Intel SDKs give developers tools for AI, quantum software
Intel launched new SDKs for AI and quantum computing and added tools underpinned by the Intel Developer Cloud in an effort to attract developers to build on its hardware.
Intel's focus on building up its developer community continues with a deeper push into open computing and the release of new tools to support developers in building artificial intelligence and quantum computing software.
The company reminded Intel Innovation conference attendees Wednesday that it has been a major contributor to the Linux kernel for more than 10 years and recently helped integrate the OneDNN performance library to TensorFlow, contributing performance improvements for the millions of developers who use the AI framework.
Though Intel has invested in open source software for years, the company hadn't promoted its contributions to open source software projects to the level it has under CEO Pat Gelsinger.
"Many developers are simply unaware of Intel's contributions to the industry, especially those who are working higher in the stack," said Al Gillen, an open source analyst at IDC.
The company dedicated day two of its conference to promoting newly available SDKs and upcoming software development tools underpinned by the Intel Developer Cloud, which Intel expanded with new features this week.
The Intel Developer Cloud is now a centralized platform where all other Intel developer technologies can be showcased and promoted to and tested by developers, Gillen said.
"Over time, Intel will be looking for ways to monetize its extensive software portfolio, and DevCloud at some point will become an entry point for paid subscription software use -- in addition to the free tier of test resources, which is unlikely to go away," he said.
Intel provided a Quantum SDK to provide developers with the tools to build on upcoming quantum computing systems.
Write-once, run-everywhere tools
One of Intel's strategies to attract more developers is to better support "open accelerated computing," and achieve the "write once, run everywhere" goal, said Greg Lavender, CTO of Intel's software and advanced technology group.
The write-once, run-everywhere method will bring developers out of proprietary approaches, like Nvidia's CUDA, allow them to write once with open tools like Sycl, and then compile to multiple different hardware accelerators, he said.
Intel acquired Codeplay in June. The company's OneAPI programming model supports Intel, AMD and Nvidia hardware, as well as Sycl on RISC-V chips. With OneAPI, developers can choose the best architecture for the specific problem they need to solve, without rewriting software for the next architecture and platform.
Developers despise writing code for multiple platforms; they want to write once and have that code run everywhere so they can focus on innovating.
Paul NashawatyAnalyst, ESG
Codeplay and OneAPI will appeal to developers who want to build on their terms, using the latest silicon technologies, said Paul Nashawaty, a senior analyst at Enterprise Strategy Group (ESG).
"Developers despise writing code for multiple platforms; they want to write once and have that code run everywhere so they can focus on innovating," Nashawaty said.
Intel also launched an open forum to support collaboration and direction for the OneAPI specification, to be led by Codeplay.
The 2023 versions of Intel's OneAPI Toolkits will be available in December with support for 4th Gen Xeon (Sapphire Rapids) and Intel's latest GPUs and FPGAs. The kits will also include the open source Syclomatic compatibility tool, which assists converting Nvidia's CUDA source code to Sycl source code, allowing developers to use various computing architectures.
AI and security
Intel also released three new AI reference kits for healthcare: document automation, disease prediction and medical imaging diagnostics. Developers can find them on GitHub, alongside the four kits released in July. The kits will be updated monthly, Lavender said.
For AI, machine learning and edge developers, Intel and Red Hat launched a joint AI Developer Program. The program allows developers to learn, test and deploy models using Red Hat OpenShift Data Science and Intel's integrated AI and edge portfolio.
In December 2019, Intel acquired Habana Labs for its AI training and inference chips. Red Hat plans to support the Habana Gaudi training accelerator on its service to support "cost-efficient, high-performance, deep-learning model training and deployment" as a managed cloud service, said Chris Wright, Red Hat's CTO and senior vice president of global engineering.
While the Red Hat partnership is valuable, there is a larger ecosystem around open source that Intel is missing out on, ESG's Nashawaty said.
Intel also provided an update on Project Amber, a SaaS offering for attestation in confidential computing slated for availability in the first half of 2023. Leidos, a U.S. federal government tech contractor, took to the conference stage to briefly share a proof of concept with Project Amber for future use in mobile clinics.
Confidential computing protects data in use by performing computation in a hardware-based trusted execution environment. Intel Software Guard Extensions available on the Xeon Scalable platform is one of the technologies behind confidential computing today.
"Project Amber is interesting as the push for confidential computing grows," Nashawaty said. "With regard to AI and where they are going with quantum, you have to be able to have privacy in place. It's a big focus in the dev community."
Intel Quantum SDKs
Enterprise use of quantum computing systems isn't expected for years, but Lavender encouraged developers to begin to explore quantum software development.
The new Intel Quantum SDK provides tools to help developers learn to program quantum algorithms. The beta version is available through the Intel Developer Cloud.
Lavender also detailed progress toward post-quantum cryptography, part of Intel's efforts to thwart threats posed by quantum computers. He said the industry must be "Y2Q-ready" -- quantum-resistant by 2030.
