Skip to content

Implement vectorized absolute value and square root for all applicable vector types on PPC - #6489

Merged
knn-k merged 6 commits into
eclipse-omr:masterfrom
midronij:vops_final
May 17, 2022
Merged

knn-k merged 6 commits into
eclipse-omr:masterfrom
midronij:vops_final

Conversation

@midronij

@midronij midronij commented Apr 21, 2022 •

Copy link
Copy Markdown
Contributor

This contribution includes the following changes:

  • New IL opcodes vabs and vsqrt have been added
  • Codegen sequences for vectorized absolute value for all vector types have been implemented on PPC
  • Codegen sequences for vectorized square root for floating point types have been implemented on PPC
  • Binary encoder unit tests have been added for newly enabled PPC instructions

@midronij

Copy link
Copy Markdown
Contributor Author

@gita-omr @IBMJimmyk @aviansie-ben could you please review when you have a chance?

Comment thread compiler/il/VectorOperations.enum Resolved
@gita-omr

Copy link
Copy Markdown
Contributor

I think this PR does not depend on eclipse-openj9/openj9#14949. If you agree, please remove the Depend comment in the description and this will simplify the merge process.

@gita-omr

Copy link
Copy Markdown
Contributor

Jenkins build all

@midronij midronij changed the title WIP: Implement vectorized absolute value and square root for all applicable vector types on PPC Implement vectorized absolute value and square root for all applicable vector types on PPC Apr 27, 2022
@gita-omr

Copy link
Copy Markdown
Contributor

@midronij do you think linux_ppc-64_le_gcc failure is unrelated?

@BradleyWood

Copy link
Copy Markdown
Contributor

@midronij do you think linux_ppc-64_le_gcc failure is unrelated?

It looks like an issue in the testing infrastructure.

@midronij

midronij commented Apr 29, 2022 •

Copy link
Copy Markdown
Contributor Author

@gita-omr I believe so. The error is occurring in a test case that I didn't modify, and I can't reproduce it locally

@BradleyWood

Copy link
Copy Markdown
Contributor

Jenkins build plinux

Comment thread compiler/il/OMRILOps.hpp Resolved
Comment thread compiler/il/OMRILOps.hpp Resolved
Comment thread compiler/p/codegen/OMRTreeEvaluator.cpp Resolved
@midronij
midronij force-pushed the vops_final branch 5 times, most recently from 2841f8e to d5d416c Compare May 3, 2022 02:42
@0xdaryl

0xdaryl commented May 9, 2022

Copy link
Copy Markdown
Contributor

@midronij : please see PPC Linux CI failure.

@BradleyWood

Copy link
Copy Markdown
Contributor

Jenkins build plinux

@0xdaryl Its definitely not related to this PR.

@BradleyWood

Copy link
Copy Markdown
Contributor

Jenkins build plinux

@0xdaryl

0xdaryl commented May 9, 2022

Copy link
Copy Markdown
Contributor

I think the Power Linux failure that has been reproduced 4 times and only while testing this PR should be investigated before proceeding.

@midronij
midronij force-pushed the vops_final branch 5 times, most recently from 9f17a81 to 82d4f4c Compare May 13, 2022 18:06
@midronij

Copy link
Copy Markdown
Contributor Author

@gita-omr merge conflicts have been resolved!

@gita-omr

Copy link
Copy Markdown
Contributor

Jenkins build plinux

@midronij

Copy link
Copy Markdown
Contributor Author

@0xdaryl I've taken a look at the PPC failure that was occurring before, but I was unable to reproduce the error in any test environment. Additionally, it seemed to be occurring during a test that I didn't touch, and the most recent run of the CI build(done just now) passed. Can we proceed, or is further investigation needed?

midronij added 6 commits May 16, 2022 12:44
Signed-off-by: midronij <jackie.midroni@ibm.com>
Signed-off-by: midronij <jackie.midroni@ibm.com>
Signed-off-by: midronij <jackie.midroni@ibm.com>
Signed-off-by: midronij <jackie.midroni@ibm.com>
Signed-off-by: midronij <jackie.midroni@ibm.com>
…structions

Signed-off-by: midronij <jackie.midroni@ibm.com>
@gita-omr

Copy link
Copy Markdown
Contributor

@0xdaryl I've taken a look at the PPC failure that was occurring before, but I was unable to reproduce the error in any test environment. Additionally, it seemed to be occurring during a test that I didn't touch, and the most recent run of the CI build(done just now) passed. Can we proceed, or is further investigation needed?

@knn-k fyi

if (et == TR::Float || et == TR::Double)
return true;
else
return false;

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.

return (et == TR::Float || et == TR::Double); would be simpler for these lines.
I think it can be done in a separate PR.

@knn-k

knn-k commented May 17, 2022

Copy link
Copy Markdown
Contributor

Jenkins build all

@knn-k knn-k self-assigned this May 17, 2022
@knn-k

knn-k commented May 17, 2022

Copy link
Copy Markdown
Contributor

Jenkins build zos_390-64

@knn-k

knn-k commented May 17, 2022

Copy link
Copy Markdown
Contributor

No failure on plinux today.
The failure on zos_390-64 was an intermittent "git clone" failure, and it disappeared when I reran the build.
Merging.

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.

6 participants