https://github.com/lukel97 commented:

Should we just emit a vscale intrinsic instead? I think vlenb would be defined 
then as `vscale * 8 = vlenb`, and it would automatically get all the value 
tracking information in vplan.

`vscale` isn't really defined for zvl32b but this code uses 
`llvm::RISCV::RVVBytesPerBlock` which has the same problem anyway. 

https://github.com/llvm/llvm-project/pull/219924
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to