[webkit-dev] Announce the ARM port of JIT

Zoltan Herczeg zherczeg at inf.u-szeged.hu
Fri Apr 3 21:32:41 PDT 2009


Hi,

after I read my mail again I am not sure it is clear what I wanted to say.

current operation:
 - push instruction. It will either
   a) push the current instruction
   b) push he constant pool, and push the current instruction
 - utility function: the next two instruction will be tightly coupled
   if the constant pool is pushed after the first instruction,
   push it NOW (implemented)

planned operation:
 - we can extend our utility function for MORE than two instructions.
   -> we can use hard coded offsets from that time as the x86 port.

Zoltan

> Hi,
>
> Thanks :)
>
> we put all those constants into the constant pool. ARM supports position
> independent code efficently, since its program counter is a regular
> register as well. We can load, modify or even jump to the addresses which
> are stored in the constant pool. Constant pools are dynamically put into
> the jit code, except when we need tightly coupled instruction pairs. Those
> must not be broken by a constant pool.
>
> Zoltan
>
>> Congratulations!
>> any improvement on X86 hard coded offset for code patch? I mean X86 uses
>> single pointer then a few constants closely coupled to X86 to patch
>> code,
>> which is a nightmare for porting.
>> rgds
>> joe
>
>
> _______________________________________________
> webkit-dev mailing list
> webkit-dev at lists.webkit.org
> http://lists.webkit.org/mailman/listinfo.cgi/webkit-dev
>



More information about the webkit-dev mailing list