Page MenuHome
Differential D1224

Cycles: use native saturate function for CUDA
ClosedPublic
Actions

Authored by Sv. Lockal (lockal) on Apr 7 2015, 10:47 PM.

Details

Reviewers
Sergey Sharybin (sergey)
Commits
rBSae7d84dbc193: Cycles: Use native saturate function for CUDA
rC20a43dbf8eeb: Use native saturate function for CUDA
rBae7d84dbc193: Cycles: Use native saturate function for CUDA
Summary

nvcc can't optimize clamp(x, 0.0f, 1.0f) into single instruction, and uses 4 instructions instead. By using builtin saturate function we could make code a little bit more clean and optimized.

Diff Detail

Repository
rB Blender

Event Timeline

Sv. Lockal (lockal) updated this revision to Diff 3925.Apr 7 2015, 10:47 PM
Sv. Lockal (lockal) retitled this revision from to Cycles: use native saturate function for CUDA.
Sv. Lockal (lockal) updated this object.
Sv. Lockal (lockal) added a reviewer: Sergey Sharybin (sergey).
Sv. Lockal (lockal) added a project: Cycles.
Sv. Lockal (lockal) added a comment.Apr 7 2015, 10:51 PM

Common diff in PTX (same for SASS), 92 replacements:

< 	mov.f32 	%f6781, 0f00000000;
< 	max.ftz.f32 	%f6782, %f1500, %f6781;
< 	mov.f32 	%f6783, 0f3F800000;
< 	min.ftz.f32 	%f10110, %f6782, %f6783;
---
> 	cvt.ftz.sat.f32.f32	%f10029, %f1500;
Sergey Sharybin (sergey) edited edge metadata.Apr 8 2015, 12:10 AM

Generally looks fine, some inlined question about uchar conversion.

It's also a good idea to include some number into the commit message, like how much % of speedup you've got for some benchmark scene.

intern/cycles/kernel/kernel_film.h
40

Did you test if saturating the whole vector and then converting to uchar gives better performance?

Thomas Dinges (dingto) added a subscriber: Thomas Dinges (dingto).Apr 14 2015, 2:47 PM
Closed by commit rBae7d84dbc193: Cycles: Use native saturate function for CUDA (authored by Sergey Sharybin (sergey)). · Explain WhyApr 27 2015, 9:49 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents
Changeset List

PathSize
intern/
cycles/
kernel/
closure/
16 lines
2 lines
8 lines
6 lines
geom/
4 lines
10 lines
4 lines
2 lines
svm/
2 lines
4 lines
2 lines
6 lines
2 lines
8 lines
2 lines
2 lines
render/
4 lines
2 lines
util/
9 lines

Diff 4088

View Options

intern/cycles/kernel/closure/bsdf_microfacet.h

Loading...
View Options

intern/cycles/kernel/closure/bsdf_oren_nayar.h

Loading...
View Options

intern/cycles/kernel/closure/bsdf_toon.h

Loading...
View Options

intern/cycles/kernel/closure/bssrdf.h

Loading...
View Options

intern/cycles/kernel/geom/geom_curve.h

Loading...
View Options

intern/cycles/kernel/kernel_film.h

Loading...
View Options

intern/cycles/kernel/kernel_globals.h

Loading...
View Options

intern/cycles/kernel/kernel_passes.h

Loading...
View Options

intern/cycles/kernel/svm/svm_brick.h

Loading...
View Options

intern/cycles/kernel/svm/svm_closure.h

Loading...
View Options

intern/cycles/kernel/svm/svm_gradient.h

Loading...
View Options

intern/cycles/kernel/svm/svm_image.h

Loading...
View Options

intern/cycles/kernel/svm/svm_math_util.h

Loading...
View Options

intern/cycles/kernel/svm/svm_mix.h

Loading...
View Options

intern/cycles/kernel/svm/svm_musgrave.h

Loading...
View Options

intern/cycles/kernel/svm/svm_ramp.h

Loading...
View Options

intern/cycles/render/buffers.cpp

Loading...
View Options

intern/cycles/render/nodes.cpp

Loading...
View Options

intern/cycles/util/util_math.h

Loading...