[llvm-dev] Help required, Target Specifc SDNode getting optimized by ISel DAG

vivek pandya via llvm-dev llvm-dev at lists.llvm.org
Thu Feb 23 02:22:29 PST 2017


Hi Krzysztof,

Thanks for reply, I checked that and it seems to be ok.
How ever I noticed that my build is have this problem on other
architectures too. Only x86 is generating return instruction. Also epilogue
loads are also missing.

I think I have done some thing wrong to some target independent code.

-Vivek

>
>
> Hi Vivek,
> Make sure that you return the correct chain from LowerReturn.
>
> -Krzysztof
>
>
> On 2/22/2017 7:17 AM, vivek pandya via llvm-dev wrote:
> > Hello LLVM Devs,
> >
> > I am trying to compile following simple program for an experimental
> backend
> >
> > int main()
> > {
> >    printf("Hello World \n");
> >     return 0;
> > }
> >
> > I have added code to lower return instruction in the backend however in
> > assembly I don't find return instruction generated.
> > In the debug-only=isel dump I can see that my target specific SDNode for
> > RET_FLAG is added but it gets removed during DAG
> > optimization phase.
> >
> >
> > Initial selection DAG: BB#0 'main:entry'
> > SelectionDAG has 18 nodes:
> >   t5: i32 = GlobalAddress<i32 (i8*, ...)* @printf> 0
> >         t0: ch = EntryToken
> >       t4: ch = store<ST4[%retval]> t0, Constant:i32<0>,
> > FrameIndex:i32<0>, undef:i32
> >     t8: ch,glue = callseq_start t4, TargetConstant:i32<0>
> >
> >   t10: ch,glue = CopyToReg t8, Register:i32 %R5, GlobalAddress:i32<[14 x
> > i8]* @.str> 0
> >   t12: ch,glue = DummyArchISD::CALL t10, TargetGlobalAddress:i32<i32
> > (i8*, ...)* @printf> 0, Register:i32 %R5, t10:1
> >   t13: ch,glue = callseq_end t12, TargetConstant:i32<0>,
> > TargetConstant:i32<0>, t12:1
> >     t15: i32,ch,glue = CopyFromReg t13, Register:i32 %R3, t13:1
> >
> >   t16: ch,glue = CopyToReg t15:1, Register:i32 %R3, Constant:i32<0>
> >
> >   t17: ch = DummyArchISD::RET_FLAG t16, Register:i32 %R3, t16:1
> >
> >   t15: i32,ch,glue = CopyFromReg t13, Register:i32 %R3, t13:1
> >
> >
> > Optimized lowered selection DAG: BB#0 'main:entry'
> > SelectionDAG has 15 nodes:
> >         t0: ch = EntryToken
> >       t4: ch = store<ST4[%retval]> t0, Constant:i32<0>,
> > FrameIndex:i32<0>, undef:i32
> >     t8: ch,glue = callseq_start t4, TargetConstant:i32<0>
> >
> >   t10: ch,glue = CopyToReg t8, Register:i32 %R5, GlobalAddress:i32<[14 x
> > i8]* @.str> 0
> >   t12: ch,glue = DummyArchISD::CALL t10, TargetGlobalAddress:i32<i32
> > (i8*, ...)* @printf> 0, Register:i32 %R5, t10:1
> >   t13: ch,glue = callseq_end t12, TargetConstant:i32<0>,
> > TargetConstant:i32<0>, t12:1
> >   t15: i32,ch,glue = CopyFromReg t13, Register:i32 %R3, t13:1
> >
> >
> >
> > Can someone please guide me how to furhter debug this situation? Or any
> > hints.
> >
> > Sincerely,
> > Vivek
> >
> >
> > _______________________________________________
> > LLVM Developers mailing list
> > llvm-dev at lists.llvm.org
> > http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
> >
>
> --
> Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
> hosted by The Linux Foundation
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20170223/c33939fa/attachment.html>


More information about the llvm-dev mailing list