Revert "radv: detect command buffers that do no work and drop them (v2)"

This just keeps popping up minor problems and regressions we should
revisit in a more sustainable manner later.

This also reverts:
Revert "radv: query cmds should mark a cmd buffer as having draws."
Revert "radv: also fixup event emission to not get culled."

This reverts commit d1640e79328af4a63c056e3ccab299a3bffbede7.
This reverts commit 8b47b97215af7157bc15676167cab73aa5a61a76.
This reverts commit b4b19afebeac84ddce3f1f8b7ffe0da6474fe29a.

Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
Signed-off-by: Dave Airlie <airlied@redhat.com>

6 files changed, 8 insertions, 32 deletions

diff --git a/src/amd/vulkan/radv_cmd_buffer.c b/src/amd/vulkan/radv_cmd_buffer.c
index ef44859ccf0..5aea0e9de4b 100644
--- a/src/amd/vulkan/radv_cmd_buffer.c
+++ b/src/amd/vulkan/radv_cmd_buffer.c
@@ -1278,7 +1278,6 @@ radv_cmd_buffer_flush_state(struct radv_cmd_buffer *cmd_buffer, bool instanced_o
 	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws,
 							   cmd_buffer->cs, 4096);
 
-	cmd_buffer->no_draws = false;
 	if ((cmd_buffer->state.vertex_descriptors_dirty || cmd_buffer->state.vb_dirty) &&
 	    cmd_buffer->state.pipeline->num_vertex_attribs) {
 		unsigned vb_offset;
@@ -1600,7 +1599,6 @@ static void  radv_reset_cmd_buffer(struct radv_cmd_buffer *cmd_buffer)
 	cmd_buffer->record_fail = false;
 
 	cmd_buffer->ring_offsets_idx = -1;
-	cmd_buffer->no_draws = true;
 }
 
 VkResult radv_ResetCommandBuffer(
@@ -2447,7 +2445,6 @@ void radv_CmdDrawIndexedIndirectCountAMD(
 static void
 radv_flush_compute_state(struct radv_cmd_buffer *cmd_buffer)
 {
-	cmd_buffer->no_draws = false;
 	radv_emit_compute_pipeline(cmd_buffer);
 	radv_flush_descriptors(cmd_buffer, cmd_buffer->state.compute_pipeline,
 			       VK_SHADER_STAGE_COMPUTE_BIT);
@@ -2890,7 +2887,6 @@ static void write_event(struct radv_cmd_buffer *cmd_buffer,
 	uint64_t va = cmd_buffer->device->ws->buffer_get_va(event->bo);
 
 	cmd_buffer->device->ws->cs_add_buffer(cs, event->bo, 8);
-	cmd_buffer->no_draws = false;
 
 	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 12);
 
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
index 6f2fac2a9fe..7900ece9c89 100644
--- a/src/amd/vulkan/radv_device.c
+++ b/src/amd/vulkan/radv_device.c
@@ -1461,18 +1461,8 @@ VkResult radv_QueueSubmit(
 		struct radeon_winsys_cs **cs_array;
 		bool can_patch = true;
 		uint32_t advance;
-		int draw_cmd_buffers_count = 0;
 
-		for (uint32_t j = 0; j < pSubmits[i].commandBufferCount; j++) {
-			RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer,
-					 pSubmits[i].pCommandBuffers[j]);
-			assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
-			if (cmd_buffer->no_draws == true)
-				continue;
-			draw_cmd_buffers_count++;
-		}
-
-		if (!draw_cmd_buffers_count) {
+		if (!pSubmits[i].commandBufferCount) {
 			if (pSubmits[i].waitSemaphoreCount || pSubmits[i].signalSemaphoreCount) {
 				ret = queue->device->ws->cs_submit(ctx, queue->queue_idx,
 								   &queue->device->empty_cs[queue->queue_family_index],
@@ -1491,27 +1481,24 @@ VkResult radv_QueueSubmit(
 			continue;
 		}
 
-		cs_array = malloc(sizeof(struct radeon_winsys_cs *) * draw_cmd_buffers_count);
+		cs_array = malloc(sizeof(struct radeon_winsys_cs *) *
+					        pSubmits[i].commandBufferCount);
 
-		int draw_cmd_buffer_idx = 0;
 		for (uint32_t j = 0; j < pSubmits[i].commandBufferCount; j++) {
 			RADV_FROM_HANDLE(radv_cmd_buffer, cmd_buffer,
 					 pSubmits[i].pCommandBuffers[j]);
 			assert(cmd_buffer->level == VK_COMMAND_BUFFER_LEVEL_PRIMARY);
-			if (cmd_buffer->no_draws == true)
-				continue;
 
-			cs_array[draw_cmd_buffer_idx] = cmd_buffer->cs;
-			draw_cmd_buffer_idx++;
+			cs_array[j] = cmd_buffer->cs;
 			if ((cmd_buffer->usage_flags & VK_COMMAND_BUFFER_USAGE_SIMULTANEOUS_USE_BIT))
 				can_patch = false;
 		}
 
-		for (uint32_t j = 0; j < draw_cmd_buffers_count; j += advance) {
+		for (uint32_t j = 0; j < pSubmits[i].commandBufferCount; j += advance) {
 			advance = MIN2(max_cs_submission,
-				       draw_cmd_buffers_count - j);
+				       pSubmits[i].commandBufferCount - j);
 			bool b = j == 0;
-			bool e = j + advance == draw_cmd_buffers_count;
+			bool e = j + advance == pSubmits[i].commandBufferCount;
 
 			if (queue->device->trace_bo)
 				*queue->device->trace_id_ptr = 0;
diff --git a/src/amd/vulkan/radv_meta_buffer.c b/src/amd/vulkan/radv_meta_buffer.c
index 4857d3dc54f..cd2973fa4a9 100644
--- a/src/amd/vulkan/radv_meta_buffer.c
+++ b/src/amd/vulkan/radv_meta_buffer.c
@@ -523,7 +523,6 @@ void radv_CmdUpdateBuffer(
 	assert(!(dataSize & 3));
 	assert(!(va & 3));
 
-	cmd_buffer->no_draws = false;
 	if (dataSize < 4096) {
 		cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, dst_buffer->bo, 8);
 
diff --git a/src/amd/vulkan/radv_private.h b/src/amd/vulkan/radv_private.h
index 099dba3f643..ac21b075c2c 100644
--- a/src/amd/vulkan/radv_private.h
+++ b/src/amd/vulkan/radv_private.h
@@ -746,8 +746,6 @@ struct radv_cmd_buffer {
 	uint32_t gsvs_ring_size_needed;
 
 	int ring_offsets_idx; /* just used for verification */
-
-	bool no_draws;
 };
 
 struct radv_image;
diff --git a/src/amd/vulkan/radv_query.c b/src/amd/vulkan/radv_query.c
index fd5d0659a9e..a29a05d4b84 100644
--- a/src/amd/vulkan/radv_query.c
+++ b/src/amd/vulkan/radv_query.c
@@ -211,7 +211,6 @@ void radv_CmdCopyQueryPoolResults(
 
 	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, pool->bo, 8);
 	cmd_buffer->device->ws->cs_add_buffer(cmd_buffer->cs, dst_buffer->bo, 8);
-	cmd_buffer->no_draws = false;
 
 	for(unsigned i = 0; i < queryCount; ++i, dest_va += stride) {
 		unsigned query = firstQuery + i;
@@ -311,7 +310,6 @@ void radv_CmdBeginQuery(
 	va += pool->stride * query;
 
 	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 8);
-	cmd_buffer->no_draws = false;
 
 	switch (pool->type) {
 	case VK_QUERY_TYPE_OCCLUSION:
@@ -345,7 +343,6 @@ void radv_CmdEndQuery(
 	va += pool->stride * query;
 
 	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 8);
-	cmd_buffer->no_draws = false;
 
 	switch (pool->type) {
 	case VK_QUERY_TYPE_OCCLUSION:
@@ -397,7 +394,6 @@ void radv_CmdWriteTimestamp(
 	uint64_t query_va = va + pool->stride * query;
 
 	cmd_buffer->device->ws->cs_add_buffer(cs, pool->bo, 5);
-	cmd_buffer->no_draws = false;
 
 	MAYBE_UNUSED unsigned cdw_max = radeon_check_space(cmd_buffer->device->ws, cs, 12);
 
diff --git a/src/amd/vulkan/si_cmd_buffer.c b/src/amd/vulkan/si_cmd_buffer.c
index e20e3bd101e..e2ba413b835 100644
--- a/src/amd/vulkan/si_cmd_buffer.c
+++ b/src/amd/vulkan/si_cmd_buffer.c
@@ -923,7 +923,7 @@ static void si_emit_cp_dma_clear_buffer(struct radv_cmd_buffer *cmd_buffer,
 static void si_cp_dma_prepare(struct radv_cmd_buffer *cmd_buffer, uint64_t byte_count,
 			      uint64_t remaining_size, unsigned *flags)
 {
-	cmd_buffer->no_draws = false;
+
 	/* Flush the caches for the first copy only.
 	 * Also wait for the previous CP DMA operations.
 	 */