[LLVMdev] lowering and non legal types in fast-isel

Reed Kotler reed.kotler at imgtec.com
Tue Sep 9 14:28:01 PDT 2014


How would you return a pair of registers then?

For example, If I lower 64 bit integer and materialize a constant, i 
need to return a pair of 32 registers.

On 09/09/2014 01:49 PM, Eric Christopher wrote:
> You can do everything from legalize it yourself, or just add code to 
> select it and legalize it in your backend during instruction 
> selection. fast-isel now has support for completely avoiding the 
> target independent version of the code.
>
> -eric
>
> On Sun, Aug 31, 2014 at 10:41 AM, Reed Kotler <Reed.Kotler at imgtec.com 
> <mailto:Reed.Kotler at imgtec.com>> wrote:
>
>     I understand that but falling back makes the compilation slower.
>
>     I'm wondering what could be done to remove this restriction about
>     fast-isel not being able to
>     handle non legal types.
>
>
>     ________________________________________
>     From: Anton Korobeynikov [anton at korobeynikov.info
>     <mailto:anton at korobeynikov.info>]
>     Sent: Sunday, August 31, 2014 12:55 AM
>     To: Reed Kotler
>     Cc: LLVMdev at cs.uiuc.edu <mailto:LLVMdev at cs.uiuc.edu>
>     Subject: Re: [LLVMdev] lowering and non legal types in fast-isel
>
>     Fastisel should fallback to usual selector in the cases it cannot
>     handle. So, for illegal types as well.
>
>     On Sun, Aug 31, 2014 at 1:42 AM, reed kotler <rkotler at mips.com
>     <mailto:rkotler at mips.com>> wrote:
>     > Fast-isel is not equipped in general to deal with non legal types.
>     >
>     > It would seem that an llvm assembler pass run after clang  but
>     before llvm
>     > could do the lowering though.
>     >
>     > Any thoughts?
>     >
>     > Reed
>     >
>     > _______________________________________________
>     > LLVM Developers mailing list
>     > LLVMdev at cs.uiuc.edu <mailto:LLVMdev at cs.uiuc.edu>
>     http://llvm.cs.uiuc.edu
>     > http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>
>
>     --
>     With best regards, Anton Korobeynikov
>     Faculty of Mathematics and Mechanics, Saint Petersburg State
>     University
>
>     _______________________________________________
>     LLVM Developers mailing list
>     LLVMdev at cs.uiuc.edu <mailto:LLVMdev at cs.uiuc.edu>
>     http://llvm.cs.uiuc.edu
>     http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20140909/86f51b8d/attachment.html>


More information about the llvm-dev mailing list