RFR: 8266530: HotSpot changes for JEP 306 [v2]

[Joe Darcy]

On Thu, 6 May 2021 19:58:00 GMT, Harold Seigel <hseigel at openjdk.org> wrote:

> 
> 
> Additional changes may be needed to these hotspot files that contain ACC_STRICT:
> cpu/x86/templateTable_x86.cpp: __ testl(rcx, JVM_ACC_STRICT);
> cpu/x86/templateTable_x86.cpp: __ testl(rcx, JVM_ACC_STRICT);
> share/ci/ciFlags.hpp: bool is_strict () const { return (_flags & JVM_ACC_STRICT ) != 0; }
> share/include/jvm_constants.h: JVM_ACC_STRICT |
> share/utilities/accessFlags.hpp: bool is_strict () const { return (_flags & JVM_ACC_STRICT ) != 0; }
> 
> Also, there are multiple hotspot tests that contain ACC_STRICT.

As of JDK-7175279: "Don't use x87 FPU on x86-64", HotSpot shouldn't be able to have non-strict execution on any 64-bit platform. Non-strict execution is only possible using the x87 CPU instructions. For a 32-bit x86 port, either long strictfp-compliant mul/divide idioms need to be used or the SSE2 or later instructions.

The nature of the spec shouldn't invalidate any existing tests, other than ones probing at "abstract && strictfp" for new class file versions.

Thanks,

-Joe

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

PR: https://git.openjdk.java.net/jdk/pull/3891