On The Heels Of An Impressive Launch

[Michael]

For whatever it is, I'm told "the ball is rolling" from the company... so something will be stated within a few hours.

[elanthis]

I want to bitch about articles like this on principle, but this may actually have been quite helpful. I was just about to lay down a ton of cash for an ICC license pack, and if this is EKOPath being released for free soon than Michael may have just saved us close to $10k.

Guess we'll have to wait a bit and see.

[V!NCENT]

It was (again according to the marketing dept) voted '2004 supercomputing product of the year'

Supercomputing product of the year? xD

Shit... definately not a great source hahahah

[XorEaxEax]

I didn't have any information on it so I took the marketing dept as truth

Heh, yeah, marketing isn't always based on absolute facts

It was (again according to the marketing dept) voted '2004 supercomputing product of the year', but that was a 'few' years back.

Likely the open sourcing of ekopath (let's hope we're not jumping the gun here) is partly because of competition and the hopes that open sourcing will add mindshare/publicity regarding their ekopath compiler suite, which is just fine with me.

[XorEaxEax]

And LLVM/CLang, while very neat and interesting, has yet to impress me with either code speed or compile speed.

Interesting, given that compile speed is perhaps LLVM/Clang's main claim to fame (in code speed it still loses out to GCC in my benchmarks and most public ones I've seen), so I gather Pathscale compilers are really fast ? Mainly I'm interested in the performance though, how would you compare ekopath vs icc in this regard?

[V!NCENT]

[citation needed]

And no Pathscale marketing material without further sources to back up any claims doesn't count.

You're absolutely correct. My information gathering works like this:

If (knowlegdeAlready)
{
    if (sourceValueInBrain < sourceValueNotInBrain)
    {
        rememberSource ();
        saveToBrain ();
        acknowledge ();
    }
    else
    {
        discusionWhoring ();
    }
}
else
{ 
    rememberSource ();
    saveToBrain ();
}

I didn't have any information on it so I took the marketing dept as truth

[yogi_berra]

Q posted what a Dirndl is. A simple google helps too.

Actually, Q posted a shot of a crossdresser that may have wearing a dirndl. This is a dirndl:

[chuckatkins]

Popular, yes. Number 1, unlikely.

given that this beast is a number one choice for supercomputers

I would hardly say it's the number one choice. It is a popular choice, yes, but number 1, doubtful. It's competing very heavily with both PGI and Intel, both of which have a very strong HPC presence.

Personally I like working with the PathScale compilers very much. In addition to pruducing good fast code, the compile times tend to be very fast as well. The Intel compiler also does tend to produce ~ 30% faster code than gcc for me (mostly numerical and imaging processing C++ code) however at the expense of greatly increased compile times. The speed of compile time is not so relevant to an end user as they don't need to do it very often but for a developer it can have a significant impact on productivity. The PGI compiler has also proven very effective over gcc but I don't have enough first hand experience with it to fairly judge it (I have several nightly builds running with it but I don't use it day-to-day).

As far as open source compilers go, however, I've never been able to build my code bases with Open64 without the compiler crashing. And LLVM/CLang, while very neat and interesting, has yet to impress me with either code speed or compile speed.

[AnonymousCoward]

Well... given that this beast is a number one choice for supercomputers,

[citation needed]

And no Pathscale marketing material without further sources to back up any claims doesn't count.

[V!NCENT]

Well... given that this beast is a number one choice for supercomputers, I think there will be a very high probability that this can compile Linux.

According to Wikipedia it is "Compatible with GNU/gcc tool chain and popular third-party debuggers".