"Award Winning" Intel Parallel Studio XE

Intel Parallel Studio XE, in the category of "Best HPC software product or technology," was honored in the annual HPCwire Readers Choice Awards. The awards are an annual feature of the publication and constitute prestigious recognition from the high performance computing community. The awards were announced and presented during the 2011 Supercomputing Conference, held in Seattle, WA. As HPCwire proclaimed, "The annual awards are highly coveted as prestigious recognition of achievement by the HPC community."

Intel Parallel Studio XE 2011 combines enhanced optimizing compilers, libraries, error checkers and performance analyzers in a single integrated suite that enables developers to write faster, more reliable and secure code on Windows and Linux.

I'm very pleased to see this additional recognition for our Intel Parallel Studio XE, which is used by a very large number of developers around the world. We just recently offered an extended version of Intel Parallel Studio XE, for developers on Cluster computers (characterized in part by their use of MPI), called Intel Cluster Studio XE. That suite contains all the award winning features of Intel Parallel Studio XE plus addition MPI support for performance, error checking and tuning in a distributed environment.

For more complete information about compiler optimizations, see our Optimization Notice.

Comments


lol

Yes, many things could do a Trillion (Billion to the British) Floating-Point operations if you gave them enough time! A little over 33 weeks was about right for an Intel 8087 from 1980.

I'll have to work hard myself. I dislike acronyms that end in "S" but are not plural... so I prefer FLOP/s to FLOPS.

Yes, we demo-ed Knights Corner registering more than 1 TeraFLOP/s sustained on a DGEMM over a wide variety of block sizes. It was exciting. I need to be sure we say "TeraFLOP/s" and not "TeraFLOP."


That's right, "FLOPS (FLoating-point OPerationS)" definition was the part I had an issue with. I guess, I was just too annoyed by recent avalanche of headlines claiming "Intel unveils 1 teraflop chip". Intel 8087 was better than 1 teraflop and it was quarter of a century ago..


I had to go have a look at http://software.intel.com/en-us/articles/intel-parallel-universe-magazine/ and see what you were talking about!

I had used "FLOP" for Floating Point Operations, and "FLOPS" for FLOP/second. In those cases, I made two choices: (1) treat it like an aconym and use all CAPS (I only recently started using Fortran instead of FORTRAN despite the Fortran-90 committee's declaration) and (2) use "S" for plural which is common, avoiding the apostrophe s (as in FLOP's) that some use to my annoyance.

I mention FORTRAN, because that is a similar acronym: FORmula TRANslation, using multiple letters from multiple words. FORTRAN is now often spelled Fortran (in fact the standards now specify that spelling) and FLOP/S are often spelled FLOPS or Flops.

As an aside, I like to avoid "flop" and "Flops" because I don't like declaring anything I work on a flop. Call it a character flaw or arrogance if you must! :-)

However, on a careful read I see that we added a definition of the term (which I apparently did not define myself):
FLOPS (FLoating-point OPerationS)

Wrong! It should be "FLoating-point OPerations per Second."

Good catch.

If that is your main concern. I agree.

Thanks for the note.


Hi James,

This is a comment on unrelated subject, but I couldn't find a better channel to convey my message. I just got to read The Parallel Universe issue 7 and noticed your unorthodox usage of FLOPS acronym. A sanity check with Wikipedia confirms that FLOPS stands for "floating point operations per second", which is the same as FLOP/s. This would normally fall below my nitpicking threshold, but FLOPS is one of most frequently used HPC acronyms. I'm not sure what would be a grammatically pure version of FLOP plural, maybe FLOPs or flops would do.

Best regards,
Paul