JDK-8345146 : [PPC64] Make intrinsic conversions between bit representations of half precision values and floats
Type:Enhancement
Component:hotspot
Sub-Component:compiler
Affected Version:24
Priority:P3
Status:Resolved
Resolution:Fixed
CPU:ppc
Submitted:2024-11-27
Updated:2025-02-27
Resolved:2024-12-05
The Version table provides details related to the release that this issue/RFE will be addressed.
Unresolved : Release in which this issue/RFE will be addressed. Resolved: Release in which this issue/RFE has been resolved. Fixed : Release in which this issue/RFE has been fixed. The release containing this fix may be available for download as an Early Access Release or a General Availability Release.
JDK-8289552 should get ported to PPC64. Power9 has instructions for half precision float conversions.
Comments
[OpenJDK 21u maintainer comment]
This is a fairly new change. Deferring to the next update as it's not a regression.
15-01-2025
[jdk21u-fix-request] Approval Request from Martin
JDK-8289551 is a JDK20 feature. It's very slow without the intrinsics. That impacts some of the jtreg tests, too. Therefore, I'm requesting JDK21u backport. Performance numbers are shown in the original PR. x86_64 and aarch64 have the intrinsics backported even to JDK20 (JDK-8302976). The backport is almost clean, has been reviewed and tested for several weeks.
15-01-2025
A pull request was submitted for review.
Branch: master
URL: https://git.openjdk.org/jdk21u-dev/pull/1295
Date: 2024-12-27 18:35:12 +0000