Template for modulo operations #40
Conversation
|
As a small note on the fast mod mul. Locally on my windows computer using PyPy the code runs around a factor of 10 faster with modmul compared to doing |
| or calculate matrix multiplication mod MOD. | ||
| """ | ||
|
|
||
| def fast_modder(MOD): |
There was a problem hiding this comment.
Perhaps call this make_modmul?
|
|
||
| def fast_modder(MOD): | ||
| """ Returns function modmul(a,b) that quickly calculates a * b % MOD, assuming 0 <= a,b < MOD """ | ||
| import sys, platform |
There was a problem hiding this comment.
I'm tempted to move this up, to the top of the file, even though this is specific.
| def redu(x): | ||
| return x if x < MOD else x - MOD | ||
|
|
||
| def matrix_modmul(A, B): |
There was a problem hiding this comment.
We have something for this elsewhere, so let's remove that too.
| import sys, platform | ||
| impl = platform.python_implementation() | ||
| maxs = sys.maxsize | ||
| if 'PyPy' in impl and MOD <= maxs and MOD ** 2 > maxs: |
There was a problem hiding this comment.
You can actually check impl == 'PyPy'. Seems to work on PyPy3 as well.
bjorn-martinsson
force-pushed
the
modmul_template
branch
from
8399462 to
ccb5210
Compare
|
With the new version of PyPy on Windows, what are your thoughts on this moving forward? |
Once CF updates (assuming it ever updates) then we can remove all the fast modmul stuff from the template. But I still think this template has its uses even after that. |
| """ Calculate n choose k in O(1) time """ | ||
| if k < 0 or k > n: | ||
| return 0 | ||
| return modmul(modmul(fac[n], fac_inv[k]), fac_inv[n - k]) |
There was a problem hiding this comment.
fac_inv should be inv_fac here. 🙂
|
This is probably not necessary anymore since 64bit has become the standard |
One of the most common codes that I've looked up over time is a mod template that I wrote many years ago for a private repository. It is useful enough to be used around once every second competition.
I cleaned up my old code, and added some extra features to it. I think this could be a super useful template to have on PyRival!
Note that I have yet to add tests to it, I will do that tomorrow.
@cheran-senthil What do you think about this template? I think it is both very simple and super useful to have.