[wwwdocs, patch] gcc-12/changes.html: OpenMP (depobj/mutexinoutset for depend), OpenACC (-Wopenacc-parallelism)

[wwwdocs, patch] gcc-12/changes.html: OpenMP (depobj/mutexinoutset for depend), OpenACC (-Wopenacc-parallelism)

[Tobias Burnus]

[-- Attachment #1: Type: text/plain, Size: 442 bytes --]

Hi all,

this patch documents a new OpenMP feature (many more to be added) in GCC 12.
And it documents a new flag for OpenACC.

Comments? Wording suggestions? I think for OpenMP, the sentence will be
modified several times before the release :-)

Tobias

-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstrasse 201, 80634 München Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Frank Thürauf

[-- Attachment #2: gcc12.diff --]
[-- Type: text/x-patch, Size: 911 bytes --]

diff --git a/htdocs/gcc-12/changes.html b/htdocs/gcc-12/changes.html
index 8f257e87..1022b5eb 100644
--- a/htdocs/gcc-12/changes.html
+++ b/htdocs/gcc-12/changes.html
@@ -40,6 +40,18 @@ a work-in-progress.</p>
 <!-- .................................................................. -->
 <h2 id="languages">New Languages and Language specific improvements</h2>
 
+<ul>
+  <li>For Fortran, OpenMP 5.0 support has been extended for following features
+      which were before only available in C and C++:  <code>depobj</code>
+      and <code>mutexinoutset</code> can now also be used with the
+      <code>depend</code> clause.
+  </li>
+  <li>The new warning flag <code>-Wopenacc-parallelism</code> was added for
+      OpenACC, which warns about potentially suboptimal choices related to
+      OpenACC parallelism.
+  </li>
+</ul>
+
 <!-- <h3 id="ada">Ada</h3> -->
 
 <!-- <h3 id="brig">BRIG (HSAIL)</h3> -->

Re: [wwwdocs, patch] gcc-12/changes.html: OpenMP (depobj/mutexinoutset for depend), OpenACC (-Wopenacc-parallelism)

[Gerald Pfeifer]

On Mon, 26 Apr 2021, Tobias Burnus wrote:
> Comments? Wording suggestions? I think for OpenMP, the sentence will be 
> modified several times before the release :-)

Can I take this as a promise? :-)

+<ul>
+  <li>For Fortran, OpenMP 5.0 support has been extended for following features
+      which were before only available in C and C++:  <code>depobj</code>
+      and <code>mutexinoutset</code> can now also be used with the
+      <code>depend</code> clause.
+  </li>

I've been staring at this multiple times thinking on how to make it a
bit stronger/clearer.

How do you feel about

   "OpenMP 5.0 support for Fortran ... by the following features 
   which were available in C and C++ before:"

Or
 
  "Fortran gained support for the following OpenMP 5.0 features..."

?

(In marketing, which this kinds is, I suggest not saying "only" since
that risks diminishing the value we've had all along. Rather it's "even 
more/better now". ;-)

+  <li>The new warning flag <code>-Wopenacc-parallelism</code> was added for
+      OpenACC, which warns about potentially suboptimal choices related to
+      OpenACC parallelism.
+  </li>

Maybe break this as "... It warns about"?  That feels a bit more natural
to me than the "which" referring to an earlier term, though that may very
well be my subjective preference. ;-)


Please consider the above, and then it's okay, however you want to go
about it.

Gerald

Re: [wwwdocs, patch] gcc-12/changes.html: OpenMP (depobj/mutexinoutset for depend), OpenACC (-Wopenacc-parallelism)

[Tobias Burnus]

[-- Attachment #1: Type: text/plain, Size: 1211 bytes --]

Hi Gerald, hi all,

On 26.04.21 21:04, Gerald Pfeifer wrote:
> On Mon, 26 Apr 2021, Tobias Burnus wrote:
>> I think for OpenMP, the sentence will be
>> modified several times before the release :-)
> Can I take this as a promise? :-)

Given that Stage1 has just begun and several patches are
either waiting for review or to be finished and submitted: Yes :-)

> How do you feel about
>     "OpenMP 5.0 support for Fortran ... by the following features
>     which were available in C and C++ before:"
> Or
>    "Fortran gained support for the following OpenMP 5.0 features..."
I went for the former: it reduces the word smithing for non-Fortran-only
changes.

Thanks for the suggestions.

Tobias

PS: When looking at https://gcc.gnu.org/mirrors.html, I wondered whether
3 of 5 http mirrors links could move to https and given the trend that
FTP is phased out (on the server side and in the webbrowsers), whether
those ftp:// mirrors which also offer https:// should list https in
addition (or instead?).

-----------------
Mentor Graphics (Deutschland) GmbH, Arnulfstrasse 201, 80634 München Registergericht München HRB 106955, Geschäftsführer: Thomas Heurung, Frank Thürauf

[-- Attachment #2: committed.diff --]
[-- Type: text/x-patch, Size: 1223 bytes --]

commit 4debbf1dda6f26bfbc3dc93b7111b34bfacb252d
Author: Tobias Burnus <tobias@codesourcery.com>
Date:   Tue Apr 27 12:03:38 2021 +0200

    gcc-12/changes.html: OpenMP, OpenACC
    
    * htdocs/gcc-12/changes.html (languages): OpenMP/Fortran - depobj/mutexinoutset
      for depend; OpenACC - -Wopenacc-parallelism.

diff --git a/htdocs/gcc-12/changes.html b/htdocs/gcc-12/changes.html
index 8f257e87..e0ac986e 100644
--- a/htdocs/gcc-12/changes.html
+++ b/htdocs/gcc-12/changes.html
@@ -40,6 +40,18 @@ a work-in-progress.</p>
 <!-- .................................................................. -->
 <h2 id="languages">New Languages and Language specific improvements</h2>
 
+<ul>
+  <li>OpenMP 5.0 support for Fortran has been extended by the following features
+      which were available in C and C++ before:  <code>depobj</code>
+      and <code>mutexinoutset</code> can now also be used with the
+      <code>depend</code> clause.
+  </li>
+  <li>The new warning flag <code>-Wopenacc-parallelism</code> was added for
+      OpenACC. It warns about potentially suboptimal choices related to
+      OpenACC parallelism.
+  </li>
+</ul>
+
 <!-- <h3 id="ada">Ada</h3> -->
 
 <!-- <h3 id="brig">BRIG (HSAIL)</h3> -->

Re: [wwwdocs, patch] gcc-12/changes.html: OpenMP (depobj/mutexinoutset for depend), OpenACC (-Wopenacc-parallelism)

[Gerald Pfeifer]

On Tue, 27 Apr 2021, Tobias Burnus wrote:
> Thanks for the suggestions.

You're always welcome!

> PS: When looking at https://gcc.gnu.org/mirrors.html, I wondered whether
> 3 of 5 http mirrors links could move to https and given the trend that
> FTP is phased out (on the server side and in the webbrowsers), whether
> those ftp:// mirrors which also offer https:// should list https in
> addition (or instead?).

That's a good observation and recommendation; thank you!

I started to work on this; happy for you (or others) to join me over
the coming days. :)

Gerald