Blender Git Commit Log

Git Commits -> Revision 89aa75d

Revision 89aa75d by Sergey Sharybin (cycles-x)

April 1, 2021, 17:14 (GMT)

Cycles: Move adaptive sampling convergence test to own kernel

This change makes it so the check is only done after all contribution
is known at the sample. It also makes it so path tracing kernel is
faster because it doesn't do per-bounce convergence test.

Makes it so that CPU samples distribution works similar to the
master branch.

The timing on CPU is way closer to the master branch, but more accurate
measurement with proper compilation flags is needed.

GPU adaptive sampling is temporarily broken.

Commit Details:

Full Hash: 89aa75d1f51af157b7bca29032514ba3596e6099
Parent Commit: 9ef790d
Lines Changed: +84, -33

13 Modified Paths:

/intern/cycles/device/cpu/kernel.cpp (+1, -0) (Diff)
/intern/cycles/device/cpu/kernel.h (+11, -0) (Diff)
/intern/cycles/integrator/path_trace.cpp (+4, -3) (Diff)
/intern/cycles/integrator/path_trace_work.h (+4, -1) (Diff)
/intern/cycles/integrator/path_trace_work_cpu.cpp (+17, -3) (Diff)
/intern/cycles/integrator/path_trace_work_cpu.h (+1, -1) (Diff)
/intern/cycles/integrator/path_trace_work_gpu.cpp (+1, -1) (Diff)
/intern/cycles/integrator/path_trace_work_gpu.h (+1, -1) (Diff)
/intern/cycles/kernel/device/cpu/kernel_arch.h (+9, -0) (Diff)
/intern/cycles/kernel/device/cpu/kernel_arch_impl.h (+16, -0) (Diff)
/intern/cycles/kernel/device/cuda/kernel.cu (+1, -3) (Diff)
/intern/cycles/kernel/kernel_accumulate.h (+0, -12) (Diff)
/intern/cycles/kernel/kernel_adaptive_sampling.h (+18, -8) (Diff)

Blender Git Commit Log

Git Commits -> Revision 89aa75d

Commit Details:

13 Modified Paths:

Navigation

Stats

RSS Feeds

Blender Git Commit Log

Git Commits -> Revision 89aa75d Go

Commit Details:

13 Modified Paths:

Navigation

Stats

RSS Feeds

Git Commits -> Revision 89aa75d