public inbox for gcc@gcc.gnu.org
 help / color / mirror / Atom feed
* Learning hardware optimization patch
@ 2022-11-15 18:42 Sławomir Lach
  2022-11-15 18:44 ` Sławomir Lach
  0 siblings, 1 reply; 2+ messages in thread
From: Sławomir Lach @ 2022-11-15 18:42 UTC (permalink / raw)
  To: gcc

I am wonder if this will be good choice to allow gcc to generate AI data about 
best optimization approach on user computer.

Yes. GCC team will provide source code for some examples and (once low 
hardware consumption) made gcc to compile these sources with various 
optimization patch. It will take avg run time of each patch. Of course - it 
will run compilation once there is many free resources on computer.

Why? To allow user select --best-optimiation-speed-for-my-machine or another 
optimization criteria, which will depends on AI data sheet.

What do you think?

Another idea is to allow adding feature point to program. Program will report 
that user is using some feature. In next step, linker could made optimization 
based on this - by simply, compiles only functionality, which is used by user. 
Rest functions will be put into shared objects and special code to load this 
shared object could been injected to executable.

I am not very good in programming, but have had crazy ideas about lot of 
stuff.

Best regards,
Lach Sławomir.



^ permalink raw reply	[flat|nested] 2+ messages in thread

* Re: Learning hardware optimization patch
  2022-11-15 18:42 Learning hardware optimization patch Sławomir Lach
@ 2022-11-15 18:44 ` Sławomir Lach
  0 siblings, 0 replies; 2+ messages in thread
From: Sławomir Lach @ 2022-11-15 18:44 UTC (permalink / raw)
  To: gcc

Dnia wtorek, 15 listopada 2022 19:42:13 CET Sławomir Lach pisze:
> I am wonder if this will be good choice to allow gcc to generate AI data
> about best optimization approach on user computer.
> 
> Yes. GCC team will provide source code for some examples and (once low
> hardware consumption) made gcc to compile these sources with various
> optimization patch. It will take avg run time of each patch. Of course - it
> will run compilation once there is many free resources on computer.
> 

Not patch, but path - sorry. Applying different patch, which have optimization 
is bad idea, because it forces developer to carry out many path :-D .

> Why? To allow user select --best-optimiation-speed-for-my-machine or another
> optimization criteria, which will depends on AI data sheet.
> 
> What do you think?
> 
> Another idea is to allow adding feature point to program. Program will
> report that user is using some feature. In next step, linker could made
> optimization based on this - by simply, compiles only functionality, which
> is used by user. Rest functions will be put into shared objects and special
> code to load this shared object could been injected to executable.
> 
> I am not very good in programming, but have had crazy ideas about lot of
> stuff.
> 
> Best regards,
> Lach Sławomir.





^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2022-11-15 18:44 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2022-11-15 18:42 Learning hardware optimization patch Sławomir Lach
2022-11-15 18:44 ` Sławomir Lach

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).