ISD::ZERO_EXTEND / SIGN_EXTEND of sub-vectors

[Jonas Paulsson via llvm-commits]

Hi,
>
> Probably of interest to other targets, yes; I assume nobody got around 
> to it because SIGN_EXTEND_VECTOR_INREG is a relatively recent addition.
>
> ...
> Yes, you have to call getSignExtendVectorInReg() to get the right node.
>
>
>
Thanks Eli - that seemed to do the trick! :-)

I just added a few lines to convert the node to SIGN/ZERO 
_EXTEND_VECTOR_INREG - see attachment.

I am thinking that the WidenVT should be legal, and since it seems 
GetWidenedVector() is supposed to always return a legal vector, I have 
neither added a check for legality of InOp. All regression tests pass. 
What do you think?

Thanks again,

Jonas




-------------- next part --------------
A non-text attachment was scrubbed...
Name: WidenVecRes_Convert.patch
Type: text/x-patch
Size: 776 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20170124/17fb74f3/attachment.bin>