Re: Building math-vector-fortran.h in glibc

[Martin Liška <mliska at suse dot cz>]

On 1/18/19 5:08 PM, Steve Ellcey wrote:
> Martin,

Hey.

> 
> I have been working with libmvec on aarch64 and I took your glibc script
> to create math-vector-fortran.h and incorporated it into the glibc
> math Makefile to try it out.  The patch is attached to this email.
> 
> The main thing I noticed is that your script was specifically grepping
> for __DECL_SIMD_x86_64 and when I created a math-vector.h file for
> Aarch64 I used __DECL_SIMD_AARCH64 so I had to change the script.
> 
> I don't think we want a different script for each target so we need to
> find some way to deal with this.  Either use a common name on all
> platforms, which I think is OK since this should be an internal
> macro name that is not used by end users or find a way to generalize
> the match regardless of the name.
> 
> The other thought I had was, instead of scanning math-vector.h,
> have the script preprocess math.h itself and then look for simd
> attributes after preprocessing.  I think that would make the scanner
> more complex though.
> 
> Steve Ellcey
> sellcey@marvell.com
> 

Thank you for working on that. I tend to write the header files by hand
for both x86_64 and aarch64. Would be probably easier than maintaining
a python script.

I'm still waiting for final version of target filter directive, as discussed here:
https://gcc.gnu.org/ml/gcc-patches/2019-01/msg01168.html

Btw. is there a chance we can catch the upcoming glibc release with the header files?

Thanks,
Martin