Journal Journal: Intel Adds Vertex Shader Hardware Support
The new drivers are Windows-only, but the existence of this functionality provides some hope that their Free Software drivers will also gain vertex shader support at some point.
The new drivers are Windows-only, but the existence of this functionality provides some hope that their Free Software drivers will also gain vertex shader support at some point.
While the Open Graphics Project is aiming to build a fairly simple fixed-function device, I wonder if this is the best approach. It seems that it would be simpler to start with something like the OpenRISC core and:
You'd get a chip that is heavily optimised for doing graphical operations, but still a general purpose design, a lot like modern GPUs. The other benefit is that there are a lot more open source developers capable of writing software OpenGL implementations for this kind of device than there are able to make meaningful contributions to GPU design.
For bonus points, you could add a few stock OpenRISC cores, a memory controller, USB controller, etc to the design, and have a completely open source system-on-a-chip design.
Apple launches a DRM-free version of iTunes, and the Slashdot coverage is an article bitching that it adds your name, email address, and date of purchase in the metadata.
The rest of the world seems to be having a slightly more positive reaction, judging by how badly lamped the iTunes store was when I tried checking it last night (song previews were failing to load, individual pages took thirty seconds or more, and the two albums I upgraded last night are still downloading).
Album upgrades cost 25% of the original price, which isn't too bad for doubling the quality and ditching the DRM. I'd bough 12 albums from iTMS (including one EP) before the DRM started irritating me by being incompatible with devices I wanted to use to play back music. Of those, only two were available DRM-free.
I plan on buying a few albums DRM-free in the next week or so. I hope other people will too, so that Apple can point at the proportion of DRM-free downloads and say to the other labels 'look how much money you are losing by not selling your tracks DRM-free.' Then all they need to do is get the price down to EmuSic levels (or, at least, the same price they have in the US, for those of us in the UK).
I had another chat with my hosting company's CEO today, and he agreed to give me one of their standard plans for slightly less than it usually costs with twice as much bandwidth and a second IP. While this is still a bit more than some other companies, knowing that if stuff goes wrong I have email and IM addresses for the CEO, CTO and head tech and can bitch at them until someone fixes it probably makes it worth while. They extended my existing contract by a month to give me time to shop around for a new host, and then agreed to the suggestion I felt fair after I'd looked at some of their competitors.
Mac Mini Colo are now back on my 'would recommend' list. Anyone looking for hosting that falls somewhere between low end dedicated servers and high-end virtual private servers would do well to check them out.
In a year, I will probably move to a Xen-based solution, but at the moment they tend not to have enough RAM or disk space. Anyone looking at starting a hosting business should take note of this. There is a serious shortage of VPS solutions with decent amounts of disk space and RAM. Get some 750GB disks (maybe RAID-1, two per machine), and 5-6GB of RAM, and you can put 10 VMs with decent specs in each node. CPU is rarely the limiting factor for budget hosting, and live migration can help out a lot here if you choose to go for a SAN backend and just have a lot of disks somewhere in your rack.
I am flexible in terms of storage space; I currently have an 80GB disk, but I'm using less than 5GB of it. I'd like to start using a bit more, but not having more bandwidth makes this difficult.
Can anyone recommend a hosting company that would offer this kind of set-up for under about $50/month? DedicatedNow seemed like they had a reasonable offer, but they had some quite unreasonable demands, such as needing a copy of my passport faxed to them to complete the transaction (and I don't do that for any business, particularly not one I have had no prior dealings with).
In unrelated news, it looks like I'm getting a Google-sponsored minion for the summer. The SoC results were announced this morning, and the project I am mentoring got accepted. Should be fun.
The NetBSD and OpenBSD articles got between 15-30K unique readers in the first seven days. I don't have figures for after that. I've been contacted by a few people who said they started using NetBSD as a result of the article. The moral of this story? If you want free publicity, try being polite.
Today, I got a reply, saying:
Thank you for your communication of the
1/2/2007, By E-Mail
which will receive attention
Now, normally, I would expect a reply of this nature to be sent by email, since it's basically an auto-responder. In this case, however, it was sent via a different mechanism. Second class mail.
Yes, my MP employs someone to act as an email auto-responder, replying by postcard to all emails. Isn't technology great?
Ping Wales have the full story. From the article:
A straw poll of businesses in Wales which use Macs reveals that no business customers have been informed of the change to the repair service.
Apparently customers were not the only ones kept in the dark over this:
Apple resellers in the region told Ping Wales that they hadn't been informed of the change either, adding that this did explain why the number of repairs coming in had increased significantly over the last few weeks.
Can anyone recommend a good supplier for Opteron machines at a reasonable price? I need a minimum of two CPUs (or cores), a reasonably sized (but not necessarily fast) hard disk, a couple of GB of RAM. It's going to run headless, so no graphics card is required (the Sun workstation line looks nice, but I object to spending money on an nVidia Quadro card that is never going to be used).
On the more geeky side, there are a couple of large SGI SMP boxes (one 60 CPU, one 30) for me to play with and a Linux cluster.
Today we warned the technicians that we might need to use a big chunk of disk space. We were told that there might not be enough, since they only had about 200GB free on the RAID array we wanted to use (the one local to the 60 CPU box). Back home, our technicians complained when I dumped a 2GB DVD image in my home directory...
OS X sucks at system calls. Due to the Mach+BSD kernel design, system calls are very expensive on OS X. On Mach, the system call cost is around 10x the cost on BSD - on OS X it's some combination of the two depending on what you do. The next thing to realise is that all of the POSIX thread synchronisation mechanisms are implemented using Mach-level synchronisation primitives. This means every time you lock a mutex (for example) you need to dive down through the BSD layer to the Mach layer. This gives you the horrible overhead of checking the permissions on a Mach port (something that sane Microkernels like L4 abandoned), and is expensive. This makes locking operations on OS X much more expensive than on BSD-like kernels. This, in turn, can make threaded code much slower. If you are Adobe, and you are rendering an entire image transformation in a small number of threads, then you will only lock at the beginning and the end of the operations, so it will be nice and fast. If you are doing low-level parallel operations for scientific computing (when not bitching about the OS X kernel), then you should really try harder to persuade someone to buy you a nice Solaris box.
This is why I like Objective-C, and why using languages like C++ feels so painful in comparison. Oh, and for reference, enumerating the classes was around half a dozen lines. Sure, this kind of thing is possible in C++, but only if you build a huge amount of extra stuff on top.
This is all well and good, but FPGAs don't offer the same feature density as even ASICs (although they are re-writable), so this raises the question of whether it would be commercially viable to do a run of ASICs based on an open core, with a motherboard also based on an open design. Would you buy a motherboard / CPU that could run, for example, NetBSD and was entirely open source? How much of a premium would you be willing to pay for such a thing?
I really like the design of the Alpha CPU, and I would be interested in a machine that had a similar core design, and maybe willing to pay around a 50% premium over x86 for the elegance of the system - particularly if I could also download the core design to an FPGA, modify it, and submit changes back for inclusion in the next revision.
One of the real advantages of open source software is that it can easily be compiled for multiple architectures, so once you've written a GCC back-end and a boot loader for your system you suddenly have a huge amount of usable software.
I am rapidly coming to the conclusion that the ideal CPU would have a very simple instruction set, and not even do out-of-order execution. Code running on it should be compiled first into some kind of byte-code, and any re-scheduling should occur in the bytecode JIT - move as much complexity into software as possible, since it's much easier to upgrade (and to configure at run-time).
The opposite of a correct statement is a false statement. But the opposite of a profound truth may well be another profound truth. -- Niels Bohr