RFR: JDK-8163469: aarch32: add support for ARMv6K CPU

Alex Kashchenko akashche at redhat.com
Sun Aug 14 16:18:37 UTC 2016


Hi,

On 08/10/2016 03:06 PM, Alex Kashchenko wrote:
> Hi Andrey,
>
> On 08/09/2016 02:32 PM, Andrey Petushkov wrote:
>> Dear All,
>>
>> Could you please review the changes made in order to support ARMv6K
>> (e.g. ARM1176 found in RPi1)
>> The following is done: fixed support of emitting mov+3*orr when
>> movw+movt is not available. Replaced use of some of the instructions
>> unavailable in v6K with those which are available
>> In addition the CPU detection code is rewritten in a way to probe the
>> CPU instead of relying on the information from the OS. This is
>> necessary to support some of the emulation engines which only emulate
>> the CPU but not OS and hence do not provide the correct cpuinfo/auxv data
>>
>> http://cr.openjdk.java.net/~apetushkov/8163469/webrev/
>> <http://cr.openjdk.java.net/~apetushkov/8163469/webrev/>
>>
>> Thanks,
>> Andrey
>>
>> PS. Alex, please excuse me for the delay, vacation time is the reason
>
> Thanks for this change, it is great to have RPi Zero/A+/B+ supported!
>
> The build worked for me on B+, currently doing bootstrap rebuild, QEMU
> testing will follow.
>

QEMU worked flawlessly for me, with both full and userspace-only 
emulations (so it allows me to drop the workarounds I used before). Also 
userspace-only emulation seems to be noticeably faster than before 
(possibly due to correct features detection) - comparable with slower 
dev boards.

-- 
-Alex


More information about the aarch32-port-dev mailing list