I know that in order to unlock the supercomputer potential of ARM architecture, Cray had to develop appropriate software, including ARMv8 compilers and other development tools, as well as runtime libraries for mathematical, scientific, and communication applications. According to Cray, in two-thirds of 135 tests, its arm compiler provides at least 20% better performance than LLVM and GNU open compilers. Servers on ThunderX2 can be mixed with servers on Intel Xeon-SP, Intel Xeon Phi and Nvidia Tesla. And because it is very interesting that is something which has developed at Cray for ARM? How is the coordination of different units of software across platforms of different firms and manufacturers? What about GNU?
You are viewing a single comment's thread from:
In such a cutting edge field I would imagine each use case regarding 'the coordination of different units of software across platforms of different firms and manufacturers' is going to be different and require a unique approach until a protocol is developed in which all software companies / firms and manufacturers can agree upon. In regard to GNU please clarify your question for me and I'll give it a shot! Thanks for your thoughts.
About GNU, the question was rhetorical. If Cray supercomputers have decided to use their ARM compiler, why compare it to the GNU compiler?