[Mesa-dev] [PATCH] radv: only enable PERFECT_ZPASS_COUNTS for precision occlusion queries

[Samuel Pitoiset]

This unnecessary when the precision bit flag is not set, and this
might hurt performance. The Vulkan explains that not setting
VK_QUERY_CONTROL_PRECISE_BIT might be more efficient on some
implementations.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
---
 src/amd/vulkan/radv_cmd_buffer.c |  4 +++-
 src/amd/vulkan/radv_private.h    |  1 +
 src/amd/vulkan/radv_query.c      | 36 +++++++++++++++++++++++++++++++-----
 3 files changed, 35 insertions(+), 6 deletions(-)

diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c
index 526b618f2a..c0f60ac0d2 100644
--- a/src/amd/vulkan/radv_cmd_buffer.c
+++ b/src/amd/vulkan/radv_cmd_buffer.c
@@ -1217,9 +1217,11 @@ void radv_set_db_count_control(struct radv_cmd_buffer *cmd_buffer)
 	} else {
 		const struct radv_subpass *subpass = cmd_buffer->state.subpass;
 		uint32_t sample_rate = subpass ? util_logbase2(subpass->max_sample_count) : 0;
+		bool perfect = cmd_buffer->state.perfect_occlusion_queries_enabled;
 
 		if (cmd_buffer->device->physical_device->rad_info.chip_class >= CIK) {
-			db_count_control = S_028004_PERFECT_ZPASS_COUNTS(1) |
+			db_count_control =
+				S_028004_PERFECT_ZPASS_COUNTS(perfect) |
 				S_028004_SAMPLE_RATE(sample_rate) |
 				S_028004_ZPASS_ENABLE(1) |
 				S_028004_SLICE_EVEN_ENABLE(1) |
diff --git a/src/amd/vulkan/radv_private.h b/src/amd/vulkan/radv_private.h
index ca3beba2d4..9e655af844 100644
--- a/src/amd/vulkan/radv_private.h
+++ b/src/amd/vulkan/radv_private.h
@@ -946,6 +946,7 @@ struct radv_cmd_state {
 	uint32_t                                     last_primitive_reset_index;
 	enum radv_cmd_flush_bits                     flush_bits;
 	unsigned                                     active_occlusion_queries;
+	bool                                         perfect_occlusion_queries_enabled;
 	float					     offset_scale;
 	uint32_t                                      trace_id;
 	uint32_t                                      last_ia_multi_vgt_param;
diff --git a/src/amd/vulkan/radv_query.c b/src/amd/vulkan/radv_query.c
index cc943d5de0..859a4a1d68 100644
--- a/src/amd/vulkan/radv_query.c
+++ b/src/amd/vulkan/radv_query.c
@@ -1079,7 +1079,8 @@ void radv_CmdResetQueryPool(
 
 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) {
@@ -1087,8 +1088,27 @@ static void emit_begin_query(struct radv_cmd_buffer *cmd_buffer,
 		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));
@@ -1119,8 +1139,14 @@ static void emit_end_query(struct radv_cmd_buffer *cmd_buffer,
 		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) {
+			/* Reset the perfect occlusion queries hint now that no
+			 * queries are active.
+			 */
+			cmd_buffer->state.perfect_occlusion_queries_enabled = false;
+
 			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));
@@ -1177,7 +1203,7 @@ void radv_CmdBeginQuery(
 
 	va += pool->stride * query;
 
-	emit_begin_query(cmd_buffer, va, pool->type);
+	emit_begin_query(cmd_buffer, va, pool->type, flags);
 
 	/*
 	 * For multiview we have to emit a query for each bit in the mask,
@@ -1193,7 +1219,7 @@ void radv_CmdBeginQuery(
 		for (unsigned i = 0; i < util_bitcount(cmd_buffer->state.subpass->view_mask); i++) {
 			va += pool->stride;
 			avail_va += 4;
-			emit_begin_query(cmd_buffer, va, pool->type);
+			emit_begin_query(cmd_buffer, va, pool->type, flags);
 			emit_end_query(cmd_buffer, va, avail_va, pool->type);
 		}
 	}
-- 
2.16.3