On Thu, Jun 30, 2022 at 10:56 PM Warren Weckesser <
warren.weckes...@gmail.com> wrote:
> On 6/30/22, Ewout ter Hoeven wrote:
> > A function to get the minimum and maximum values of an array
> simultaneously
> > could be very useful, from both a convenience and performance point of
> view.
> > Especially when arrays get larger the performance benefit could be
> > significant, and even more if the array doesn't fit in L2/L3 cache or
> even
> > memory.
> >
> > There are many cases where not either the minimum or the maximum of an
> array
> > is required, but both. Think of clipping an array, getting it's range,
> > checking for outliers, normalizing, making a plot like a histogram, etc.
> >
> > This function could be called aminmax() for example, and also be called
> like
> > ndarray.minmax(). It should return a tuple (min, max) with the minimum
> and
> > maximum values of the array, identical to calling (ndarray.min(),
> > ndarray.max()).
> >
> > With such a function, numpy.ptp() and the special cases of
> numpy.quantile(a,
> > q=[0,1]) and numpy.percentile(a, q=[0,100]) could also potentially be
> > speeded up, among others.
> >
> > Potentially argmin and argmax could get the same treatment, being called
> > argminmax().
> >
> > There is also a very extensive post on Stack Overflow (a bit old already)
> > with discussion and benchmarks:
> >
> https://stackoverflow.com/questions/12200580/numpy-function-for-simultaneous-max-and-min
>
>
> FYI, I have a fairly simple gufunc implementation of `minmax` in
> ufunclab (https://github.com/WarrenWeckesser/ufunclab), along with
> `arg_minmax`, `min_argmin` and `max_argmax`. See README.md starting
> here: https://github.com/WarrenWeckesser/ufunclab#minmax
>
> For those familiar with C and gufunc implementation details, you can
> find the implementations in
>
> https://github.com/WarrenWeckesser/ufunclab/blob/main/src/minmax/minmax_gufunc.c.src
> .
> You'll see that, as far as gufuncs go, these are not very
> sophisticated. They do not include implementations for all the NumPy
> data types, and I haven't yet spent much time on optimization.
>
Thanks for sharing Warren. While that is interesting code, for inclusion in
NumPy purposes I'd much prefer to see something along the suggestions by
Eric and Marten on https://github.com/numpy/numpy/issues/9836 about making
it easier to combine existing ufuncs. Adding 500 LoC for a `minmax`
(separate from the discussion on whether we want such fused operators) does
not seem healthy. It's also a new potential source of bugs, because
`minmax` isn't going to yield exactly the same numerical values as `min`
and `max` used separately if you do it as a gufunc, and any bug fixes will
need to be made in two places. Also, given that `min` and `max` use SIMD
instructions, the gufunc `minmax` as you have it now is probably
substantially slower - see benchmarks in
https://github.com/numpy/numpy/pull/20131.
Cheers,
Ralf
___
NumPy-Discussion mailing list -- numpy-discussion@python.org
To unsubscribe send an email to numpy-discussion-le...@python.org
https://mail.python.org/mailman3/lists/numpy-discussion.python.org/
Member address: arch...@mail-archive.com
