Intel has released the first service pack for the latest version of Parallel Studio, bringing performance improvements and refinements to the parallel models, among other things. The new version of the Intel Compiler creates code that is 11% faster compared to the previous Windows version (there’s a 6% improvement on Linux), and Intel says the compiler now generates (C++ Integer) code that is up to 47% faster than its nearest rival compiler on Windows.
For Fortran programmers, there’s more in store too: The new version of the Intel Compiler creates code that is 22% faster on Windows and 27% faster on Linux, Intel says (based on the Polyhedron benchmark). There is also support for coarrays in Fortran for distributed memory systems. I wrote about Fortan coarrays last summer, if you want a bit more background on them.
There is also support for C++ Variadic templates and Lambda support, plus support for OpenMP 3.1.
For parallel programmers, there are also improved versions of Intel Threading Building Blocks template library (version 4.0) and an updated Intel Cilk Plus compiler extension. The press pack for the release has a quote that says “moving code to a GPU is a project but moving to a Many Integrated Core Architecture takes a few minutes”. The tools Intel has released aim to deliver parallel models that are highly scalable to make this transition as easy as possible, and to help programmers to future-proof their code.
For more information, see what’s new in Intel Parallel Studio XE 2011, or see James Reinders’ presentation announcing the new service pack (PDF).