Commit: 4b14b33ea887e685937b7757af0c2093093b7c7e Author: Xavier Hallade Date: Wed Oct 26 10:35:18 2022 +0200 Branches: master https://developer.blender.org/rB4b14b33ea887e685937b7757af0c2093093b7c7e
Cycles: use packed float3 back for oneAPI This fixes a 15% performance regression silently introduced by 79ab76e156d4bde937335be784cdf220294600d5 that aligned the compact float3 on 16 bytes for oneAPI. Current change is minimalist, there are further cleanup opportunities such as removing packed_float3 definition for oneAPI but for some reason, it cuts the recovered speedup in half, so we're starting with this small fix for now. Reviewed by: brecht Differential Revision: https://developer.blender.org/D16340 =================================================================== M intern/cycles/util/types_float3.h =================================================================== diff --git a/intern/cycles/util/types_float3.h b/intern/cycles/util/types_float3.h index 87c6b1d3654..34430945c38 100644 --- a/intern/cycles/util/types_float3.h +++ b/intern/cycles/util/types_float3.h @@ -10,7 +10,12 @@ CCL_NAMESPACE_BEGIN #ifndef __KERNEL_NATIVE_VECTOR_TYPES__ +# ifdef __KERNEL_ONEAPI__ +/* Define float3 as packed for oneAPI. */ +struct float3 +# else struct ccl_try_align(16) float3 +# endif { # ifdef __KERNEL_GPU__ /* Compact structure for GPU. */ _______________________________________________ Bf-blender-cvs mailing list Bf-blender-cvs@blender.org List details, subscription details or unsubscribe: https://lists.blender.org/mailman/listinfo/bf-blender-cvs