Blender Git Commit Log

Git Commits -> Revision 47da8dc

Revision 47da8dc by Stefan Werner (master)
March 14, 2019, 10:45 (GMT)
Cycles: Improved thread order for better CUDA performance.

This patch puts threads that render the same pixel closer together,
as opposed to threads that render the same sample. Thus threads
within a warp are more coherent in memory access and control flow,
leading to performance improvements.

Example benchmarks on a Quadro RTX4000 (WDDM) on Windows 10:
Koro: 4:23 -> 3:46
BMW: 1:18 -> 1:25
Barbershop Interior: 17:52 -> 14:55
Classroom: 4:37 -> 3:45

Performance differences on OpenCL/AMD were hit and miss, some scenes
became faster, others lost significantly. Therefore, this is kept as
CUDA only change for now.

Commit Details:

Full Hash: 47da8dcbcad4ccc5349bc303394e1d01d1c822c5
Parent Commit: 4887baf
Lines Changed: +6, -0

1 Modified Path:

/intern/cycles/kernel/kernel_work_stealing.h (+6, -0) (Diff)
By: Miika HämäläinenLast update: Nov-07-2014 14:18MiikaHweb | 2003-2021