diff options
| author | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2016-10-02 15:48:39 +0300 |
|---|---|---|
| committer | Sergey Sharybin <sergey.vfx@gmail.com> | 2016-10-19 15:26:38 +0300 |
| commit | fd0dea585cfcb50a8b056f0d1101505f12b508bc (patch) | |
| tree | 67d79231d48fa3af86e505022c27a1638f42a876 /intern/cycles/kernel/kernel_accumulate.h | |
| parent | 4d14bd10c02db2b8dc53d6a6dc5f2d5d72fcbc88 (diff) | |
Fix Cycles CUDA performance on CUDA 8.0.
Mostly this is making inlining match CUDA 7.5 in a few performance critical
places. The end result is that performance is now better than before, possibly
due to less register spilling or other CUDA 8.0 compiler improvements.
On benchmarks scenes, there are 3% to 35% render time reductions. Stack memory
usage is reduced a little too.
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D2269
Diffstat (limited to 'intern/cycles/kernel/kernel_accumulate.h')
| -rw-r--r-- | intern/cycles/kernel/kernel_accumulate.h | 8 |
1 files changed, 1 insertions, 7 deletions
diff --git a/intern/cycles/kernel/kernel_accumulate.h b/intern/cycles/kernel/kernel_accumulate.h index 0e13b22bd2c..2a215240248 100644 --- a/intern/cycles/kernel/kernel_accumulate.h +++ b/intern/cycles/kernel/kernel_accumulate.h @@ -54,13 +54,7 @@ ccl_device_inline void bsdf_eval_init(BsdfEval *eval, ClosureType type, float3 v #endif } -/* TODO(sergey): This is just a workaround for annoying 6.5 compiler bug. */ -#if !defined(__KERNEL_CUDA__) || __CUDA_ARCH__ < 500 -ccl_device_inline -#else -ccl_device_noinline -#endif -void bsdf_eval_accum(BsdfEval *eval, ClosureType type, float3 value) +ccl_device_inline void bsdf_eval_accum(BsdfEval *eval, ClosureType type, float3 value) { #ifdef __PASSES__ if(eval->use_light_pass) { |