Albert Steppi
Albert Steppi
> * If the input is `float32`, we get a `float32` result, but internally the function works with `float64`. Is this intended (probably standard)? Yeah, this is standard throughout special...
~"Linux tests / free-threaded" seem to be failing everywhere right now.~ scooped
Hi @miflynn, thanks for the request. I did a quick check and extending the Jacobi elliptic functions to $m < 0$ would actually be pretty straightforward. For negative $m$ we...
> hi @steppi, can i give this a go ( if you are not looking into it right now ) ? I can go through a bit of the code...
Thanks @WarrenWeckesser for pointing this out. This is also on me for letting this get through review. I guess the right idiom here might be something like ```C++ T* u...
> It looks like this occurs in C++ functions with `void` return type. They could be changed to return an `int` status code (eg. 0 means OK, -1 means memory...
> In general, it does not feell right to alloc for a special function/ufunc if it is running over a long array. > > Especially functions like `kmn` doing some...
> > > > Chiming in here. We shouldn't change the `void` return type to return an errir code, as that actually messes up the direct mapping to ufuncs that...
> If it is possible we should not deal with errors, since I did not have ufuncs in mind I did a casual choice but now it is not a...
> In most cases it over-allocates that's why it is asking 100 a fixed number. But uses input dependent many entries. Ah, that’s right. Thanks for pointing that out again....