Disallow Accelerate as a LAPACK backend for NumPy

classic Classic list List threaded Threaded
6 messages Options
Reply | Threaded
Open this post in threaded view
|

Disallow Accelerate as a LAPACK backend for NumPy

mattip
Administrator
Apple has dropped support for Accelerate. It has bugs that have not been
fixed, and is closed source so we cannot fix them ourselves. We have
been getting a handful of reports from users who end up building NumPy
on macOS, and inadvertently link to Accelerate, then end up with wrong
linalg results. In PR 14880 https://github.com/numpy/numpy/pull/14880 I
propose to disallow finding it when building NumPy. At this time it will
remain in distutils as one of the backends to support users, but how do
people feel about a future PR to totally remove it?


Matti

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Disallow Accelerate as a LAPACK backend for NumPy

Ian Henriksen
Extra data point here: SciPy already dropped support for Accelerate as of version 1.2.0.

Best,

Ian Henriksen


On Mon, Nov 11, 2019 at 6:40 PM Matti Picus <[hidden email]> wrote:
Apple has dropped support for Accelerate. It has bugs that have not been
fixed, and is closed source so we cannot fix them ourselves. We have
been getting a handful of reports from users who end up building NumPy
on macOS, and inadvertently link to Accelerate, then end up with wrong
linalg results. In PR 14880 https://github.com/numpy/numpy/pull/14880 I
propose to disallow finding it when building NumPy. At this time it will
remain in distutils as one of the backends to support users, but how do
people feel about a future PR to totally remove it?


Matti

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Disallow Accelerate as a LAPACK backend for NumPy

Derek Homeier
On 12 Nov 2019, at 3:27 am, Ian Henriksen <[hidden email]> wrote:

>
> Extra data point here: SciPy already dropped support for Accelerate as of version 1.2.0.
>
> Best,
>
> Ian Henriksen
>
>
> On Mon, Nov 11, 2019 at 6:40 PM Matti Picus <[hidden email]> wrote:
> Apple has dropped support for Accelerate. It has bugs that have not been
> fixed, and is closed source so we cannot fix them ourselves. We have
> been getting a handful of reports from users who end up building NumPy
> on macOS, and inadvertently link to Accelerate, then end up with wrong
> linalg results. In PR 14880 https://github.com/numpy/numpy/pull/14880 I
> propose to disallow finding it when building NumPy. At this time it will
> remain in distutils as one of the backends to support users, but how do
> people feel about a future PR to totally remove it?

+1 from this side - when switching the packaged version of Scipy to OpenBLAS
(already taking Numpy along the way) I noticed barely any performance penalties
(if - with some benchmarks - it wasn’t actually faster than Accelerate).

Cheers,
                                Derek

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Disallow Accelerate as a LAPACK backend for NumPy

mattip
Administrator
In reply to this post by mattip
    On Tue, Nov 12, 2019 at 12:41 AM Matti Picus <[hidden email]>
    wrote:

    Apple has dropped support for Accelerate. It has bugs that have not been
    fixed, and is closed source so we cannot fix them ourselves. We have
    been getting a handful of reports from users who end up building NumPy
    on macOS, and inadvertently link to Accelerate, then end up with wrong
    linalg results. In PR 14880https://github.com/numpy/numpy/pull/14880  I
    propose to disallow finding it when building NumPy. At this time it will
    remain in distutils as one of the backends to support users, but how do
    people feel about a future PR to totally remove it?

Someone pointed out that Apple has not officially dropped support as far as it can be determined. Sorry for the bad information. However, I still stand by the "has bugs that have not been fixed, and is closed source". An alternative to dropping automatic support for it would be to find a channel for engaging with Apple to report and fix the bugs.

Matti

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Disallow Accelerate as a LAPACK backend for NumPy

ralfgommers


On Fri, Nov 15, 2019 at 5:27 AM Matti Picus <[hidden email]> wrote:
    On Tue, Nov 12, 2019 at 12:41 AM Matti Picus <[hidden email]>
    wrote:

    Apple has dropped support for Accelerate. It has bugs that have not been
    fixed, and is closed source so we cannot fix them ourselves. We have
    been getting a handful of reports from users who end up building NumPy
    on macOS, and inadvertently link to Accelerate, then end up with wrong
    linalg results. In PR 14880https://github.com/numpy/numpy/pull/14880  I
    propose to disallow finding it when building NumPy. At this time it will
    remain in distutils as one of the backends to support users, but how do
    people feel about a future PR to totally remove it?

Someone pointed out that Apple has not officially dropped support as far as it can be determined. Sorry for the bad information. However, I still stand by the "has bugs that have not been fixed, and is closed source". An alternative to dropping automatic support for it would be to find a channel for engaging with Apple to report and fix the bugs.

That's been tried, repeatedly. I would suggest not to spend time on that. Apple knows, they have just decided it's not important to them.

Spending time on contributing to either OpenBLAS or BLIS/libFLAME seems like a more useful activity.

Cheers,
Ralf


_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion
Reply | Threaded
Open this post in threaded view
|

Re: Disallow Accelerate as a LAPACK backend for NumPy

Ilhan Polat
We have a wiki page with all the details on scipy repo for the rationale of why we wanted to drop it. There is no need to discuss further about the situation of Accelerate. 

On Fri, Nov 15, 2019, 15:41 Ralf Gommers <[hidden email]> wrote:


On Fri, Nov 15, 2019 at 5:27 AM Matti Picus <[hidden email]> wrote:
    On Tue, Nov 12, 2019 at 12:41 AM Matti Picus <[hidden email]>
    wrote:

    Apple has dropped support for Accelerate. It has bugs that have not been
    fixed, and is closed source so we cannot fix them ourselves. We have
    been getting a handful of reports from users who end up building NumPy
    on macOS, and inadvertently link to Accelerate, then end up with wrong
    linalg results. In PR 14880https://github.com/numpy/numpy/pull/14880  I
    propose to disallow finding it when building NumPy. At this time it will
    remain in distutils as one of the backends to support users, but how do
    people feel about a future PR to totally remove it?

Someone pointed out that Apple has not officially dropped support as far as it can be determined. Sorry for the bad information. However, I still stand by the "has bugs that have not been fixed, and is closed source". An alternative to dropping automatic support for it would be to find a channel for engaging with Apple to report and fix the bugs.

That's been tried, repeatedly. I would suggest not to spend time on that. Apple knows, they have just decided it's not important to them.

Spending time on contributing to either OpenBLAS or BLIS/libFLAME seems like a more useful activity.

Cheers,
Ralf

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion

_______________________________________________
NumPy-Discussion mailing list
[hidden email]
https://mail.python.org/mailman/listinfo/numpy-discussion