> Stavros Macrakis <macrakis at alum.mit.edu> writes:
>
> Stavros> cache) bandwidth is the limiting factor, not processor speed. I think there
> Stavros> are old published papers with timings of calculations.
Last year I posted the following in this list:
> In "Computer Algebra. Systems and Algorithms for Algebraic
> Computation", by Davenport, Siret and Tournier, 1988, it says
> on page 17:
> "The following example takes 4 minutes on the Honeywell computer
> DPS8-70... the reader can already see that factorisation is an
> expensive process"
> here is the example, without any change, in Maxima 5.9.3.1cvs (clisp),
> run in my laptop which is already a relic:
> (%i3) for n:2 thru 10 do print(factor((u^n+v^n)^n - (u^n-v^n)^n));
> 4*u^2*v^2
> 2*v^3*(v^6+3*u^6)
> 8*u^4*v^4*(v^8+u^8)
> 2*v^5*(v^20+10*u^10*v^10+5*u^20)
> 4*u^6*v^6*(v^12+3*u^12)*(3*v^12+u^12)
> 2*v^7*(v^42+21*u^14*v^28+35*u^28*v^14+7*u^42)
> 16*u^8*v^8*(v^16+u^16)*(v^32+6*u^16*v^16+u^32)
> 2*v^9*(v^18+3*u^18)*(v^54+33*u^18*v^36+27*u^36*v^18+3*u^54)
> 4*u^10*v^10*(v^40+10*u^20*v^20+5*u^40)*(5*v^40+10*u^20*v^20+u^40)
> Evaluation took 2.75 seconds (2.81 elapsed) using 8.960 MB.
Now that I'm working on an AMD Athlon(tm) XP 2000+, 1GB RAM, I get
the following results, using Maxima 5.11.0cvs
clisp 2.38: Evaluation took 4.16 seconds (4.27 elapsed) using 8.987 MB.
GCL 2.6.7 : Evaluation took 0.89 seconds (1.62 elapsed)
SBCL 0.9.8: Evaluation took 1.48 seconds (1.77 elapsed) using 6.737 MB.
CMUCL 19a : Evaluation took 1.21 seconds (1.43 elapsed) using 6.857 MB.
It might be interesting to create a wiki page with benchmarks like this
and the results for different systems and versions.
Regards,
Jaime Villate