RFR: 8253690: [PPC64] Use flag kind diagnostic for platform specific flags

Thomas Stuefe stuefe at openjdk.java.net
Wed Sep 30 07:01:49 UTC 2020


On Tue, 29 Sep 2020 20:49:01 GMT, Martin Doerr <mdoerr at openjdk.org> wrote:

> Current platform implementation (globals_ppc.hpp) uses regular product flags for almost everything.
> Most platform specific flags were never intended for official support. They are only there to diagnose issues and find
> workarounds. So flag kind "diagnostic" fits better for them.
> 
> Note that I rearranged a couple of lines when looking at the diff.
> My actual change is what is described here: https://bugs.openjdk.java.net/browse/JDK-8253692

Hi Martin,

make sure you check up on the places the switches are used and pass UnlockDiagnosticVMOptions.

Cheers, Thomas

src/hotspot/cpu/ppc/globals_ppc.hpp line 95:

> 93:                                                                             \
> 94:   /* Power 8: Configure Data Stream Control Register. */                    \
> 95:   product(uint64_t, DSCR_PPC64, (uint64_t)-1,                               \

Has nothing to do with this issue. But I leave it up to you.

src/hotspot/cpu/ppc/globals_ppc.hpp line 116:

> 114:                                                                             \
> 115:   /* special instructions */                                                \
> 116:   product(bool, SuperwordUseVSX, false,                                     \

Why leave this one out?

src/hotspot/cpu/ppc/globals_ppc.hpp line 150:

> 148:                                                                             \
> 149:   product(bool, ZapMemory, false, "Write 0x0101... to empty memory."        \
> 150:           " Use this to ease debugging.")                                   \

Future cleanup: this feels like it should be in shared code. The usual way to do this is to zap in DEBUG.

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

PR: https://git.openjdk.java.net/jdk/pull/413


More information about the hotspot-compiler-dev mailing list