AlexIndustrial/mesa
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

i965: Use a WC map and memcpy for the batch instead of pwrite.

Browse Source 
We'd like to eliminate the malloc'd shadow copy eventually, but there
are still unresolved performance problems.  In the meantime, let's at
least get rid of pwrite.

On Apollolake, improves Synmark OglBatch6 performance by:
1.53581% +/- 0.269589% (n=108).

Reviewed-by: Matt Turner <mattst88@gmail.com>
Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk>
This commit is contained in:
Kenneth Graunke
2017-09-08 15:00:14 -07:00
parent 343aa09a22
commit 717e753912
1 changed files with 8 additions and 10 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/mesa/drivers/dri/i965/intel_batchbuffer.c
+8 -10
View File
@@ -629,18 +629,16 @@ do_flush_locked(struct brw_context *brw, int in_fence_fd, int *out_fence_fd)
struct intel_batchbuffer *batch = &brw->batch;
int ret = 0;
if (devinfo->has_llc) {
brw_bo_unmap(batch->bo);
} else {
ret = brw_bo_subdata(batch->bo, 0, 4 * USED_BATCH(*batch), batch->map);
if (ret == 0 && batch->state_batch_offset != batch->bo->size) {
ret = brw_bo_subdata(batch->bo,
batch->state_batch_offset,
batch->bo->size - batch->state_batch_offset,
(char *)batch->map + batch->state_batch_offset);
}
if (batch->cpu_map) {
void *bo_map = brw_bo_map(brw, batch->bo, MAP_WRITE);
memcpy(bo_map, batch->cpu_map, 4 * USED_BATCH(*batch));
memcpy(bo_map + batch->state_batch_offset,
(char *) batch->cpu_map + batch->state_batch_offset,
batch->bo->size - batch->state_batch_offset);
}
brw_bo_unmap(batch->bo);
if (!brw->screen->no_hw) {
/* The requirement for using I915_EXEC_NO_RELOC are:
*