AMD Announces Server and Data Center Optimized Developer Software For APUs

AMD Announces Server and Data Center Optimized Developer Software For APUs

AMD is banking big on a heterogeneous server future, one where CPUs and GPUs work in unison to offer unparalleled performance. And speaking of parallels, that's just where AMD thinks its APUs can shine. It should come as a surprise to no one at this point that if a process can take advantage of a highly parallel processor, then running it on the GPU is a no-brainer; otherwise, the process can be kept on the CPU. AMD calls this magic HSA, or heterogeneous system architecture.

To aide developers for the future of HSA, AMD has been working with a couple of developers to release tools that are going to make the transition to CPU+GPU computing much easier. Project Sumatra, for example, is a joint project between AMD and Oracle which will allow Java developers to take advantage of GPU compute, while GCC/HSA is the fruits of AMD working with SUSE to enable OpenMP APIs to work with the popular Linux compiler.

Other updated software includes PGI Accelerator Compiler (OpenACC directives for C, C++ and Fortran under Linux and Windows); clMath (AMD OpenCL math libraries that take advantage of APUs), AccelerEyes' ArrayFire for OpenCL (another "fast" math library), and CodeXL 1.3, AMD's updated developer suite for Windows and Linux, which importantly brings remote debugging to the table.

What does the industry think about all of these updates? Paul Teich, CTO and Senior Analyst of Moor Insights & Strategy believes it could be big: "Combining AMD’s upcoming HSA compliant APUs with new east-west fabric architectures, like AMD’s Freedom Fabric and HP’s Moonshot, is a potential game changer for many server-side HPC and big data analytics workloads."

One thing's for sure: GPUs are fantastic for certain workloads - especially big number-crunchers - so it does seem like AMD is setting itself up right to make a big impact on the server market in the near-future.

Login or Register to Comment
Post a Comment
Username:   Password: