RFR: 8279676: Dubious YMM register clearing in x86_64 arraycopy stubs

Aleksey Shipilev shade at openjdk.java.net
Tue Jan 11 08:58:54 UTC 2022


See discussion in the bug.

Additional testing, TR3970X (Zen 2, AVX2):
 - [x] Linux x86_64 fastdebug `hotspot_compiler_arraycopy`
 - [x] Linux x86_64 fastdebug `tier1`
 - [x] Linux x86_64 fastdebug `tier2`
 - [x] Linux x86_64 fastdebug `tier3`

Additional testing, i5-11500 (Rocket Lake, AVX-512):
 - [x] Linux x86_64 fastdebug `hotspot_compiler_arraycopy`
 - [x] Linux x86_64 fastdebug `tier1`
 - [x] Linux x86_64 fastdebug `tier2`
 - [x] Linux x86_64 fastdebug `tier3`

Targeted arraycopy microbenchmarks find a barely noticeable improvement on TR 3970X and i5-11500 with UseAVX=2.

-------------

Commit messages:
 - Drop x86_32 changes
 - Handle AVX-512 in x86_32
 - Fix

Changes: https://git.openjdk.java.net/jdk/pull/7016/files
 Webrev: https://webrevs.openjdk.java.net/?repo=jdk&pr=7016&range=00
  Issue: https://bugs.openjdk.java.net/browse/JDK-8279676
  Stats: 10 lines in 1 file changed: 0 ins; 10 del; 0 mod
  Patch: https://git.openjdk.java.net/jdk/pull/7016.diff
  Fetch: git fetch https://git.openjdk.java.net/jdk pull/7016/head:pull/7016

PR: https://git.openjdk.java.net/jdk/pull/7016


More information about the hotspot-compiler-dev mailing list