I did a regression report on V 1.3.8 on one Linux running
fricas --version
FriCAS 1.3.8
based on sbcl 2.2.5
And new Linux running
>fricas --version
FriCAS 1.3.8
based on ecl 21.2.1
And noticed a timeout on this one integration problem. I build both Fricas on each Linux from same source tar file.
setSimplifyDenomsFlag(true)
integrate(1/(-x^3+1)/(b*x^3+a)^(1/3),x)
On the first system it finished much faster, in less than 2 minutes, on the second it finished in 6 minutes.
Both Linux are different Virtual boxes, running with similar Linux OS (Manjaro) and similar and plenty amount of RAM. All virtual boxes are on windows 10.
So the only difference I could think of is the lisp used?
I repeated the above more than one time, and the timings remain very different. The second one is almost three times slower.
So my question: is there a preferred lisp to use? and how to tell fricas when building it to use it?
I am not now sure why one used one lisp and the other used different one. I do not remember now setting this when configuring the build from source for Fricas.
May be Linux had default lisp and it picked it up during the build?
--Nasser