Intel Grabs Codeplay as It Repositions for a Diverse Chip Future
Intel snatched up one more company in its software-maker buying spree, agreeing on Wednesday to buy Codeplay Software for an undisclosed sum.
Codeplay is known for development tools around the open source SYCL parallel-programming model, which includes tools, runtimes and execution models so standard C++ code can be adapted for concurrent execution across CPUs, GPUs and other processors.
SYCL + OneAPI
Intel intends to pair up Codeplay’s tools with its homegrown OneAPI parallel programming framework so developers can extract the most computing power from heterogeneous computing environments.
In the past, program execution relied on CPUs, but in the last 15 years new types of processors such GPUs, FPGAs and ASICs have excelled for applications such as AI and analytics. The CPU now offloads programs to run on these alternative processors, which is automated by parallel programming tools that include OneAPI, OpenCL and CUDA. These frameworks ensure coders don’t have to worry about the hardware.
Standard C++ by default doesn’t have the hooks for parallel execution of code, and hardware and software companies have stepped in with their libraries. Nvidia has its set of standard C++ libraries called libuc++, which enables parallel execution of code across CPUs, the company’s GPUs, and other chips. But Nvidia, Intel and other companies are working together on proposals to bring parallelism to standard C++ releases.
Intel’s OneAPI already has support for SYCL with a toolset called oneAPI Data Parallel C++ (DPC++), and Codeplay has another SYCL tool called ComputeCpp. (Intel recently announced SYCLomatic, a tool that allows coders to export code written in Nvidia CUDA over to SYCL.)
Codeplay has won contracts from the likes of the U.S. Department of Energy to implement the SYCL open standard for their Frontier supercomputer, which was ranked as the world’s fastest supercomputer on the Top500 list. The supercomputer, which is the world’s first exascale system, uses AMD’s Epyc chips. Codeplay supports a range of CPUs and GPUs from AMD, Intel and Nvidia.
The transaction will help Intel further its “oneAPI vision and reinforce our commitment to open ecosystems,” a company spokeswoman said in an email.
Intel is changing course from its 50-year CPU heritage with a stable of alternative chips for accelerated computing. Intel plans to ship the Ponte Vecchio GPU for supercomputing in the second half of this year, and spent $16.7 billion to buy Altera in 2016, and $2 billion for AI chip company Habana in 2019.
Intel has also shifted its chip design to cram CPUs and accelerators in a single chip package, which aligns with its long-term manufacturing strategy. Intel has opened up its factories to make RISC-V and ARM-based chips, so the Codeplay acquisition will help developers adapt code for parallel execution on chips that marry x86, ARM and RISC-V instruction set architectures in a single package.
Intel employs about 17,000 software developers and is looking to build software services over its chip offerings. The company’s CEO Pat Gelsinger has made it clear that the company is looking to make strategic software acquisitions.
Intel’s recent acquisitions include AI provider Granulate and open source software company Linutronix. At Intel’s Vision conference last month, the company showcased new software services including Project Endgame, a cloud-gaming service running on its GPUs, and Project Amber, which is an attestation layer for secure communications in the cloud.