FP16 isNaN, isFinite, isInfinite intrinsics without Reinterpret nodes #1242
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This patch adds intrinsic support for FP16 isNaN, isFinite and isInfinite methods and also adds aarch64 backend for these intrinsics. Tested all FP16 related tests successfully on aarch64.
This patch removes the ReinterpretS2HF nodes in the mid-end during the generation of isNaNHF,isFiniteHF and isInfiniteHF nodes. Performance results for this patch on an aarch64 machine - Benchmark Gain over baseline Gain over default FP16Ops.isFiniteHF 1.29 1.85 FP16Ops.isInfiniteHF 1.28 1.90 FP16Ops.isNaNHF 65504 1.45 1.58 The baseline patch generates floating point FP16 instructions and the default is where no FP16 intrinsics are used and FP32 instructions are generated. Gain : thrpt of this patch / thrpt of either baseline or default Tested FP16Ops.isInfiniteHF test on x86 and the performance is 2.6x better over the default case (which converts FP16 to FP32 and uses the vfpclass instruction). The JMH tests are added in this patch.
|
👋 Welcome back bkilambi! A progress list of the required criteria for merging this PR into |
|
❗ This change is not yet ready to be integrated. |
Bhavana-Kilambi
changed the title
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch removes the ReinterpretS2HF nodes in the mid-end during the generation of isNaNHF,isFiniteHF and isInfiniteHF nodes.
Performance results for this patch on an aarch64 machine -
The baseline patch generates floating point FP16 instructions and the default is where no FP16 intrinsics are used and FP32 instructions are generated.
Gain : thrpt of this patch / thrpt of either baseline or default
Tested FP16Ops.isInfiniteHF test on x86 and the performance is 2.6x better over the default case (which converts FP16 to FP32 and uses the vfpclass instruction).
The JMH tests are added in this patch.
Progress
Reviewing
Using
gitCheckout this PR locally:
$ git fetch https://git.openjdk.org/valhalla.git pull/1242/head:pull/1242$ git checkout pull/1242Update a local copy of the PR:
$ git checkout pull/1242$ git pull https://git.openjdk.org/valhalla.git pull/1242/headUsing Skara CLI tools
Checkout this PR locally:
$ git pr checkout 1242View PR using the GUI difftool:
$ git pr show -t 1242Using diff file
Download this PR as a diff file:
https://git.openjdk.org/valhalla/pull/1242.diff