RFR: 8345635: JVMCI compiler should not be disabled by java -Xshare:dump

Doug Simon dnsimon at openjdk.org
Wed Dec 11 20:29:36 UTC 2024


On Wed, 11 Dec 2024 20:23:40 GMT, Ioi Lam <iklam at openjdk.org> wrote:

>> src/hotspot/share/cds/cdsConfig.cpp line 432:
>> 
>>> 430: 
>>> 431:   if (is_dumping_static_archive()) {
>>> 432:     if (!mode_flag_cmd_line JVMCI_ONLY(&& !UseJVMCICompiler && !UseGraalJIT)) {
>> 
>> Since `UseGraalJIT` [implies](https://github.com/openjdk/jdk/blob/08bdeedfd355b61f7cdabbe943657691e5af2c82/src/hotspot/share/jvmci/jvmci_globals.hpp#L50) `UseJVMCICompiler` (indirectly via [EnableJVMCIProduct](https://github.com/openjdk/jdk/blob/08bdeedfd355b61f7cdabbe943657691e5af2c82/src/hotspot/share/jvmci/jvmci_globals.hpp#L55C17-L55C35)), you can omit it here.
>
> `CDSConfig::check_vm_args_consistency()` is called before `CompilerConfig::check_args_consistency()`, so I need to catch all "user facing" arguments that can enable the `EnableJVMCI` flag. That's why I am checking `UseGraalJIT` and `UseJVMCICompiler` directly.
> 
> Are there any other flags that may cause `EnableJVMCI` to true?

Yes: `-XX:+EnableJVMCIProduct` and of course `-XX:+EnableJVMCI` itself.

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

PR Review Comment: https://git.openjdk.org/jdk/pull/22687#discussion_r1880902310


More information about the hotspot-runtime-dev mailing list