Programming Matters

GPU’s vs. quad processors

With dual processors everywhere and quad-core coming down the pipe - what will the future of custom GPU’s be on graphics cards? Something in my purest mentality has always disliked GPU languages from the get go. Pixel shaders may be one thing, but vertex shaders especially. It violates the concept that venerable old c delivered back in the day. Code portability and reuse. Now usually I don’t go insanely old-school to justify my points but here it makes perfect sense. I have already encountered plenty of situations where visualization was not enough, the CPU had to have access to the data in order to do analysis on it. The fact that operations could occur at one end of the pipe in a very proprietary way is more than a time waster from the standpoint of coding. It also isolates the results from being used by the CPU side code. In the simple case of shadow volumes with extruded verts jumping through shader hoops and trapping the data down the pipe seem to not be worth it IMHO. Back in the day (before 3D cards…before tight compilers) I would often write machine code to get the job done faster. But at the very least in this case the results where in fact something that other code could get access to.

I don’t mind custom processors for specialized purposes. Often the cost of these processors goes down nicely over time. But custom processors you are expected to program in a very atypical way. Time and quad processors will surely take care of that.

AdSense and curiosity

Never being a big believer in something for nothing Google and AdSense of course makes me skeptical. I do feel that inevitably this is the right way to go. People who create content shouldn’t have to worry about advertising it too. As a lark I’ve started playing around with AdSense just to get you the data. As part of the terms of use I am not allowed to report profitability information here, however I am simply dying to know. I do personally know people who have made thousands a month off of this technology. This of course is a skill-set in itself. As the web grows and matures I am sure that people who have popular sites and want to support there efforts with AdSense will be able to do so. But for now it does seem that engineering these efforts beats out honest attempts. But once again, the best way to know is to take a look. The question isn’t whether this is truly ready for prime time - the question is how far off it is. Some day I have no doubt that honest content authors will make comfortable livings paid for by AdSense alone. It’s exciting to watch this gather steam as inevitably it must.

Virtual machines

Lately I have spent a lot of time playing with virtual machines like VM-Ware and Microsoft provide. I’ve been curious about this for years. I remember a game I worked on in 2000 where we considered using VM-Ware to host Linux and Win2000 on the same box. We didn’t end up going with the solution but the debates were interesting. It certainly seems like Virtual Machines are progressing but having played in depth with VMWare I still have some major wishes. One is that it actually mirrored your bios and configuration. Having a generic VM seems to defeat some of the purpose. Also at some point if you could have mitigated access to 3D hardware. I still need to boot to new configurations when playing with XP vs. Linux 3D code and once again - this seems to defeat the purpose. We’ll see where it goes over time. But I certainly think it is here to stay - it earns it’s keep even at what still seems like early stages.

MSDN subscriptions and more

If it weren’t for MSDN subscriptions I’d definitely be taking a close look at Linux as a primary server OS and possibly even desktop OS. I never thought I’d say that. However the quality of Linux distros is so high now and the ubiquitous nature of the web so prevalent that I hardly notice when I’m running under Linux. Not so five years ago. But now - I am very intrigued. Projects like ReactOS (an open source Windows clone) blur the line further. And of course mono.net etc. bring all the “latest” technologies to the Linux world. I’m sure the “next wave” of amazing advances in OS are going to once again shake up the box a bit. But so far everything I’ve seen from longhorn doesn’t inspire me with the inevitable KISS principles that have proven to really make the difference to end users. I like the idea of native 3D support in the OS GUI. But I’m still not seeing the compelling factors here. The OS situation will continue to evolve. But the fight over features vs. clean simplicity is one that is still to be resolved.

With all of the latest talk about virtual machines no one can miss the impact they are bound to have soon. With dual core and quad core processors virtual machines will just be good common sense from a number of perspectives. One close relative of this that I have been waiting to come out of the wings for a while on is the truly transparent remote desktop. A world where you can go anywhere, log into your desktop and get the same experience as though you were right there. The only thing holding us back is bandwidth - but like everything else that fixes itself over time with minimal effort. The thought of running virtual pcs half way around the world on a remote desktop where the experience is truly seemless will change everything we know about computing. Web sites will be able to do things however they wish - they will merely be interacting with dumb terminals - pushing images back forth.

In other news - Google brilliance. Using KISS(keep it simple stupid) they implement only the bang for the buck features and go lightyears beyond anyone else in short order. I still underestimate them periodically. Google talk - yet another IM client. When I first saw it I shook my head and thought this is surely a sign of Google going downhill. After several months I have begun to realize how wrong I was. Ties to gmail were perfect but obvious. Chat history in gmail was more clever but still hinted at…. well yes I could have though of that. But taking your entire gmail list and showing the invited status of each of your contacts - brilliant. It creates the impression that it is already bigger than it is. I see a list in Google Talk that includes 20 people who are currently invited. I can’t help but think it is just a question of time until all of those people are using Google Talk. Microsoft proved more clever and agile than IBM back in the day. It certainly seems as though Google is proving itself more clever and agile than Microsoft. This will be no quick fix for Microsoft. By the time you are a force of nature - it is too late to reroute the butterfly wings that started you in the first place.