Thought I'd share with the mailing list a little comparison I did
for polynomial GCD: FriCAS vs Fermat ( http://home.bway.net/lewis/ )

(My FriCAS is compiled with gmp and algebra optimizations ((speed 3) 
(safety 0)))

The following expression.input contains a 2.4MB large rational function

(If you want to reproduce the following results:
https://burp.github.io/expression.input.gz
https://burp.github.io/expression.fermat.gz
)

Run on a Intel Xeon from 2013:

)read expression.input )quiet
Time: 12.78 (IN) + 68.30 (EV) + 8.81 (OT) = 89.88 sec
(memory about 7.5GB maximum)

fermat6.37 < expression.fermat
CPU time: 1:20.180
(memory about 95MB)

So apart from the memory requirements, in terms of CPU time FriCAS seems to 
be on-par with Fermat.

I thought that maybe limiting the SBCL memory would slow it down, but I also
tested it on my laptop where I limited the SBCL memory somewhat more. As it
turns out, FriCAS is even faster there with a maximum memory size of ~1GB:

Run on Intel i7 from 2015:

)read expression.input )quiet
Time: 13.43 (IN) + 61.54 (EV) + 9.51 (OT) = 84.47 sec
(memory about 1.3GB maximum)

fermat6.36 < expression.fermat
CPU time: 2:4.706
(memory about 80MB)

fermat6.19 < expression.fermat
CPU time: 1:44.028 (There's clearly a regression in Fermat)
(memory about 70MB)

I don't understand the large difference between the Fermat runs on the 
older Intel and the newer, but since it's just using the binary and was not 
compiled from source it seems likely due to bad optimizations. In any case 
FriCAS does extremely well.

-- 
You received this message because you are subscribed to the Google Groups 
"FriCAS - computer algebra system" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion on the web visit 
https://groups.google.com/d/msgid/fricas-devel/ee7de10b-17e9-4081-9d34-141201511062n%40googlegroups.com.

Reply via email to