ENH: jit the 1d quadrature routines #352
Conversation
|
This is now ready for review/merge |
| @@ -573,7 +586,7 @@ def qnwbeta(n, a=1.0, b=1.0): | |||
| return _make_multidim_func(_qnwbeta1, n, a, b) | |||
|
|
|||
|
|
|||
| def qnwgamma(n, a=None): | |||
| def qnwgamma(n, a=1.0, b=1.0, tol=3e-14): | |||
There was a problem hiding this comment.
@sglyon the function signature seems to no longer correspond to the docstring.
|
Nice -- thanks @sglyon. |
|
Does this need any additional tests. |
|
@sglyon do you by chance have any performance comparison. No worries if you don't but if you had a ROM then I would include that in the release notes. |
|
I don't have performance comparisons. I did this not because of potential performance gains from calling it, but rather so I can call these routines inside other Good catch regarding the docstring, I'll fix that up |
|
Thanks @sglyon - ok cool - that's a good reason. When (!) I get performance tracking up and running then it will pick up these kind of commits anyway from the git history. |
|
Ok, docstrings updated! |
|
thanks @sglyon. |
This just jits the 1-dimensional quadrature routines.
The multi-dim ones are implemented using higher order functions, so they still will not work in nopython mode.