Skip to content

InlinedIndexOf improvement for P10 - #13390

Merged
zl-wang merged 1 commit into
eclipse-openj9:masterfrom
mnalam-p:indexOf
Oct 22, 2021
Merged

zl-wang merged 1 commit into
eclipse-openj9:masterfrom
mnalam-p:indexOf

Conversation

@mnalam-p

Copy link
Copy Markdown
Contributor

This implementation uses fewer branch and new P10 instructions to
improve the performance on a P10 machine.

@mnalam-p

Copy link
Copy Markdown
Contributor Author

Depends on: eclipse-omr/omr#6161

@mnalam-p

Copy link
Copy Markdown
Contributor Author

@zl-wang @ymanton kindly review

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

do you have performance data for this special-casing? i.e. for a general case, does this special-casing help or hurt the performance?

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

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

I don't have the performance data right now, but I did ran the two versions (one with special checking and another one without). In summary, in general case, the bumble bench performance was very similar. However, I did not ran the two version to becnhmark the special case (discussed in the previous implementation). I kept it because a) in general it didn't hurt performance b) it has a shorter code path for the special case.
I can redo the benchmark for the special case.

This implementation uses fewer branch and new P10 instructions to
improve the performance on a P10 machine.
@zl-wang

zl-wang commented Oct 22, 2021

Copy link
Copy Markdown
Contributor

Jenkins test sanity.functional plinux,aix jdk8

@zl-wang
zl-wang merged commit f774136 into eclipse-openj9:master Oct 22, 2021
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants