NeroReflex/mesa
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity

gallium/radeon: Do not use u_upload_mgr for buffer downloads

Browse source 
Instead create a staging texture with pipe_buffer_create and
PIPE_USAGE_STAGING.

u_upload_mgr sets the usage of its staging buffer to PIPE_USAGE_STREAM.
But since 150ac07b85 CPU -> GPU streaming buffers
are created in VRAM. Therefore the staging texture (in VRAM) does not offer any
performance improvements for buffer downloads.

Signed-off-by: Niels Ole Salscheider <niels_ole@salscheider-online.de>
Reviewed-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Marek Olšák <marek.olsak@amd.com>
This commit is contained in:
Niels Ole Salscheider 2014-08-14 20:22:26 +02:00 • committed by Marek Olšák
parent 498dc676ea
commit 5ae9bdafd4
1 changed files with 8 additions and 12 deletions
Download patch file Download diff file Expand all files Collapse all files

20
src/gallium/drivers/radeon/r600_buffer_common.c
View file

@ -303,26 +303,22 @@ static void *r600_buffer_transfer_map(struct pipe_context *ctx,
!(usage & PIPE_TRANSFER_WRITE) &&
rbuffer->domains == RADEON_DOMAIN_VRAM &&
r600_can_dma_copy_buffer(rctx, 0, box->x, box->width)) {
unsigned offset;
struct r600_resource *staging = NULL;
u_upload_alloc(rctx->uploader, 0,
box->width + (box->x % R600_MAP_BUFFER_ALIGNMENT),
&offset, (struct pipe_resource**)&staging, (void**)&data);
struct r600_resource *staging;
staging = (struct r600_resource*) pipe_buffer_create(
ctx->screen, PIPE_BIND_TRANSFER_READ, PIPE_USAGE_STAGING,
box->width + (box->x % R600_MAP_BUFFER_ALIGNMENT));
if (staging) {
data += box->x % R600_MAP_BUFFER_ALIGNMENT;
/* Copy the VRAM buffer to the staging buffer. */
rctx->dma_copy(ctx, &staging->b.b, 0,
offset + box->x % R600_MAP_BUFFER_ALIGNMENT,
box->x % R600_MAP_BUFFER_ALIGNMENT,
0, 0, resource, level, box);
/* Just do the synchronization. The buffer is mapped already. */
r600_buffer_map_sync_with_rings(rctx, staging, PIPE_TRANSFER_READ);
data = r600_buffer_map_sync_with_rings(rctx, staging, PIPE_TRANSFER_READ);
data += box->x % R600_MAP_BUFFER_ALIGNMENT;
return r600_buffer_get_transfer(ctx, resource, level, usage, box,
ptransfer, data, staging, offset);
ptransfer, data, staging, 0);
}
}