On 10/18/2017 08:53 PM, Bas Nieuwenhuizen wrote:
I'd prefer not to. The current size is already huge when you consider
that a lot of applications use pretty small command buffers, adding
another 12k per command buffer is a bit much. I'd prefer not having
that overhead, since the GL_vs_VK benchmarks were IIRC not really
representative.
Yeah, good point, and I don't think this can improve any real applications.
On Wed, Oct 18, 2017 at 2:47 PM, Samuel Pitoiset
<[email protected]> wrote:
This reduces the number of conflicts when a bunch of buffers
are added to the CS list. It improves the benchmark #3 (from
GL_vs_VK) by +3/4%.
Signed-off-by: Samuel Pitoiset <[email protected]>
---
src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c
b/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c
index 53d428681c..4adbe27b2e 100644
--- a/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c
+++ b/src/amd/vulkan/winsys/amdgpu/radv_amdgpu_cs.c
@@ -59,7 +59,7 @@ struct radv_amdgpu_cs {
bool failed;
bool is_chained;
- int buffer_hash_table[1024];
+ int buffer_hash_table[4096];
unsigned hw_ip;
unsigned num_virtual_buffers;
--
2.14.2
_______________________________________________
mesa-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
_______________________________________________
mesa-dev mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/mesa-dev
