RFR(S): 8206107: [x86_32] jck tests for ldc2_w bytecode fail

[Boris Ulasevich]

Thank you!

On 24.12.2018 16:09, Dmitry Samersoff wrote:
> Boris,
> 
> The fix looks good to me.
> 
> -Dmitry
> 
> On 10.12.2018 14:43, Tobias Hartmann wrote:
>> Forwarding this to hotspot-runtime-dev.
>>
>> On 10.12.18 12:35, Boris Ulasevich wrote:
>>> Hi all,
>>>
>>> Please review following fix in x86_32 ldc2_w template generator:
>>>
>>> http://cr.openjdk.java.net/~bulasevich/8206107/webrev.00
>>> http://bugs.openjdk.java.net/browse/JDK-8206107
>>>
>>> We need to reorder instructions to avoid usage of broken eax value:
>>>    const Register obj = rax;
>>>    const Register off = rbx;
>>>    const Address field(obj, off, Address::times_1, 0*wordSize);
>>>    ..
>>>    __ movptr(rax, field);                        //    mov (%eax,%ebx,1),%eax
>>>    NOT_LP64(__ movptr(rdx, field.plus_disp(4))); // => mov 0x4(%eax,%ebx,1),%edx
>>>
>>> thanks,
>>> Boris