffer;
- VkResult result;
unsigned ring;
cmd_buffer = vk_alloc(&pool->alloc, sizeof(*cmd_buffer), 8,
VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
cmd_buffer->cs = device->ws->cs_create(device->ws, ring);
if (!cmd_buffer->cs) {
- result = VK_ERROR_OUT_OF_HOST_MEMORY;
- goto fail;
+ vk_free(&cmd_buffer->pool->alloc, cmd_buffer);
+ return vk_error(VK_ERROR_OUT_OF_HOST_MEMORY);
}
*pCommandBuffer = radv_cmd_buffer_to_handle(cmd_buffer);
list_inithead(&cmd_buffer->upload.list);
return VK_SUCCESS;
-
-fail:
- vk_free(&cmd_buffer->pool->alloc, cmd_buffer);
-
- return result;
}
static void
vk_free(&cmd_buffer->pool->alloc, cmd_buffer);
}
-static void radv_reset_cmd_buffer(struct radv_cmd_buffer *cmd_buffer)
+static VkResult
+radv_reset_cmd_buffer(struct radv_cmd_buffer *cmd_buffer)
{
cmd_buffer->device->ws->cs_reset(cmd_buffer->cs);
free(up);
}
+ cmd_buffer->push_constant_stages = 0;
cmd_buffer->scratch_size_needed = 0;
cmd_buffer->compute_scratch_size_needed = 0;
cmd_buffer->esgs_ring_size_needed = 0;
cmd_buffer->upload.upload_bo, 8);
cmd_buffer->upload.offset = 0;
- cmd_buffer->record_fail = false;
+ cmd_buffer->record_result = VK_SUCCESS;
cmd_buffer->ring_offsets_idx = -1;
+
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ void *fence_ptr;
+ radv_cmd_buffer_upload_alloc(cmd_buffer, 8, 0,
+ &cmd_buffer->gfx9_fence_offset,
+ &fence_ptr);
+ cmd_buffer->gfx9_fence_bo = cmd_buffer->upload.upload_bo;
+ }
+
+ return cmd_buffer->record_result;
}
static bool
RADEON_FLAG_CPU_ACCESS);
if (!bo) {
- cmd_buffer->record_fail = true;
+ cmd_buffer->record_result = VK_ERROR_OUT_OF_DEVICE_MEMORY;
return false;
}
upload = malloc(sizeof(*upload));
if (!upload) {
- cmd_buffer->record_fail = true;
+ cmd_buffer->record_result = VK_ERROR_OUT_OF_HOST_MEMORY;
device->ws->buffer_destroy(bo);
return false;
}
cmd_buffer->upload.map = device->ws->buffer_map(cmd_buffer->upload.upload_bo);
if (!cmd_buffer->upload.map) {
- cmd_buffer->record_fail = true;
+ cmd_buffer->record_result = VK_ERROR_OUT_OF_DEVICE_MEMORY;
return false;
}
return true;
}
+static void
+radv_emit_write_data_packet(struct radeon_winsys_cs *cs, uint64_t va,
+ unsigned count, const uint32_t *data)
+{
+ radeon_emit(cs, PKT3(PKT3_WRITE_DATA, 2 + count, 0));
+ radeon_emit(cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
+ S_370_WR_CONFIRM(1) |
+ S_370_ENGINE_SEL(V_370_ME));
+ radeon_emit(cs, va);
+ radeon_emit(cs, va >> 32);
+ radeon_emit_array(cs, data, count);
+}
+
void radv_cmd_buffer_trace_emit(struct radv_cmd_buffer *cmd_buffer)
{
struct radv_device *device = cmd_buffer->device;
return;
va = device->ws->buffer_get_va(device->trace_bo);
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY)
+ va += 4;
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 7);
++cmd_buffer->state.trace_id;
device->ws->cs_add_buffer(cs, device->trace_bo, 8);
- radeon_emit(cs, PKT3(PKT3_WRITE_DATA, 3, 0));
- radeon_emit(cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
- S_370_WR_CONFIRM(1) |
- S_370_ENGINE_SEL(V_370_ME));
- radeon_emit(cs, va);
- radeon_emit(cs, va >> 32);
- radeon_emit(cs, cmd_buffer->state.trace_id);
+ radv_emit_write_data_packet(cs, va, 1, &cmd_buffer->state.trace_id);
radeon_emit(cs, PKT3(PKT3_NOP, 0, 0));
radeon_emit(cs, AC_ENCODE_TRACE_POINT(cmd_buffer->state.trace_id));
}
+static void
+radv_cmd_buffer_after_draw(struct radv_cmd_buffer *cmd_buffer)
+{
+ if (cmd_buffer->device->debug_flags & RADV_DEBUG_SYNC_SHADERS) {
+ enum radv_cmd_flush_bits flags;
+
+ /* Force wait for graphics/compute engines to be idle. */
+ flags = RADV_CMD_FLAG_PS_PARTIAL_FLUSH |
+ RADV_CMD_FLAG_CS_PARTIAL_FLUSH;
+
+ si_cs_emit_cache_flush(cmd_buffer->cs, false,
+ cmd_buffer->device->physical_device->rad_info.chip_class,
+ NULL, 0,
+ radv_cmd_buffer_uses_mec(cmd_buffer),
+ flags);
+ }
+
+ radv_cmd_buffer_trace_emit(cmd_buffer);
+}
+
+static void
+radv_save_pipeline(struct radv_cmd_buffer *cmd_buffer,
+ struct radv_pipeline *pipeline, enum ring_type ring)
+{
+ struct radv_device *device = cmd_buffer->device;
+ struct radeon_winsys_cs *cs = cmd_buffer->cs;
+ uint32_t data[2];
+ uint64_t va;
+
+ if (!device->trace_bo)
+ return;
+
+ va = device->ws->buffer_get_va(device->trace_bo);
+
+ switch (ring) {
+ case RING_GFX:
+ va += 8;
+ break;
+ case RING_COMPUTE:
+ va += 16;
+ break;
+ default:
+ assert(!"invalid ring type");
+ }
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(device->ws,
+ cmd_buffer->cs, 6);
+
+ data[0] = (uintptr_t)pipeline;
+ data[1] = (uintptr_t)pipeline >> 32;
+
+ device->ws->cs_add_buffer(cs, device->trace_bo, 8);
+ radv_emit_write_data_packet(cs, va, 2, data);
+}
+
+static void
+radv_save_descriptors(struct radv_cmd_buffer *cmd_buffer)
+{
+ struct radv_device *device = cmd_buffer->device;
+ struct radeon_winsys_cs *cs = cmd_buffer->cs;
+ uint32_t data[MAX_SETS * 2] = {};
+ uint64_t va;
+
+ if (!device->trace_bo)
+ return;
+
+ va = device->ws->buffer_get_va(device->trace_bo) + 24;
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(device->ws,
+ cmd_buffer->cs, 4 + MAX_SETS * 2);
+
+ for (int i = 0; i < MAX_SETS; i++) {
+ struct radv_descriptor_set *set = cmd_buffer->state.descriptors[i];
+ if (!set)
+ continue;
+
+ data[i * 2] = (uintptr_t)set;
+ data[i * 2 + 1] = (uintptr_t)set >> 32;
+ }
+
+ device->ws->cs_add_buffer(cs, device->trace_bo, 8);
+ radv_emit_write_data_packet(cs, va, MAX_SETS * 2, data);
+}
+
static void
radv_emit_graphics_blend_state(struct radv_cmd_buffer *cmd_buffer,
struct radv_pipeline *pipeline)
8);
radeon_set_context_reg(cmd_buffer->cs, R_028808_CB_COLOR_CONTROL, pipeline->graphics.blend.cb_color_control);
radeon_set_context_reg(cmd_buffer->cs, R_028B70_DB_ALPHA_TO_MASK, pipeline->graphics.blend.db_alpha_to_mask);
+
+ if (cmd_buffer->device->physical_device->has_rbplus) {
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028760_SX_MRT0_BLEND_OPT, 8);
+ radeon_emit_array(cmd_buffer->cs, pipeline->graphics.blend.sx_mrt_blend_opt, 8);
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028754_SX_PS_DOWNCONVERT, 3);
+ radeon_emit(cmd_buffer->cs, 0); /* R_028754_SX_PS_DOWNCONVERT */
+ radeon_emit(cmd_buffer->cs, 0); /* R_028758_SX_BLEND_OPT_EPSILON */
+ radeon_emit(cmd_buffer->cs, 0); /* R_02875C_SX_BLEND_OPT_CONTROL */
+ }
}
static void
x >= 4096 ? 0xffff : x * 16;
}
-static uint32_t
-shader_stage_to_user_data_0(gl_shader_stage stage, bool has_gs, bool has_tess)
-{
- switch (stage) {
- case MESA_SHADER_FRAGMENT:
- return R_00B030_SPI_SHADER_USER_DATA_PS_0;
- case MESA_SHADER_VERTEX:
- if (has_tess)
- return R_00B530_SPI_SHADER_USER_DATA_LS_0;
- else
- return has_gs ? R_00B330_SPI_SHADER_USER_DATA_ES_0 : R_00B130_SPI_SHADER_USER_DATA_VS_0;
- case MESA_SHADER_GEOMETRY:
- return R_00B230_SPI_SHADER_USER_DATA_GS_0;
- case MESA_SHADER_COMPUTE:
- return R_00B900_COMPUTE_USER_DATA_0;
- case MESA_SHADER_TESS_CTRL:
- return R_00B430_SPI_SHADER_USER_DATA_HS_0;
- case MESA_SHADER_TESS_EVAL:
- if (has_gs)
- return R_00B330_SPI_SHADER_USER_DATA_ES_0;
- else
- return R_00B130_SPI_SHADER_USER_DATA_VS_0;
- default:
- unreachable("unknown shader");
- }
-}
-
-static struct ac_userdata_info *
+struct ac_userdata_info *
radv_lookup_user_sgpr(struct radv_pipeline *pipeline,
gl_shader_stage stage,
int idx)
int idx, uint64_t va)
{
struct ac_userdata_info *loc = radv_lookup_user_sgpr(pipeline, stage, idx);
- uint32_t base_reg = shader_stage_to_user_data_0(stage, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(stage, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
if (loc->sgpr_idx == -1)
return;
assert(loc->num_sgprs == 2);
radv_cayman_emit_msaa_sample_locs(cmd_buffer->cs, num_samples);
- if (pipeline->shaders[MESA_SHADER_FRAGMENT]->info.fs.uses_sample_positions) {
+ /* GFX9: Flush DFSM when the AA mode changes. */
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+ radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_FLUSH_DFSM) | EVENT_INDEX(0));
+ }
+ if (pipeline->shaders[MESA_SHADER_FRAGMENT]->info.info.ps.needs_sample_positions) {
uint32_t offset;
struct ac_userdata_info *loc = radv_lookup_user_sgpr(pipeline, MESA_SHADER_FRAGMENT, AC_UD_PS_SAMPLE_POS_OFFSET);
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_FRAGMENT, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(MESA_SHADER_FRAGMENT, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
if (loc->sgpr_idx == -1)
return;
assert(loc->num_sgprs == 1);
raster->pa_su_sc_mode_cntl);
}
+static inline void
+radv_emit_prefetch(struct radv_cmd_buffer *cmd_buffer, uint64_t va,
+ unsigned size)
+{
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK)
+ si_cp_dma_prefetch(cmd_buffer, va, size);
+}
+
static void
radv_emit_hw_vs(struct radv_cmd_buffer *cmd_buffer,
struct radv_pipeline *pipeline,
struct ac_vs_output_info *outinfo)
{
struct radeon_winsys *ws = cmd_buffer->device->ws;
- uint64_t va = ws->buffer_get_va(shader->bo);
+ uint64_t va = ws->buffer_get_va(shader->bo) + shader->bo_offset;
unsigned export_count;
ws->cs_add_buffer(cmd_buffer->cs, shader->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, shader->code_size);
export_count = MAX2(1, outinfo->param_exports);
radeon_set_context_reg(cmd_buffer->cs, R_0286C4_SPI_VS_OUT_CONFIG,
radeon_set_context_reg(cmd_buffer->cs, R_02881C_PA_CL_VS_OUT_CNTL,
pipeline->graphics.pa_cl_vs_out_cntl);
- radeon_set_context_reg(cmd_buffer->cs, R_028AB4_VGT_REUSE_OFF,
- S_028AB4_REUSE_OFF(outinfo->writes_viewport_index));
+ if (cmd_buffer->device->physical_device->rad_info.chip_class <= VI)
+ radeon_set_context_reg(cmd_buffer->cs, R_028AB4_VGT_REUSE_OFF,
+ S_028AB4_REUSE_OFF(outinfo->writes_viewport_index));
}
static void
struct ac_es_output_info *outinfo)
{
struct radeon_winsys *ws = cmd_buffer->device->ws;
- uint64_t va = ws->buffer_get_va(shader->bo);
+ uint64_t va = ws->buffer_get_va(shader->bo) + shader->bo_offset;
ws->cs_add_buffer(cmd_buffer->cs, shader->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, shader->code_size);
radeon_set_context_reg(cmd_buffer->cs, R_028AAC_VGT_ESGS_RING_ITEMSIZE,
outinfo->esgs_itemsize / 4);
struct radv_shader_variant *shader)
{
struct radeon_winsys *ws = cmd_buffer->device->ws;
- uint64_t va = ws->buffer_get_va(shader->bo);
+ uint64_t va = ws->buffer_get_va(shader->bo) + shader->bo_offset;
uint32_t rsrc2 = shader->rsrc2;
ws->cs_add_buffer(cmd_buffer->cs, shader->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, shader->code_size);
radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B520_SPI_SHADER_PGM_LO_LS, 2);
radeon_emit(cmd_buffer->cs, va >> 8);
struct radv_shader_variant *shader)
{
struct radeon_winsys *ws = cmd_buffer->device->ws;
- uint64_t va = ws->buffer_get_va(shader->bo);
+ uint64_t va = ws->buffer_get_va(shader->bo) + shader->bo_offset;
ws->cs_add_buffer(cmd_buffer->cs, shader->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, shader->code_size);
radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B420_SPI_SHADER_PGM_LO_HS, 4);
radeon_emit(cmd_buffer->cs, va >> 8);
else
radv_emit_hw_vs(cmd_buffer, pipeline, vs, &vs->info.vs.outinfo);
- radeon_set_context_reg(cmd_buffer->cs, R_028A84_VGT_PRIMITIVEID_EN, 0);
+ radeon_set_context_reg(cmd_buffer->cs, R_028A84_VGT_PRIMITIVEID_EN, pipeline->graphics.vgt_primitiveid_en);
}
loc = radv_lookup_user_sgpr(pipeline, MESA_SHADER_TESS_CTRL, AC_UD_TCS_OFFCHIP_LAYOUT);
if (loc->sgpr_idx != -1) {
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_TESS_CTRL, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(MESA_SHADER_TESS_CTRL, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
assert(loc->num_sgprs == 4);
assert(!loc->indirect);
radeon_set_sh_reg_seq(cmd_buffer->cs, base_reg + loc->sgpr_idx * 4, 4);
loc = radv_lookup_user_sgpr(pipeline, MESA_SHADER_TESS_EVAL, AC_UD_TES_OFFCHIP_LAYOUT);
if (loc->sgpr_idx != -1) {
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_TESS_EVAL, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(MESA_SHADER_TESS_EVAL, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
assert(loc->num_sgprs == 1);
assert(!loc->indirect);
loc = radv_lookup_user_sgpr(pipeline, MESA_SHADER_VERTEX, AC_UD_VS_LS_TCS_IN_LAYOUT);
if (loc->sgpr_idx != -1) {
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_VERTEX, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(MESA_SHADER_VERTEX, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
assert(loc->num_sgprs == 1);
assert(!loc->indirect);
S_028B90_CNT(MIN2(gs_num_invocations, 127)) |
S_028B90_ENABLE(gs_num_invocations > 0));
- va = ws->buffer_get_va(gs->bo);
+ va = ws->buffer_get_va(gs->bo) + gs->bo_offset;
ws->cs_add_buffer(cmd_buffer->cs, gs->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, gs->code_size);
+
radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B220_SPI_SHADER_PGM_LO_GS, 4);
radeon_emit(cmd_buffer->cs, va >> 8);
radeon_emit(cmd_buffer->cs, va >> 40);
assert (pipeline->shaders[MESA_SHADER_FRAGMENT]);
ps = pipeline->shaders[MESA_SHADER_FRAGMENT];
-
- va = ws->buffer_get_va(ps->bo);
+ va = ws->buffer_get_va(ps->bo) + ps->bo_offset;
ws->cs_add_buffer(cmd_buffer->cs, ps->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, ps->code_size);
radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B020_SPI_SHADER_PGM_LO_PS, 4);
radeon_emit(cmd_buffer->cs, va >> 8);
radeon_set_context_reg(cmd_buffer->cs, R_0286D0_SPI_PS_INPUT_ADDR,
ps->config.spi_ps_input_addr);
- if (ps->info.fs.force_persample)
+ if (ps->info.info.ps.force_persample)
spi_baryc_cntl |= S_0286E0_POS_FLOAT_LOCATION(2);
radeon_set_context_reg(cmd_buffer->cs, R_0286D8_SPI_PS_IN_CONTROL,
radeon_set_context_reg(cmd_buffer->cs, R_028238_CB_TARGET_MASK, blend->cb_target_mask);
radeon_set_context_reg(cmd_buffer->cs, R_02823C_CB_SHADER_MASK, blend->cb_shader_mask);
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ /* optimise this? */
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+ radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_FLUSH_DFSM) | EVENT_INDEX(0));
+ }
+
if (pipeline->graphics.ps_input_cntl_num) {
radeon_set_context_reg_seq(cmd_buffer->cs, R_028644_SPI_PS_INPUT_CNTL_0, pipeline->graphics.ps_input_cntl_num);
for (unsigned i = 0; i < pipeline->graphics.ps_input_cntl_num; i++) {
}
static void
-radv_emit_graphics_pipeline(struct radv_cmd_buffer *cmd_buffer,
- struct radv_pipeline *pipeline)
+radv_emit_graphics_pipeline(struct radv_cmd_buffer *cmd_buffer)
{
+ struct radv_pipeline *pipeline = cmd_buffer->state.pipeline;
+
if (!pipeline || cmd_buffer->state.emitted_pipeline == pipeline)
return;
cmd_buffer->state.emitted_pipeline->graphics.can_use_guardband !=
pipeline->graphics.can_use_guardband)
cmd_buffer->state.dirty |= RADV_CMD_DIRTY_DYNAMIC_SCISSOR;
+
+ radeon_set_context_reg(cmd_buffer->cs, R_028B54_VGT_SHADER_STAGES_EN, pipeline->graphics.vgt_shader_stages_en);
+
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK) {
+ radeon_set_uconfig_reg_idx(cmd_buffer->cs, R_030908_VGT_PRIMITIVE_TYPE, 1, pipeline->graphics.prim);
+ } else {
+ radeon_set_config_reg(cmd_buffer->cs, R_008958_VGT_PRIMITIVE_TYPE, pipeline->graphics.prim);
+ }
+ radeon_set_context_reg(cmd_buffer->cs, R_028A6C_VGT_GS_OUT_PRIM_TYPE, pipeline->graphics.gs_out);
+
+ radv_save_pipeline(cmd_buffer, pipeline, RING_GFX);
+
cmd_buffer->state.emitted_pipeline = pipeline;
}
cmd_buffer->state.pipeline->graphics.ms.pa_sc_mode_cntl_0 | S_028A48_VPORT_SCISSOR_ENABLE(count ? 1 : 0));
}
+static void
+radv_emit_line_width(struct radv_cmd_buffer *cmd_buffer)
+{
+ unsigned width = cmd_buffer->state.dynamic.line_width * 8;
+
+ radeon_set_context_reg(cmd_buffer->cs, R_028A08_PA_SU_LINE_CNTL,
+ S_028A08_WIDTH(CLAMP(width, 0, 0xFFF)));
+}
+
+static void
+radv_emit_blend_constants(struct radv_cmd_buffer *cmd_buffer)
+{
+ struct radv_dynamic_state *d = &cmd_buffer->state.dynamic;
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028414_CB_BLEND_RED, 4);
+ radeon_emit_array(cmd_buffer->cs, (uint32_t *)d->blend_constants, 4);
+}
+
+static void
+radv_emit_stencil(struct radv_cmd_buffer *cmd_buffer)
+{
+ struct radv_dynamic_state *d = &cmd_buffer->state.dynamic;
+
+ radeon_set_context_reg_seq(cmd_buffer->cs,
+ R_028430_DB_STENCILREFMASK, 2);
+ radeon_emit(cmd_buffer->cs,
+ S_028430_STENCILTESTVAL(d->stencil_reference.front) |
+ S_028430_STENCILMASK(d->stencil_compare_mask.front) |
+ S_028430_STENCILWRITEMASK(d->stencil_write_mask.front) |
+ S_028430_STENCILOPVAL(1));
+ radeon_emit(cmd_buffer->cs,
+ S_028434_STENCILTESTVAL_BF(d->stencil_reference.back) |
+ S_028434_STENCILMASK_BF(d->stencil_compare_mask.back) |
+ S_028434_STENCILWRITEMASK_BF(d->stencil_write_mask.back) |
+ S_028434_STENCILOPVAL_BF(1));
+}
+
+static void
+radv_emit_depth_bounds(struct radv_cmd_buffer *cmd_buffer)
+{
+ struct radv_dynamic_state *d = &cmd_buffer->state.dynamic;
+
+ radeon_set_context_reg(cmd_buffer->cs, R_028020_DB_DEPTH_BOUNDS_MIN,
+ fui(d->depth_bounds.min));
+ radeon_set_context_reg(cmd_buffer->cs, R_028024_DB_DEPTH_BOUNDS_MAX,
+ fui(d->depth_bounds.max));
+}
+
+static void
+radv_emit_depth_biais(struct radv_cmd_buffer *cmd_buffer)
+{
+ struct radv_raster_state *raster = &cmd_buffer->state.pipeline->graphics.raster;
+ struct radv_dynamic_state *d = &cmd_buffer->state.dynamic;
+ unsigned slope = fui(d->depth_bias.slope * 16.0f);
+ unsigned bias = fui(d->depth_bias.bias * cmd_buffer->state.offset_scale);
+
+ if (G_028814_POLY_OFFSET_FRONT_ENABLE(raster->pa_su_sc_mode_cntl)) {
+ radeon_set_context_reg_seq(cmd_buffer->cs,
+ R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 5);
+ radeon_emit(cmd_buffer->cs, fui(d->depth_bias.clamp)); /* CLAMP */
+ radeon_emit(cmd_buffer->cs, slope); /* FRONT SCALE */
+ radeon_emit(cmd_buffer->cs, bias); /* FRONT OFFSET */
+ radeon_emit(cmd_buffer->cs, slope); /* BACK SCALE */
+ radeon_emit(cmd_buffer->cs, bias); /* BACK OFFSET */
+ }
+}
+
static void
radv_emit_fb_color_state(struct radv_cmd_buffer *cmd_buffer,
int index,
struct radv_color_buffer_info *cb)
{
bool is_vi = cmd_buffer->device->physical_device->rad_info.chip_class >= VI;
- radeon_set_context_reg_seq(cmd_buffer->cs, R_028C60_CB_COLOR0_BASE + index * 0x3c, 11);
- radeon_emit(cmd_buffer->cs, cb->cb_color_base);
- radeon_emit(cmd_buffer->cs, cb->cb_color_pitch);
- radeon_emit(cmd_buffer->cs, cb->cb_color_slice);
- radeon_emit(cmd_buffer->cs, cb->cb_color_view);
- radeon_emit(cmd_buffer->cs, cb->cb_color_info);
- radeon_emit(cmd_buffer->cs, cb->cb_color_attrib);
- radeon_emit(cmd_buffer->cs, cb->cb_dcc_control);
- radeon_emit(cmd_buffer->cs, cb->cb_color_cmask);
- radeon_emit(cmd_buffer->cs, cb->cb_color_cmask_slice);
- radeon_emit(cmd_buffer->cs, cb->cb_color_fmask);
- radeon_emit(cmd_buffer->cs, cb->cb_color_fmask_slice);
- if (is_vi) { /* DCC BASE */
- radeon_set_context_reg(cmd_buffer->cs, R_028C94_CB_COLOR0_DCC_BASE + index * 0x3c, cb->cb_dcc_base);
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028C60_CB_COLOR0_BASE + index * 0x3c, 11);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_base);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_base >> 32);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_attrib2);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_view);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_info);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_attrib);
+ radeon_emit(cmd_buffer->cs, cb->cb_dcc_control);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_cmask);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_cmask >> 32);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_fmask);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_fmask >> 32);
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028C94_CB_COLOR0_DCC_BASE + index * 0x3c, 2);
+ radeon_emit(cmd_buffer->cs, cb->cb_dcc_base);
+ radeon_emit(cmd_buffer->cs, cb->cb_dcc_base >> 32);
+
+ radeon_set_context_reg(cmd_buffer->cs, R_0287A0_CB_MRT0_EPITCH + index * 4,
+ cb->gfx9_epitch);
+ } else {
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028C60_CB_COLOR0_BASE + index * 0x3c, 11);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_base);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_pitch);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_slice);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_view);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_info);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_attrib);
+ radeon_emit(cmd_buffer->cs, cb->cb_dcc_control);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_cmask);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_cmask_slice);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_fmask);
+ radeon_emit(cmd_buffer->cs, cb->cb_color_fmask_slice);
+
+ if (is_vi) { /* DCC BASE */
+ radeon_set_context_reg(cmd_buffer->cs, R_028C94_CB_COLOR0_DCC_BASE + index * 0x3c, cb->cb_dcc_base);
+ }
}
}
VkImageLayout layout)
{
uint32_t db_z_info = ds->db_z_info;
+ uint32_t db_stencil_info = ds->db_stencil_info;
- if (!radv_layout_has_htile(image, layout))
+ if (!radv_layout_has_htile(image, layout,
+ radv_image_queue_family_mask(image,
+ cmd_buffer->queue_family_index,
+ cmd_buffer->queue_family_index))) {
db_z_info &= C_028040_TILE_SURFACE_ENABLE;
-
- if (!radv_layout_can_expclear(image, layout))
- db_z_info &= C_028040_ALLOW_EXPCLEAR & C_028044_ALLOW_EXPCLEAR;
+ db_stencil_info |= S_028044_TILE_STENCIL_DISABLE(1);
+ }
radeon_set_context_reg(cmd_buffer->cs, R_028008_DB_DEPTH_VIEW, ds->db_depth_view);
- radeon_set_context_reg(cmd_buffer->cs, R_028014_DB_HTILE_DATA_BASE, ds->db_htile_data_base);
-
- radeon_set_context_reg_seq(cmd_buffer->cs, R_02803C_DB_DEPTH_INFO, 9);
- radeon_emit(cmd_buffer->cs, ds->db_depth_info); /* R_02803C_DB_DEPTH_INFO */
- radeon_emit(cmd_buffer->cs, db_z_info); /* R_028040_DB_Z_INFO */
- radeon_emit(cmd_buffer->cs, ds->db_stencil_info); /* R_028044_DB_STENCIL_INFO */
- radeon_emit(cmd_buffer->cs, ds->db_z_read_base); /* R_028048_DB_Z_READ_BASE */
- radeon_emit(cmd_buffer->cs, ds->db_stencil_read_base); /* R_02804C_DB_STENCIL_READ_BASE */
- radeon_emit(cmd_buffer->cs, ds->db_z_write_base); /* R_028050_DB_Z_WRITE_BASE */
- radeon_emit(cmd_buffer->cs, ds->db_stencil_write_base); /* R_028054_DB_STENCIL_WRITE_BASE */
- radeon_emit(cmd_buffer->cs, ds->db_depth_size); /* R_028058_DB_DEPTH_SIZE */
- radeon_emit(cmd_buffer->cs, ds->db_depth_slice); /* R_02805C_DB_DEPTH_SLICE */
-
radeon_set_context_reg(cmd_buffer->cs, R_028ABC_DB_HTILE_SURFACE, ds->db_htile_surface);
- radeon_set_context_reg(cmd_buffer->cs, R_028B78_PA_SU_POLY_OFFSET_DB_FMT_CNTL,
- ds->pa_su_poly_offset_db_fmt_cntl);
-}
-/*
- * To hw resolve multisample images both src and dst need to have the same
- * micro tiling mode. However we don't always know in advance when creating
- * the images. This function gets called if we have a resolve attachment,
- * and tests if the attachment image has the same tiling mode, then it
- * checks if the generated framebuffer data has the same tiling mode, and
- * updates it if not.
- */
-static void radv_set_optimal_micro_tile_mode(struct radv_device *device,
- struct radv_attachment_info *att,
- uint32_t micro_tile_mode)
-{
- struct radv_image *image = att->attachment->image;
- uint32_t tile_mode_index;
- if (image->surface.nsamples <= 1)
- return;
- if (image->surface.micro_tile_mode != micro_tile_mode) {
- radv_image_set_optimal_micro_tile_mode(device, image, micro_tile_mode);
- }
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028014_DB_HTILE_DATA_BASE, 3);
+ radeon_emit(cmd_buffer->cs, ds->db_htile_data_base);
+ radeon_emit(cmd_buffer->cs, ds->db_htile_data_base >> 32);
+ radeon_emit(cmd_buffer->cs, ds->db_depth_size);
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028038_DB_Z_INFO, 10);
+ radeon_emit(cmd_buffer->cs, db_z_info); /* DB_Z_INFO */
+ radeon_emit(cmd_buffer->cs, db_stencil_info); /* DB_STENCIL_INFO */
+ radeon_emit(cmd_buffer->cs, ds->db_z_read_base); /* DB_Z_READ_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_z_read_base >> 32); /* DB_Z_READ_BASE_HI */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_read_base); /* DB_STENCIL_READ_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_read_base >> 32); /* DB_STENCIL_READ_BASE_HI */
+ radeon_emit(cmd_buffer->cs, ds->db_z_write_base); /* DB_Z_WRITE_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_z_write_base >> 32); /* DB_Z_WRITE_BASE_HI */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_write_base); /* DB_STENCIL_WRITE_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_write_base >> 32); /* DB_STENCIL_WRITE_BASE_HI */
+
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028068_DB_Z_INFO2, 2);
+ radeon_emit(cmd_buffer->cs, ds->db_z_info2);
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_info2);
+ } else {
+ radeon_set_context_reg(cmd_buffer->cs, R_028014_DB_HTILE_DATA_BASE, ds->db_htile_data_base);
- if (att->cb.micro_tile_mode != micro_tile_mode) {
- tile_mode_index = image->surface.tiling_index[0];
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_02803C_DB_DEPTH_INFO, 9);
+ radeon_emit(cmd_buffer->cs, ds->db_depth_info); /* R_02803C_DB_DEPTH_INFO */
+ radeon_emit(cmd_buffer->cs, db_z_info); /* R_028040_DB_Z_INFO */
+ radeon_emit(cmd_buffer->cs, db_stencil_info); /* R_028044_DB_STENCIL_INFO */
+ radeon_emit(cmd_buffer->cs, ds->db_z_read_base); /* R_028048_DB_Z_READ_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_read_base); /* R_02804C_DB_STENCIL_READ_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_z_write_base); /* R_028050_DB_Z_WRITE_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_stencil_write_base); /* R_028054_DB_STENCIL_WRITE_BASE */
+ radeon_emit(cmd_buffer->cs, ds->db_depth_size); /* R_028058_DB_DEPTH_SIZE */
+ radeon_emit(cmd_buffer->cs, ds->db_depth_slice); /* R_02805C_DB_DEPTH_SLICE */
- att->cb.cb_color_attrib &= C_028C74_TILE_MODE_INDEX;
- att->cb.cb_color_attrib |= S_028C74_TILE_MODE_INDEX(tile_mode_index);
- att->cb.micro_tile_mode = micro_tile_mode;
}
+
+ radeon_set_context_reg(cmd_buffer->cs, R_028B78_PA_SU_POLY_OFFSET_DB_FMT_CNTL,
+ ds->pa_su_poly_offset_db_fmt_cntl);
}
void
radeon_emit(cmd_buffer->cs, 0);
}
+/*
+ *with DCC some colors don't require CMASK elimiation before being
+ * used as a texture. This sets a predicate value to determine if the
+ * cmask eliminate is required.
+ */
+void
+radv_set_dcc_need_cmask_elim_pred(struct radv_cmd_buffer *cmd_buffer,
+ struct radv_image *image,
+ bool value)
+{
+ uint64_t pred_val = value;
+ uint64_t va = cmd_buffer->device->ws->buffer_get_va(image->bo);
+ va += image->offset + image->dcc_pred_offset;
+
+ if (!image->surface.dcc_size)
+ return;
+
+ cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, image->bo, 8);
+
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_WRITE_DATA, 4, 0));
+ radeon_emit(cmd_buffer->cs, S_370_DST_SEL(V_370_MEM_ASYNC) |
+ S_370_WR_CONFIRM(1) |
+ S_370_ENGINE_SEL(V_370_PFP));
+ radeon_emit(cmd_buffer->cs, va);
+ radeon_emit(cmd_buffer->cs, va >> 32);
+ radeon_emit(cmd_buffer->cs, pred_val);
+ radeon_emit(cmd_buffer->cs, pred_val >> 32);
+}
+
void
radv_set_color_clear_regs(struct radv_cmd_buffer *cmd_buffer,
struct radv_image *image,
uint32_t reg = R_028C8C_CB_COLOR0_CLEAR_WORD0 + idx * 0x3c;
cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, image->bo, 8);
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_COPY_DATA, 4, 0));
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_COPY_DATA, 4, cmd_buffer->state.predicating));
radeon_emit(cmd_buffer->cs, COPY_DATA_SRC_SEL(COPY_DATA_MEM) |
COPY_DATA_DST_SEL(COPY_DATA_REG) |
COPY_DATA_COUNT_SEL);
radeon_emit(cmd_buffer->cs, reg >> 2);
radeon_emit(cmd_buffer->cs, 0);
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_PFP_SYNC_ME, 0, 0));
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_PFP_SYNC_ME, 0, cmd_buffer->state.predicating));
radeon_emit(cmd_buffer->cs, 0);
}
int i;
struct radv_framebuffer *framebuffer = cmd_buffer->state.framebuffer;
const struct radv_subpass *subpass = cmd_buffer->state.subpass;
- int dst_resolve_micro_tile_mode = -1;
- if (subpass->has_resolve) {
- uint32_t a = subpass->resolve_attachments[0].attachment;
- const struct radv_image *image = framebuffer->attachments[a].attachment->image;
- dst_resolve_micro_tile_mode = image->surface.micro_tile_mode;
- }
- for (i = 0; i < subpass->color_count; ++i) {
+ /* this may happen for inherited secondary recording */
+ if (!framebuffer)
+ return;
+
+ for (i = 0; i < 8; ++i) {
+ if (i >= subpass->color_count || subpass->color_attachments[i].attachment == VK_ATTACHMENT_UNUSED) {
+ radeon_set_context_reg(cmd_buffer->cs, R_028C70_CB_COLOR0_INFO + i * 0x3C,
+ S_028C70_FORMAT(V_028C70_COLOR_INVALID));
+ continue;
+ }
+
int idx = subpass->color_attachments[i].attachment;
struct radv_attachment_info *att = &framebuffer->attachments[idx];
- if (dst_resolve_micro_tile_mode != -1) {
- radv_set_optimal_micro_tile_mode(cmd_buffer->device,
- att, dst_resolve_micro_tile_mode);
- }
cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, att->attachment->bo, 8);
assert(att->attachment->aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT);
radv_load_color_clear_regs(cmd_buffer, att->attachment->image, i);
}
- for (i = subpass->color_count; i < 8; i++)
- radeon_set_context_reg(cmd_buffer->cs, R_028C70_CB_COLOR0_INFO + i * 0x3C,
- S_028C70_FORMAT(V_028C70_COLOR_INVALID));
-
if(subpass->depth_stencil_attachment.attachment != VK_ATTACHMENT_UNUSED) {
int idx = subpass->depth_stencil_attachment.attachment;
VkImageLayout layout = subpass->depth_stencil_attachment.layout;
struct radv_attachment_info *att = &framebuffer->attachments[idx];
struct radv_image *image = att->attachment->image;
cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, att->attachment->bo, 8);
+ MAYBE_UNUSED uint32_t queue_mask = radv_image_queue_family_mask(image,
+ cmd_buffer->queue_family_index,
+ cmd_buffer->queue_family_index);
+ /* We currently don't support writing decompressed HTILE */
+ assert(radv_layout_has_htile(image, layout, queue_mask) ==
+ radv_layout_is_htile_compressed(image, layout, queue_mask));
radv_emit_fb_ds_state(cmd_buffer, &att->ds, image, layout);
}
radv_load_depth_clear_regs(cmd_buffer, image);
} else {
- radeon_set_context_reg_seq(cmd_buffer->cs, R_028040_DB_Z_INFO, 2);
- radeon_emit(cmd_buffer->cs, S_028040_FORMAT(V_028040_Z_INVALID)); /* R_028040_DB_Z_INFO */
- radeon_emit(cmd_buffer->cs, S_028044_FORMAT(V_028044_STENCIL_INVALID)); /* R_028044_DB_STENCIL_INFO */
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9)
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028038_DB_Z_INFO, 2);
+ else
+ radeon_set_context_reg_seq(cmd_buffer->cs, R_028040_DB_Z_INFO, 2);
+
+ radeon_emit(cmd_buffer->cs, S_028040_FORMAT(V_028040_Z_INVALID)); /* DB_Z_INFO */
+ radeon_emit(cmd_buffer->cs, S_028044_FORMAT(V_028044_STENCIL_INVALID)); /* DB_STENCIL_INFO */
}
radeon_set_context_reg(cmd_buffer->cs, R_028208_PA_SC_WINDOW_SCISSOR_BR,
S_028208_BR_X(framebuffer->width) |
S_028208_BR_Y(framebuffer->height));
+
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_EVENT_WRITE, 0, 0));
+ radeon_emit(cmd_buffer->cs, EVENT_TYPE(V_028A90_BREAK_BATCH) | EVENT_INDEX(0));
+ }
}
void radv_set_db_count_control(struct radv_cmd_buffer *cmd_buffer)
static void
radv_cmd_buffer_flush_dynamic_state(struct radv_cmd_buffer *cmd_buffer)
{
- struct radv_dynamic_state *d = &cmd_buffer->state.dynamic;
+ if (G_028810_DX_RASTERIZATION_KILL(cmd_buffer->state.pipeline->graphics.raster.pa_cl_clip_cntl))
+ return;
- if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_DYNAMIC_LINE_WIDTH) {
- unsigned width = cmd_buffer->state.dynamic.line_width * 8;
- radeon_set_context_reg(cmd_buffer->cs, R_028A08_PA_SU_LINE_CNTL,
- S_028A08_WIDTH(CLAMP(width, 0, 0xFFF)));
- }
+ if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_DYNAMIC_VIEWPORT))
+ radv_emit_viewport(cmd_buffer);
- if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS) {
- radeon_set_context_reg_seq(cmd_buffer->cs, R_028414_CB_BLEND_RED, 4);
- radeon_emit_array(cmd_buffer->cs, (uint32_t*)d->blend_constants, 4);
- }
+ if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_DYNAMIC_SCISSOR | RADV_CMD_DIRTY_DYNAMIC_VIEWPORT))
+ radv_emit_scissor(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_DYNAMIC_LINE_WIDTH)
+ radv_emit_line_width(cmd_buffer);
+
+ if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_DYNAMIC_BLEND_CONSTANTS)
+ radv_emit_blend_constants(cmd_buffer);
if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE |
RADV_CMD_DIRTY_DYNAMIC_STENCIL_WRITE_MASK |
- RADV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK)) {
- radeon_set_context_reg_seq(cmd_buffer->cs, R_028430_DB_STENCILREFMASK, 2);
- radeon_emit(cmd_buffer->cs, S_028430_STENCILTESTVAL(d->stencil_reference.front) |
- S_028430_STENCILMASK(d->stencil_compare_mask.front) |
- S_028430_STENCILWRITEMASK(d->stencil_write_mask.front) |
- S_028430_STENCILOPVAL(1));
- radeon_emit(cmd_buffer->cs, S_028434_STENCILTESTVAL_BF(d->stencil_reference.back) |
- S_028434_STENCILMASK_BF(d->stencil_compare_mask.back) |
- S_028434_STENCILWRITEMASK_BF(d->stencil_write_mask.back) |
- S_028434_STENCILOPVAL_BF(1));
- }
+ RADV_CMD_DIRTY_DYNAMIC_STENCIL_COMPARE_MASK))
+ radv_emit_stencil(cmd_buffer);
if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_PIPELINE |
- RADV_CMD_DIRTY_DYNAMIC_DEPTH_BOUNDS)) {
- radeon_set_context_reg(cmd_buffer->cs, R_028020_DB_DEPTH_BOUNDS_MIN, fui(d->depth_bounds.min));
- radeon_set_context_reg(cmd_buffer->cs, R_028024_DB_DEPTH_BOUNDS_MAX, fui(d->depth_bounds.max));
- }
+ RADV_CMD_DIRTY_DYNAMIC_DEPTH_BOUNDS))
+ radv_emit_depth_bounds(cmd_buffer);
if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_PIPELINE |
- RADV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS)) {
- struct radv_raster_state *raster = &cmd_buffer->state.pipeline->graphics.raster;
- unsigned slope = fui(d->depth_bias.slope * 16.0f);
- unsigned bias = fui(d->depth_bias.bias * cmd_buffer->state.offset_scale);
-
- if (G_028814_POLY_OFFSET_FRONT_ENABLE(raster->pa_su_sc_mode_cntl)) {
- radeon_set_context_reg_seq(cmd_buffer->cs, R_028B7C_PA_SU_POLY_OFFSET_CLAMP, 5);
- radeon_emit(cmd_buffer->cs, fui(d->depth_bias.clamp)); /* CLAMP */
- radeon_emit(cmd_buffer->cs, slope); /* FRONT SCALE */
- radeon_emit(cmd_buffer->cs, bias); /* FRONT OFFSET */
- radeon_emit(cmd_buffer->cs, slope); /* BACK SCALE */
- radeon_emit(cmd_buffer->cs, bias); /* BACK OFFSET */
- }
- }
+ RADV_CMD_DIRTY_DYNAMIC_DEPTH_BIAS))
+ radv_emit_depth_biais(cmd_buffer);
cmd_buffer->state.dirty = 0;
}
gl_shader_stage stage)
{
struct ac_userdata_info *desc_set_loc = &pipeline->shaders[stage]->info.user_sgprs_locs.descriptor_sets[idx];
- uint32_t base_reg = shader_stage_to_user_data_0(stage, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(stage, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
- if (desc_set_loc->sgpr_idx == -1)
+ if (desc_set_loc->sgpr_idx == -1 || desc_set_loc->indirect)
return;
assert(!desc_set_loc->indirect);
static void
radv_emit_descriptor_set_userdata(struct radv_cmd_buffer *cmd_buffer,
- struct radv_pipeline *pipeline,
VkShaderStageFlags stages,
struct radv_descriptor_set *set,
unsigned idx)
{
- if (stages & VK_SHADER_STAGE_FRAGMENT_BIT)
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
- idx, set->va,
- MESA_SHADER_FRAGMENT);
-
- if (stages & VK_SHADER_STAGE_VERTEX_BIT)
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
- idx, set->va,
- MESA_SHADER_VERTEX);
-
- if ((stages & VK_SHADER_STAGE_GEOMETRY_BIT) && radv_pipeline_has_gs(pipeline))
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
- idx, set->va,
- MESA_SHADER_GEOMETRY);
-
- if ((stages & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) && radv_pipeline_has_tess(pipeline))
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
- idx, set->va,
- MESA_SHADER_TESS_CTRL);
-
- if ((stages & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) && radv_pipeline_has_tess(pipeline))
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
- idx, set->va,
- MESA_SHADER_TESS_EVAL);
+ if (cmd_buffer->state.pipeline) {
+ radv_foreach_stage(stage, stages) {
+ if (cmd_buffer->state.pipeline->shaders[stage])
+ emit_stage_descriptor_set_userdata(cmd_buffer, cmd_buffer->state.pipeline,
+ idx, set->va,
+ stage);
+ }
+ }
- if (stages & VK_SHADER_STAGE_COMPUTE_BIT)
- emit_stage_descriptor_set_userdata(cmd_buffer, pipeline,
+ if (cmd_buffer->state.compute_pipeline && (stages & VK_SHADER_STAGE_COMPUTE_BIT))
+ emit_stage_descriptor_set_userdata(cmd_buffer, cmd_buffer->state.compute_pipeline,
idx, set->va,
MESA_SHADER_COMPUTE);
}
radv_flush_push_descriptors(struct radv_cmd_buffer *cmd_buffer)
{
struct radv_descriptor_set *set = &cmd_buffer->push_descriptors.set;
- uint32_t *ptr = NULL;
unsigned bo_offset;
- if (!radv_cmd_buffer_upload_alloc(cmd_buffer, set->size, 32,
- &bo_offset,
- (void**) &ptr))
+ if (!radv_cmd_buffer_upload_data(cmd_buffer, set->size, 32,
+ set->mapped_ptr,
+ &bo_offset))
return;
set->va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->upload.upload_bo);
set->va += bo_offset;
+}
- memcpy(ptr, set->mapped_ptr, set->size);
+static void
+radv_flush_indirect_descriptor_sets(struct radv_cmd_buffer *cmd_buffer)
+{
+ uint32_t size = MAX_SETS * 2 * 4;
+ uint32_t offset;
+ void *ptr;
+
+ if (!radv_cmd_buffer_upload_alloc(cmd_buffer, size,
+ 256, &offset, &ptr))
+ return;
+
+ for (unsigned i = 0; i < MAX_SETS; i++) {
+ uint32_t *uptr = ((uint32_t *)ptr) + i * 2;
+ uint64_t set_va = 0;
+ struct radv_descriptor_set *set = cmd_buffer->state.descriptors[i];
+ if (set)
+ set_va = set->va;
+ uptr[0] = set_va & 0xffffffff;
+ uptr[1] = set_va >> 32;
+ }
+
+ uint64_t va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->upload.upload_bo);
+ va += offset;
+
+ if (cmd_buffer->state.pipeline) {
+ if (cmd_buffer->state.pipeline->shaders[MESA_SHADER_VERTEX])
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_VERTEX,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
+
+ if (cmd_buffer->state.pipeline->shaders[MESA_SHADER_FRAGMENT])
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_FRAGMENT,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
+
+ if (radv_pipeline_has_gs(cmd_buffer->state.pipeline))
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_GEOMETRY,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
+
+ if (radv_pipeline_has_tess(cmd_buffer->state.pipeline))
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_TESS_CTRL,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
+
+ if (radv_pipeline_has_tess(cmd_buffer->state.pipeline))
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_TESS_EVAL,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
+ }
+
+ if (cmd_buffer->state.compute_pipeline)
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.compute_pipeline, MESA_SHADER_COMPUTE,
+ AC_UD_INDIRECT_DESCRIPTOR_SETS, va);
}
static void
radv_flush_descriptors(struct radv_cmd_buffer *cmd_buffer,
- struct radv_pipeline *pipeline,
VkShaderStageFlags stages)
{
unsigned i;
+
if (!cmd_buffer->state.descriptors_dirty)
return;
if (cmd_buffer->state.push_descriptors_dirty)
radv_flush_push_descriptors(cmd_buffer);
+ if ((cmd_buffer->state.pipeline && cmd_buffer->state.pipeline->need_indirect_descriptor_sets) ||
+ (cmd_buffer->state.compute_pipeline && cmd_buffer->state.compute_pipeline->need_indirect_descriptor_sets)) {
+ radv_flush_indirect_descriptor_sets(cmd_buffer);
+ }
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
+ cmd_buffer->cs,
+ MAX_SETS * MESA_SHADER_STAGES * 4);
+
for (i = 0; i < MAX_SETS; i++) {
- if (!(cmd_buffer->state.descriptors_dirty & (1 << i)))
+ if (!(cmd_buffer->state.descriptors_dirty & (1u << i)))
continue;
struct radv_descriptor_set *set = cmd_buffer->state.descriptors[i];
if (!set)
continue;
- radv_emit_descriptor_set_userdata(cmd_buffer, pipeline, stages, set, i);
+ radv_emit_descriptor_set_userdata(cmd_buffer, stages, set, i);
}
cmd_buffer->state.descriptors_dirty = 0;
cmd_buffer->state.push_descriptors_dirty = false;
+
+ radv_save_descriptors(cmd_buffer);
+
+ assert(cmd_buffer->cs->cdw <= cdw_max);
}
static void
va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->upload.upload_bo);
va += offset;
- if (stages & VK_SHADER_STAGE_VERTEX_BIT)
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_VERTEX,
- AC_UD_PUSH_CONSTANTS, va);
-
- if (stages & VK_SHADER_STAGE_FRAGMENT_BIT)
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_FRAGMENT,
- AC_UD_PUSH_CONSTANTS, va);
-
- if ((stages & VK_SHADER_STAGE_GEOMETRY_BIT) && radv_pipeline_has_gs(pipeline))
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_GEOMETRY,
- AC_UD_PUSH_CONSTANTS, va);
-
- if ((stages & VK_SHADER_STAGE_TESSELLATION_CONTROL_BIT) && radv_pipeline_has_tess(pipeline))
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_TESS_CTRL,
- AC_UD_PUSH_CONSTANTS, va);
-
- if ((stages & VK_SHADER_STAGE_TESSELLATION_EVALUATION_BIT) && radv_pipeline_has_tess(pipeline))
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_TESS_EVAL,
- AC_UD_PUSH_CONSTANTS, va);
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
+ cmd_buffer->cs, MESA_SHADER_STAGES * 4);
- if (stages & VK_SHADER_STAGE_COMPUTE_BIT)
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_COMPUTE,
- AC_UD_PUSH_CONSTANTS, va);
+ radv_foreach_stage(stage, stages) {
+ if (pipeline->shaders[stage]) {
+ radv_emit_userdata_address(cmd_buffer, pipeline, stage,
+ AC_UD_PUSH_CONSTANTS, va);
+ }
+ }
cmd_buffer->push_constant_stages &= ~stages;
+ assert(cmd_buffer->cs->cdw <= cdw_max);
}
static void radv_emit_primitive_reset_state(struct radv_cmd_buffer *cmd_buffer,
if (primitive_reset_en != cmd_buffer->state.last_primitive_reset_en) {
cmd_buffer->state.last_primitive_reset_en = primitive_reset_en;
- radeon_set_context_reg(cmd_buffer->cs, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN,
- primitive_reset_en);
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_set_uconfig_reg(cmd_buffer->cs, R_03092C_VGT_MULTI_PRIM_IB_RESET_EN,
+ primitive_reset_en);
+ } else {
+ radeon_set_context_reg(cmd_buffer->cs, R_028A94_VGT_MULTI_PRIM_IB_RESET_EN,
+ primitive_reset_en);
+ }
}
if (primitive_reset_en) {
}
}
-static void
-radv_cmd_buffer_flush_state(struct radv_cmd_buffer *cmd_buffer,
- bool indexed_draw, bool instanced_draw,
- bool indirect_draw,
- uint32_t draw_vertex_count)
+static bool
+radv_cmd_buffer_update_vertex_descriptors(struct radv_cmd_buffer *cmd_buffer)
{
- struct radv_pipeline *pipeline = cmd_buffer->state.pipeline;
struct radv_device *device = cmd_buffer->device;
- uint32_t ia_multi_vgt_param;
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
- cmd_buffer->cs, 4096);
-
- if ((cmd_buffer->state.vertex_descriptors_dirty || cmd_buffer->state.vb_dirty) &&
- cmd_buffer->state.pipeline->num_vertex_attribs) {
+ if ((cmd_buffer->state.pipeline != cmd_buffer->state.emitted_pipeline || cmd_buffer->state.vb_dirty) &&
+ cmd_buffer->state.pipeline->vertex_elements.count &&
+ cmd_buffer->state.pipeline->shaders[MESA_SHADER_VERTEX]->info.info.vs.has_vertex_buffers) {
+ struct radv_vertex_elements_info *velems = &cmd_buffer->state.pipeline->vertex_elements;
unsigned vb_offset;
void *vb_ptr;
uint32_t i = 0;
- uint32_t num_attribs = cmd_buffer->state.pipeline->num_vertex_attribs;
+ uint32_t count = velems->count;
uint64_t va;
/* allocate some descriptor state for vertex buffers */
- radv_cmd_buffer_upload_alloc(cmd_buffer, num_attribs * 16, 256,
- &vb_offset, &vb_ptr);
+ if (!radv_cmd_buffer_upload_alloc(cmd_buffer, count * 16, 256,
+ &vb_offset, &vb_ptr))
+ return false;
- for (i = 0; i < num_attribs; i++) {
+ for (i = 0; i < count; i++) {
uint32_t *desc = &((uint32_t *)vb_ptr)[i * 4];
uint32_t offset;
- int vb = cmd_buffer->state.pipeline->va_binding[i];
+ int vb = velems->binding[i];
struct radv_buffer *buffer = cmd_buffer->state.vertex_bindings[vb].buffer;
uint32_t stride = cmd_buffer->state.pipeline->binding_stride[vb];
device->ws->cs_add_buffer(cmd_buffer->cs, buffer->bo, 8);
va = device->ws->buffer_get_va(buffer->bo);
- offset = cmd_buffer->state.vertex_bindings[vb].offset + cmd_buffer->state.pipeline->va_offset[i];
+ offset = cmd_buffer->state.vertex_bindings[vb].offset + velems->offset[i];
va += offset + buffer->offset;
desc[0] = va;
desc[1] = S_008F04_BASE_ADDRESS_HI(va >> 32) | S_008F04_STRIDE(stride);
if (cmd_buffer->device->physical_device->rad_info.chip_class <= CIK && stride)
- desc[2] = (buffer->size - offset - cmd_buffer->state.pipeline->va_format_size[i]) / stride + 1;
+ desc[2] = (buffer->size - offset - velems->format_size[i]) / stride + 1;
else
desc[2] = buffer->size - offset;
- desc[3] = cmd_buffer->state.pipeline->va_rsrc_word3[i];
+ desc[3] = velems->rsrc_word3[i];
}
va = device->ws->buffer_get_va(cmd_buffer->upload.upload_bo);
va += vb_offset;
- radv_emit_userdata_address(cmd_buffer, pipeline, MESA_SHADER_VERTEX,
+ radv_emit_userdata_address(cmd_buffer, cmd_buffer->state.pipeline, MESA_SHADER_VERTEX,
AC_UD_VS_VERTEX_BUFFERS, va);
}
+ cmd_buffer->state.vb_dirty = false;
+
+ return true;
+}
+
+static void
+radv_cmd_buffer_flush_state(struct radv_cmd_buffer *cmd_buffer,
+ bool indexed_draw, bool instanced_draw,
+ bool indirect_draw,
+ uint32_t draw_vertex_count)
+{
+ uint32_t ia_multi_vgt_param;
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
+ cmd_buffer->cs, 4096);
+
+ if (!radv_cmd_buffer_update_vertex_descriptors(cmd_buffer))
+ return;
- cmd_buffer->state.vertex_descriptors_dirty = false;
- cmd_buffer->state.vb_dirty = 0;
if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_PIPELINE)
- radv_emit_graphics_pipeline(cmd_buffer, pipeline);
+ radv_emit_graphics_pipeline(cmd_buffer);
if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_RENDER_TARGETS)
radv_emit_framebuffer_state(cmd_buffer);
- if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_DYNAMIC_VIEWPORT))
- radv_emit_viewport(cmd_buffer);
-
- if (cmd_buffer->state.dirty & (RADV_CMD_DIRTY_DYNAMIC_SCISSOR | RADV_CMD_DIRTY_DYNAMIC_VIEWPORT))
- radv_emit_scissor(cmd_buffer);
-
ia_multi_vgt_param = si_get_ia_multi_vgt_param(cmd_buffer, instanced_draw, indirect_draw, draw_vertex_count);
if (cmd_buffer->state.last_ia_multi_vgt_param != ia_multi_vgt_param) {
- if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK)
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9)
+ radeon_set_uconfig_reg_idx(cmd_buffer->cs, R_030960_IA_MULTI_VGT_PARAM, 4, ia_multi_vgt_param);
+ else if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK)
radeon_set_context_reg_idx(cmd_buffer->cs, R_028AA8_IA_MULTI_VGT_PARAM, 1, ia_multi_vgt_param);
else
radeon_set_context_reg(cmd_buffer->cs, R_028AA8_IA_MULTI_VGT_PARAM, ia_multi_vgt_param);
cmd_buffer->state.last_ia_multi_vgt_param = ia_multi_vgt_param;
}
- if (cmd_buffer->state.dirty & RADV_CMD_DIRTY_PIPELINE) {
- radeon_set_context_reg(cmd_buffer->cs, R_028B54_VGT_SHADER_STAGES_EN, pipeline->graphics.vgt_shader_stages_en);
-
- if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK) {
- radeon_set_uconfig_reg_idx(cmd_buffer->cs, R_030908_VGT_PRIMITIVE_TYPE, 1, cmd_buffer->state.pipeline->graphics.prim);
- } else {
- radeon_set_config_reg(cmd_buffer->cs, R_008958_VGT_PRIMITIVE_TYPE, cmd_buffer->state.pipeline->graphics.prim);
- }
- radeon_set_context_reg(cmd_buffer->cs, R_028A6C_VGT_GS_OUT_PRIM_TYPE, cmd_buffer->state.pipeline->graphics.gs_out);
- }
-
radv_cmd_buffer_flush_dynamic_state(cmd_buffer);
radv_emit_primitive_reset_state(cmd_buffer, indexed_draw);
- radv_flush_descriptors(cmd_buffer, cmd_buffer->state.pipeline,
- VK_SHADER_STAGE_ALL_GRAPHICS);
+ radv_flush_descriptors(cmd_buffer, VK_SHADER_STAGE_ALL_GRAPHICS);
radv_flush_constants(cmd_buffer, cmd_buffer->state.pipeline,
VK_SHADER_STAGE_ALL_GRAPHICS);
radv_subpass_barrier(cmd_buffer, &subpass->start_barrier);
for (unsigned i = 0; i < subpass->color_count; ++i) {
- radv_handle_subpass_image_transition(cmd_buffer,
- subpass->color_attachments[i]);
+ if (subpass->color_attachments[i].attachment != VK_ATTACHMENT_UNUSED)
+ radv_handle_subpass_image_transition(cmd_buffer,
+ subpass->color_attachments[i]);
}
for (unsigned i = 0; i < subpass->input_count; ++i) {
cmd_buffer->state.dirty |= RADV_CMD_DIRTY_RENDER_TARGETS;
}
-static void
+static VkResult
radv_cmd_state_setup_attachments(struct radv_cmd_buffer *cmd_buffer,
struct radv_render_pass *pass,
const VkRenderPassBeginInfo *info)
if (pass->attachment_count == 0) {
state->attachments = NULL;
- return;
+ return VK_SUCCESS;
}
state->attachments = vk_alloc(&cmd_buffer->pool->alloc,
sizeof(state->attachments[0]),
8, VK_SYSTEM_ALLOCATION_SCOPE_OBJECT);
if (state->attachments == NULL) {
- /* FIXME: Propagate VK_ERROR_OUT_OF_HOST_MEMORY to vkEndCommandBuffer */
- abort();
+ cmd_buffer->record_result = VK_ERROR_OUT_OF_HOST_MEMORY;
+ return cmd_buffer->record_result;
}
for (uint32_t i = 0; i < pass->attachment_count; ++i) {
if ((att_aspects & VK_IMAGE_ASPECT_DEPTH_BIT) &&
att->load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) {
clear_aspects |= VK_IMAGE_ASPECT_DEPTH_BIT;
+ if ((att_aspects & VK_IMAGE_ASPECT_STENCIL_BIT) &&
+ att->stencil_load_op == VK_ATTACHMENT_LOAD_OP_DONT_CARE)
+ clear_aspects |= VK_IMAGE_ASPECT_STENCIL_BIT;
}
if ((att_aspects & VK_IMAGE_ASPECT_STENCIL_BIT) &&
att->stencil_load_op == VK_ATTACHMENT_LOAD_OP_CLEAR) {
}
state->attachments[i].pending_clear_aspects = clear_aspects;
+ state->attachments[i].cleared_views = 0;
if (clear_aspects && info) {
assert(info->clearValueCount > i);
state->attachments[i].clear_value = info->pClearValues[i];
state->attachments[i].current_layout = att->initial_layout;
}
+
+ return VK_SUCCESS;
}
VkResult radv_AllocateCommandBuffers(
list_del(&cmd_buffer->pool_link);
list_addtail(&cmd_buffer->pool_link, &pool->cmd_buffers);
- radv_reset_cmd_buffer(cmd_buffer);
+ result = radv_reset_cmd_buffer(cmd_buffer);
cmd_buffer->_loader_data.loaderMagic = ICD_LOADER_MAGIC;
cmd_buffer->level = pAllocateInfo->level;
pCommandBuffers[i] = radv_cmd_buffer_to_handle(cmd_buffer);
- result = VK_SUCCESS;
} else {
result = radv_create_cmd_buffer(device, pool, pAllocateInfo->level,
&pCommandBuffers[i]);
VkCommandBufferResetFlags flags)
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
- radv_reset_cmd_buffer(cmd_buffer);
- return VK_SUCCESS;
+ return radv_reset_cmd_buffer(cmd_buffer);
}
static void emit_gfx_buffer_state(struct radv_cmd_buffer *cmd_buffer)
device->ws->cs_add_buffer(cmd_buffer->cs, device->gfx_init, 8);
radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDIRECT_BUFFER_CIK, 2, 0));
radeon_emit(cmd_buffer->cs, va);
- radeon_emit(cmd_buffer->cs, (va >> 32) & 0xffff);
+ radeon_emit(cmd_buffer->cs, va >> 32);
radeon_emit(cmd_buffer->cs, device->gfx_init_size_dw & 0xffff);
} else
si_init_config(cmd_buffer);
const VkCommandBufferBeginInfo *pBeginInfo)
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
- radv_reset_cmd_buffer(cmd_buffer);
+ VkResult result;
+
+ result = radv_reset_cmd_buffer(cmd_buffer);
+ if (result != VK_SUCCESS)
+ return result;
memset(&cmd_buffer->state, 0, sizeof(cmd_buffer->state));
cmd_buffer->state.last_primitive_reset_en = -1;
+ cmd_buffer->usage_flags = pBeginInfo->flags;
/* setup initial configuration into command buffer */
if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY) {
struct radv_subpass *subpass =
&cmd_buffer->state.pass->subpasses[pBeginInfo->pInheritanceInfo->subpass];
- radv_cmd_state_setup_attachments(cmd_buffer, cmd_buffer->state.pass, NULL);
+ result = radv_cmd_state_setup_attachments(cmd_buffer, cmd_buffer->state.pass, NULL);
+ if (result != VK_SUCCESS)
+ return result;
+
radv_cmd_buffer_set_subpass(cmd_buffer, subpass, false);
}
radv_cmd_buffer_trace_emit(cmd_buffer);
- return VK_SUCCESS;
+ return result;
}
void radv_CmdBindVertexBuffers(
/* We have to defer setting up vertex buffer since we need the buffer
* stride from the pipeline. */
- assert(firstBinding + bindingCount < MAX_VBS);
+ assert(firstBinding + bindingCount <= MAX_VBS);
for (uint32_t i = 0; i < bindingCount; i++) {
vb[firstBinding + i].buffer = radv_buffer_from_handle(pBuffers[i]);
vb[firstBinding + i].offset = pOffsets[i];
- cmd_buffer->state.vb_dirty |= 1 << (firstBinding + i);
}
+
+ cmd_buffer->state.vb_dirty = true;
}
void radv_CmdBindIndexBuffer(
VkIndexType indexType)
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
+ RADV_FROM_HANDLE(radv_buffer, index_buffer, buffer);
- cmd_buffer->state.index_buffer = radv_buffer_from_handle(buffer);
- cmd_buffer->state.index_offset = offset;
cmd_buffer->state.index_type = indexType; /* vk matches hw */
+ cmd_buffer->state.index_va = cmd_buffer->device->ws->buffer_get_va(index_buffer->bo);
+ cmd_buffer->state.index_va += index_buffer->offset + offset;
+
+ int index_size_shift = cmd_buffer->state.index_type ? 2 : 1;
+ cmd_buffer->state.max_index_count = (index_buffer->size - offset) >> index_size_shift;
cmd_buffer->state.dirty |= RADV_CMD_DIRTY_INDEX_BUFFER;
- cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, cmd_buffer->state.index_buffer->bo, 8);
+ cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, index_buffer->bo, 8);
}
struct radeon_winsys *ws = cmd_buffer->device->ws;
cmd_buffer->state.descriptors[idx] = set;
- cmd_buffer->state.descriptors_dirty |= (1 << idx);
+ cmd_buffer->state.descriptors_dirty |= (1u << idx);
if (!set)
return;
+ assert(!(set->layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR));
+
for (unsigned j = 0; j < set->layout->buffer_count; ++j)
if (set->descriptors[j])
ws->cs_add_buffer(cmd_buffer->cs, set->descriptors[j], 7);
if (!set->mapped_ptr) {
cmd_buffer->push_descriptors.capacity = 0;
- cmd_buffer->record_fail = true;
+ cmd_buffer->record_result = VK_ERROR_OUT_OF_HOST_MEMORY;
return false;
}
return true;
}
+void radv_meta_push_descriptor_set(
+ struct radv_cmd_buffer* cmd_buffer,
+ VkPipelineBindPoint pipelineBindPoint,
+ VkPipelineLayout _layout,
+ uint32_t set,
+ uint32_t descriptorWriteCount,
+ const VkWriteDescriptorSet* pDescriptorWrites)
+{
+ RADV_FROM_HANDLE(radv_pipeline_layout, layout, _layout);
+ struct radv_descriptor_set *push_set = &cmd_buffer->meta_push_descriptors;
+ unsigned bo_offset;
+
+ assert(set == 0);
+ assert(layout->set[set].layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR);
+
+ push_set->size = layout->set[set].layout->size;
+ push_set->layout = layout->set[set].layout;
+
+ if (!radv_cmd_buffer_upload_alloc(cmd_buffer, push_set->size, 32,
+ &bo_offset,
+ (void**) &push_set->mapped_ptr))
+ return;
+
+ push_set->va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->upload.upload_bo);
+ push_set->va += bo_offset;
+
+ radv_update_descriptor_sets(cmd_buffer->device, cmd_buffer,
+ radv_descriptor_set_to_handle(push_set),
+ descriptorWriteCount, pDescriptorWrites, 0, NULL);
+
+ cmd_buffer->state.descriptors[set] = push_set;
+ cmd_buffer->state.descriptors_dirty |= (1u << set);
+}
+
void radv_CmdPushDescriptorSetKHR(
VkCommandBuffer commandBuffer,
VkPipelineBindPoint pipelineBindPoint,
descriptorWriteCount, pDescriptorWrites, 0, NULL);
cmd_buffer->state.descriptors[set] = push_set;
- cmd_buffer->state.descriptors_dirty |= (1 << set);
+ cmd_buffer->state.descriptors_dirty |= (1u << set);
cmd_buffer->state.push_descriptors_dirty = true;
}
descriptorUpdateTemplate, pData);
cmd_buffer->state.descriptors[set] = push_set;
- cmd_buffer->state.descriptors_dirty |= (1 << set);
+ cmd_buffer->state.descriptors_dirty |= (1u << set);
cmd_buffer->state.push_descriptors_dirty = true;
}
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
- if (cmd_buffer->queue_family_index != RADV_QUEUE_TRANSFER)
+ if (cmd_buffer->queue_family_index != RADV_QUEUE_TRANSFER) {
+ if (cmd_buffer->device->physical_device->rad_info.chip_class == SI)
+ cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_CS_PARTIAL_FLUSH | RADV_CMD_FLAG_PS_PARTIAL_FLUSH | RADV_CMD_FLAG_WRITEBACK_GLOBAL_L2;
si_emit_cache_flush(cmd_buffer);
+ }
- if (!cmd_buffer->device->ws->cs_finalize(cmd_buffer->cs) ||
- cmd_buffer->record_fail)
+ if (!cmd_buffer->device->ws->cs_finalize(cmd_buffer->cs))
return VK_ERROR_OUT_OF_DEVICE_MEMORY;
- return VK_SUCCESS;
+
+ return cmd_buffer->record_result;
}
static void
cmd_buffer->state.emitted_compute_pipeline = pipeline;
compute_shader = pipeline->shaders[MESA_SHADER_COMPUTE];
- va = ws->buffer_get_va(compute_shader->bo);
+ va = ws->buffer_get_va(compute_shader->bo) + compute_shader->bo_offset;
ws->cs_add_buffer(cmd_buffer->cs, compute_shader->bo, 8);
+ radv_emit_prefetch(cmd_buffer, va, compute_shader->code_size);
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
cmd_buffer->cs, 16);
S_00B81C_NUM_THREAD_FULL(compute_shader->info.cs.block_size[2]));
assert(cmd_buffer->cs->cdw <= cdw_max);
+ radv_save_pipeline(cmd_buffer, pipeline, RING_COMPUTE);
}
+static void radv_mark_descriptor_sets_dirty(struct radv_cmd_buffer *cmd_buffer)
+{
+ for (unsigned i = 0; i < MAX_SETS; i++) {
+ if (cmd_buffer->state.descriptors[i])
+ cmd_buffer->state.descriptors_dirty |= (1u << i);
+ }
+}
void radv_CmdBindPipeline(
VkCommandBuffer commandBuffer,
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
RADV_FROM_HANDLE(radv_pipeline, pipeline, _pipeline);
- for (unsigned i = 0; i < MAX_SETS; i++) {
- if (cmd_buffer->state.descriptors[i])
- cmd_buffer->state.descriptors_dirty |= (1 << i);
- }
+ radv_mark_descriptor_sets_dirty(cmd_buffer);
switch (pipelineBindPoint) {
case VK_PIPELINE_BIND_POINT_COMPUTE:
break;
case VK_PIPELINE_BIND_POINT_GRAPHICS:
cmd_buffer->state.pipeline = pipeline;
- cmd_buffer->state.vertex_descriptors_dirty = true;
+ if (!pipeline)
+ break;
+
cmd_buffer->state.dirty |= RADV_CMD_DIRTY_PIPELINE;
cmd_buffer->push_constant_stages |= pipeline->active_stages;
cmd_buffer->state.dirty |= RADV_CMD_DIRTY_DYNAMIC_STENCIL_REFERENCE;
}
-
void radv_CmdExecuteCommands(
VkCommandBuffer commandBuffer,
uint32_t commandBufferCount,
primary->state.dirty |= RADV_CMD_DIRTY_DYNAMIC_ALL;
primary->state.last_primitive_reset_en = -1;
primary->state.last_primitive_reset_index = 0;
+ radv_mark_descriptor_sets_dirty(primary);
}
}
VkCommandPoolResetFlags flags)
{
RADV_FROM_HANDLE(radv_cmd_pool, pool, commandPool);
+ VkResult result;
list_for_each_entry(struct radv_cmd_buffer, cmd_buffer,
&pool->cmd_buffers, pool_link) {
- radv_reset_cmd_buffer(cmd_buffer);
+ result = radv_reset_cmd_buffer(cmd_buffer);
+ if (result != VK_SUCCESS)
+ return result;
}
return VK_SUCCESS;
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
cmd_buffer->cs, 2048);
+ MAYBE_UNUSED VkResult result;
cmd_buffer->state.framebuffer = framebuffer;
cmd_buffer->state.pass = pass;
cmd_buffer->state.render_area = pRenderPassBegin->renderArea;
- radv_cmd_state_setup_attachments(cmd_buffer, pass, pRenderPassBegin);
+
+ result = radv_cmd_state_setup_attachments(cmd_buffer, pass, pRenderPassBegin);
+ if (result != VK_SUCCESS)
+ return;
radv_cmd_buffer_set_subpass(cmd_buffer, pass->subpasses, true);
assert(cmd_buffer->cs->cdw <= cdw_max);
radv_cmd_buffer_clear_subpass(cmd_buffer);
}
+static void radv_emit_view_index(struct radv_cmd_buffer *cmd_buffer, unsigned index)
+{
+ struct radv_pipeline *pipeline = cmd_buffer->state.pipeline;
+ for (unsigned stage = 0; stage < MESA_SHADER_STAGES; ++stage) {
+ if (!pipeline->shaders[stage])
+ continue;
+ struct ac_userdata_info *loc = radv_lookup_user_sgpr(pipeline, stage, AC_UD_VIEW_INDEX);
+ if (loc->sgpr_idx == -1)
+ continue;
+ uint32_t base_reg = radv_shader_stage_to_user_data_0(stage, radv_pipeline_has_gs(pipeline), radv_pipeline_has_tess(pipeline));
+ radeon_set_sh_reg(cmd_buffer->cs, base_reg + loc->sgpr_idx * 4, index);
+
+ }
+ if (pipeline->gs_copy_shader) {
+ struct ac_userdata_info *loc = &pipeline->gs_copy_shader->info.user_sgprs_locs.shader_data[AC_UD_VIEW_INDEX];
+ if (loc->sgpr_idx != -1) {
+ uint32_t base_reg = R_00B130_SPI_SHADER_USER_DATA_VS_0;
+ radeon_set_sh_reg(cmd_buffer->cs, base_reg + loc->sgpr_idx * 4, index);
+ }
+ }
+}
+
+static void
+radv_cs_emit_draw_packet(struct radv_cmd_buffer *cmd_buffer,
+ uint32_t vertex_count)
+{
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_DRAW_INDEX_AUTO, 1, cmd_buffer->state.predicating));
+ radeon_emit(cmd_buffer->cs, vertex_count);
+ radeon_emit(cmd_buffer->cs, V_0287F0_DI_SRC_SEL_AUTO_INDEX |
+ S_0287F0_USE_OPAQUE(0));
+}
+
void radv_CmdDraw(
VkCommandBuffer commandBuffer,
uint32_t vertexCount,
radv_cmd_buffer_flush_state(cmd_buffer, false, (instanceCount > 1), false, vertexCount);
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 10);
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 20 * MAX_VIEWS);
- struct ac_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.pipeline, MESA_SHADER_VERTEX,
- AC_UD_VS_BASE_VERTEX_START_INSTANCE);
- if (loc->sgpr_idx != -1) {
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_VERTEX, radv_pipeline_has_gs(cmd_buffer->state.pipeline),
- radv_pipeline_has_tess(cmd_buffer->state.pipeline));
- radeon_set_sh_reg_seq(cmd_buffer->cs, base_reg + loc->sgpr_idx * 4, 3);
- radeon_emit(cmd_buffer->cs, firstVertex);
- radeon_emit(cmd_buffer->cs, firstInstance);
+ assert(cmd_buffer->state.pipeline->graphics.vtx_base_sgpr);
+ radeon_set_sh_reg_seq(cmd_buffer->cs, cmd_buffer->state.pipeline->graphics.vtx_base_sgpr,
+ cmd_buffer->state.pipeline->graphics.vtx_emit_num);
+ radeon_emit(cmd_buffer->cs, firstVertex);
+ radeon_emit(cmd_buffer->cs, firstInstance);
+ if (cmd_buffer->state.pipeline->graphics.vtx_emit_num == 3)
radeon_emit(cmd_buffer->cs, 0);
- }
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_NUM_INSTANCES, 0, 0));
+
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_NUM_INSTANCES, 0, cmd_buffer->state.predicating));
radeon_emit(cmd_buffer->cs, instanceCount);
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_DRAW_INDEX_AUTO, 1, 0));
- radeon_emit(cmd_buffer->cs, vertexCount);
- radeon_emit(cmd_buffer->cs, V_0287F0_DI_SRC_SEL_AUTO_INDEX |
- S_0287F0_USE_OPAQUE(0));
+ if (!cmd_buffer->state.subpass->view_mask) {
+ radv_cs_emit_draw_packet(cmd_buffer, vertexCount);
+ } else {
+ unsigned i;
+ for_each_bit(i, cmd_buffer->state.subpass->view_mask) {
+ radv_emit_view_index(cmd_buffer, i);
+
+ radv_cs_emit_draw_packet(cmd_buffer, vertexCount);
+ }
+ }
assert(cmd_buffer->cs->cdw <= cdw_max);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ radv_cmd_buffer_after_draw(cmd_buffer);
+}
+
+
+static void
+radv_cs_emit_draw_indexed_packet(struct radv_cmd_buffer *cmd_buffer,
+ uint64_t index_va,
+ uint32_t index_count)
+{
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_DRAW_INDEX_2, 4, false));
+ radeon_emit(cmd_buffer->cs, cmd_buffer->state.max_index_count);
+ radeon_emit(cmd_buffer->cs, index_va);
+ radeon_emit(cmd_buffer->cs, (index_va >> 32UL) & 0xFF);
+ radeon_emit(cmd_buffer->cs, index_count);
+ radeon_emit(cmd_buffer->cs, V_0287F0_DI_SRC_SEL_DMA);
}
void radv_CmdDrawIndexed(
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
int index_size = cmd_buffer->state.index_type ? 4 : 2;
- uint32_t index_max_size = (cmd_buffer->state.index_buffer->size - cmd_buffer->state.index_offset) / index_size;
uint64_t index_va;
radv_cmd_buffer_flush_state(cmd_buffer, true, (instanceCount > 1), false, indexCount);
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 15);
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 26 * MAX_VIEWS);
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDEX_TYPE, 0, 0));
- radeon_emit(cmd_buffer->cs, cmd_buffer->state.index_type);
+ if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ radeon_set_uconfig_reg_idx(cmd_buffer->cs, R_03090C_VGT_INDEX_TYPE,
+ 2, cmd_buffer->state.index_type);
+ } else {
+ radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDEX_TYPE, 0, 0));
+ radeon_emit(cmd_buffer->cs, cmd_buffer->state.index_type);
+ }
- struct ac_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.pipeline, MESA_SHADER_VERTEX,
- AC_UD_VS_BASE_VERTEX_START_INSTANCE);
- if (loc->sgpr_idx != -1) {
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_VERTEX, radv_pipeline_has_gs(cmd_buffer->state.pipeline),
- radv_pipeline_has_tess(cmd_buffer->state.pipeline));
- radeon_set_sh_reg_seq(cmd_buffer->cs, base_reg + loc->sgpr_idx * 4, 3);
- radeon_emit(cmd_buffer->cs, vertexOffset);
- radeon_emit(cmd_buffer->cs, firstInstance);
+ assert(cmd_buffer->state.pipeline->graphics.vtx_base_sgpr);
+ radeon_set_sh_reg_seq(cmd_buffer->cs, cmd_buffer->state.pipeline->graphics.vtx_base_sgpr,
+ cmd_buffer->state.pipeline->graphics.vtx_emit_num);
+ radeon_emit(cmd_buffer->cs, vertexOffset);
+ radeon_emit(cmd_buffer->cs, firstInstance);
+ if (cmd_buffer->state.pipeline->graphics.vtx_emit_num == 3)
radeon_emit(cmd_buffer->cs, 0);
- }
+
radeon_emit(cmd_buffer->cs, PKT3(PKT3_NUM_INSTANCES, 0, 0));
radeon_emit(cmd_buffer->cs, instanceCount);
- index_va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->state.index_buffer->bo);
- index_va += firstIndex * index_size + cmd_buffer->state.index_buffer->offset + cmd_buffer->state.index_offset;
- radeon_emit(cmd_buffer->cs, PKT3(PKT3_DRAW_INDEX_2, 4, false));
- radeon_emit(cmd_buffer->cs, index_max_size);
- radeon_emit(cmd_buffer->cs, index_va);
- radeon_emit(cmd_buffer->cs, (index_va >> 32UL) & 0xFF);
- radeon_emit(cmd_buffer->cs, indexCount);
- radeon_emit(cmd_buffer->cs, V_0287F0_DI_SRC_SEL_DMA);
+ index_va = cmd_buffer->state.index_va;
+ index_va += firstIndex * index_size;
+ if (!cmd_buffer->state.subpass->view_mask) {
+ radv_cs_emit_draw_indexed_packet(cmd_buffer, index_va, indexCount);
+ } else {
+ unsigned i;
+ for_each_bit(i, cmd_buffer->state.subpass->view_mask) {
+ radv_emit_view_index(cmd_buffer, i);
+
+ radv_cs_emit_draw_indexed_packet(cmd_buffer, index_va, indexCount);
+ }
+ }
assert(cmd_buffer->cs->cdw <= cdw_max);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ radv_cmd_buffer_after_draw(cmd_buffer);
+}
+
+static void
+radv_cs_emit_indirect_draw_packet(struct radv_cmd_buffer *cmd_buffer,
+ bool indexed,
+ uint32_t draw_count,
+ uint64_t count_va,
+ uint32_t stride)
+{
+ struct radeon_winsys_cs *cs = cmd_buffer->cs;
+ unsigned di_src_sel = indexed ? V_0287F0_DI_SRC_SEL_DMA
+ : V_0287F0_DI_SRC_SEL_AUTO_INDEX;
+ bool draw_id_enable = cmd_buffer->state.pipeline->shaders[MESA_SHADER_VERTEX]->info.info.vs.needs_draw_id;
+ uint32_t base_reg = cmd_buffer->state.pipeline->graphics.vtx_base_sgpr;
+ assert(base_reg);
+
+ if (draw_count == 1 && !count_va && !draw_id_enable) {
+ radeon_emit(cs, PKT3(indexed ? PKT3_DRAW_INDEX_INDIRECT :
+ PKT3_DRAW_INDIRECT, 3, false));
+ radeon_emit(cs, 0);
+ radeon_emit(cs, (base_reg - SI_SH_REG_OFFSET) >> 2);
+ radeon_emit(cs, ((base_reg + 4) - SI_SH_REG_OFFSET) >> 2);
+ radeon_emit(cs, di_src_sel);
+ } else {
+ radeon_emit(cs, PKT3(indexed ? PKT3_DRAW_INDEX_INDIRECT_MULTI :
+ PKT3_DRAW_INDIRECT_MULTI,
+ 8, false));
+ radeon_emit(cs, 0);
+ radeon_emit(cs, (base_reg - SI_SH_REG_OFFSET) >> 2);
+ radeon_emit(cs, ((base_reg + 4) - SI_SH_REG_OFFSET) >> 2);
+ radeon_emit(cs, (((base_reg + 8) - SI_SH_REG_OFFSET) >> 2) |
+ S_2C3_DRAW_INDEX_ENABLE(draw_id_enable) |
+ S_2C3_COUNT_INDIRECT_ENABLE(!!count_va));
+ radeon_emit(cs, draw_count); /* count */
+ radeon_emit(cs, count_va); /* count_addr */
+ radeon_emit(cs, count_va >> 32);
+ radeon_emit(cs, stride); /* stride */
+ radeon_emit(cs, di_src_sel);
+ }
}
static void
RADV_FROM_HANDLE(radv_buffer, buffer, _buffer);
RADV_FROM_HANDLE(radv_buffer, count_buffer, _count_buffer);
struct radeon_winsys_cs *cs = cmd_buffer->cs;
- unsigned di_src_sel = indexed ? V_0287F0_DI_SRC_SEL_DMA
- : V_0287F0_DI_SRC_SEL_AUTO_INDEX;
+
uint64_t indirect_va = cmd_buffer->device->ws->buffer_get_va(buffer->bo);
indirect_va += offset + buffer->offset;
uint64_t count_va = 0;
cmd_buffer->device->ws->cs_add_buffer(cs, buffer->bo, 8);
- struct ac_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.pipeline, MESA_SHADER_VERTEX,
- AC_UD_VS_BASE_VERTEX_START_INSTANCE);
- uint32_t base_reg = shader_stage_to_user_data_0(MESA_SHADER_VERTEX, radv_pipeline_has_gs(cmd_buffer->state.pipeline),
- radv_pipeline_has_tess(cmd_buffer->state.pipeline));
- assert(loc->sgpr_idx != -1);
radeon_emit(cs, PKT3(PKT3_SET_BASE, 2, 0));
radeon_emit(cs, 1);
radeon_emit(cs, indirect_va);
radeon_emit(cs, indirect_va >> 32);
- radeon_emit(cs, PKT3(indexed ? PKT3_DRAW_INDEX_INDIRECT_MULTI :
- PKT3_DRAW_INDIRECT_MULTI,
- 8, false));
- radeon_emit(cs, 0);
- radeon_emit(cs, ((base_reg + loc->sgpr_idx * 4) - SI_SH_REG_OFFSET) >> 2);
- radeon_emit(cs, ((base_reg + (loc->sgpr_idx + 1) * 4) - SI_SH_REG_OFFSET) >> 2);
- radeon_emit(cs, (((base_reg + (loc->sgpr_idx + 2) * 4) - SI_SH_REG_OFFSET) >> 2) |
- S_2C3_DRAW_INDEX_ENABLE(1) |
- S_2C3_COUNT_INDIRECT_ENABLE(!!count_va));
- radeon_emit(cs, draw_count); /* count */
- radeon_emit(cs, count_va); /* count_addr */
- radeon_emit(cs, count_va >> 32);
- radeon_emit(cs, stride); /* stride */
- radeon_emit(cs, di_src_sel);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ if (!cmd_buffer->state.subpass->view_mask) {
+ radv_cs_emit_indirect_draw_packet(cmd_buffer, indexed, draw_count, count_va, stride);
+ } else {
+ unsigned i;
+ for_each_bit(i, cmd_buffer->state.subpass->view_mask) {
+ radv_emit_view_index(cmd_buffer, i);
+
+ radv_cs_emit_indirect_draw_packet(cmd_buffer, indexed, draw_count, count_va, stride);
+ }
+ }
+ radv_cmd_buffer_after_draw(cmd_buffer);
}
static void
radv_cmd_buffer_flush_state(cmd_buffer, false, false, true, 0);
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
- cmd_buffer->cs, 14);
+ cmd_buffer->cs, 24 * MAX_VIEWS);
radv_emit_indirect_draw(cmd_buffer, buffer, offset,
countBuffer, countBufferOffset, maxDrawCount, stride, false);
uint32_t stride)
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
- int index_size = cmd_buffer->state.index_type ? 4 : 2;
- uint32_t index_max_size = (cmd_buffer->state.index_buffer->size - cmd_buffer->state.index_offset) / index_size;
uint64_t index_va;
radv_cmd_buffer_flush_state(cmd_buffer, true, false, true, 0);
- index_va = cmd_buffer->device->ws->buffer_get_va(cmd_buffer->state.index_buffer->bo);
- index_va += cmd_buffer->state.index_buffer->offset + cmd_buffer->state.index_offset;
+ index_va = cmd_buffer->state.index_va;
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 21);
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 31 * MAX_VIEWS);
radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDEX_TYPE, 0, 0));
radeon_emit(cmd_buffer->cs, cmd_buffer->state.index_type);
radeon_emit(cmd_buffer->cs, index_va >> 32);
radeon_emit(cmd_buffer->cs, PKT3(PKT3_INDEX_BUFFER_SIZE, 0, 0));
- radeon_emit(cmd_buffer->cs, index_max_size);
+ radeon_emit(cmd_buffer->cs, cmd_buffer->state.max_index_count);
radv_emit_indirect_draw(cmd_buffer, buffer, offset,
countBuffer, countBufferOffset, maxDrawCount, stride, true);
radv_flush_compute_state(struct radv_cmd_buffer *cmd_buffer)
{
radv_emit_compute_pipeline(cmd_buffer);
- radv_flush_descriptors(cmd_buffer, cmd_buffer->state.compute_pipeline,
- VK_SHADER_STAGE_COMPUTE_BIT);
+ radv_flush_descriptors(cmd_buffer, VK_SHADER_STAGE_COMPUTE_BIT);
radv_flush_constants(cmd_buffer, cmd_buffer->state.compute_pipeline,
VK_SHADER_STAGE_COMPUTE_BIT);
si_emit_cache_flush(cmd_buffer);
MESA_SHADER_COMPUTE, AC_UD_CS_GRID_SIZE);
if (loc->sgpr_idx != -1) {
assert(!loc->indirect);
- assert(loc->num_sgprs == 3);
- radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B900_COMPUTE_USER_DATA_0 + loc->sgpr_idx * 4, 3);
+ uint8_t grid_used = cmd_buffer->state.compute_pipeline->shaders[MESA_SHADER_COMPUTE]->info.info.cs.grid_components_used;
+ assert(loc->num_sgprs == grid_used);
+ radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B900_COMPUTE_USER_DATA_0 + loc->sgpr_idx * 4, grid_used);
radeon_emit(cmd_buffer->cs, x);
- radeon_emit(cmd_buffer->cs, y);
- radeon_emit(cmd_buffer->cs, z);
+ if (grid_used > 1)
+ radeon_emit(cmd_buffer->cs, y);
+ if (grid_used > 2)
+ radeon_emit(cmd_buffer->cs, z);
}
radeon_emit(cmd_buffer->cs, PKT3(PKT3_DISPATCH_DIRECT, 3, 0) |
radeon_emit(cmd_buffer->cs, 1);
assert(cmd_buffer->cs->cdw <= cdw_max);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ radv_cmd_buffer_after_draw(cmd_buffer);
}
void radv_CmdDispatchIndirect(
struct ac_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.compute_pipeline,
MESA_SHADER_COMPUTE, AC_UD_CS_GRID_SIZE);
if (loc->sgpr_idx != -1) {
- for (unsigned i = 0; i < 3; ++i) {
+ uint8_t grid_used = cmd_buffer->state.compute_pipeline->shaders[MESA_SHADER_COMPUTE]->info.info.cs.grid_components_used;
+ for (unsigned i = 0; i < grid_used; ++i) {
radeon_emit(cmd_buffer->cs, PKT3(PKT3_COPY_DATA, 4, 0));
radeon_emit(cmd_buffer->cs, COPY_DATA_SRC_SEL(COPY_DATA_MEM) |
COPY_DATA_DST_SEL(COPY_DATA_REG));
}
assert(cmd_buffer->cs->cdw <= cdw_max);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ radv_cmd_buffer_after_draw(cmd_buffer);
}
void radv_unaligned_dispatch(
struct ac_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.compute_pipeline,
MESA_SHADER_COMPUTE, AC_UD_CS_GRID_SIZE);
if (loc->sgpr_idx != -1) {
- radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B900_COMPUTE_USER_DATA_0 + loc->sgpr_idx * 4, 3);
+ uint8_t grid_used = cmd_buffer->state.compute_pipeline->shaders[MESA_SHADER_COMPUTE]->info.info.cs.grid_components_used;
+ radeon_set_sh_reg_seq(cmd_buffer->cs, R_00B900_COMPUTE_USER_DATA_0 + loc->sgpr_idx * 4, grid_used);
radeon_emit(cmd_buffer->cs, blocks[0]);
- radeon_emit(cmd_buffer->cs, blocks[1]);
- radeon_emit(cmd_buffer->cs, blocks[2]);
+ if (grid_used > 1)
+ radeon_emit(cmd_buffer->cs, blocks[1]);
+ if (grid_used > 2)
+ radeon_emit(cmd_buffer->cs, blocks[2]);
}
radeon_emit(cmd_buffer->cs, PKT3(PKT3_DISPATCH_DIRECT, 3, 0) |
PKT3_SHADER_TYPE_S(1));
S_00B800_PARTIAL_TG_EN(1));
assert(cmd_buffer->cs->cdw <= cdw_max);
- radv_cmd_buffer_trace_emit(cmd_buffer);
+ radv_cmd_buffer_after_draw(cmd_buffer);
}
void radv_CmdEndRenderPass(
cmd_buffer->state.framebuffer = NULL;
}
-
+/*
+ * For HTILE we have the following interesting clear words:
+ * 0x0000030f: Uncompressed.
+ * 0xfffffff0: Clear depth to 1.0
+ * 0x00000000: Clear depth to 0.0
+ */
static void radv_initialize_htile(struct radv_cmd_buffer *cmd_buffer,
struct radv_image *image,
- const VkImageSubresourceRange *range)
+ const VkImageSubresourceRange *range,
+ uint32_t clear_word)
{
assert(range->baseMipLevel == 0);
assert(range->levelCount == 1 || range->levelCount == VK_REMAINING_ARRAY_LAYERS);
cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB |
RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
- radv_fill_buffer(cmd_buffer, image->bo, offset, size, 0xffffffff);
+ radv_fill_buffer(cmd_buffer, image->bo, offset, size, clear_word);
cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB_META |
RADV_CMD_FLAG_CS_PARTIAL_FLUSH |
struct radv_image *image,
VkImageLayout src_layout,
VkImageLayout dst_layout,
+ unsigned src_queue_mask,
+ unsigned dst_queue_mask,
const VkImageSubresourceRange *range,
VkImageAspectFlags pending_clears)
{
if (dst_layout == VK_IMAGE_LAYOUT_DEPTH_STENCIL_ATTACHMENT_OPTIMAL &&
(pending_clears & vk_format_aspects(image->vk_format)) == vk_format_aspects(image->vk_format) &&
cmd_buffer->state.render_area.offset.x == 0 && cmd_buffer->state.render_area.offset.y == 0 &&
- cmd_buffer->state.render_area.extent.width == image->extent.width &&
- cmd_buffer->state.render_area.extent.height == image->extent.height) {
+ cmd_buffer->state.render_area.extent.width == image->info.width &&
+ cmd_buffer->state.render_area.extent.height == image->info.height) {
/* The clear will initialize htile. */
return;
} else if (src_layout == VK_IMAGE_LAYOUT_UNDEFINED &&
- radv_layout_has_htile(image, dst_layout)) {
+ radv_layout_has_htile(image, dst_layout, dst_queue_mask)) {
/* TODO: merge with the clear if applicable */
- radv_initialize_htile(cmd_buffer, image, range);
- } else if (!radv_layout_has_htile(image, src_layout) &&
- radv_layout_has_htile(image, dst_layout)) {
- radv_initialize_htile(cmd_buffer, image, range);
- } else if ((radv_layout_has_htile(image, src_layout) &&
- !radv_layout_has_htile(image, dst_layout)) ||
- (radv_layout_is_htile_compressed(image, src_layout) &&
- !radv_layout_is_htile_compressed(image, dst_layout))) {
+ radv_initialize_htile(cmd_buffer, image, range, 0);
+ } else if (!radv_layout_is_htile_compressed(image, src_layout, src_queue_mask) &&
+ radv_layout_is_htile_compressed(image, dst_layout, dst_queue_mask)) {
+ radv_initialize_htile(cmd_buffer, image, range, 0xffffffff);
+ } else if (radv_layout_is_htile_compressed(image, src_layout, src_queue_mask) &&
+ !radv_layout_is_htile_compressed(image, dst_layout, dst_queue_mask)) {
VkImageSubresourceRange local_range = *range;
local_range.aspectMask = VK_IMAGE_ASPECT_DEPTH_BIT;
local_range.baseMipLevel = 0;
if (image->surface.htile_size)
radv_handle_depth_image_transition(cmd_buffer, image, src_layout,
- dst_layout, range, pending_clears);
+ dst_layout, src_queue_mask,
+ dst_queue_mask, range,
+ pending_clears);
if (image->cmask.size)
radv_handle_cmask_image_transition(cmd_buffer, image, src_layout,
cmd_buffer->device->ws->cs_add_buffer(cs, event->bo, 8);
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 12);
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 18);
/* TODO: this is overkill. Probably should figure something out from
* the stage mask. */
- if (cmd_buffer->device->physical_device->rad_info.chip_class == CIK) {
- radeon_emit(cs, PKT3(PKT3_EVENT_WRITE_EOP, 4, 0));
- radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_BOTTOM_OF_PIPE_TS) |
- EVENT_INDEX(5));
- radeon_emit(cs, va);
- radeon_emit(cs, (va >> 32) | EOP_DATA_SEL(1));
- radeon_emit(cs, 2);
- radeon_emit(cs, 0);
- }
-
- radeon_emit(cs, PKT3(PKT3_EVENT_WRITE_EOP, 4, 0));
- radeon_emit(cs, EVENT_TYPE(EVENT_TYPE_BOTTOM_OF_PIPE_TS) |
- EVENT_INDEX(5));
- radeon_emit(cs, va);
- radeon_emit(cs, (va >> 32) | EOP_DATA_SEL(1));
- radeon_emit(cs, value);
- radeon_emit(cs, 0);
+ si_cs_emit_write_event_eop(cs,
+ cmd_buffer->state.predicating,
+ cmd_buffer->device->physical_device->rad_info.chip_class,
+ false,
+ EVENT_TYPE_BOTTOM_OF_PIPE_TS, 0,
+ 1, va, 2, value);
assert(cmd_buffer->cs->cdw <= cdw_max);
}
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 7);
- radeon_emit(cs, PKT3(PKT3_WAIT_REG_MEM, 5, 0));
- radeon_emit(cs, WAIT_REG_MEM_EQUAL | WAIT_REG_MEM_MEM_SPACE(1));
- radeon_emit(cs, va);
- radeon_emit(cs, va >> 32);
- radeon_emit(cs, 1); /* reference value */
- radeon_emit(cs, 0xffffffff); /* mask */
- radeon_emit(cs, 4); /* poll interval */
-
+ si_emit_wait_fence(cs, false, va, 1, 0xffffffff);
assert(cmd_buffer->cs->cdw <= cdw_max);
}
projects / mesa.git / blobdiff