From a3abb020e37a072eb71fd30de9ab125d1c16623a Mon Sep 17 00:00:00 2001 From: Brecht Van Lommel Date: Sun, 2 Oct 2016 14:48:39 +0200 Subject: 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 --- intern/cycles/kernel/kernel_accumulate.h | 8 +------- 1 file changed, 1 insertion(+), 7 deletions(-) (limited to 'intern/cycles/kernel/kernel_accumulate.h') diff --git a/intern/cycles/kernel/kernel_accumulate.h b/intern/cycles/kernel/kernel_accumulate.h index dd5752f4c91..623c1dcaaa1 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 } } -/* 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) { -- cgit v1.2.3