PathScale Open-Sources The EKOPath 4 Compiler Suite

[HokTar]

I really don't want to be the whiny little boy here, but the others seem to get replies so I'd like to point out that my previous post likely went under your radar:

@codestr0m:

What about Fortran? Do you happen to have any recent benchmarks against gfortran and ifort? I remember seeing a few from about 2006.

@Michael:

Does the HIMENO benchmark use the C or Fortran source?

Thanks for your answers and keep up the good work! (Both of you, obviously. )

[Prescience500]

I've heard a couple times that the compiler isn't likely to speed up mesa. That's surprising considering how heavily CPU bound the drivers are. I can't wait to see benchmarks of it, though. I hope it yields the same kind of performance gains that this article's benchmarks showed.

[bridgman]

I've heard a couple times that the compiler isn't likely to speed up mesa. That's surprising considering how heavily CPU bound the drivers are. I can't wait to see benchmarks of it, though. I hope it yields the same kind of performance gains that this article's benchmarks showed.

I guess the implication was that the kind of work mesa is doing when CPU bound (pushing data around, dancing across data structures etc..) is not necessarily the type of work that the new compiler optimizes (heavy floating point calculations etc..). Just a guess though...

[XorEaxEax]

One of the most powerful optimizations added to gcc in the past years is PGO (profile guided optimization), anyone know if it is it supported on ekopath4?

[ninez]

At the time of reading this, the PathScale press release should be available on their web-site and/or PRNewsWire.

maybe an hour or two later

No and/or no.

actually, if you visit their website you can grab a nightly build. So, the compiler suite is technically available already. and there IS a press release on the PathScale website too.

the link to the nightly build of the compiler is here;

http://c591116.r16.cf2.rackcdn.com/ekopath/nightly/Linux/ekopath-2011-06-12-installer.run

on the official website the link is under "NIGHTLY DOWNLOADS'" - which is in bold black lettering.

i've already compiled FFTW and a few other libraries, to test it out. so far, so good

cheerz

[ahlaht]

Python 3.2

I have been trying to speed up Python programs with this, but so far they only run slower.

Tried with CFLAGS "-O2", "-O3", "-march=core -O3" and some others. I'm not sure what to use for -march since "native" is not accepted. I found out by accident that "core" works.

-Ofast didn't compile.

I guess this doesn't help Python. Or I could be doing something wrong...

[deanjo]

michael

are there any benchmarks on things that might be useful on a desktop?? ie mp3 encoding video encoding etc

i don't remember seeing any.

You probably won't see much, if any improvements on libraries like those given that they have a lot of hand written assembler for the areas where optimization is important .

[bug1]

Encumbered by the GPL ?

"PathDB has been released under a permissive license, in response to requests from the open source community for a modern debugger that is not encumbered by the GPLv3."

en?cum?ber
   /ɛnˈkʌmbər/ Show Spelled[en-kuhm-ber] Show IPA
?verb (used with object)
1. to impede or hinder; hamper; retard: Red tape encumbers all our attempts at action.
2. to block up or fill with what is obstructive or superfluous: a mind encumbered with trivial and useless information.
3. to burden or weigh down: She was encumbered with a suitcase and several packages.

Why does PathScale go out of their way to insult the GPLv3 in this PathDB release note, while in the same document describign the release of the compiler under that very license ?


Thanks for the contribution.

[deanjo]

"PathDB has been released under a permissive license, in response to requests from the open source community for a modern debugger that is not encumbered by the GPLv3."

en?cum?ber
   /ɛnˈkʌmbər/ Show Spelled[en-kuhm-ber] Show IPA
?verb (used with object)
1. to impede or hinder

That is what the GPL 3 does however. It impedes and hinders by giving conditions of use of the source code. That is what any restrictive license does.

[bug1]

Im sure we have all seen an _ideological_ debate about restrictive vs permission open source license, whether restrictions are along term benefit etc, so lets not go there.

I wonder how in a _practical_ sense how the GPLv3 would have encumbered their debugger ?


The release demonstrates a major PR failure by PathScales. Giving away stuff should be good PR, but criticizing the license under which your giving it away in the same release note is just ridiculously stupid.

Not a PR guy, but i would have at least made them separate announcements.