if (cmd_buffer->upload.upload_bo)
radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
- cmd_buffer->upload.upload_bo, 8);
+ cmd_buffer->upload.upload_bo);
cmd_buffer->upload.offset = 0;
cmd_buffer->record_result = VK_SUCCESS;
- cmd_buffer->ring_offsets_idx = -1;
-
for (unsigned i = 0; i < VK_PIPELINE_BIND_POINT_RANGE_SIZE; i++) {
cmd_buffer->descriptors[i].dirty = 0;
cmd_buffer->descriptors[i].valid = 0;
}
if (cmd_buffer->device->physical_device->rad_info.chip_class >= GFX9) {
+ unsigned num_db = cmd_buffer->device->physical_device->rad_info.num_render_backends;
+ unsigned eop_bug_offset;
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;
+
+ /* Allocate a buffer for the EOP bug on GFX9. */
+ radv_cmd_buffer_upload_alloc(cmd_buffer, 16 * num_db, 0,
+ &eop_bug_offset, &fence_ptr);
+ cmd_buffer->gfx9_eop_bug_va =
+ radv_buffer_get_va(cmd_buffer->upload.upload_bo);
+ cmd_buffer->gfx9_eop_bug_va += eop_bug_offset;
}
cmd_buffer->status = RADV_CMD_BUFFER_STATUS_INITIAL;
return false;
}
- radv_cs_add_buffer(device->ws, cmd_buffer->cs, bo, 8);
+ radv_cs_add_buffer(device->ws, cmd_buffer->cs, bo);
if (cmd_buffer->upload.upload_bo) {
upload = malloc(sizeof(*upload));
MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cmd_buffer->cs, 7);
++cmd_buffer->state.trace_id;
- radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
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));
cmd_buffer->device->physical_device->rad_info.chip_class,
ptr, va,
radv_cmd_buffer_uses_mec(cmd_buffer),
- flags);
+ flags, cmd_buffer->gfx9_eop_bug_va);
}
if (unlikely(cmd_buffer->device->trace_bo))
data[0] = (uintptr_t)pipeline;
data[1] = (uintptr_t)pipeline >> 32;
- radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
radv_emit_write_data_packet(cs, va, 2, data);
}
radv_get_descriptors_state(cmd_buffer, bind_point);
descriptors_state->sets[idx] = set;
- if (set)
- descriptors_state->valid |= (1u << idx);
- else
- descriptors_state->valid &= ~(1u << idx);
+
+ descriptors_state->valid |= (1u << idx); /* active descriptors */
descriptors_state->dirty |= (1u << idx);
}
data[i * 2 + 1] = (uintptr_t)set >> 32;
}
- radv_cs_add_buffer(device->ws, cs, device->trace_bo, 8);
radv_emit_write_data_packet(cs, va, MAX_SETS * 2, data);
}
uint32_t sh_base = pipeline->user_data_0[stage];
struct radv_userdata_locations *locs =
&pipeline->shaders[stage]->info.user_sgprs_locs;
- unsigned mask;
+ unsigned mask = locs->descriptor_sets_enabled;
- mask = descriptors_state->dirty & descriptors_state->valid;
-
- for (int i = 0; i < MAX_SETS; i++) {
- struct radv_userdata_info *loc = &locs->descriptor_sets[i];
- if (loc->sgpr_idx != -1 && !loc->indirect)
- continue;
- mask &= ~(1 << i);
- }
+ mask &= descriptors_state->dirty & descriptors_state->valid;
while (mask) {
int start, count;
unsigned sx_blend_opt_control = 0;
for (unsigned i = 0; i < subpass->color_count; ++i) {
- if (subpass->color_attachments[i].attachment == VK_ATTACHMENT_UNUSED)
+ if (subpass->color_attachments[i].attachment == VK_ATTACHMENT_UNUSED) {
+ sx_blend_opt_control |= S_02875C_MRT0_COLOR_OPT_DISABLE(1) << (i * 4);
+ sx_blend_opt_control |= S_02875C_MRT0_ALPHA_OPT_DISABLE(1) << (i * 4);
continue;
+ }
int idx = subpass->color_attachments[i].attachment;
struct radv_color_buffer_info *cb = &framebuffer->attachments[idx].cb;
}
}
+ for (unsigned i = subpass->color_count; i < 8; ++i) {
+ sx_blend_opt_control |= S_02875C_MRT0_COLOR_OPT_DISABLE(1) << (i * 4);
+ sx_blend_opt_control |= S_02875C_MRT0_ALPHA_OPT_DISABLE(1) << (i * 4);
+ }
radeon_set_context_reg_seq(cmd_buffer->cs, R_028754_SX_PS_DOWNCONVERT, 3);
radeon_emit(cmd_buffer->cs, sx_ps_downconvert);
radeon_emit(cmd_buffer->cs, sx_blend_opt_epsilon);
continue;
radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
- pipeline->shaders[i]->bo, 8);
+ pipeline->shaders[i]->bo);
}
if (radv_pipeline_has_gs(pipeline))
radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
- pipeline->gs_copy_shader->bo, 8);
+ pipeline->gs_copy_shader->bo);
if (unlikely(cmd_buffer->device->trace_bo))
radv_save_pipeline(cmd_buffer, pipeline, RING_GFX);
/**
* Set the clear depth/stencil values to the image's metadata.
*/
-void
+static void
radv_set_ds_clear_metadata(struct radv_cmd_buffer *cmd_buffer,
struct radv_image *image,
VkClearDepthStencilValue ds_clear_value,
va += image->offset + image->clear_value_offset;
- assert(radv_image_has_htile(image));
-
if (aspects & VK_IMAGE_ASPECT_STENCIL_BIT) {
++reg_count;
} else {
radeon_emit(cs, ds_clear_value.stencil);
if (aspects & VK_IMAGE_ASPECT_DEPTH_BIT)
radeon_emit(cs, fui(ds_clear_value.depth));
+}
+
+/**
+ * Update the clear depth/stencil values for this image.
+ */
+void
+radv_update_ds_clear_metadata(struct radv_cmd_buffer *cmd_buffer,
+ struct radv_image *image,
+ VkClearDepthStencilValue ds_clear_value,
+ VkImageAspectFlags aspects)
+{
+ assert(radv_image_has_htile(image));
+
+ radv_set_ds_clear_metadata(cmd_buffer, image, ds_clear_value, aspects);
radv_update_bound_fast_clear_ds(cmd_buffer, image, ds_clear_value,
aspects);
struct radv_image *image = att->attachment->image;
VkImageLayout layout = subpass->color_attachments[i].layout;
- radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo, 8);
+ radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo);
assert(att->attachment->aspect_mask & VK_IMAGE_ASPECT_COLOR_BIT);
radv_emit_fb_color_state(cmd_buffer, i, att, image, layout);
VkImageLayout layout = subpass->depth_stencil_attachment.layout;
struct radv_attachment_info *att = &framebuffer->attachments[idx];
struct radv_image *image = att->attachment->image;
- radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo, 8);
+ radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, att->attachment->bo);
MAYBE_UNUSED uint32_t queue_mask = radv_image_queue_family_mask(image,
cmd_buffer->queue_family_index,
cmd_buffer->queue_family_index);
R_028A4C_PA_SC_MODE_CNTL_1,
pa_sc_mode_cntl_1);
}
- db_count_control = 0;
- } else {
- db_count_control = S_028004_ZPASS_INCREMENT_DISABLE(1);
}
+ db_count_control = S_028004_ZPASS_INCREMENT_DISABLE(1);
} else {
const struct radv_subpass *subpass = cmd_buffer->state.subpass;
uint32_t sample_rate = subpass ? util_logbase2(subpass->max_sample_count) : 0;
struct radv_pipeline *pipeline = stages & VK_SHADER_STAGE_COMPUTE_BIT
? cmd_buffer->state.compute_pipeline
: cmd_buffer->state.pipeline;
+ VkPipelineBindPoint bind_point = stages & VK_SHADER_STAGE_COMPUTE_BIT ?
+ VK_PIPELINE_BIND_POINT_COMPUTE :
+ VK_PIPELINE_BIND_POINT_GRAPHICS;
+ struct radv_descriptor_state *descriptors_state =
+ radv_get_descriptors_state(cmd_buffer, bind_point);
struct radv_pipeline_layout *layout = pipeline->layout;
struct radv_shader_variant *shader, *prev_shader;
unsigned offset;
return;
memcpy(ptr, cmd_buffer->push_constants, layout->push_constant_size);
- memcpy((char*)ptr + layout->push_constant_size, cmd_buffer->dynamic_buffers,
+ memcpy((char*)ptr + layout->push_constant_size,
+ descriptors_state->dynamic_buffers,
16 * layout->dynamic_offset_count);
va = radv_buffer_get_va(cmd_buffer->upload.upload_bo);
cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_CS_PARTIAL_FLUSH;
}
- if (src_stage_mask & (VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT |
- VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT |
- VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT |
- VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
+ if (src_stage_mask & (VK_PIPELINE_STAGE_FRAGMENT_SHADER_BIT |
VK_PIPELINE_STAGE_EARLY_FRAGMENT_TESTS_BIT |
VK_PIPELINE_STAGE_LATE_FRAGMENT_TESTS_BIT |
VK_PIPELINE_STAGE_COLOR_ATTACHMENT_OUTPUT_BIT |
cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_PS_PARTIAL_FLUSH;
} else if (src_stage_mask & (VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT |
VK_PIPELINE_STAGE_VERTEX_INPUT_BIT |
- VK_PIPELINE_STAGE_VERTEX_SHADER_BIT)) {
+ VK_PIPELINE_STAGE_VERTEX_SHADER_BIT |
+ VK_PIPELINE_STAGE_TESSELLATION_CONTROL_SHADER_BIT |
+ VK_PIPELINE_STAGE_TESSELLATION_EVALUATION_SHADER_BIT |
+ VK_PIPELINE_STAGE_GEOMETRY_SHADER_BIT)) {
cmd_buffer->state.flush_bits |= RADV_CMD_FLAG_VS_PARTIAL_FLUSH;
}
}
static enum radv_cmd_flush_bits
radv_src_access_flush(struct radv_cmd_buffer *cmd_buffer,
- VkAccessFlags src_flags)
+ VkAccessFlags src_flags,
+ struct radv_image *image)
{
+ bool flush_CB_meta = true, flush_DB_meta = true;
enum radv_cmd_flush_bits flush_bits = 0;
uint32_t b;
+
+ if (image && !radv_image_has_CB_metadata(image))
+ flush_CB_meta = false;
+ if (image && !radv_image_has_htile(image))
+ flush_DB_meta = false;
+
for_each_bit(b, src_flags) {
switch ((VkAccessFlagBits)(1 << b)) {
case VK_ACCESS_SHADER_WRITE_BIT:
flush_bits |= RADV_CMD_FLAG_WRITEBACK_GLOBAL_L2;
break;
case VK_ACCESS_COLOR_ATTACHMENT_WRITE_BIT:
- flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_CB |
- RADV_CMD_FLAG_FLUSH_AND_INV_CB_META;
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_CB;
+ if (flush_CB_meta)
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_CB_META;
break;
case VK_ACCESS_DEPTH_STENCIL_ATTACHMENT_WRITE_BIT:
- flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB |
- RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB;
+ if (flush_DB_meta)
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
break;
case VK_ACCESS_TRANSFER_WRITE_BIT:
flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_CB |
- RADV_CMD_FLAG_FLUSH_AND_INV_CB_META |
RADV_CMD_FLAG_FLUSH_AND_INV_DB |
- RADV_CMD_FLAG_FLUSH_AND_INV_DB_META |
RADV_CMD_FLAG_INV_GLOBAL_L2;
+
+ if (flush_CB_meta)
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_CB_META;
+ if (flush_DB_meta)
+ flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
break;
default:
break;
static void radv_subpass_barrier(struct radv_cmd_buffer *cmd_buffer, const struct radv_subpass_barrier *barrier)
{
- cmd_buffer->state.flush_bits |= radv_src_access_flush(cmd_buffer, barrier->src_access_mask);
+ cmd_buffer->state.flush_bits |= radv_src_access_flush(cmd_buffer, barrier->src_access_mask,
+ NULL);
radv_stage_flush(cmd_buffer, barrier->src_stage_mask);
cmd_buffer->state.flush_bits |= radv_dst_access_flush(cmd_buffer, barrier->dst_access_mask,
NULL);
}
static void radv_handle_subpass_image_transition(struct radv_cmd_buffer *cmd_buffer,
- VkAttachmentReference att)
+ struct radv_subpass_attachment att)
{
unsigned idx = att.attachment;
struct radv_image_view *view = cmd_buffer->state.framebuffer->attachments[idx].attachment;
struct radv_device *device = cmd_buffer->device;
if (device->gfx_init) {
uint64_t va = radv_buffer_get_va(device->gfx_init);
- radv_cs_add_buffer(device->ws, cmd_buffer->cs, device->gfx_init, 8);
+ radv_cs_add_buffer(device->ws, cmd_buffer->cs, device->gfx_init);
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);
}
}
- if (pBeginInfo->flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT) {
+ if (cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_SECONDARY &&
+ (pBeginInfo->flags & VK_COMMAND_BUFFER_USAGE_RENDER_PASS_CONTINUE_BIT)) {
assert(pBeginInfo->pInheritanceInfo);
cmd_buffer->state.framebuffer = radv_framebuffer_from_handle(pBeginInfo->pInheritanceInfo->framebuffer);
cmd_buffer->state.pass = radv_render_pass_from_handle(pBeginInfo->pInheritanceInfo->renderPass);
radv_cmd_buffer_set_subpass(cmd_buffer, subpass, false);
}
- if (unlikely(cmd_buffer->device->trace_bo))
+ if (unlikely(cmd_buffer->device->trace_bo)) {
+ struct radv_device *device = cmd_buffer->device;
+
+ radv_cs_add_buffer(device->ws, cmd_buffer->cs,
+ device->trace_bo);
+
radv_cmd_buffer_trace_emit(cmd_buffer);
+ }
cmd_buffer->status = RADV_CMD_BUFFER_STATUS_RECORDING;
vb[idx].offset = pOffsets[i];
radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
- vb[idx].buffer->bo, 8);
+ vb[idx].buffer->bo);
}
if (!changed) {
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;
- radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, index_buffer->bo, 8);
+ radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs, index_buffer->bo);
}
struct radeon_winsys *ws = cmd_buffer->device->ws;
radv_set_descriptor_set(cmd_buffer, bind_point, set, idx);
- if (!set)
- return;
+ assert(set);
assert(!(set->layout->flags & VK_DESCRIPTOR_SET_LAYOUT_CREATE_PUSH_DESCRIPTOR_BIT_KHR));
if (!cmd_buffer->device->use_global_bo_list) {
for (unsigned j = 0; j < set->layout->buffer_count; ++j)
if (set->descriptors[j])
- radv_cs_add_buffer(ws, cmd_buffer->cs, set->descriptors[j], 7);
+ radv_cs_add_buffer(ws, cmd_buffer->cs, set->descriptors[j]);
}
if(set->bo)
- radv_cs_add_buffer(ws, cmd_buffer->cs, set->bo, 8);
+ radv_cs_add_buffer(ws, cmd_buffer->cs, set->bo);
}
void radv_CmdBindDescriptorSets(
unsigned dyn_idx = 0;
const bool no_dynamic_bounds = cmd_buffer->device->instance->debug_flags & RADV_DEBUG_NO_DYNAMIC_BOUNDS;
+ struct radv_descriptor_state *descriptors_state =
+ radv_get_descriptors_state(cmd_buffer, pipelineBindPoint);
for (unsigned i = 0; i < descriptorSetCount; ++i) {
unsigned idx = i + firstSet;
for(unsigned j = 0; j < set->layout->dynamic_offset_count; ++j, ++dyn_idx) {
unsigned idx = j + layout->set[i + firstSet].dynamic_offset_start;
- uint32_t *dst = cmd_buffer->dynamic_buffers + idx * 4;
+ uint32_t *dst = descriptors_state->dynamic_buffers + idx * 4;
assert(dyn_idx < dynamicOffsetCount);
struct radv_descriptor_range *range = set->dynamic_descriptors + j;
si_emit_cache_flush(cmd_buffer);
}
+ /* Make sure CP DMA is idle at the end of IBs because the kernel
+ * doesn't wait for it.
+ */
+ si_cp_dma_wait_for_idle(cmd_buffer);
+
vk_free(&cmd_buffer->pool->alloc, cmd_buffer->state.attachments);
if (!cmd_buffer->device->ws->cs_finalize(cmd_buffer->cs))
pipeline->max_waves * pipeline->scratch_bytes_per_wave);
radv_cs_add_buffer(cmd_buffer->device->ws, cmd_buffer->cs,
- pipeline->shaders[MESA_SHADER_COMPUTE]->bo, 8);
+ pipeline->shaders[MESA_SHADER_COMPUTE]->bo);
if (unlikely(cmd_buffer->device->trace_bo))
radv_save_pipeline(cmd_buffer, pipeline, RING_COMPUTE);
if (radv_pipeline_has_tess(pipeline))
cmd_buffer->tess_rings_needed = true;
-
- if (radv_pipeline_has_gs(pipeline)) {
- struct radv_userdata_info *loc = radv_lookup_user_sgpr(cmd_buffer->state.pipeline, MESA_SHADER_GEOMETRY,
- AC_UD_SCRATCH_RING_OFFSETS);
- if (cmd_buffer->ring_offsets_idx == -1)
- cmd_buffer->ring_offsets_idx = loc->sgpr_idx;
- else if (loc->sgpr_idx != -1)
- assert(loc->sgpr_idx == cmd_buffer->ring_offsets_idx);
- }
break;
default:
assert(!"invalid bind point");
if (secondary->sample_positions_needed)
primary->sample_positions_needed = true;
- if (secondary->ring_offsets_idx != -1) {
- if (primary->ring_offsets_idx == -1)
- primary->ring_offsets_idx = secondary->ring_offsets_idx;
- else
- assert(secondary->ring_offsets_idx == primary->ring_offsets_idx);
- }
primary->device->ws->cs_execute_secondary(primary->cs, secondary->cs);
radv_cmd_buffer_clear_subpass(cmd_buffer);
}
+void radv_CmdBeginRenderPass2KHR(
+ VkCommandBuffer commandBuffer,
+ const VkRenderPassBeginInfo* pRenderPassBeginInfo,
+ const VkSubpassBeginInfoKHR* pSubpassBeginInfo)
+{
+ radv_CmdBeginRenderPass(commandBuffer, pRenderPassBeginInfo,
+ pSubpassBeginInfo->contents);
+}
+
void radv_CmdNextSubpass(
VkCommandBuffer commandBuffer,
VkSubpassContents contents)
radv_cmd_buffer_clear_subpass(cmd_buffer);
}
+void radv_CmdNextSubpass2KHR(
+ VkCommandBuffer commandBuffer,
+ const VkSubpassBeginInfoKHR* pSubpassBeginInfo,
+ const VkSubpassEndInfoKHR* pSubpassEndInfo)
+{
+ radv_CmdNextSubpass(commandBuffer, pSubpassBeginInfo->contents);
+}
+
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])
+ if (!radv_get_shader(pipeline, stage))
continue;
+
struct radv_userdata_info *loc = radv_lookup_user_sgpr(pipeline, stage, AC_UD_VIEW_INDEX);
if (loc->sgpr_idx == -1)
continue;
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, PKT3(PKT3_DRAW_INDEX_2, 4, cmd_buffer->state.predicating));
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 >> 32);
: V_0287F0_DI_SRC_SEL_AUTO_INDEX;
bool draw_id_enable = radv_get_shader(cmd_buffer->state.pipeline, MESA_SHADER_VERTEX)->info.info.vs.needs_draw_id;
uint32_t base_reg = cmd_buffer->state.pipeline->graphics.vtx_base_sgpr;
+ bool predicating = cmd_buffer->state.predicating;
assert(base_reg);
/* just reset draw state for vertex data */
if (draw_count == 1 && !count_va && !draw_id_enable) {
radeon_emit(cs, PKT3(indexed ? PKT3_DRAW_INDEX_INDIRECT :
- PKT3_DRAW_INDIRECT, 3, false));
+ PKT3_DRAW_INDIRECT, 3, predicating));
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);
} else {
radeon_emit(cs, PKT3(indexed ? PKT3_DRAW_INDEX_INDIRECT_MULTI :
PKT3_DRAW_INDIRECT_MULTI,
- 8, false));
+ 8, predicating));
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);
va += info->indirect->offset + info->indirect_offset;
- radv_cs_add_buffer(ws, cs, info->indirect->bo, 8);
+ radv_cs_add_buffer(ws, cs, info->indirect->bo);
radeon_emit(cs, PKT3(PKT3_SET_BASE, 2, 0));
radeon_emit(cs, 1);
count_va += info->count_buffer->offset +
info->count_buffer_offset;
- radv_cs_add_buffer(ws, cs, info->count_buffer->bo, 8);
+ radv_cs_add_buffer(ws, cs, info->count_buffer->bo);
}
if (!state->subpass->view_mask) {
struct radv_shader_variant *compute_shader = pipeline->shaders[MESA_SHADER_COMPUTE];
unsigned dispatch_initiator = cmd_buffer->device->dispatch_initiator;
struct radeon_winsys *ws = cmd_buffer->device->ws;
+ bool predicating = cmd_buffer->state.predicating;
struct radeon_cmdbuf *cs = cmd_buffer->cs;
struct radv_userdata_info *loc;
va += info->indirect->offset + info->indirect_offset;
- radv_cs_add_buffer(ws, cs, info->indirect->bo, 8);
+ radv_cs_add_buffer(ws, cs, info->indirect->bo);
if (loc->sgpr_idx != -1) {
for (unsigned i = 0; i < 3; ++i) {
}
if (radv_cmd_buffer_uses_mec(cmd_buffer)) {
- radeon_emit(cs, PKT3(PKT3_DISPATCH_INDIRECT, 2, 0) |
+ radeon_emit(cs, PKT3(PKT3_DISPATCH_INDIRECT, 2, predicating) |
PKT3_SHADER_TYPE_S(1));
radeon_emit(cs, va);
radeon_emit(cs, va >> 32);
radeon_emit(cs, va);
radeon_emit(cs, va >> 32);
- radeon_emit(cs, PKT3(PKT3_DISPATCH_INDIRECT, 1, 0) |
+ radeon_emit(cs, PKT3(PKT3_DISPATCH_INDIRECT, 1, predicating) |
PKT3_SHADER_TYPE_S(1));
radeon_emit(cs, 0);
radeon_emit(cs, dispatch_initiator);
dispatch_initiator |= S_00B800_FORCE_START_AT_000(1);
}
- radeon_emit(cs, PKT3(PKT3_DISPATCH_DIRECT, 3, 0) |
+ radeon_emit(cs, PKT3(PKT3_DISPATCH_DIRECT, 3, predicating) |
PKT3_SHADER_TYPE_S(1));
radeon_emit(cs, blocks[0]);
radeon_emit(cs, blocks[1]);
for (unsigned i = 0; i < cmd_buffer->state.framebuffer->attachment_count; ++i) {
VkImageLayout layout = cmd_buffer->state.pass->attachments[i].final_layout;
radv_handle_subpass_image_transition(cmd_buffer,
- (VkAttachmentReference){i, layout});
+ (struct radv_subpass_attachment){i, layout});
}
vk_free(&cmd_buffer->pool->alloc, cmd_buffer->state.attachments);
cmd_buffer->state.framebuffer = NULL;
}
+void radv_CmdEndRenderPass2KHR(
+ VkCommandBuffer commandBuffer,
+ const VkSubpassEndInfoKHR* pSubpassEndInfo)
+{
+ radv_CmdEndRenderPass(commandBuffer);
+}
+
/*
* For HTILE we have the following interesting clear words:
* 0xfffff30f: Uncompressed, full depth range, for depth+stencil HTILE
assert(range->levelCount == 1 || range->levelCount == VK_REMAINING_ARRAY_LAYERS);
unsigned layer_count = radv_get_layerCount(image, range);
uint64_t size = image->surface.htile_slice_size * layer_count;
+ VkImageAspectFlags aspects = VK_IMAGE_ASPECT_DEPTH_BIT;
uint64_t offset = image->offset + image->htile_offset +
image->surface.htile_slice_size * range->baseArrayLayer;
struct radv_cmd_state *state = &cmd_buffer->state;
+ VkClearDepthStencilValue value = {};
state->flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB |
RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
state->flush_bits |= RADV_CMD_FLAG_FLUSH_AND_INV_DB_META;
- /* Initialize the depth clear registers and update the ZRANGE_PRECISION
- * value for the TC-compat bug (because ZRANGE_PRECISION is 1 by
- * default). This is only needed whean clearing Z to 0.0f.
- */
- if (radv_image_is_tc_compat_htile(image) && clear_word == 0) {
- VkImageAspectFlags aspects = VK_IMAGE_ASPECT_DEPTH_BIT;
- VkClearDepthStencilValue value = {};
-
- if (vk_format_is_stencil(image->vk_format))
- aspects |= VK_IMAGE_ASPECT_STENCIL_BIT;
+ if (vk_format_is_stencil(image->vk_format))
+ aspects |= VK_IMAGE_ASPECT_STENCIL_BIT;
- radv_set_ds_clear_metadata(cmd_buffer, image, value, aspects);
- }
+ radv_set_ds_clear_metadata(cmd_buffer, image, value, aspects);
}
static void radv_handle_depth_image_transition(struct radv_cmd_buffer *cmd_buffer,
if (!radv_image_has_htile(image))
return;
- 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->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 &&
+ if (src_layout == VK_IMAGE_LAYOUT_UNDEFINED &&
radv_layout_has_htile(image, dst_layout, dst_queue_mask)) {
/* TODO: merge with the clear if applicable */
radv_initialize_htile(cmd_buffer, image, range, 0);
}
}
-void radv_CmdPipelineBarrier(
- VkCommandBuffer commandBuffer,
- VkPipelineStageFlags srcStageMask,
- VkPipelineStageFlags destStageMask,
- VkBool32 byRegion,
- uint32_t memoryBarrierCount,
- const VkMemoryBarrier* pMemoryBarriers,
- uint32_t bufferMemoryBarrierCount,
- const VkBufferMemoryBarrier* pBufferMemoryBarriers,
- uint32_t imageMemoryBarrierCount,
- const VkImageMemoryBarrier* pImageMemoryBarriers)
+struct radv_barrier_info {
+ uint32_t eventCount;
+ const VkEvent *pEvents;
+ VkPipelineStageFlags srcStageMask;
+};
+
+static void
+radv_barrier(struct radv_cmd_buffer *cmd_buffer,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier *pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier *pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier *pImageMemoryBarriers,
+ const struct radv_barrier_info *info)
{
- RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct radeon_cmdbuf *cs = cmd_buffer->cs;
enum radv_cmd_flush_bits src_flush_bits = 0;
enum radv_cmd_flush_bits dst_flush_bits = 0;
+ for (unsigned i = 0; i < info->eventCount; ++i) {
+ RADV_FROM_HANDLE(radv_event, event, info->pEvents[i]);
+ uint64_t va = radv_buffer_get_va(event->bo);
+
+ radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo);
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 7);
+
+ si_emit_wait_fence(cs, va, 1, 0xffffffff);
+ assert(cmd_buffer->cs->cdw <= cdw_max);
+ }
+
for (uint32_t i = 0; i < memoryBarrierCount; i++) {
- src_flush_bits |= radv_src_access_flush(cmd_buffer, pMemoryBarriers[i].srcAccessMask);
+ src_flush_bits |= radv_src_access_flush(cmd_buffer, pMemoryBarriers[i].srcAccessMask,
+ NULL);
dst_flush_bits |= radv_dst_access_flush(cmd_buffer, pMemoryBarriers[i].dstAccessMask,
NULL);
}
for (uint32_t i = 0; i < bufferMemoryBarrierCount; i++) {
- src_flush_bits |= radv_src_access_flush(cmd_buffer, pBufferMemoryBarriers[i].srcAccessMask);
+ src_flush_bits |= radv_src_access_flush(cmd_buffer, pBufferMemoryBarriers[i].srcAccessMask,
+ NULL);
dst_flush_bits |= radv_dst_access_flush(cmd_buffer, pBufferMemoryBarriers[i].dstAccessMask,
NULL);
}
for (uint32_t i = 0; i < imageMemoryBarrierCount; i++) {
RADV_FROM_HANDLE(radv_image, image, pImageMemoryBarriers[i].image);
- src_flush_bits |= radv_src_access_flush(cmd_buffer, pImageMemoryBarriers[i].srcAccessMask);
+
+ src_flush_bits |= radv_src_access_flush(cmd_buffer, pImageMemoryBarriers[i].srcAccessMask,
+ image);
dst_flush_bits |= radv_dst_access_flush(cmd_buffer, pImageMemoryBarriers[i].dstAccessMask,
image);
}
- radv_stage_flush(cmd_buffer, srcStageMask);
+ radv_stage_flush(cmd_buffer, info->srcStageMask);
cmd_buffer->state.flush_bits |= src_flush_bits;
for (uint32_t i = 0; i < imageMemoryBarrierCount; i++) {
0);
}
+ /* Make sure CP DMA is idle because the driver might have performed a
+ * DMA operation for copying or filling buffers/images.
+ */
+ si_cp_dma_wait_for_idle(cmd_buffer);
+
cmd_buffer->state.flush_bits |= dst_flush_bits;
}
+void radv_CmdPipelineBarrier(
+ VkCommandBuffer commandBuffer,
+ VkPipelineStageFlags srcStageMask,
+ VkPipelineStageFlags destStageMask,
+ VkBool32 byRegion,
+ uint32_t memoryBarrierCount,
+ const VkMemoryBarrier* pMemoryBarriers,
+ uint32_t bufferMemoryBarrierCount,
+ const VkBufferMemoryBarrier* pBufferMemoryBarriers,
+ uint32_t imageMemoryBarrierCount,
+ const VkImageMemoryBarrier* pImageMemoryBarriers)
+{
+ RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
+ struct radv_barrier_info info;
+
+ info.eventCount = 0;
+ info.pEvents = NULL;
+ info.srcStageMask = srcStageMask;
+
+ radv_barrier(cmd_buffer, memoryBarrierCount, pMemoryBarriers,
+ bufferMemoryBarrierCount, pBufferMemoryBarriers,
+ imageMemoryBarrierCount, pImageMemoryBarriers, &info);
+}
+
static void write_event(struct radv_cmd_buffer *cmd_buffer,
struct radv_event *event,
struct radeon_cmdbuf *cs = cmd_buffer->cs;
uint64_t va = radv_buffer_get_va(event->bo);
- radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo, 8);
+ radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo);
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. */
+ /* Flags that only require a top-of-pipe event. */
+ VkPipelineStageFlags top_of_pipe_flags =
+ VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT;
- si_cs_emit_write_event_eop(cs,
- cmd_buffer->state.predicating,
- cmd_buffer->device->physical_device->rad_info.chip_class,
- radv_cmd_buffer_uses_mec(cmd_buffer),
- V_028A90_BOTTOM_OF_PIPE_TS, 0,
- 1, va, 2, value);
+ /* Flags that only require a post-index-fetch event. */
+ VkPipelineStageFlags post_index_fetch_flags =
+ top_of_pipe_flags |
+ VK_PIPELINE_STAGE_DRAW_INDIRECT_BIT |
+ VK_PIPELINE_STAGE_VERTEX_INPUT_BIT;
+
+ /* Make sure CP DMA is idle because the driver might have performed a
+ * DMA operation for copying or filling buffers/images.
+ */
+ si_cp_dma_wait_for_idle(cmd_buffer);
+
+ /* TODO: Emit EOS events for syncing PS/CS stages. */
+
+ if (!(stageMask & ~top_of_pipe_flags)) {
+ /* Just need to sync the PFP engine. */
+ 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_PFP));
+ radeon_emit(cs, va);
+ radeon_emit(cs, va >> 32);
+ radeon_emit(cs, value);
+ } else if (!(stageMask & ~post_index_fetch_flags)) {
+ /* Sync ME because PFP reads index and indirect buffers. */
+ 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, value);
+ } else {
+ /* Otherwise, sync all prior GPU work using an EOP event. */
+ si_cs_emit_write_event_eop(cs,
+ cmd_buffer->device->physical_device->rad_info.chip_class,
+ radv_cmd_buffer_uses_mec(cmd_buffer),
+ V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ EOP_DATA_SEL_VALUE_32BIT, va, 2, value,
+ cmd_buffer->gfx9_eop_bug_va);
+ }
assert(cmd_buffer->cs->cdw <= cdw_max);
}
const VkImageMemoryBarrier* pImageMemoryBarriers)
{
RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
- struct radeon_cmdbuf *cs = cmd_buffer->cs;
+ struct radv_barrier_info info;
- for (unsigned i = 0; i < eventCount; ++i) {
- RADV_FROM_HANDLE(radv_event, event, pEvents[i]);
- uint64_t va = radv_buffer_get_va(event->bo);
+ info.eventCount = eventCount;
+ info.pEvents = pEvents;
+ info.srcStageMask = 0;
- radv_cs_add_buffer(cmd_buffer->device->ws, cs, event->bo, 8);
+ radv_barrier(cmd_buffer, memoryBarrierCount, pMemoryBarriers,
+ bufferMemoryBarrierCount, pBufferMemoryBarriers,
+ imageMemoryBarrierCount, pImageMemoryBarriers, &info);
+}
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 7);
- si_emit_wait_fence(cs, false, va, 1, 0xffffffff);
- assert(cmd_buffer->cs->cdw <= cdw_max);
- }
+void radv_CmdSetDeviceMask(VkCommandBuffer commandBuffer,
+ uint32_t deviceMask)
+{
+ /* No-op */
+}
+
+/* VK_EXT_conditional_rendering */
+void vkCmdBeginConditionalRenderingEXT(
+ VkCommandBuffer commandBuffer,
+ const VkConditionalRenderingBeginInfoEXT* pConditionalRenderingBegin)
+{
+ RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
+ RADV_FROM_HANDLE(radv_buffer, buffer, pConditionalRenderingBegin->buffer);
+ bool inverted;
+ uint64_t va;
+ va = radv_buffer_get_va(buffer->bo) + pConditionalRenderingBegin->offset;
- for (uint32_t i = 0; i < imageMemoryBarrierCount; i++) {
- RADV_FROM_HANDLE(radv_image, image, pImageMemoryBarriers[i].image);
+ inverted = pConditionalRenderingBegin->flags & VK_CONDITIONAL_RENDERING_INVERTED_BIT_EXT;
- radv_handle_image_transition(cmd_buffer, image,
- pImageMemoryBarriers[i].oldLayout,
- pImageMemoryBarriers[i].newLayout,
- pImageMemoryBarriers[i].srcQueueFamilyIndex,
- pImageMemoryBarriers[i].dstQueueFamilyIndex,
- &pImageMemoryBarriers[i].subresourceRange,
- 0);
- }
+ /* Enable predication for this command buffer. */
+ si_emit_set_predication_state(cmd_buffer, inverted, va);
+ cmd_buffer->state.predicating = true;
- /* TODO: figure out how to do memory barriers without waiting */
- cmd_buffer->state.flush_bits |= RADV_CMD_FLUSH_AND_INV_FRAMEBUFFER |
- RADV_CMD_FLAG_INV_GLOBAL_L2 |
- RADV_CMD_FLAG_INV_VMEM_L1 |
- RADV_CMD_FLAG_INV_SMEM_L1;
+ /* Store conditional rendering user info. */
+ cmd_buffer->state.predication_type = inverted;
+ cmd_buffer->state.predication_va = va;
}
-
-void radv_CmdSetDeviceMask(VkCommandBuffer commandBuffer,
- uint32_t deviceMask)
+void vkCmdEndConditionalRenderingEXT(
+ VkCommandBuffer commandBuffer)
{
- /* No-op */
+ RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer, commandBuffer);
+
+ /* Disable predication for this command buffer. */
+ si_emit_set_predication_state(cmd_buffer, false, 0);
+ cmd_buffer->state.predicating = false;
+
+ /* Reset conditional rendering user info. */
+ cmd_buffer->state.predication_type = -1;
+ cmd_buffer->state.predication_va = 0;
}