llvm dev - May 2016 - Is it possible to avoid inserting spill/split code in certain instruction sequence in RA?

Hi all,

I am working on an out-of-tree target. I am wondering if it is possible to
force the register allocator (and/or spiller) to not break certain
instruction sequence.

For example:

phys_reg = MI1 vreg1
vreg 2 = MI2 phys_reg

Is there a way to tell RA/spiller not to insert COPY or spill between MI1
and MI2?

I am using greedy register allocator and inline spiller.

-- 
Regards,
Dongrui
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160509/03c0b3e6/attachment.html>

On 5/9/2016 1:30 PM, Dongrui She via llvm-dev wrote:
>
> I am working on an out-of-tree target. I am wondering if it is possible
> to force the register allocator (and/or spiller) to not break certain
> instruction sequence.
>
> For example:
>
> phys_reg = MI1 vreg1
> vreg 2 = MI2 phys_reg
>
> Is there a way to tell RA/spiller not to insert COPY or spill between
> MI1 and MI2?
You can make a pseudo-instruction that corresponds to the combination 
MI1/MI2, then expand it into the actual instructions in 
TII::expandPostRAPseudo.

-Krzysztof

-- 
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, 
hosted by The Linux Foundation

Hi,

Thanks for the suggestion.

For my target making pseudo instructions for this is a bit too much,
because there are many possible combinations. I tried making a bundle in
this situation, and it seems to work fine (with a little bit hacking).

Regards,
Dongrui

On Mon, May 9, 2016 at 8:35 PM, Krzysztof Parzyszek via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> On 5/9/2016 1:30 PM, Dongrui She via llvm-dev wrote:
>
>>
>> I am working on an out-of-tree target. I am wondering if it is possible
>> to force the register allocator (and/or spiller) to not break certain
>> instruction sequence.
>>
>> For example:
>>
>> phys_reg = MI1 vreg1
>> vreg 2 = MI2 phys_reg
>>
>> Is there a way to tell RA/spiller not to insert COPY or spill between
>> MI1 and MI2?
>>
>
> You can make a pseudo-instruction that corresponds to the combination
> MI1/MI2, then expand it into the actual instructions in
> TII::expandPostRAPseudo.
>
> -Krzysztof
>
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, hosted
> by The Linux Foundation
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>


-- 
Regards,
Dongrui
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160512/8c685c1e/attachment.html>