radeonsi/gfx10: implement streamout-related queries

The NGG hardware pipeline doesn't track these statistics automatically,
and in fact *cannot* track them automatically when API geometry shaders
are involved, so we accumulate statistics in the shader using atomic
adds.

This implementation accumulates statistics via the memory system and
the RW buffer descriptor setup. We could use GDS, but since these
atomics aren't latency-sensitive, that basically just trades off
L2$ bandwidth vs. export bus bandwidth. One single memory transaction
per shader workgroup doesn't seem too bad. The result ring buffer in
memory is needed either way to avoid pipeline stalls.

The shader code contains the atomic unconditionally, though the
GFX10_GS_QUERY_BUF is a null buffer when no queries are active. The
atomic is simply discarded by the shader hardware in that case.

Acked-by: Bas Nieuwenhuizen <[email protected]>

1 files changed, 7 insertions, 0 deletions

diff --git a/src/gallium/drivers/radeonsi/si_pipe.c b/src/gallium/drivers/radeonsi/si_pipe.c
index c2cee024982..91b474d4d8f 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.c
+++ b/src/gallium/drivers/radeonsi/si_pipe.c
@@ -168,6 +168,9 @@ static void si_destroy_context(struct pipe_context *context)
 
 	si_release_all_descriptors(sctx);
 
+	if (sctx->chip_class >= GFX10)
+		gfx10_destroy_query(sctx);
+
 	pipe_resource_reference(&sctx->esgs_ring, NULL);
 	pipe_resource_reference(&sctx->gsvs_ring, NULL);
 	pipe_resource_reference(&sctx->tess_rings, NULL);
@@ -239,6 +242,8 @@ static void si_destroy_context(struct pipe_context *context)
 
 	if (sctx->query_result_shader)
 		sctx->b.delete_compute_state(&sctx->b, sctx->query_result_shader);
+	if (sctx->sh_query_result_shader)
+		sctx->b.delete_compute_state(&sctx->b, sctx->sh_query_result_shader);
 
 	if (sctx->gfx_cs)
 		sctx->ws->cs_destroy(sctx->gfx_cs);
@@ -516,6 +521,8 @@ static struct pipe_context *si_create_context(struct pipe_screen *screen,
 	/* Initialize graphics-only context functions. */
 	if (sctx->has_graphics) {
 		si_init_context_texture_functions(sctx);
+		if (sctx->chip_class >= GFX10)
+			gfx10_init_query(sctx);
 		si_init_msaa_functions(sctx);
 		si_init_shader_functions(sctx);
 		si_init_state_functions(sctx);