[DOCS] Document make jobserver related changes for GCC 13.

Message ID 87f6f563-8e1a-0f14-8993-cc0bfb3d926c@suse.cz
State New, archived
Headers
Series [DOCS] Document make jobserver related changes for GCC 13. |

Commit Message

Martin Liška Aug. 18, 2022, 11:33 a.m. UTC
  Ready for master?

---
 htdocs/gcc-13/changes.html | 9 +++++++++
 1 file changed, 9 insertions(+)
  

Comments

Gerald Pfeifer Aug. 18, 2022, 8:53 p.m. UTC | #1
On Thu, 18 Aug 2022, Martin Liška wrote:
> Ready for master?

Nearly. :)

> +    Link-time optimization improvements:
> +    <ul>
> +        <li>LTO supports the newly added GNU make's jobserver that uses named pipes (<code>--jobserver-style=fifo</code>)

I believe that's just "GNU make" (instead of "GNU make's"), or maybe 
"...the newly added jobserver of GNU make...".

Not 100% sure myself.

> +        <li>If make's jobserver is active, parallel LTO WPA streaming 
>    communicates with it and so the streaming
> +        does not lead to the system overcommitting.</li>

Might we be able to simplify this to something like

  "...parallel LTO WPA streaming communicates with it and thus avoids
  system overcommitting" ?


Just two minor points; the patch as such looks fine.

Thank you,
Gerald
  
Martin Liška Aug. 22, 2022, 9:24 a.m. UTC | #2
On 8/18/22 22:53, Gerald Pfeifer wrote:
> On Thu, 18 Aug 2022, Martin Liška wrote:
>> Ready for master?
> 
> Nearly. :)
> 
>> +    Link-time optimization improvements:
>> +    <ul>
>> +        <li>LTO supports the newly added GNU make's jobserver that uses named pipes (<code>--jobserver-style=fifo</code>)
> 
> I believe that's just "GNU make" (instead of "GNU make's"), or maybe 
> "...the newly added jobserver of GNU make...".
> 
> Not 100% sure myself.
> 
>> +        <li>If make's jobserver is active, parallel LTO WPA streaming 
>>    communicates with it and so the streaming
>> +        does not lead to the system overcommitting.</li>
> 
> Might we be able to simplify this to something like
> 
>   "...parallel LTO WPA streaming communicates with it and thus avoids
>   system overcommitting" ?
> 
> 
> Just two minor points; the patch as such looks fine.

Thanks for the review, installed with your suggestions.

Martin

> 
> Thank you,
> Gerald
  

Patch

diff --git a/htdocs/gcc-13/changes.html b/htdocs/gcc-13/changes.html
index 57bd8724..95fee90d 100644
--- a/htdocs/gcc-13/changes.html
+++ b/htdocs/gcc-13/changes.html
@@ -52,6 +52,15 @@  a work-in-progress.</p>
     AddressSanitizer defaults to <code>detect_stack_use_after_return=1</code> on GNU/Linux targets.
     For compatibility, it can be disabled with <code>env ASAN_OPTIONS=detect_stack_use_after_return=0</code>.
   </li>
+  <li>
+    Link-time optimization improvements:
+    <ul>
+        <li>LTO supports the newly added GNU make's jobserver that uses named pipes (<code>--jobserver-style=fifo</code>)
+        by default.</li>
+        <li>If make's jobserver is active, parallel LTO WPA streaming communicates with it and so the streaming
+        does not lead to the system overcommitting.</li>
+    </ul>
+  </li>
 </ul>