As said (and linked to: http://www.phoronix.com/scan.php?pag...item&px=OTU0Ng) and further discussed in the forums, it's a temporary codename until the respective company makes their public announcement... If I say what it is much in advance, the company probably won't be too thrilled. At that point, there's a 4-page Phoronix article in the queue that explains what is 'dirndl.'
Originally Posted by Aleve Sicofante
Sadly phoronix has become (or always was?) a tabloid. Who can take this site seriously with stunts like this? Boasting scoops and annoying people when the article hardly lives up to the headline. Always shamelessly advertising itself via links in articles back to phoronix articles (have you ever read an article here that contained fewer than 5 links back to other phoronix articles? I think not..). Always going on about the "test suite", etc etc.
It's a shame this site is the only (sort of, there are others like osnews, ars technica, slashdot, etc. But there isn't really another place where almost everything is gathered in one place) site that gathers most linux news in one place. Until such a site appears (hint, hint) I shall be forced to visit this place every once in a week to see if i missed something off the other sites (if I can be bothered to scroll through the endless "does xxx boot faster than yyy" articles and similar garbage)... On the other hand if this tabloid/crap "benchmarking" can stop here then I would be happy to return. But I won't hold my breath.
I'm hoping it's a CPU optimization; otherwise announcing results for Intel SB and AMD Opterons would be a deliberate attempt to fool us all.
Maybe some geniuses developed a breakthrough C compiler which can automatically parallelize non-optimal code over 4 or 8 cores (like loop unrolling, but over multiple cores). Kudos if it uses SSE2 and/or finally forgets about 32-bit compatibility.
I'd be happy to see that in the GCC, but it's more realistic to expect such an improvement on a competitor.
Whoa, let's not get carried away. Would I prefer it if Phoronix went for less 'sensationalistic' headlines? Sure. Would I prefer it if Phoronix weren't so prone to 'milk' their scoops for all it's worth (and probably then some)? Sure. But overall it's a great site, free (I ain't paying for it anyways, although I click on the adverts from time to time) and with the best overall technical coverage on Linux and it's surrounding technologies, not to mention the high level of technical knowledge amongst many of the forum members and even guys like xorg/ati devs here providing very interesting information on the development/infrastructure of Linux graphic drivers among other things. In short, Phoronix is fantastic in what it offers and it's FREE!
Originally Posted by netrage
As mentioned there are things I would prefer to be different in regards to the way certain articles are presented, but I can't understand how some people can be so negative when those 'gripes' pale compared to all the great stuff Phoronix offers, and again for FREE. Compared to 99% of the so called 'technology sites' covering Linux out there Phoronix is PURE WIN in my opinion, despite it's warts.
If it's indeed the PathScale EKOPath compiler going open source, like some have alluded to, then your description of what it can do is probably not far from the truth. This kind of technology isn't entirely new; Intel's own C/C++ compilers have been claiming most of these features for a couple years now. What would be new is if we had a compiler like this that was free and open source.
Originally Posted by jijitus
And hopefully its gcc compatibility is better than icc's, so all our favorite software that compiles with gcc will compile with EKOPath.
Of course, all of this is speculation, but the story seems likely:
*PathScale and Phoronix have a working relationship; PathScale representatives post here and Michael writes articles about pscnv
*Amazing performance coming out of PathScale wouldn't exactly surprise me, it's to be expected (their whole company mission is to make compilers that produce high performance)
*PathScale already writes open source software (they like the BSD license) in the way of pscnv and that C++ library for BSD
*Someone found out that EKOPath used to cost $1795 USD for a single license back when it was a commercial product, and Michael separately said on his twitter feed that the big news is related to a piece of software that costed $1795
*I haven't seen any posts on the forums recently from the PathScale guys; they are probably on a hush order or they are too busy getting their software ready for the big unveil
The evidence points to it pretty strongly, but who knows, maybe the "news" is just that Michael learned about the -O3 and -flto options in gcc. (although that wouldn't agree with the very significant performance increases in the teaser benchmarks, since these two options tend not to affect the performance of your typical program in such a dramatic way!)
If "real-world" performance gains for CPU-intensive tasks like encoders and decoders, physics engines and software renderers comes anywhere near the benchmark performance gains with EKOPath, this will be a huge benefit to the open source ecosystem -- provided their gcc support is compatible enough to compile things like libjpeg, libvorbis, libtheora, bullet, mesa, etc.
GCC 3.2 for AMD. And it seems that the current code state of the driver prevents it from being moved up to compile with GCC 4.x.
Originally Posted by ahlaht
Tags for this Thread