On 2/20/19 2:05 PM, Zack Weinberg wrote:
> On Wed, Feb 20, 2019 at 4:43 AM Martin Liška <mliska@suse.cz> wrote:
>> On 2/19/19 8:14 PM, Zack Weinberg wrote:
>>> My first reaction is to wonder whether math-vector-fortran.h should
>>> have an extension other than .h.  I don't know what conventions there
>>> are for this sort of thing in Fortran, but if it could have some other
>>> extension then it would be easy to exclude it from this test in the
>>> makefiles.  Another possibility is to put an Emacs-style "mode"
>>> annotation on the first line of the file:  -*- f90 -*- which I could
>>> make my test program recognize and skip the file.
>>
>> Hi.
>>
>> I prefer to do the annotation instead of changing the file extension.
>> Can you please verify it's correct in the attached patch?
> 
> Yes, these annotations are correct.  You don't have to put them on a
> line by themselves, though,
> 
> +! Platform-specific Fortran SIMD declarations of math functions. -*- f90 -*-
> 
> would also work.  As long as I'm looking at this patch, I have two
> editorial notes:
> 
> In both new files, "Platform-specific Fortran SIMD declarations of
> math functions" would be clearer with different word order: suggest
> "Platform-specific declarations of SIMD math functions for Fortran."
> 
> In the generic file, "Get default empty builtin definitiones." has a
> typo ("definitiones" should be "definitions"), and having a sentence
> starting with "Get" on the last line of the file makes it look like
> there's something missing.  Suggest instead "No SIMD math functions
> are available for this platform."
> 
> zw
> 

Thank you for the corrections. I'm attaching final version of the patch
that I'm going to install.

Martin