diff options
| author | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2016-10-02 15:48:39 +0300 |
|---|---|---|
| committer | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2016-10-03 23:15:25 +0300 |
| commit | a3abb020e37a072eb71fd30de9ab125d1c16623a (patch) | |
| tree | b525be7f8a0792eedecb2b95802ede88dc3f330e /intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h | |
| parent | 49ad4215baf16d850d0e367f003ab688e4a3d08e (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/closure/bsdf_microfacet_multi_impl.h')
| -rw-r--r-- | intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h | 4 |
1 files changed, 2 insertions, 2 deletions
diff --git a/intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h b/intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h index 6ebe2f6a751..8054fa8e849 100644 --- a/intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h +++ b/intern/cycles/kernel/closure/bsdf_microfacet_multi_impl.h @@ -25,7 +25,7 @@ * energy is used. In combination with MIS, that is enough to produce an unbiased result, although * the balance heuristic isn't necessarily optimal anymore. */ -ccl_device_inline float3 MF_FUNCTION_FULL_NAME(mf_eval)( +ccl_device_forceinline float3 MF_FUNCTION_FULL_NAME(mf_eval)( float3 wi, float3 wo, const bool wo_outside, @@ -168,7 +168,7 @@ ccl_device_inline float3 MF_FUNCTION_FULL_NAME(mf_eval)( * escaped the surface in wo. The function returns the throughput between wi and wo. * Without reflection losses due to coloring or fresnel absorption in conductors, the sampling is optimal. */ -ccl_device float3 MF_FUNCTION_FULL_NAME(mf_sample)(float3 wi, float3 *wo, const float3 color, const float alpha_x, const float alpha_y, ccl_addr_space uint *lcg_state +ccl_device_forceinline float3 MF_FUNCTION_FULL_NAME(mf_sample)(float3 wi, float3 *wo, const float3 color, const float alpha_x, const float alpha_y, ccl_addr_space uint *lcg_state #ifdef MF_MULTI_GLASS , const float eta #elif defined(MF_MULTI_GLOSSY) |