is -45Elo for Minic ...
Does this happened to someone else ?
Any countermeasure (apart from falling back to gcc9)?
Thanks for your inputs !
g++10 versus g++9
Moderators: hgm, Rebel, chrisw
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
-
- Posts: 253
- Joined: Mon Nov 16, 2020 12:13 pm
- Full name: Manuel Rivera
-
- Posts: 2488
- Joined: Tue Aug 30, 2016 8:19 pm
- Full name: Rasmus Althoff
Re: g++10 versus g++9
How many games?
No, just checked at a thousand games. However, the GCC-10 generated binary is 4.5% faster for me.Does this happened to someone else ?
Run your engine with a single worker thread. Use a few positions to calculate to some depth so that it takes at least 10 seconds. Check whether reported PV and node counts match for the binaries generated with GCC-9 and GCC-10. It's not uncommon for bugs to only manifest upon a compiler update.Any countermeasure (apart from falling back to gcc9)?
Rasmus Althoff
https://www.ct800.net
https://www.ct800.net
-
- Posts: 1871
- Joined: Sat Nov 25, 2017 2:28 pm
- Location: France
Re: g++10 versus g++9
1000 games, so accurate enough to be a real issue.Ras wrote: ↑Wed Jan 13, 2021 3:18 pmHow many games?
No, just checked at a thousand games. However, the GCC-10 generated binary is 4.5% faster for me.Does this happened to someone else ?
Run your engine with a single worker thread. Use a few positions to calculate to some depth so that it takes at least 10 seconds. Check whether reported PV and node counts match for the binaries generated with GCC-9 and GCC-10. It's not uncommon for bugs to only manifest upon a compiler update.Any countermeasure (apart from falling back to gcc9)?
No difference in terms of nodes count (or any other statistic Minic is able to print), so no obvious bug.
But 10% lower nps.
Anyway Minic was always slow and affected by small changes. I guess there are too much bad design that the compiler cannot always handle easily and optimize. It's propably more on my side than one gcc.
-
- Posts: 2559
- Joined: Fri Nov 26, 2010 2:00 pm
- Location: Czech Republic
- Full name: Martin Sedlak
Re: g++10 versus g++9
10% lower nps - that's at most 10 elo at hyperbullet, realistically about 5.
so I wonder where you get -45 I suspect an issue with testing.
Martin Sedlak
-
- Posts: 5566
- Joined: Tue Feb 28, 2012 11:56 pm
Re: g++10 versus g++9
Add - O3 to the compiler flags. (Hopefully this is a stupid suggestion.)