[9] RFR(XS): 8031389: On x86 C1 emits two relocations for polls

Tobias Hartmann tobias.hartmann at oracle.com
Thu Jun 5 07:09:00 UTC 2014


Thank you, Vladimir.

Best,
Tobias

On 04.06.2014 20:31, Vladimir Kozlov wrote:
> Looks reasonable.
>
> Vladimir
>
> On 6/4/14 3:51 AM, Tobias Hartmann wrote:
>> Hi,
>>
>> please review the following patch for 8031389.
>>
>> *Problem*
>> C1 emits two poll_type relocations at safepoint polls because
>> LIR_Assembler::safepoint_poll(..) adds a relocation and calls
>> LIR_Assembler::add_debug_info_for_branch(..) which adds a relocation as
>> well.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8031389
>>
>> *Solution*
>> The implementation of add_debug_info_for_branch(..) is changed to not
>> emit a poll relocation. This also improves the semantics because the
>> method is used for normal branches as well (see
>> LIR_Assembler::emit_opBranch(..)).
>>
>> The implementation of LIR_Assembler::safepoint_poll(...) for sparc, arm
>> and ppc is changed to always emit a poll relocation.
>>
>> Webrev: http://cr.openjdk.java.net/~thartmann/8031389/webrev.00/
>>
>> *Tests*
>> JPRT
>>
>> Thanks,
>> Tobias



More information about the hotspot-compiler-dev mailing list