Add broadcasting to kernel generator #1797

t4c1 · 2020-03-24T11:33:48Z

Summary

Adds broadcasting to kernel generator. If an expression is broadcasted in a dimmension it can be used as if it has any size in that dimmension when used in an operation that accepts more than one argument. The size is determined from other arguments to same expression.

Also contains a bugfix for determining view of the result.

Tests

Added tests for broadcasting and bug.

Side Effects

None.

Release notes

Added broadcasting to kernel generator.

Checklist

Math issue Implement OpenCL kernel generator #1342
Copyright holder: Tadej Ciglarič

The copyright holder is typically you or your assignee, such as a university or company. By submitting this pull request, the copyright holder is agreeing to the license the submitted work under the following licenses:
- Code: BSD 3-clause (https://opensource.org/licenses/BSD-3-Clause)
- Documentation: CC-BY 4.0 (https://creativecommons.org/licenses/by/4.0/)
the basic tests are passing
- unit tests pass (to run, use: ./runTests.py test/unit)
- header checks pass, (make test-headers)
- dependencies checks pass, (make test-math-dependencies)
- docs build, (make doxygen)
- code passes the built in C++ standards checks (make cpplint)
the code is written in idiomatic C++ and changes are documented in the doxygen
the new changes are tested

…xp1~20180509124008.99 (branches/release_50)

…stable/2017-11-14)

stan-buildbot · 2020-03-25T21:29:14Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.93	4.84	1.02	1.86% faster
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.96	-3.92% slower
eight_schools/eight_schools.stan	0.09	0.09	1.03	2.78% faster
gp_regr/gp_regr.stan	0.22	0.22	1.01	1.46% faster
irt_2pl/irt_2pl.stan	6.47	6.46	1.0	0.17% faster
performance.compilation	88.6	86.77	1.02	2.07% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.52	7.53	1.0	-0.08% slower
pkpd/one_comp_mm_elim_abs.stan	20.14	20.24	0.99	-0.51% slower
sir/sir.stan	91.96	91.29	1.01	0.73% faster
gp_regr/gen_gp_data.stan	0.05	0.05	1.0	-0.13% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.95	2.96	1.0	-0.33% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.31	0.31	0.98	-2.54% slower
arK/arK.stan	1.77	1.74	1.02	1.72% faster
arma/arma.stan	0.67	0.67	0.99	-0.53% slower
garch/garch.stan	0.52	0.51	1.01	0.6% faster
Mean result: 1.00253158827

Jenkins Console Log
Blue Ocean
Commit hash: 9d46311

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

SteveBronder

Nice! Couple lil' comments and one API question

SteveBronder · 2020-03-30T18:36:04Z

stan/math/opencl/kernel_generator/broadcast.hpp

+  inline kernel_parts generate(const std::string& i, const std::string& j,
+                               const std::string& var_name_arg) const {


[seperate pr]

Reading this over wrt modify_argument_indices it's kind of confusing to read i and j here as the names of the index but in modify_argument_indices as the actual index values. It may be good here to change the argument names to something like i_name and j_name. Or maybe like row_idx_name and col_idx_name

stan/math/opencl/kernel_generator/broadcast.hpp

test/unit/math/opencl/kernel_generator/broadcast_test.cpp

Co-Authored-By: Steve Bronder <Stevo15025@gmail.com>

…xp1~20180509124008.99 (branches/release_50)

…tatcomp/math into cl_kernel_generator_broadcast

…gs/RELEASE_500/final)

stan-buildbot · 2020-04-01T16:09:36Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.86	4.87	1.0	-0.15% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	0.98	-1.78% slower
eight_schools/eight_schools.stan	0.09	0.09	1.01	1.21% faster
gp_regr/gp_regr.stan	0.22	0.22	0.99	-0.87% slower
irt_2pl/irt_2pl.stan	6.48	6.45	1.01	0.59% faster
performance.compilation	88.23	87.14	1.01	1.23% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.52	7.51	1.0	0.06% faster
pkpd/one_comp_mm_elim_abs.stan	21.48	20.04	1.07	6.71% faster
sir/sir.stan	92.94	93.88	0.99	-1.02% slower
gp_regr/gen_gp_data.stan	0.05	0.05	0.98	-2.14% slower
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.95	2.96	1.0	-0.17% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.31	0.31	1.01	1.41% faster
arK/arK.stan	1.89	1.74	1.09	8.03% faster
arma/arma.stan	0.66	0.65	1.01	1.42% faster
garch/garch.stan	0.51	0.51	1.0	-0.37% slower
Mean result: 1.01035490013

Jenkins Console Log
Blue Ocean
Commit hash: 78852c4

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

SteveBronder

One doc thing that would be good to change since I figure we'll forget later but I'll approve now. If you make the change I'll re-approve rq

SteveBronder · 2020-04-01T17:25:32Z

stan/math/opencl/kernel_generator/broadcast.hpp

+}
+
+/**
+ * Brodcast an expression in colwise dimmension. The argument must have single


Can you just find/replace Bordcast with Broadcast

…stable/2017-11-14)

SteveBronder

lgtm!

stan-buildbot · 2020-04-01T21:54:38Z

Name	Old Result	New Result	Ratio	Performance change( 1 - new / old )
gp_pois_regr/gp_pois_regr.stan	4.83	4.85	0.99	-0.52% slower
low_dim_corr_gauss/low_dim_corr_gauss.stan	0.02	0.02	1.01	1.28% faster
eight_schools/eight_schools.stan	0.09	0.1	0.93	-7.58% slower
gp_regr/gp_regr.stan	0.22	0.22	0.99	-1.31% slower
irt_2pl/irt_2pl.stan	6.45	6.45	1.0	0.1% faster
performance.compilation	89.37	87.09	1.03	2.54% faster
low_dim_gauss_mix_collapse/low_dim_gauss_mix_collapse.stan	7.52	7.57	0.99	-0.59% slower
pkpd/one_comp_mm_elim_abs.stan	20.46	20.65	0.99	-0.93% slower
sir/sir.stan	90.95	93.57	0.97	-2.89% slower
gp_regr/gen_gp_data.stan	0.05	0.05	1.01	0.68% faster
low_dim_gauss_mix/low_dim_gauss_mix.stan	2.95	2.96	1.0	-0.49% slower
pkpd/sim_one_comp_mm_elim_abs.stan	0.31	0.34	0.92	-8.99% slower
arK/arK.stan	1.73	1.73	1.0	-0.01% slower
arma/arma.stan	0.66	0.66	0.99	-0.62% slower
garch/garch.stan	0.52	0.51	1.02	1.75% faster
Mean result: 0.989293267294

Jenkins Console Log
Blue Ocean
Commit hash: 4154ff4

Machine information

ProductName: Mac OS X ProductVersion: 10.11.6 BuildVersion: 15G22010

CPU:
Intel(R) Xeon(R) CPU E5-1680 v2 @ 3.00GHz

G++:
Configured with: --prefix=/Applications/Xcode.app/Contents/Developer/usr --with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

Clang:
Apple LLVM version 7.0.2 (clang-700.1.81)
Target: x86_64-apple-darwin15.6.0
Thread model: posix

t4c1 and others added 10 commits March 23, 2020 14:23

tmp

b2c1cd7

Merge branch 'develop' into cl_kernel_generator_broadcast

d534196

finished brodcast

a5e2258

[Jenkins] auto-formatting by clang-format version 5.0.2-svn328729-1~e…

0e23ab3

…xp1~20180509124008.99 (branches/release_50)

fix cpplint

f82ec8a

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

d63ab94

…stable/2017-11-14)

typo

7fe7b85

fixed error handling

089ef37

Merge commit '6c432ff239a1a2ea15eaaebc89e1423c51c6a088' into HEAD

1e394c9

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

9d46311

…stable/2017-11-14)

rok-cesnovar self-requested a review March 30, 2020 11:29

SteveBronder requested changes Mar 30, 2020

View reviewed changes

t4c1 and others added 7 commits March 31, 2020 08:37

update comments according to review

e384822

Co-Authored-By: Steve Bronder <Stevo15025@gmail.com>

Merge commit '6a6c1f9aa6b5c41d2bb56d14566cda3beb6488ba' into HEAD

9a4bc79

[Jenkins] auto-formatting by clang-format version 5.0.2-svn328729-1~e…

fda087f

…xp1~20180509124008.99 (branches/release_50)

added helpers rowwise_broadcast and colwise_broadcast

f49a83a

Merge branch 'cl_kernel_generator_broadcast' of https://github.com/bs…

3086f10

…tatcomp/math into cl_kernel_generator_broadcast

Merge commit '004c1525aa0cbd8b04fe7731aff764b330e0872a' into HEAD

579c105

[Jenkins] auto-formatting by clang-format version 5.0.0-3~16.04.1 (ta…

78852c4

…gs/RELEASE_500/final)

SteveBronder previously approved these changes Apr 1, 2020

View reviewed changes

t4c1 dismissed SteveBronder’s stale review via 51b091b April 1, 2020 17:51

t4c1 and others added 3 commits April 1, 2020 19:51

fix doc typo

51b091b

Merge commit 'ad4600f08636fe26d3781679bd88f05aabb44674' into HEAD

e88db6b

[Jenkins] auto-formatting by clang-format version 6.0.0 (tags/google/…

4154ff4

…stable/2017-11-14)

SteveBronder approved these changes Apr 1, 2020

View reviewed changes

t4c1 merged commit b225daf into stan-dev:develop Apr 2, 2020

SteveBronder mentioned this pull request Apr 16, 2020

Stan Math 3.2 release #1826

Closed

bbbales2 mentioned this pull request Apr 20, 2020

Stanc3 release for Cmdstan 2.23 stan-dev/stanc3#498

Closed

t4c1 deleted the cl_kernel_generator_broadcast branch July 10, 2020 07:09

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add broadcasting to kernel generator #1797

Add broadcasting to kernel generator #1797

t4c1 commented Mar 24, 2020 •

edited

Loading

stan-buildbot commented Mar 25, 2020

SteveBronder left a comment

SteveBronder Mar 30, 2020

stan-buildbot commented Apr 1, 2020

SteveBronder left a comment

SteveBronder Apr 1, 2020

SteveBronder left a comment

stan-buildbot commented Apr 1, 2020

		inline kernel_parts generate(const std::string& i, const std::string& j,
		const std::string& var_name_arg) const {

Add broadcasting to kernel generator #1797

Add broadcasting to kernel generator #1797

Conversation

t4c1 commented Mar 24, 2020 • edited Loading

Summary

Tests

Side Effects

Release notes

Checklist

stan-buildbot commented Mar 25, 2020

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder Mar 30, 2020

Choose a reason for hiding this comment

stan-buildbot commented Apr 1, 2020

SteveBronder left a comment

Choose a reason for hiding this comment

SteveBronder Apr 1, 2020

Choose a reason for hiding this comment

SteveBronder left a comment

Choose a reason for hiding this comment

stan-buildbot commented Apr 1, 2020

t4c1 commented Mar 24, 2020 •

edited

Loading