static void emit_begin_query(struct radv_cmd_buffer *cmd_buffer,
uint64_t va,
- VkQueryType query_type)
+ VkQueryType query_type,
+ VkQueryControlFlags flags)
{
struct radeon_winsys_cs *cs = cmd_buffer->cs;
switch (query_type) {
radeon_check_space(cmd_buffer->device->ws, cs, 7);
++cmd_buffer->state.active_occlusion_queries;
- if (cmd_buffer->state.active_occlusion_queries == 1)
+ if (cmd_buffer->state.active_occlusion_queries == 1) {
+ if (flags & VK_QUERY_CONTROL_PRECISE_BIT) {
+ /* This is the first occlusion query, enable
+ * the hint if the precision bit is set.
+ */
+ cmd_buffer->state.perfect_occlusion_queries_enabled = true;
+ }
+
radv_set_db_count_control(cmd_buffer);
+ } else {
+ if ((flags & VK_QUERY_CONTROL_PRECISE_BIT) &&
+ !cmd_buffer->state.perfect_occlusion_queries_enabled) {
+ /* This is not the first query, but this one
+ * needs to enable precision, DB_COUNT_CONTROL
+ * has to be updated accordingly.
+ */
+ cmd_buffer->state.perfect_occlusion_queries_enabled = true;
+
+ radv_set_db_count_control(cmd_buffer);
+ }
+ }
radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
radeon_emit(cs, EVENT_TYPE(V_028A90_ZPASS_DONE) | EVENT_INDEX(1));
radeon_check_space(cmd_buffer->device->ws, cs, 14);
cmd_buffer->state.active_occlusion_queries--;
- if (cmd_buffer->state.active_occlusion_queries == 0)
+ if (cmd_buffer->state.active_occlusion_queries == 0) {
radv_set_db_count_control(cmd_buffer);
+ /* Reset the perfect occlusion queries hint now that no
+ * queries are active.
+ */
+ cmd_buffer->state.perfect_occlusion_queries_enabled = false;
+ }
+
radeon_emit(cs, PKT3(PKT3_EVENT_WRITE, 2, 0));
radeon_emit(cs, EVENT_TYPE(V_028A90_ZPASS_DONE) | EVENT_INDEX(1));
radeon_emit(cs, va + 8);
va += pool->stride * query;
- emit_begin_query(cmd_buffer, va, pool->type);
+ emit_begin_query(cmd_buffer, va, pool->type, flags);
}
* currently be active, which means the BO is already in the list.
*/
emit_end_query(cmd_buffer, va, avail_va, pool->type);
+
+ /*
+ * For multiview we have to emit a query for each bit in the mask,
+ * however the first query we emit will get the totals for all the
+ * operations, so we don't want to get a real value in the other
+ * queries. This emits a fake begin/end sequence so the waiting
+ * code gets a completed query value and doesn't hang, but the
+ * query returns 0.
+ */
+ if (cmd_buffer->state.subpass && cmd_buffer->state.subpass->view_mask) {
+ uint64_t avail_va = va + pool->availability_offset + 4 * query;
+
+
+ for (unsigned i = 1; i < util_bitcount(cmd_buffer->state.subpass->view_mask); i++) {
+ va += pool->stride;
+ avail_va += 4;
+ emit_begin_query(cmd_buffer, va, pool->type, 0);
+ emit_end_query(cmd_buffer, va, avail_va, pool->type);
+ }
+ }
}
void radv_CmdWriteTimestamp(
radv_cs_add_buffer(cmd_buffer->device->ws, cs, pool->bo, 5);
- MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 28);
-
- switch(pipelineStage) {
- case VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT:
- radeon_emit(cs, PKT3(PKT3_COPY_DATA, 4, 0));
- radeon_emit(cs, COPY_DATA_COUNT_SEL | COPY_DATA_WR_CONFIRM |
- COPY_DATA_SRC_SEL(COPY_DATA_TIMESTAMP) |
- COPY_DATA_DST_SEL(V_370_MEM_ASYNC));
- radeon_emit(cs, 0);
- radeon_emit(cs, 0);
- radeon_emit(cs, query_va);
- radeon_emit(cs, query_va >> 32);
-
- 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, avail_va);
- radeon_emit(cs, avail_va >> 32);
- radeon_emit(cs, 1);
- break;
- default:
- si_cs_emit_write_event_eop(cs,
- false,
- cmd_buffer->device->physical_device->rad_info.chip_class,
- mec,
- V_028A90_BOTTOM_OF_PIPE_TS, 0,
- 3, query_va, 0, 0);
- si_cs_emit_write_event_eop(cs,
- false,
- cmd_buffer->device->physical_device->rad_info.chip_class,
- mec,
- V_028A90_BOTTOM_OF_PIPE_TS, 0,
- 1, avail_va, 0, 1);
- break;
- }
+ int num_queries = 1;
+ if (cmd_buffer->state.subpass && cmd_buffer->state.subpass->view_mask)
+ num_queries = util_bitcount(cmd_buffer->state.subpass->view_mask);
+
+ MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 28 * num_queries);
+ for (unsigned i = 0; i < num_queries; i++) {
+ switch(pipelineStage) {
+ case VK_PIPELINE_STAGE_TOP_OF_PIPE_BIT:
+ radeon_emit(cs, PKT3(PKT3_COPY_DATA, 4, 0));
+ radeon_emit(cs, COPY_DATA_COUNT_SEL | COPY_DATA_WR_CONFIRM |
+ COPY_DATA_SRC_SEL(COPY_DATA_TIMESTAMP) |
+ COPY_DATA_DST_SEL(V_370_MEM_ASYNC));
+ radeon_emit(cs, 0);
+ radeon_emit(cs, 0);
+ radeon_emit(cs, query_va);
+ radeon_emit(cs, query_va >> 32);
+
+ 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, avail_va);
+ radeon_emit(cs, avail_va >> 32);
+ radeon_emit(cs, 1);
+ break;
+ default:
+ si_cs_emit_write_event_eop(cs,
+ false,
+ cmd_buffer->device->physical_device->rad_info.chip_class,
+ mec,
+ V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ 3, query_va, 0, 0);
+ si_cs_emit_write_event_eop(cs,
+ false,
+ cmd_buffer->device->physical_device->rad_info.chip_class,
+ mec,
+ V_028A90_BOTTOM_OF_PIPE_TS, 0,
+ 1, avail_va, 0, 1);
+ break;
+ }
+ query_va += pool->stride;
+ avail_va += 4;
+ }
assert(cmd_buffer->cs->cdw <= cdw_max);
}
projects / mesa.git / blobdiff