benchmarks for sparse polynomial methods

[1e9abhi1e10]

@oscarbenjamin I want to know whether I create a new file or add a benchmark in the benchmarks/polys.py file. Also, should I add all benchmarks which were I commented on the Issue.

[oscarbenjamin]

It would be good to have a polynomial benchmarks file if there is not already somewhere that has polynomial benchmarks.

Also, should I add all benchmarks which were I commented on the

Yes, and there should be many different benchmarks for them like comparing different methods for the same operation e.g. the prem function vs the Poly.prem method vs the PolyElement.prem method etc:

In [1]: f, g = x**2*y + x*z + 1, x*z + y

In [2]: %time prem(f, g, x)
CPU times: user 13.2 ms, sys: 196 µs, total: 13.4 ms
Wall time: 12.3 ms
Out[2]: 
 3      2    2
y  - y⋅z  + z 

In [3]: fp, gp = Poly(f, [x, y, z]), Poly(g, [x, y, z])

In [4]: %time fp.prem(gp)
CPU times: user 3.71 ms, sys: 0 ns, total: 3.71 ms
Wall time: 3.39 ms
Out[4]: Poly(y**3 - y*z**2 + z**2, x, y, z, domain='ZZ')

In [5]: R = ZZ[x,y,z]

In [6]: fpe, gpe = R.from_sympy(f), R.from_sympy(g)

In [7]: %time fpe.prem(gpe)
CPU times: user 1.86 ms, sys: 0 ns, total: 1.86 ms
Wall time: 1.88 ms
Out[7]: y**3 - y*z**2 + z**2

[1e9abhi1e10]

Moving ahead should we also add benchmarks for the pdiv, pquo and pexquo?

[oscarbenjamin]

Moving ahead should we also add benchmarks for the pdiv, pquo and pexquo?

Yes.

[oscarbenjamin]

Moving ahead should we also add benchmarks for the pdiv, pquo and pexquo?

Yes.

Actually thinking about this we don't need benchmarks for pquo, pexquo and pdiv because they all use the same code (and that is unlikely to change in future). Just one of these is fine.