[cfe-dev] Floating-point traps on x86-64

[Dallman, John via cfe-dev]

Cameron wrote:

> The scalar constrained intrinsics would prevent vectorization, so you hopefully wouldn't see a vector
> sqrt/etc instruction with scalar input(s). I don't think the X86 backend itself would select a vector
> instruction for a scalar operation (I could be wrong).

I fear it would, because they’re faster.

> If your user/frontend is issuing a vector instruction, the user/frontend would be responsible for
> inserting safe values as needed (or masking on targets that support it).

Our code is all C, with no assembler, so it has no way to issue vector instructions.

> Only a small number of people are currently working on this. New contributors are welcome...

The learning curve before I could do anything useful makes it very hard to justify spending working
hours on this. I frankly lack the enthusiasm for spending lots of my own time on this.

--
John Dallman

-----------------
Siemens Industry Software Limited is a limited company registered in England and Wales.
Registered number: 3476850.
Registered office: Faraday House, Sir William Siemens Square, Frimley, Surrey, GU16 8QD.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20190311/b542586f/attachment.html>