freedreno: avoid no-op flushes by re-using last-fence

Noticed that with webgl (in chromium, at least) we end up generating a
lot of no-op submits just to get a fence.  Tracking the last fence and
returning that if there is no rendering since last flush avoids this.

Signed-off-by: Rob Clark <robdclark@gmail.com>

src/gallium/drivers/freedreno/a6xx/fd6_blitter.c

@@ -28,6 +28,7 @@
 #include "util/u_dump.h"
 
 #include "freedreno_blitter.h"
+#include "freedreno_fence.h"
 #include "freedreno_resource.h"
 
 #include "fd6_blitter.h"
@@ -486,6 +487,8 @@ fd6_blit(struct pipe_context *pctx, const struct pipe_blit_info *info)
 		return;
 	}
 
+	fd_fence_ref(pctx->screen, &ctx->last_fence, NULL);
+
 	batch = fd_bc_alloc_batch(&ctx->screen->batch_cache, ctx, true);
 
 	fd6_emit_restore(batch, batch->draw);

src/gallium/drivers/freedreno/freedreno_context.c

@@ -49,13 +49,23 @@ fd_context_flush(struct pipe_context *pctx, struct pipe_fence_handle **fencep,
 
 	DBG("%p: flush: flags=%x\n", ctx->batch, flags);
 
+	/* if no rendering since last flush, ie. app just decided it needed
+	 * a fence, re-use the last one:
+	 */
+	if (ctx->last_fence) {
+		fd_fence_ref(pctx->screen, &fence, ctx->last_fence);
+		goto out;
+	}
+
 	if (!batch)
 		return;
 
 	/* Take a ref to the batch's fence (batch can be unref'd when flushed: */
 	fd_fence_ref(pctx->screen, &fence, batch->fence);
 
-	if (flags & PIPE_FLUSH_FENCE_FD)
+	/* TODO is it worth trying to figure out if app is using fence-fd's, to
+	 * avoid requesting one every batch?
+	 */
 	batch->needs_out_fence_fd = true;
 
 	if (!ctx->screen->reorder) {
@@ -66,9 +76,12 @@ fd_context_flush(struct pipe_context *pctx, struct pipe_fence_handle **fencep,
 		fd_bc_flush(&ctx->screen->batch_cache, ctx);
 	}
 
+out:
 	if (fencep)
 		fd_fence_ref(pctx->screen, fencep, fence);
 
+	fd_fence_ref(pctx->screen, &ctx->last_fence, fence);
+
 	fd_fence_ref(pctx->screen, &fence, NULL);
 }
 
@@ -137,6 +150,8 @@ fd_context_destroy(struct pipe_context *pctx)
 
 	DBG("");
 
+	fd_fence_ref(pctx->screen, &ctx->last_fence, NULL);
+
 	if (ctx->screen->reorder && util_queue_is_initialized(&ctx->flush_queue))
 		util_queue_destroy(&ctx->flush_queue);
 

src/gallium/drivers/freedreno/freedreno_context.h

@@ -230,6 +230,12 @@ struct fd_context {
 	 */
 	struct fd_batch *batch;
 
+	/* NULL if there has been rendering since last flush.  Otherwise
+	 * keeps a reference to the last fence so we can re-use it rather
+	 * than having to flush no-op batch.
+	 */
+	struct pipe_fence_handle *last_fence;
+
 	/* Are we in process of shadowing a resource? Used to detect recursion
 	 * in transfer_map, and skip unneeded synchronization.
 	 */

src/gallium/drivers/freedreno/freedreno_draw.c

@@ -34,6 +34,7 @@
 
 #include "freedreno_draw.h"
 #include "freedreno_context.h"
+#include "freedreno_fence.h"
 #include "freedreno_state.h"
 #include "freedreno_resource.h"
 #include "freedreno_query_acc.h"
@@ -98,6 +99,8 @@ fd_draw_vbo(struct pipe_context *pctx, const struct pipe_draw_info *info)
 		return;
 	}
 
+	fd_fence_ref(pctx->screen, &ctx->last_fence, NULL);
+
 	/* Upload a user index buffer. */
 	struct pipe_resource *indexbuf = NULL;
 	unsigned index_offset = 0;
@@ -382,6 +385,8 @@ fd_clear(struct pipe_context *pctx, unsigned buffers,
 	if (!fd_render_condition_check(pctx))
 		return;
 
+	fd_fence_ref(pctx->screen, &ctx->last_fence, NULL);
+
 	if (ctx->in_blit) {
 		fd_batch_reset(batch);
 		fd_context_all_dirty(ctx);

src/gallium/drivers/freedreno/freedreno_resource.c

@@ -35,6 +35,7 @@
 
 #include "freedreno_resource.h"
 #include "freedreno_batch_cache.h"
+#include "freedreno_fence.h"
 #include "freedreno_screen.h"
 #include "freedreno_surface.h"
 #include "freedreno_context.h"
@@ -1070,6 +1071,8 @@ void
 fd_blitter_pipe_begin(struct fd_context *ctx, bool render_cond, bool discard,
 		enum fd_render_stage stage)
 {
+	fd_fence_ref(ctx->base.screen, &ctx->last_fence, NULL);
+
 	util_blitter_save_fragment_constant_buffer_slot(ctx->blitter,
 			ctx->constbuf[PIPE_SHADER_FRAGMENT].cb);
 	util_blitter_save_vertex_buffer_slot(ctx->blitter, ctx->vtx.vertexbuf.vb);