RFR: 8325937: runtime/handshake/HandshakeDirectTest.java causes "monitor end should be strictly below the frame pointer" assertion failure on AArch64

[Coleen Phillimore]

On Thu, 3 Oct 2024 00:51:22 GMT, David Holmes <dholmes at openjdk.org> wrote:

> So that implies "all" we need is a storeload barrier after the write to arm the poll. That might be semantically clearer than adding a full fence - though probably adds some redundancy with the SMB.

I looked at adding it to where we arm the safepoint, but then the handshake all case would have a store load barrier per thread (where we loop and arm all the threads).

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

PR Comment: https://git.openjdk.org/jdk/pull/21295#issuecomment-2391230066