llvm dev - Jan 2019 - [FPEnv] Rust/Go/Swift/Flang/other llvm IRBuilder needs?

I've got a ticket open where I would very much like some input from
maintainers of other languages that rely on llvm and use the IRBuilder. See:

Teach the IRBuilder about constrained fadd and friends:
https://reviews.llvm.org/D53157

I'm adding support to functions like CreateFAdd() the ability to optionally
emit the constrained math intrinsics required to support strict floating point
mode. This makes the changes in clang relatively easy, but the concern is that
there may be concerns that haven't been raised in the diff. And I'd like
to tailor the interface so that it meets the needs of IRBuilder users, but
ideally that wants input from actual IRBuilder users.

So, can I get some input from other front end authors and maintainers?

Apologies for the cross posting, but I don't know how to more precisely get
the attention of the relevant people.
--
Kevin P. Neal
SAS/C and SAS/C++ Compiler
Host Research and Development
SAS Institute, Inc.




-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190116/35ff5945/attachment.html>

For us, constrained fadd and friends aren't particular relevant. We 
don't need to model changes to the floating point environment, and chose 
to use explicit control flow for fault handling (e.g. div-by-zero).  We 
might someday explore using constrained operations for performance 
reasons, but up until now, haven't really had the need to.  So, no 
opinion, sorry.

Philip

On 1/16/19 10:02 AM, Kevin Neal via llvm-dev wrote:
> I’ve got a ticket open where I would very much like some input from 
> maintainers of other languages that rely on llvm and use the 
> IRBuilder. See:
> Teach the IRBuilder about constrained fadd and friends:
> _https://reviews.llvm.org/D53157_
> I’m adding support to functions like CreateFAdd() the ability to 
> optionally emit the constrained math intrinsics required to support 
> strict floating point mode. This makes the changes in clang relatively 
> easy, but the concern is that there may be concerns that haven’t been 
> raised in the diff. And I’d like to tailor the interface so that it 
> meets the needs of IRBuilder users, but ideally that wants input from 
> actual IRBuilder users.
> So, can I get some input from other front end authors and maintainers?
> Apologies for the cross posting, but I don’t know how to more 
> precisely get the attention of the relevant people.
> --
> Kevin P. Neal
> SAS/C and SAS/C++ Compiler
> Host Research and Development
> SAS Institute, Inc.
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190116/359d3e03/attachment.html>

Hi Kevin,
Same situation for our Go compiler (gollvm). We also use explicit checks
for division by zero, and I can't envision a need to have more fine-grained
control over rounding/exception-mode for fadd etc.
Cheers, Than

On Wed, Jan 16, 2019 at 5:39 PM Philip Reames via llvm-dev <
llvm-dev at lists.llvm.org> wrote:
> For us, constrained fadd and friends aren't particular relevant.  We
don't
> need to model changes to the floating point environment, and chose to use
> explicit control flow for fault handling (e.g. div-by-zero).  We might
> someday explore using constrained operations for performance reasons, but
> up until now, haven't really had the need to.  So, no opinion, sorry.
>
> Philip
> On 1/16/19 10:02 AM, Kevin Neal via llvm-dev wrote:
>
> I’ve got a ticket open where I would very much like some input from
> maintainers of other languages that rely on llvm and use the IRBuilder.
See:
>
> Teach the IRBuilder about constrained fadd and friends:
> *https://reviews.llvm.org/D53157* <https://reviews.llvm.org/D53157>
>
> I’m adding support to functions like CreateFAdd() the ability to
> optionally emit the constrained math intrinsics required to support strict
> floating point mode. This makes the changes in clang relatively easy, but
> the concern is that there may be concerns that haven’t been raised in the
> diff. And I’d like to tailor the interface so that it meets the needs of
> IRBuilder users, but ideally that wants input from actual IRBuilder users.
>
> So, can I get some input from other front end authors and maintainers?
>
> Apologies for the cross posting, but I don’t know how to more precisely
> get the attention of the relevant people.
> --
> Kevin P. Neal
> SAS/C and SAS/C++ Compiler
> Host Research and Development
> SAS Institute, Inc.
>
>
>
>
>
> _______________________________________________
> LLVM Developers mailing listllvm-dev at
lists.llvm.orghttp://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
> _______________________________________________
> LLVM Developers mailing list
> llvm-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20190117/1c172e7f/attachment.html>