RFR: 8265711: C1: Intrinsify Class.getModifier method [v2]
thartmann at openjdk.java.net
Tue Apr 27 13:31:40 UTC 2021
On Tue, 27 Apr 2021 09:40:11 GMT, Yi Yang <yyang at openjdk.org> wrote:
>> It's relatively a common case to get modifiers from a constant Class instance, i.e. ThirdPartyClass.class.getModifiers(). Currently, C1 Canonicalizer missed the opportunity of replacing Class.getModifiers intrinsic calls with compile-time constants.
> Yi Yang has updated the pull request with a new target base due to a merge or a rebase. The incremental webrev excludes the unrelated changes brought in by the merge/rebase. The pull request contains three additional commits since the last revision:
> - add missing stuff; change according to review comment
> - Merge branch 'master' into canon_opt
> - canonicalize well-chosen intrinsic methods
src/hotspot/share/c1/c1_LIRGenerator.cpp line 1330:
> 1328: LabelObj* L_done = new LabelObj();
> 1330: LIR_Opr temp = new_register(T_METADATA);
`temp` could be renamed to `klass`.
src/hotspot/share/c1/c1_LIRGenerator.cpp line 1339:
> 1338: __ branch_destination(L_not_prim->label());
> 1339: __ move(new LIR_Address(receiver.result(), java_lang_Class::klass_offset(), T_ADDRESS), temp, NULL);
Why are you reloading the klass here?
Also, as @rwestrel explained in the RFR for (https://mail.openjdk.java.net/pipermail/hotspot-compiler-dev/2018-September/030745.html), we need to be really careful when emitting branches in C1's LIR.
More information about the hotspot-compiler-dev