radeon: Reduce number of radeon_bo_map calls that require kernel entry.

DMA buffers are often remapped many times for singel cs. To reduce number of
mapping calls ending to kernel dma buffers are mapped when allocated and unmapped
when full or in cs flush. This makes indifual mapping calls in other parts of
code simple increment/decrement reference count which is faster than entering
kernel.

This improves Anholt's openarena benchmark from 36 fps to 44 fps.

1 files changed, 13 insertions, 3 deletions

diff --git a/src/mesa/drivers/dri/radeon/radeon_dma.c b/src/mesa/drivers/dri/radeon/radeon_dma.c
index d31e4e47ddb..22499bc38d1 100644
--- a/src/mesa/drivers/dri/radeon/radeon_dma.c
+++ b/src/mesa/drivers/dri/radeon/radeon_dma.c
@@ -184,6 +184,8 @@ void radeonRefillCurrentDmaRegion(radeonContextPtr rmesa, int size)
 	radeon_print(RADEON_DMA, RADEON_NORMAL, "%s size %d minimum_size %d\n",
 			__FUNCTION__, size, rmesa->dma.minimum_size);
 
+	if (!is_empty_list(&rmesa->dma.reserved))
+		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
 
 	if (is_empty_list(&rmesa->dma.free)
 	      || last_elem(&rmesa->dma.free)->bo->size < size) {
@@ -211,7 +213,7 @@ again_alloc:
 
 	rmesa->dma.current_used = 0;
 	rmesa->dma.current_vertexptr = 0;
-	
+
 	if (radeon_cs_space_check_with_bo(rmesa->cmdbuf.cs,
 					  first_elem(&rmesa->dma.reserved)->bo,
 					  RADEON_GEM_DOMAIN_GTT, 0))
@@ -221,6 +223,7 @@ again_alloc:
         /* Cmd buff have been flushed in radeon_revalidate_bos */
 		goto again_alloc;
 	}
+	radeon_bo_map(first_elem(&rmesa->dma.reserved)->bo, 1);
 }
 
 /* Allocates a region from rmesa->dma.current.  If there isn't enough
@@ -332,6 +335,10 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa)
 		/* request updated cs processing information from kernel */
 		legacy_track_pending(rmesa->radeonScreen->bom, 0);
 	}
+
+	if (!is_empty_list(&rmesa->dma.reserved))
+		radeon_bo_unmap(first_elem(&rmesa->dma.reserved)->bo);
+
 	/* move waiting bos to free list.
 	   wait list provides gpu time to handle data before reuse */
 	foreach_s(dma_bo, temp, &rmesa->dma.wait) {
@@ -349,8 +356,11 @@ void radeonReleaseDmaRegions(radeonContextPtr rmesa)
 		   FREE(dma_bo);
 		   continue;
 		}
-		if (!radeon_bo_is_idle(dma_bo->bo))
+		if (!radeon_bo_is_idle(dma_bo->bo)) {
+			if (rmesa->radeonScreen->driScreen->dri2.enabled)
+				break;
 			continue;
+		}
 		remove_from_list(dma_bo);
 		dma_bo->expire_counter = expire_at;
 		insert_at_tail(&rmesa->dma.free, dma_bo);
@@ -388,7 +398,7 @@ void rcommon_flush_last_swtcl_prim( GLcontext *ctx  )
 {
 	radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
 	struct radeon_dma *dma = &rmesa->dma;
-		
+
 	if (RADEON_DEBUG & RADEON_IOCTL)
 		fprintf(stderr, "%s\n", __FUNCTION__);
 	dma->flush = NULL;