Wrap her, syr, her2, syr2 blas routines #3984
Conversation
|
Admittedly these routines don't seem particularly vital, and we'd probably like the various packed matrix-vector products more, but since I hadn't used f2py before I though I would start with some that seemed easier than gbmv. |
|
It would be good to add these subroutines to the list at the top of blas.py so they are shown in the docs. |
|
Pushed a commit that adds them to the list in blas.py |
| integer, intent(in), optional :: offx = 0 | ||
| integer, intent(in), optional :: incx = 1 | ||
|
|
||
| integer, intent(in), optional :: n = (len(x)-1-offx)/abs(incx)+1 |
There was a problem hiding this comment.
Does this need a depend(x, offs) --- I vaguely remember some problems from missing these. Edit: yeah, you have it in syr2 below
There was a problem hiding this comment.
Yes, all variables used need the correspoding depends() statement, or
the computations may be done in an incorrect order.
There was a problem hiding this comment.
Furthermore, if a depends on b and b depends on c, c needs depend(a, b) --- it seems f2py does not follow the dependency chain. cf ev-br@a38faa5
Conflicts: scipy/linalg/tests/test_blas.py
|
Updated. |
|
Changes Unknown when pulling 6ad3686 on ewmoore:wrap_blas into * on scipy:master*. |
|
LGTM, I'll merge this. Meanwhile, which routines do you think need |
Wrap her, syr, her2, syr2 blas routines
As a start toward including the rest of the currently unwrapped blas routines, I've wrapped these.
I'd like some feedback here on what we actually want out of these wrappers. Specifically, what we have doesn't seem particularly consistent from function to function (some have offx, some don't, etc.)
There also seem to be some issues (cf. #3983) and it looks to me like
xandyingershouldn't really have the overwrite flag, since those are input only. I haven't changed that, since it would change the signature which we probably can't from a back compat standpoint.