New release? #442
Comments
|
It is difficult to spend any more time on this activity because my
institute does not appreciate SLEEF project AT ALL.
…On 3/13/2022 1:50 AM, Francesco Biscani wrote:
Hello,
First of all, thanks for this great library!
Is there an ETA for the release of version 3.6? Apologies if this is
the wrong communication channel for this sort of question...
—
Reply to this email directly, view it on GitHub
<#442>, or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AD2A7F7ATEXRB3UJQQTWAA3U7TDNDANCNFSM5QSDIBYA>.
Triage notifications on the go with GitHub Mobile for iOS
<https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675>
or Android
<https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub>.
You are receiving this because you are subscribed to this
thread.Message ID: ***@***.***>
|
|
I am really sorry to hear that :( |
|
There is no other OSS implementation of trigonometric functions that can fulfill the triad of performance, accuracy, and platform independence in the way Sleef does. It would be very sad if Sleef just fades away. |
|
Before version 2.80, SLEEF has been developed intermittently. The way
SLEEF is developed is just reverted to that state.
As you know, this is a fundamental problem of OSS development. Usually,
development of software like SLEEF would require several 10 times of the
amount I received as donations so far. It is impossible to continue the
pace of development with that amount.
…On 3/22/2022 4:19 AM, carlkl wrote:
There is no other OSS implementation of trigonometric functions that
can fulfill the triad of performance, accuracy, and platform
independence in the way Sleef does. It would be very sad if Sleef just
fades away.
—
Reply to this email directly, view it on GitHub
<#442 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AD2A7F4FPEVO6DR3LL3EJEDVBDDVNANCNFSM5QSDIBYA>.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
Maybe Sleef development needs something like that: https://chanzuckerberg.com/eoss/ NumPy and OpenBLAS as an example: |
|
The original plan was to gain recognition from my institute for
maintaining software that has a significant impact on society.
I succeeded in publishing a good paper on SLEEF and received donations
from large companies, so I thought that people in my institute would
understand the importance of SLEEF.
In reality, however, i found that people in my institute do not
distinguish between SLEEF and some silly proof-of-concept codes for
student researches.
They frequently tell us that it is important to implement the results of
the researches to directly contribute to the society, but it turned out
that they simply say it like that and have no interest in actually doing it.
…On 3/22/2022 8:58 PM, carlkl wrote:
Maybe Sleef development needs something like that:
https://chanzuckerberg.com/eoss/
NumPy and OpenBLAS as an example:
https://chanzuckerberg.com/eoss/proposals/strengthening-numpys-foundations-growing-beyond-code/
—
Reply to this email directly, view it on GitHub
<#442 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AD2A7F5X2GYSBLEF63G7KHDVBGYXJANCNFSM5QSDIBYA>.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
Sad to hear. However, it makes no sense to comment on the research policy of an institute, it's on there own. If maintenance work is not supported (because it is does not qualify as research work?) further 3rd party donations may be the answer. Keeping up the motivation for a bigger OSS project can be quite tedious. BTW: In the last two years I tried to promote Sleef here and there in the Python/Numpy community, but there was not much more than some courteous interest. And I didn't yet found time to get my hands dirty to integrate into a python project worth mentioning (last year I started with Sleef and https://github.com/pydata/numexpr but didn't go far). At least I can say a lot of developers in this community are aware of Sleef and know that this is a very interesting library when it comes to SIMD vectorization. On the other hand pytorch is the only python project I'm aware of that is using Sleef in production. Numpy decided to use to integrate the OSS part of SVML into its codebase (https://github.com/numpy/SVML). This works only for avx-512. And some parts of numpy are now written with universal intrinsics: see https://numpy.org/devdocs/reference/simd/index.html and https://github.com/numpy/numpy/blob/main/numpy/core/src/umath/loops_trigonometric.dispatch.c.src, but Sleef was never considered to integrate in into Numpy. In comparison: Sleef is feature complete (C99 trig functions) and could yield better results, especially for the full numeric range IMHO. Maybe a new Sleef release and some more promotion could help to gain interest in this community and for gain a grant.(Numfocus i.e.) Just my two cents. |
|
I think that's a kind of idealism, and in reality, everyone cares about
their own career and salary. If you turn your back on that, people will
eventually leave OSS development. Even the GNU project grew because of
the support of universities.
Looking at the current situation, it is very difficult to make a living
only by developing OSS.
The fact that institutes have thrown away their development support
basically means that the developers have to use their spare time to
maintain the software.
…On 3/23/2022 2:51 AM, carlkl wrote:
Sad to hear. However, it makes no sense to comment on the research
policy of an institute, it's on there own. If maintenance work is not
supported (because it is does not qualify as research work?) further
3rd party donations may be the answer. Keeping up the motivation for a
bigger OSS project can be quite tedious.
BTW: In the last two years I tried to promote Sleef here and there in
the Python/Numpy community, but there was not much more than some
courteous interest. And I didn't yet found time to get my hands dirty
to integrate into a python project worth mentioning (last year I
started with Sleef and https://github.com/pydata/numexpr but didn't go
far). At least I can say a lot of developers in this community are
aware of Sleef and know that this is a very interesting library when
it comes to SIMD vectorization. On the other hand pytorch is the only
python project I'm aware of that is using Sleef in production.
Numpy decided to use to integrate the OSS part of SVML into its
codebase (https://github.com/numpy/SVML). This works only for avx-512.
And some parts of numpy are now written with universal intrinsics: see
https://numpy.org/devdocs/reference/simd/index.html and
https://github.com/numpy/numpy/blob/main/numpy/core/src/umath/loops_trigonometric.dispatch.c.src,
but Sleef was never considered to integrate in into Numpy. In
comparison: Sleef is feature complete (C99 trig functions) and could
yield better results, especially for the full numeric range IMHO.
Maybe a new Sleef release and some more promotion could help to gain
interest in this community and for gain a grant.(Numfocus i.e.)
Just my two cents.
—
Reply to this email directly, view it on GitHub
<#442 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AD2A7F2ICPLRJ6FPXVTJ3ETVBICCDANCNFSM5QSDIBYA>.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
Thanks for this insight. I faced the same situation: OSS work in spare time means that good ideas come to a premature end due to slow progress. And small grants don't really help in the end. |
|
The 3.6 release was tagged today. |
Hello,
First of all, thanks for this great library!
Is there an ETA for the release of version 3.6? Apologies if this is the wrong communication channel for this sort of question...
The text was updated successfully, but these errors were encountered: