# Is applying a ring homomorphism faster than symbolic substitution?

I'm working on a project where I need to do composition of polynomials; something like

P(Q1 + Q2)


where P, Q1, and Q2 are univariate polynomials with several hundred terms, and large integer coefficients (on the order of 10^10 or so). I've been doing this with the .subs() method which, I think, moves things to the symbolic ring and does the substitutions there. (I think this because when I get errors, they have to do with coercing to or from the symbolic ring.) But it occurred to me I could also define a ring homomorphism sending the variable of P to Q1 + Q2, and then apply the homomorphism to P.

So my question: will this be worth my while, or are the ring homomorphism methods too slow?

edit retag close merge delete

Sort by » oldest newest most voted

Hi Niles!

Let's try.

Substitution:

sage: R.<x> = QQ[]
sage: p = R.random_element(50)
sage: q = R.random_element(50)
sage: %time r=p(q)
CPU times: user 0.11 s, sys: 0.00 s, total: 0.11 s
Wall time: 0.11 s
sage: p = R.random_element(100)
sage: q = R.random_element(100)
sage: %time r=p(q)
CPU times: user 1.10 s, sys: 0.02 s, total: 1.12 s
Wall time: 1.12 s
sage: p = R.random_element(200)
sage: q = R.random_element(200)
sage: %time r=p(q)
CPU times: user 25.27 s, sys: 0.27 s, total: 25.54 s
Wall time: 25.61 s


Homomorphism:

sage: p = R.random_element(50)
sage: q = R.random_element(50)
sage: f = R.hom([q])
# Test that the result is the same
sage: p(q)==f(p)
True
sage: %time r=f(p)
CPU times: user 0.45 s, sys: 0.00 s, total: 0.45 s
Wall time: 0.45 s
sage: p = R.random_element(100)
sage: q = R.random_element(100)
sage: f = R.hom([q])
sage: %time r=f(p)
CPU times: user 16.73 s, sys: 0.00 s, total: 16.74 s
Wall time: 16.79 s


So, substitution is considerably faster than a homomorphism. I find that's a shame! If other people agree, we should open a ticket.

By the way: I tried the same using a multivariate ring with one variable. The degree 50 example didn't even finish.

more

Thanks Simon :) I guess I could have done something like this myself . . . do you think there's any scenario in which ring homomorphisms beat symbolic substitution? (e.g. maybe when the coefficients are also polynomials?)

( 2011-02-06 09:25:26 -0500 )edit