[cfe-dev] Proposed changes to vectorize_width #pragma

[Michael Kruse via cfe-dev]

I see the motivation, but there are different requirements for
LLVM-internals and user-facing extensions, which is why e.g. clang
does not implement a #pragma ivdep.

The definitions looks fine to me, as long as it is documented without
referring to compiler internals.

Michael

Am Di., 1. Dez. 2020 um 07:46 Uhr schrieb David Sherwood
<David.Sherwood at arm.com>:
>
> Hi,
>
> So by adding support for scalable vectorisation widths we are effectively
> updating the pragma to mirror the existing VectorType class in LLVM,
> which is defined by a ElementCount and an element Type. The
> ElementCount is a tuple consisting of a minimum number of elements
> and a scalable flag. The meaning of 'scalable' as used in the vectorize_width
> pragma as identical to that of ElementCount. Using one of my examples
> in the initial proposal then this pragma
>
> #pragma clang loop vectorize_width(4, scalable)
>
> would mean the same in LLVM as a VectorType like this:
>
> <vscale x 4 x eltty>
>
> where eltty depends upon the types used in the loop. The 'vscale' parameter
> is defined by the target - it is at least 1 and does not have to be a power of 2.
>
> Kind Regards,
> David.
>
> -----Original Message-----
> From: Michael Kruse <llvm at meinersbur.de>
> Sent: 30 November 2020 18:01
> To: Sjoerd Meijer <Sjoerd.Meijer at arm.com>
> Cc: David Sherwood <David.Sherwood at arm.com>; cfe-dev at lists.llvm.org; Sander De Smalen <Sander.DeSmalen at arm.com>
> Subject: Re: [cfe-dev] Proposed changes to vectorize_width #pragma
>
> Am Mi., 25. Nov. 2020 um 10:15 Uhr schrieb Sjoerd Meijer via cfe-dev
> <cfe-dev at lists.llvm.org>:
> > I think that leaves us with 2 options that can express the same things, i.e. change or introduce:
> >
> > 1)
> > vectorize_width(VF, fixed|scalable)
> > vectorize_width(fixed|scalable)
> > vectorize_width(VF)
> >
> > 2)
> > vectorize_style(fixed|scalable)
>
> Another proposal:
>
> 3)
> vectorize_width(VF) // For fixed vector width.
> vectorize_width_at_least(MinVF) // For SVE; alternatives:
> vectorize_dynamic, vectorize_scalable.
>
> What are the intended semantics? Does scalable mean "width of MinVF or
> more", "any multiple of MinVF",  "power-of-2 multiple of MinVF", "any
> width of at least MinVF allowed by ARM's SVE"?
>
>
> Michael