[llvm] r182219 - [Sparc] Rearrange integer registers' allocation order so that register allocator will use I and G registers before using L and O registers.

Venkatraman Govindaraju venkatra at cs.wisc.edu
Mon May 20 14:49:19 PDT 2013


Thank you for letting me know that.

I do not have a strong preference on one over another.

Thanks,
Venkatraman


On Mon, May 20, 2013 at 11:26 AM, Jakob Stoklund Olesen <stoklund at 2pi.dk> wrote:
>
> On May 19, 2013, at 1:07 PM, Venkatraman Govindaraju <venkatra at cs.wisc.edu> wrote:
>
>> def IntRegs : RegisterClass<"SP", [i32, i64], 32,
>> -                            (add L0, L1, L2, L3, L4, L5, L6,
>> -                                 L7, I0, I1, I2, I3, I4, I5,
>> -                                 O0, O1, O2, O3, O4, O5, O7,
>> +                            (add I0, I1, I2, I3, I4, I5,
>>                                  G1,
>> -                                 // Non-allocatable regs:
>> -                                 G2, G3, G4, // FIXME: OK for use only in
>> +                                 G2, G3, G4, // OK for use only in
>>                                              // applications, not libraries.
>> +                                 G5,         // OK for use in 64 bit mode.
>> +                                 L0, L1, L2, L3, L4, L5, L6, L7,
>> +                                 O0, O1, O2, O3, O4, O5, O7,
>> +                                 // Non-allocatable regs:
>>                                  O6, // stack ptr
>>                                  I6, // frame ptr
>>                                  I7, // return address
>>                                  G0, // constant zero
>> -                                 G5, G6, G7 // reserved for kernel
>> +                                 G6, G7 // reserved for kernel
>>                                  )>;
>
> Hi Venkat,
>
> The register allocator automatically filters out the reserved registers now, so it isn’t necessary to put them at the end of the register classes any more.
>
> If you prefer to keep them there, that’s fine.
>
> Thanks,
> /jakob
>




More information about the llvm-commits mailing list