On 18/10/18 9:51 am, Bas Nieuwenhuizen wrote:
On Thu, Oct 18, 2018 at 12:04 AM Timothy Arceri <[email protected]> wrote:
Totals from affected shaders:
SGPRS: 2856 -> 2856 (0.00 %)
VGPRS: 3236 -> 3248 (0.37 %)
Spilled SGPRs: 0 -> 0 (0.00 %)
Spilled VGPRs: 0 -> 0 (0.00 %)
Private memory VGPRs: 0 -> 0 (0.00 %)
Scratch size: 0 -> 0 (0.00 %) dwords per thread
Code Size: 236560 -> 233548 (-1.27 %) bytes
LDS: 0 -> 0 (0.00 %) blocks
Max Waves: 277 -> 283 (2.17 %)
Wait states: 0 -> 0 (0.00 %)
Interesting that both max waves and VGPRs increased.
Yeah it was just one of those changes where the new NIR increased VGPR
use in a larger number of shaders compared to the number that reduced
enough to bump the max waves. However as I tried to indicate below the
increase of VGPRs is something that could likely be improved on the LLVM
side, the NIR itself looks much better.
Reviewed-by: Bas Nieuwenhuizen <[email protected]>
Even in the cases were we have increased VGPR use it appears
the NIR is improved significantly.
---
src/amd/vulkan/radv_shader.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/src/amd/vulkan/radv_shader.c b/src/amd/vulkan/radv_shader.c
index 3e3eb96a531..3b3422c8da6 100644
--- a/src/amd/vulkan/radv_shader.c
+++ b/src/amd/vulkan/radv_shader.c
@@ -127,6 +127,10 @@ radv_optimize_nir(struct nir_shader *shader, bool
optimize_conservatively)
NIR_PASS_V(shader, nir_lower_vars_to_ssa);
NIR_PASS_V(shader, nir_lower_pack);
+
+ NIR_PASS(progress, shader, nir_opt_copy_prop_vars);
+ NIR_PASS(progress, shader, nir_opt_dead_write_vars);
+
NIR_PASS_V(shader, nir_lower_alu_to_scalar);
NIR_PASS_V(shader, nir_lower_phis_to_scalar);
--
2.17.1
_______________________________________________
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
