Some more improvements for OFPModules

[HechtiDerLachs]

This are some small things I encountered during my work on tuning direct images. I will add specific examples and timings.

[HechtiDerLachs]

Timings. On this branch:

julia> R, (x, y) = QQ[:x, :y]
(Multivariate polynomial ring in 2 variables over QQ, QQMPolyRingElem[x, y])

julia> @time F = FreeMod(R, 100000)
  0.000012 seconds (16 allocations: 736 bytes)
Free module of rank 100000 over R

julia> @time phi = hom(F, F, gens(F))
  0.009471 seconds (800.02 k allocations: 40.436 MiB)
Module homomorphism
  from F
  to F

julia> @time compose(phi, phi)
  0.032749 seconds (1.50 M allocations: 65.605 MiB)
Module homomorphism
  from F
  to F

On master:

julia> R, (x, y) = QQ[:x, :y]
(Multivariate polynomial ring in 2 variables over QQ, QQMPolyRingElem[x, y])

julia> @time F = FreeMod(R, 100000)
  0.221484 seconds (601.90 k allocations: 19.946 MiB, 67.16% gc time, 2.20% compilation time)
Free module of rank 100000 over R

julia> @time phi = hom(F, F, gens(F))
  0.008930 seconds (800.04 k allocations: 40.438 MiB)
Module homomorphism
  from F
  to F

julia> @time compose(phi, phi)
  0.268555 seconds (3.80 M allocations: 167.070 MiB, 59.42% gc time)
Module homomorphism
  from F
  to F

In the examples I have, this makes an actual difference.

[jankoboehm]

As along as images_of_generators is available in the generality we discussed it looks fine.

[jankoboehm]

What is the culprit for the failing test?

[lgoettgens]

What is the culprit for the failing test?

some random segfault in the low-level julia IO code. I restarted the job

[HechtiDerLachs]

images_of_generators has been made available for FreeModuleHom and SubQuoHom, so that should be fine.