RFR: 8316702: Only evaluate buffer when IGVPrintLevelOption >= 5
Xin Liu
xliu at openjdk.org
Fri Sep 22 05:42:10 UTC 2023
On Fri, 22 Sep 2023 01:53:24 GMT, Xin Liu <xliu at openjdk.org> wrote:
> This is a follow-up of JDK-8310220. hotspot C2 only emit IR per bytecode to IGV when level = 5.
>
> Currently, we conservatively check C->should_print_igv(1) in Parse::do_one_bytecode(). It's wasteful when level>=1 and <= 4. We evaluate `jio_snprintf(buffer, sizeof(buffer), "Bytecode %d: %s", bci(), Bytecodes::name(bc()))` for each bytecode but emit nothing.
>
> We should use the same level(5) in those 2 places.
The only failure is [gc/TestAllocHumongousFragment#aggressive](https://github.com/navyxliu/jdk/actions/runs/6269158038#user-content-gc_testallochumongousfragment#aggressive) on macos-x86. I don't think it's related to this chagne.
-------------
PR Comment: https://git.openjdk.org/jdk/pull/15878#issuecomment-1730830146
More information about the hotspot-compiler-dev
mailing list