SoylentNews is people
SoylentNews
Write AI code once, run anywhere—it's not Java, it's Intel's oneAPI
Saturday afternoon (Nov. 16) at Supercomputing 2019, Intel launched a new programming model called oneAPI. Intel describes the necessity of tightly coupling middleware and frameworks directly to specific hardware as one of the largest pain points of AI/Machine Learning development. The oneAPI model is intended to abstract that tight coupling away, allowing developers to focus on their actual project and re-use the same code when the underlying hardware changes.
This sort of "write once, run anywhere" mantra is reminiscent of Sun's early pitches for the Java language. However, Bill Savage, general manager of compute performance for Intel, told Ars that's not an accurate characterization. Although each approach addresses the same basic problem—tight coupling to machine hardware making developers' lives more difficult and getting in the way of code re-use—the approaches are very different.
[...] When we questioned Savage about oneAPI's design and performance expectations, he distanced it firmly from Java, pointing out that there is no bytecode involved. Instead, oneAPI is a set of libraries that tie hardware-agnostic API calls directly to heavily optimized, low-level code that drives the actual hardware available in the local environment. So instead of "Java for Artificial Intelligence," the high-level takeaway is more along the lines of "OpenGL/DirectX for Artificial Intelligence."
For even higher-performance coding inside tight loops, oneAPI also introduces a new language variant called "Data Parallel C++" allowing even very low-level optimized code to target multiple architectures. Data Parallel C++ leverages and extends SYCL, a "single source" abstraction layer for OpenCL programming.
In its current version, a oneAPI developer still needs to target the basic hardware type he or she is coding for—for example, CPUs, GPUs, or FPGAs. Beyond that basic targeting, oneAPI keeps the code optimized for any supported hardware variant. This would, for example, allow users of a oneAPI-developed project to run the same code on either Nvidia's Tesla v100 or Intel's own newly released Ponte Vecchio GPU.
Related: Intel Xe High Performance Computing GPUs will use Chiplets
(Score: 1, Interesting) by Anonymous Coward on Thursday November 21 2019, @12:15PM (2 children)
One of my colleagues is using one of Intel's AI frameworks at the moment. It appears to be a serious cluster-fsck with improper thread locking, outdated documentation and poor performance plastered over by shiny marketing claims*. Whenever he works around one hurdle something else pops up to bite him in the ass. I don't think Intel is the right company to write software in this area.
* Not unlike their CPUs when you think of it
(Score: 4, Touché) by takyon on Thursday November 21 2019, @12:28PM (1 child)
Maybe oneAPI is what they need to fix the situation.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 2, Funny) by Anonymous Coward on Thursday November 21 2019, @03:19PM
oneAPI to bring them all and in the darkness bind them.
(Score: 0) by Anonymous Coward on Thursday November 21 2019, @02:59PM (1 child)
I think that is going to be kind of difficult now that Intel has a reputation for being backdoor central. Why would you cluster GPU's with software from a vendor who has contempt for security?
(Score: 3, Interesting) by takyon on Thursday November 21 2019, @03:15PM
The Aurora supercomputer doesn't need to mitigate the latest hardware attacks. Amazon, on the other hand...
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]
(Score: 3, Informative) by Mojibake Tengu on Thursday November 21 2019, @03:41PM
And what kind of high-level software do you expect from Intel freaks who can gladly sell you an 83W TDP workstation CPU bundled with a barely 65W cooler?
https://ark.intel.com/content/www/us/en/ark/products/191042/intel-xeon-e-2274g-processor-8m-cache-4-00-ghz.html [intel.com]
https://diit.cz/clanek/intel-k-83w-xeonu-balil-65w-chladic [diit.cz]
Respect Authorities. Know your social status. Woke responsibly.
(Score: 0) by Anonymous Coward on Thursday November 21 2019, @04:01PM (2 children)
OneAPI to rule them all.
(Score: 0) by Anonymous Coward on Thursday November 21 2019, @07:06PM (1 child)
If you're coming late to the party, at least look at all the comments first [soylentnews.org] before making that joke.
(Score: 0) by Anonymous Coward on Saturday November 23 2019, @09:25AM
Is that like... reading the article?
(Score: 1, Insightful) by Anonymous Coward on Thursday November 21 2019, @04:34PM (1 child)
"necessity of tightly coupling middleware and frameworks"
when will the rube goldberg contraption of "modern" software fall over?
(Score: 2) by MostCynical on Thursday November 21 2019, @06:49PM
"Middleware" is what happens when you leave all the important stuff out of your product
Modern systems won't fall over while the developers are able to keep adding to the "middleware"
The salesman will tell you how the core system is so fast and lean...
"I guess once you start doubting, there's no end to it." -Batou, Ghost in the Shell: Stand Alone Complex
(Score: 0) by Anonymous Coward on Friday November 22 2019, @12:24AM (1 child)
Still waiting on my Canyon Lake...
(Score: 2) by takyon on Friday November 22 2019, @01:42AM
It, heheh, already launched [wikipedia.org]. I heard it was a fantastic dual-core.
[SIG] 10/28/2017: Soylent Upgrade v14 [soylentnews.org]