gcc math functions for OpenMP vectoization

gcc math functions for OpenMP vectoization

[Feltgen, Eric]

Hi there,


my name is Eric, I'm a german student at RWTH Aachen University currently researching OpenMP.


For my research, I'm also looking at math functions provided by compilers like GCC. When writing vectorizable code, it is important to use math functions which also allow for vectorization. I've read that gcc includes a set of math functions which are SIMD compatible. Where can I find resources on them?


It would be awesome if you could tell me where to find documentation.


Thanks in advance, I highly appreciate your support.


Best regards,


Eric

Re: gcc math functions for OpenMP vectoization

[Jakub Jelinek]

On Fri, Jun 05, 2020 at 01:58:55PM +0000, Feltgen, Eric wrote:
> my name is Eric, I'm a german student at RWTH Aachen University currently researching OpenMP.
> 
> 
> For my research, I'm also looking at math functions provided by compilers like GCC. When writing vectorizable code, it is important to use math functions which also allow for vectorization. I've read that gcc includes a set of math functions which are SIMD compatible. Where can I find resources on them?
> 
> 
> It would be awesome if you could tell me where to find documentation.

It is glibc that provides them, not GCC.
See https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD
for the header file and the functions are then implemented (mostly?) in assembly in
e.g. /usr/src/libc/sysdeps/x86_64/fpu/ directory.  There is similar support
for aarch64.
All that is provided by GCC is -fopenmp #pragma omp declare simd support
and support for corresponding simd attribute (which is what is used if
-fopenmp is not enabled).

	Jakub

Re: gcc math functions for OpenMP vectoization

[Tobias Burnus]

On 6/5/20 4:11 PM, Jakub Jelinek via Gcc wrote:
> It is glibc that provides them, not GCC.
> See https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD

Minor addition: That header file is included in math.h, i.e. automatically available.
For Fortran/gfortran there is math-vector-fortran.h (also provided by glibc)
which has the same functions and a similar effect.

Cheers,

Tobias

-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstraße 201, 80634 München / Germany
Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Alexander Walter

Re: gcc math functions for OpenMP vectoization

[Toon Moene]

On 6/5/20 6:10 PM, Tobias Burnus wrote:

> On 6/5/20 4:11 PM, Jakub Jelinek via Gcc wrote:

>> It is glibc that provides them, not GCC.
>> See 
>> https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD 
>>
> 
> Minor addition: That header file is included in math.h, i.e. 
> automatically available.
> For Fortran/gfortran there is math-vector-fortran.h (also provided by 
> glibc)
> which has the same functions and a similar effect.

I wonder if there are Linux distributions where this actually effected 
already.

I know for sure that it is not in Debian Testing (as of two weeks ago) 
and Red Hat Fedora 30 (similarly).

Do you know of any ?

Kind regards,

-- 
Toon Moene - e-mail: toon@moene.org - phone: +31 346 214290
Saturnushof 14, 3738 XG  Maartensdijk, The Netherlands

Re: gcc math functions for OpenMP vectoization

[Richard Biener]

On June 5, 2020 7:58:20 PM GMT+02:00, Toon Moene <toon@moene.org> wrote:
>On 6/5/20 6:10 PM, Tobias Burnus wrote:
>
>> On 6/5/20 4:11 PM, Jakub Jelinek via Gcc wrote:
>
>>> It is glibc that provides them, not GCC.
>>> See 
>>>
>https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD
>
>>>
>> 
>> Minor addition: That header file is included in math.h, i.e. 
>> automatically available.
>> For Fortran/gfortran there is math-vector-fortran.h (also provided by
>
>> glibc)
>> which has the same functions and a similar effect.
>
>I wonder if there are Linux distributions where this actually effected 
>already.
>
>I know for sure that it is not in Debian Testing (as of two weeks ago) 
>and Red Hat Fedora 30 (similarly).
>
>Do you know of any ?

It works in openSUSE Tumbleweed at least.

Richard. 

>
>Kind regards,

Re: gcc math functions for OpenMP vectoization

[Jonathan Wakely]

On Fri, 5 Jun 2020 at 18:59, Toon Moene <toon@moene.org> wrote:
>
> On 6/5/20 6:10 PM, Tobias Burnus wrote:
>
> > On 6/5/20 4:11 PM, Jakub Jelinek via Gcc wrote:
>
> >> It is glibc that provides them, not GCC.
> >> See
> >> https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD
> >>
> >
> > Minor addition: That header file is included in math.h, i.e.
> > automatically available.
> > For Fortran/gfortran there is math-vector-fortran.h (also provided by
> > glibc)
> > which has the same functions and a similar effect.
>
> I wonder if there are Linux distributions where this actually effected
> already.
>
> I know for sure that it is not in Debian Testing (as of two weeks ago)
> and Red Hat Fedora 30 (similarly).

Fedora 30 is EOL.

> Do you know of any ?

Fedora 31 and 32.

Re: gcc math functions for OpenMP vectoization

[Florian Weimer]

* Toon Moene:

> On 6/5/20 6:10 PM, Tobias Burnus wrote:
>
>> On 6/5/20 4:11 PM, Jakub Jelinek via Gcc wrote:
>
>>> It is glibc that provides them, not GCC.
>>> See 
>>> https://sourceware.org/git/?p=glibc.git;a=blob;f=sysdeps/x86/fpu/bits/math-vector.h;h=0801905da7b85e2f43fb6b682a7b84c5eec469d4;hb=HEAD 
>>>
>> 
>> Minor addition: That header file is included in math.h, i.e. 
>> automatically available.
>> For Fortran/gfortran there is math-vector-fortran.h (also provided by 
>> glibc)
>> which has the same functions and a similar effect.
>
> I wonder if there are Linux distributions where this actually effected 
> already.
>
> I know for sure that it is not in Debian Testing (as of two weeks ago) 
> and Red Hat Fedora 30 (similarly).
>
> Do you know of any ?

I think we backported the header file into Red Hat Enterprise Linux
8.2.  So CentOS 8 should have it as well.

The system compiler is too old to use it, but GCC Toolset 9 comes with
a Fortran compiler that should pick it up (but I don't use Fortran
myself).