Skip to content

ARROW-9079: [C++] Write benchmark for arithmetic kernels #7417

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
kszucs wants to merge 3 commits into from
Closed

ARROW-9079: [C++] Write benchmark for arithmetic kernels #7417

kszucs wants to merge 3 commits into from

Conversation

@kszucs
Copy link
Member

@kszucs kszucs commented Jun 12, 2020 •
edited
Loading

Quickly wanted to add a benchmark for the Add function to verify that no significant regressions were introduced by #7341

Before:

---------------------------------------------------------------------------------------
Benchmark                                Time           CPU Iterations UserCounters...
---------------------------------------------------------------------------------------
AddArrayArrayKernel/32768/10000         18 us         18 us      35892 null_percent=0.01 size=32.768k   1.67854GB/s
AddArrayArrayKernel/32768/100           19 us         19 us      37540 null_percent=1 size=32.768k   1.61941GB/s
AddArrayArrayKernel/32768/10            20 us         20 us      37049 null_percent=10 size=32.768k   1.55599GB/s
AddArrayArrayKernel/32768/2             20 us         20 us      35394 null_percent=50 size=32.768k   1.54512GB/s
AddArrayArrayKernel/32768/1             19 us         19 us      37901 null_percent=100 size=32.768k   1.63153GB/s

After:

---------------------------------------------------------------------------------------
Benchmark                                Time           CPU Iterations UserCounters...
---------------------------------------------------------------------------------------
AddArrayArrayKernel/32768/10000         19 us         19 us      36704 null_percent=0.01 size=32.768k   1.64619GB/s
AddArrayArrayKernel/32768/100           18 us         18 us      37194 null_percent=1 size=32.768k   1.67588GB/s
AddArrayArrayKernel/32768/10            18 us         18 us      36341 null_percent=10 size=32.768k   1.65205GB/s
AddArrayArrayKernel/32768/2             18 us         18 us      37502 null_percent=50 size=32.768k     1.662GB/s
AddArrayArrayKernel/32768/1             18 us         18 us      38622 null_percent=100 size=32.768k   1.66593GB/s

cc @wesm

kszucs
kszucs commented Jun 12, 2020
View reviewed changes
@github-actions
Copy link

github-actions bot commented Jun 12, 2020

https://issues.apache.org/jira/browse/ARROW-9079

@wesm
Copy link
Member

wesm commented Jun 12, 2020

Thanks for working on this. I'll check the benchmarks on MSVC also

@wesm
Copy link
Member

wesm commented Jun 12, 2020

There don't seem to be issues on MSVC

https://gist.github.com/wesm/45be57393b2d9186f87faae228f12380/revisions

wesm
wesm reviewed Jun 12, 2020
View reviewed changes
@wesm
Copy link
Member

wesm commented Jun 13, 2020

+1. I fixed a few lingering issues that jumped out at me, will merge this once build passes

wesm
wesm reviewed Jun 13, 2020
View reviewed changes
cpp/src/arrow/compute/kernels/scalar_arithmetic_benchmark.cc
bench->Unit(benchmark::kMicrosecond);

for (const auto size : kMemorySizes) {
for (const auto size : {kL1Size, kL2Size}) {
Copy link
Member

@wesm wesm Jun 13, 2020

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I have a processor with 22MB L3 cache so generating that much random data is quite expensive. If we want to benchmark arrays that big we should generate a smaller sample of random data and repeat/tile it to make the bigger array.

@wesm wesm closed this in 96279ee Jun 15, 2020
@kszucs
Copy link
Member Author

kszucs commented Jun 15, 2020

Thanks!

@asfimport asfimport mentioned this pull request Jun 15, 2020
Closed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@wesm wesm wesm left review comments

@bkietz bkietz bkietz left review comments

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@kszucs @wesm @bkietz