RFR(S): 8016696: PPC64 (part 4): add relocation for trampoline stubs

[Vladimir Kozlov]

Albert ran JPRT testing with this Part 4 changes and it passed.

Vladimir

On 6/17/13 9:28 AM, Vladimir Kozlov wrote:
> I don't see problems with it. I asked Roland to look in also.
>
> Can you push it before Part 3 (which may need some time to adjust closed
> sources)?
>
> Thanks,
> Vladimir
>
> On 6/17/13 5:24 AM, Lindenmaier, Goetz wrote:
>> Hi,
>>
>> On PPC, we use special stubs we call trampoline stubs for calls to
>>
>> target addresses that can not be encoded as offset in the branch
>>
>> instruction.
>>
>> A trampoline allows to encode a small branch in the code, even if
>>
>> there is the chance that this branch can not reach all possible code
>>
>> locations. If the relocation finds that a branch is too far for the
>>
>> instruction in the code, it can patch it to jump to the trampoline
>>
>> where is sufficient space for a far branch.
>>
>> To implement this, we extended the relocations by a new kind,
>>
>> trampoline_stub_Relocation:
>>
>> http://cr.openjdk.java.net/~goetz/webrevs/8016696-trampoline/
>>
>> Aditionally to the code in 0004_opto-trampoline_relocations.patch
>>
>> I added debug printing functionality.
>>
>> Could I get a review on this change, please?
>>
>> Thanks,
>>
>>    Goetz.
>>