Why don't they replace Finder with something that works at least on par with Explorer from Windows 95? Adding "move/cut" feature in 2009 will be a huge step forward. Or make iPhoto actually usable, so it will see and recognize JPEGs on your own file system without "importing"? Or how about making Aperture just slow, as opposed to pathetically slow as it is right now? How about fixing Spaces to exclude icons of other desktops from Alt-Tab list? How about making the whole thing to feel like it's been made for non-retards?
Catch up to freaking Gnome first, dammit. Then talk about parallel computing...
I'd like to see more real details before they claim that they've figured it out. I love Apple, I use a Mac exclusively and I'd love to see my computer rock, but I'd holding my breath until I see some actual details.
Take a look at the link I posted below - OpenCL can be accomplished by using a thread scheduler that's "aware" of asymmetric processing capabilities across the different cores (and treating the GPU as a usable processing core in non-OpenGL code). I suspect they'll be using ULE or something similar to it to pull this off.
Must be in the same sense that Altivec was a "breakthrough" in SIMD architecture, or the PPC 601 was a "breakthrough" in CPU design. Apple's history is littered with breakthroughs like this. I'll believe it when I see it.
As things stand right now, typical SMP scalability benchmarks put OS X solidly behind even Vista, and trailing Linux and FreeBSD by a mile (I wish I could find a cite for those -- they were doing the rounds a few months back, I think).
Everything is relative. "Breakthrough" scalability and parallel computing on OS X, but more importantly compared to previous versions of OS X.
The fact that OS X is currently in last place (I read the same article you're referring to - I think it was on OS News but I'm not sure) is the very reason why they can improve so drastically. Basically, what Apple meant to say is, they're going to be backporting all the FreeBSD scheduler patches to OS X - perhaps even using the ULE Scheduler which would be 1) easy enough to switch to without too much development on Apple's behalf and 2) totally capable of delivering the promised improvements.
This is mostly a fallacy. The FreeBSD code is limited almost exclusively to userspace stuff. The Darwin kernel descends much more directly from NeXT's mach-based product. It grew SMP support only very recently, and via a very different path from that taken in the monolithic BSD community.
Honestly, I don't understand why Apple continues to work on their own kernel. It has near zero competitive advantage relative to FreeBSD or Linux, and much cost a fortune to maintain. The real value of their product has always been above the kernel anyway: I have to believe that switching the kernel out for a more robust Unix variant can only be a good thing for the Mac.
Switching to the FreeBSD kernel seems like it would have been a good idea back in the Rhapsody days, but Tevanian probably wouldn't allow it. By now Apple has added so many OS X-specific features like IOKit and DRM that it would be a lot of work to switch. Costs a fortune to maintain, costs a fortune to switch.
Catch up to freaking Gnome first, dammit. Then talk about parallel computing...
Come on, macwhores, let the downmodding begin.