[LLVMdev] x86: inline an LLVM IR function?

[Mikael Lyngvig]

Thank you!


-- Mikael


2013/12/1 Joey Gouly <joel.gouly at gmail.com>

> You should use "opt" with -inline, or -O2/3 to optimise this.
> llc only runs backend optimisations.
>
>
>
> On 1 December 2013 18:51, Mikael Lyngvig <mikael at lyngvig.org> wrote:
>
>> Sorry for what may possibly be a rather stupid question, but how on earth
>> do you make LLC inline a function?  I've got this code:
>>
>>    attributes 0 = { alwaysinline nounwind }
>>
>>    define internal i32 @lambda(i32 %a, i32 %x) #0 {
>>       %1 = add i32 %a, %x
>>       ret i32 %1
>>    }
>>
>>    define i32 @foo(i32 %a) nounwind {
>>       %1 = call i32 @lambda(i32 %a, i32 10)
>>       ret i32 %1
>>    }
>>
>> And no matter what I do, the function @lambda ends up being called with a call instruction.  I've tried with -O2, -O3, and with and without the alwaysinline attribute.
>>
>> How do I get LLC or Clang to inline the above @lambda function?
>>
>> Thanks in advance!
>>
>>
>> -- Mikael
>>
>>
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> 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/20131201/81face6d/attachment.html>