RFR: Store cpu features in AOTCodeCache header [v5]

Ashutosh Mehra asmehra at openjdk.org
Fri Jul 25 14:41:13 UTC 2025


On Thu, 24 Jul 2025 21:50:54 GMT, Vladimir Kozlov <kvn at openjdk.org> wrote:

>> Ashutosh Mehra has updated the pull request with a new target base due to a merge or a rebase. The pull request now contains 11 commits:
>> 
>>  - Merge branch 'premain' into aot-cache-feature-flags
>>  - Add test to check cpu feature incompatibility
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Ignore CPU_HT when storing cpu features in AOTCodeCache
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Fix formatting of log messages
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Restore changes deleted by mistake
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Use FormatBuffer to obtain cpu features names string
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Merge branch 'premain' into aot-cache-feature-flags
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Remove UseSSE and UseAVX checks
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - Merge branch 'premain' into aot-cache-feature-flags
>>  - Address review comments
>>    
>>    Signed-off-by: Ashutosh Mehra <asmehra at redhat.com>
>>  - ... and 1 more: https://git.openjdk.org/leyden/compare/a529aa73...5f872c11
>
> I submitted leyden/premain testing for current 04 version.

@vnkozlov 

> Look on test/hotspot/jtreg/compiler/arraycopy/stress/TestStressArrayCopy.java how to get CPU features using WB instead of running separate processes.
> 
> Also add  * @requires os.simpleArch == "x64" if you want to run it only on x64.
> 

Thanks for the pointers. I updated the tests as per your suggestion. I wanted to add a similar test for aarch64 and I picked up crc32 flag for that. But it turns out the command line option to disable this feature -XX:-UseCRC32 is not reflected in the `VM_Version::_features` bitmap. I think this  should be treated as a bug and fixed in mainline. I have commented out the aarch64 specific test for now.
I also adding a diagnostic option to disable cpu feature check.
These changes should not affect your testing though.

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

PR Comment: https://git.openjdk.org/leyden/pull/84#issuecomment-3118169979


More information about the leyden-dev mailing list