androm3da wrote:

> I'm going to defer to @aankit-ca though I suppose Ankit won't have enough 
> history/context to know why it was chosen the way it was?
> 
> I would hope that either @SundeepKushwaha or @quic-seaswara will know why we 
> want small data on by default, because I don't. I suspect it gives better 
> code size performance on benchmarks and that's why?
> 
> > it depends on linker support which LLD does not have
> 
> I'll keep fixing hexagon lld bugs as we encounter them. But it's worth 
> mentioning that the vast majority of hexagon codegen put to use in real 
> devices gets linked with `ld.eld`.
> 
> I will concede that I like the uniformity when the llvm project comes with 
> its own linker. Though, at the same time, there's some [risk of a monoculture 
> there too](https://lists.llvm.org/pipermail/llvm-dev/2019-June/133308.html).

@aankit-ca can you weigh in on this PR?

https://github.com/llvm/llvm-project/pull/187344
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to