Plans for new sparse compilation backend for PyData/Sparse

Hameer Abbasi <[email protected]> Wed, 3 Jan 2024 14:34:17 +0100
Newsgroups gmane.comp.python.scientific.devel,gmane.comp.python.numeric.general
Message-ID <[email protected]>
--===============0104302331157273780==
Content-Type: multipart/alternative;
	boundary="Apple-Mail=_65ACEEF1-B62B-4FE8-8209-921ACFD0150A"


--Apple-Mail=_65ACEEF1-B62B-4FE8-8209-921ACFD0150A
Content-Transfer-Encoding: quoted-printable
Content-Type: text/plain;
	charset=utf-8

Hello everyone,

The stated goal for sparse is to provide a NumPy-like API with a sparse =
representation of arrays. To this end, Quansight and I have been =
collaborating with researchers at MIT CSAIL <https://www.csail.mit.edu/> =
- in particular Prof. Amarasinge's group =
<https://www.csail.mit.edu/research/commit-group> and the TACO =
<https://github.com/tensor-compiler/taco/> team - to develop a =
performant and production-ready package for N-dimensional sparse arrays. =
There were several attempts made to explore this over the last couple of =
years, including a LLVM back-end =
<https://github.com/Quansight-Labs/taco/pulls?q=3Dis%3Apr+llvm> for TACO =
<https://github.com/tensor-compiler/taco/>, and a pure-C++ =
template-metaprogramming approach called XSparse =
<https://github.com/hameerabbasi/xsparse>.

To this end, we, at Quansight, are happy to announce that we have =
received funding from DARPA, together with our partners from MIT, under =
their Small Business Innovation Research (SBIR) program =
<https://www.darpa.mil/work-with-us/for-small-businesses/HR0011SB20234-06>=
 to build out sparse using state-of-the-art just-in-time compilation =
strategies to boost performance for users. Additionally, as an =
interface, we'll adopt the Array API standard =
<https://data-apis.org/array-api/latest/> which was championed by major =
libraries like NumPy, PyTorch and CuPy.

More details about the plan are posted on GitHub =
<https://github.com/pydata/sparse/discussions/618> =E2=80=94 please join =
in the discussion there, to keep it all in one place.

Best Regards,

Hameer Abbasi=

--Apple-Mail=_65ACEEF1-B62B-4FE8-8209-921ACFD0150A
Content-Transfer-Encoding: quoted-printable
Content-Type: text/html;
	charset=utf-8

<html><head><meta http-equiv=3D"content-type" content=3D"text/html; =
charset=3Dutf-8"></head><body style=3D"overflow-wrap: break-word; =
-webkit-nbsp-mode: space; line-break: after-white-space;"><p =
dir=3D"auto">Hello everyone,</p><p dir=3D"auto">The stated goal =
for&nbsp;<code class=3D"notranslate">sparse</code>&nbsp;is to provide a =
NumPy-like API with a sparse representation of arrays. To this end, =
Quansight and I have been collaborating with researchers at&nbsp;<a =
href=3D"https://www.csail.mit.edu/" rel=3D"nofollow">MIT =
CSAIL</a>&nbsp;- in particular&nbsp;<a =
href=3D"https://www.csail.mit.edu/research/commit-group" =
rel=3D"nofollow">Prof. Amarasinge's group</a>&nbsp;and the&nbsp;<a =
href=3D"https://github.com/tensor-compiler/taco/">TACO</a>&nbsp;team - =
to develop a performant and production-ready package for N-dimensional =
sparse arrays. There were several attempts made to explore this over the =
last couple of years, including a&nbsp;<a =
href=3D"https://github.com/Quansight-Labs/taco/pulls?q=3Dis%3Apr+llvm">LLV=
M back-end</a>&nbsp;for&nbsp;<a =
href=3D"https://github.com/tensor-compiler/taco/">TACO</a>, and a =
pure-C++ template-metaprogramming approach called&nbsp;<a =
href=3D"https://github.com/hameerabbasi/xsparse">XSparse</a>.</p><p =
dir=3D"auto">To this end, we, at Quansight, are happy to announce that =
we have received funding from DARPA, together with our partners from =
MIT, under their&nbsp;<a =
href=3D"https://www.darpa.mil/work-with-us/for-small-businesses/HR0011SB20=
234-06" rel=3D"nofollow">Small Business Innovation Research (SBIR) =
program</a>&nbsp;to build out&nbsp;<code =
class=3D"notranslate">sparse</code>&nbsp;using state-of-the-art =
just-in-time compilation strategies to boost performance for users. =
Additionally, as an interface, we'll adopt the&nbsp;<a =
href=3D"https://data-apis.org/array-api/latest/" rel=3D"nofollow">Array =
API standard</a>&nbsp;which was championed by major libraries like =
NumPy, PyTorch and CuPy.</p><p dir=3D"auto">More details about the plan =
are&nbsp;<a =
href=3D"https://github.com/pydata/sparse/discussions/618">posted on =
GitHub</a>&nbsp;=E2=80=94 please join in the discussion there, to keep =
it all in one place.</p><p dir=3D"auto">Best Regards,</p><p =
dir=3D"auto">Hameer Abbasi</p></body></html>=

--Apple-Mail=_65ACEEF1-B62B-4FE8-8209-921ACFD0150A--

--===============0104302331157273780==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
SciPy-Dev mailing list -- [email protected]
To unsubscribe send an email to [email protected]
https://mail.python.org/mailman3/lists/scipy-dev.python.org/
Member address: [email protected]

--===============0104302331157273780==--