Repository navigation
Override isCompilable for TR_ResolveRelocatableJ9Method - #16556
Conversation
|
FYI: @tajila, @pshipton, @fengxue-IS |
|
Given the required symmetry between Even as it stands right now your new |
|
@dsouzai Requesting your input on whether |
|
I think if |
Also updated existing use of the function. Signed-off-by: Nazim Bhuiyan <nubhuiyan@ibm.com>
TR_ResolvedRelocatableJ9Method::isCompilable should agree with what TR::CompilationInfo::isMethodIneligibleForAot returns for a method. This override is necessary because methods in java/lang/invoke/* are ineligible for AOT compilations, and therefore isCompilable should also return false for these methods during AOT. This impacts what methods are considered for inlining. Signed-off-by: Nazim Bhuiyan <nubhuiyan@ibm.com>
|
@0xdaryl based on Irwin's comment, I have not made any changes to the check for |
|
Jenkins test sanity all jdk17,jdk19 |
|
Jenkins test sanity xlinux jdk17 |
|
Jenkins test sanity plinux,aix,xmac,amac,alinux64 jdk17,jdk19 |
|
Jenkins test sanity xlinux jdk19 |
|
Jenkins test sanity plinux,aix,xmac,amac,alinux64,xlinux jdk17,jdk19 |
|
Jenkins test sanity aix jdk17 |
TR_ResolvedRelocatableJ9Method::isCompilable should match what TR::CompilationInfoPerThreadBase::isMethodIneligibleForAot returns for a method. This override is necessary because methods in java/lang/invoke/* are marked ineligible from AOT compilations, and therefore isCompilable should also return false for these methods during AOT. This impacts what methods are considered for inlining.
Fixes: #16161, #14135
Signed-off-by: Nazim Bhuiyan nubhuiyan@ibm.com