RFR: JDK-8156067: aarch32: backport "Better byte behaviour" patch from aarch64 project (jdk8)

[Alex Kashchenko]

Hi,

On 05/05/2016 10:26 AM, Edward Nevill wrote:
> On Wed, 2016-05-04 at 21:38 +0000, Sergey Nazarkin wrote:
>> Hi!
>>
>> Please review adaptation of “Better byte behaviour” patch from AARCH64 project
>>
>> -issue: https://bugs.openjdk.java.net/browse/JDK-8156067
>> -webrev: http://cr.openjdk.java.net/~snazarki/8156067/
>>
>> -original patch: http://hg.openjdk.java.net/aarch64-port/jdk8u/hotspot/rev/388e9d0905e6
>>
>> Please note this is no one-to-one copy. Few branches were omitted to make execution flow more linear and predictable.  Please correct me if I’m wrong.
>>
>> The need of new JDK issue is questionable  as well.  We may use the same “8132051” number in the description (as Andrew Hughes did) just for consistency with other u91 patches . However this issue is marked as confidential and I have no idea what about is it.
>
> Hi Sergey,
>
> Looks good. Thanks for doing this.
>
> One minor change is to change
>
> +  __ ands(rscratch1, r1, diffbit);
> +  __ andr(r0, r0, 1, Assembler::NE);  // if it is a T_BOOLEAN array, mask the stored value to 0/1
>
> to
>
> +  __ tst(r1, diffbit);
> +  __ andr(r0, r0, 1, Assembler::NE);  // if it is a T_BOOLEAN array, mask the stored value to 0/1
>
> This is mainly aesthetic. I will make this change. No need for a new webrev.
>
> I will build and run this through jtreg hotspot & langtools and then push along with the update to u91.

I met a minor problem building the patch, this bit in 
interp_masm_aarch32.cpp:

ldr(rscratch1, Address(rfp, frame::get_interpreter_frame_method_offset() 
* wordSize));

probably need to be changed to:

ldr(rscratch1, Address(rfp, frame::interpreter_frame_method_offset * 
wordSize));

or maybe some changes to frame_aarch32.hpp are missed in the patch?

-- 
-Alex