@pytorch 2.4 upstream now includes a prototype feature supporting Intel GPUs through source build using #SYCL and #oneDNN as well as a backend integrated to inductor on top of Triton - enabling a path for millions and millions of GPUs through #oneAPI for #AI.

Lots of important milestones to make this happen - including support for #UXL Foundation open AI technologies. Just a prototype, but a big step forward... thanks to all in the PyTorch community. Feedback welcome!

https://pytorch.org/blog/pytorch2-4/

#GenAI enthusiasts... the folks from Prediction Guard and the Intel Liftoff program have a fun series of 5 progressively more challenging problems. The Intel Developer Cloud is offering cycles, GPUs and more for contenstants...and I hear about 2k have signed up to test their skills. The first couple are easy.

Check it out on this vid... and have fun!

https://www.youtube.com/watch?v=9VC6EZ7b6sk

Hey... the HPC Wire Readers just chose our #oneAPI tools (since contributed to the #UXL Foundation as Best Programming Tool or Technology! We're seeing real momentum in Khronos #sycl as a neutral, open, multi-architecture language for accelerators at #SC23 and beyond

https://www.hpcwire.com/2023-readers-editors-choice-programming-tool-technology/

I'd been told that #oneAPI seemed complicated by a friend in our developer community. I countered...well not really... and

So that friend told me I should simply record that pitch. It made it clear why this is important and it was short and sweet...

It took me a couple months to get time to record it, but... here you go!

https://www.youtube.com/watch?v=qvxmeupPftU

One of Amazon's Top #C++ #programming texts -- now updated and in 2nd edition!
This book teaches us how to harness accelerated computing using C++ with SYCL and provides advice for balancing application performance, portability across compute devices, and productivity.

https://link.springer.com/chapter/10.1007/978-1-4842-9691-2_1

This is a nice article on Argonne Leadership Computing Facility's work on the Monte Carlo application OpenMC. And there are some nice complements for the compiler engineers at Intel Software for responsiveness. But underlying the story is the importance of supporting standards like #openmp and #llvm - both available in the #oneAPI compilers - to achieve performance portability.

Open standards for accelerated computing work. Accept nothing less.
https://lnkd.in/gK2Y-kTg

Earlier this week Intel Software announced the launch of Intel Developer Cloud where devs can get access to the lastest hardware, security features, developer software and more. Guy Tamir - who was a great series of <3 minute intros for Dev Tools put this intro together for folks who want to try out the developer cloud

https://youtu.be/QFR3SvcTikY?si=xaMtfAt3kE25eb5G

At the airport getting ready to return after #intelinnovation. A couple big announcements on Open accelerated computing and the new UXL Foundation, Intel Trust Authority and Secure Attestation, progress in AI both in the datacenter and client, and an amazing 288 cores in upcoming Sierra Forest.

The software and advanced tech keynote is here: https://www.youtube.com/watch?v=7SoOwVP0VLw

Lots of coverage today on the Linux Foundation launch of UXL.

Proprietary development serves the proves the possibility of new technologies. Over time, openness provides scale and market expansion. Along with cool AI domain interfaces like XLA and Triton, general accelerated programming also has made a huge step forward today with UXL

https://www.nextplatform.com/2023/09/19/the-new-uxl-foundations-has-bold-blueprint-for-open-acceleration/