Well, I finally got my SATA drive issues resolved. As briefly discussed in the comments of the last post, on looking at the detail of the drive I noticed that whilst mine was in the same model range of those affected by the bad implementation of NCQ (DiamondMax 10, model number starting with 6B), the firmware version on mine was a few notches below that supposedly affected by the bug (mine is BANC1B10, the affected version in this family is BANC1B70).
Well, my upgraded machine has been yanking my chain all evening. I’d resigned myself to losing the games drive and went to recreate the partition and format it, however XP was having none of it. The remaining 100Gb of my disk thus remains inaccessible, due to what seemed to be inexplicable format errors. Turns out now that there are actually known problems with Maxtor drives and the nforce4 chipset, for which Maxtor have issued a firmware patch, although you have to contact their support to get it.
Well, my stuff arrived today so this evening was dedicated to hardware upgrades. The new Thermaltake chassis is indeed very nice, and surprisingly light. As befitted the occasion, I ensured that I still managed to cut myself and thus christened the case with a bit of the old personal claret - not a fault of the case, I might add, actually a particularly sharp section of the motherboard blanking plate. After some swearing (whilst they may have invented ‘zero insertion force’ CPU sockets some years back, they still haven’t managed to resolve the 1500 pounds of pressure required immediately afterward to mount the heatsink) I managed to get the new bits in and working.
Well, I got time to do more testing of the multi-SceneManager today, and amazingly it all seems to work; I now have multiple independent SceneManagers playing nice together and running pretty darn fast I have to say; that’s a Quake3 level, a LOD’ed terrain and a texture shadow test all running at once on my FX5900, each in their own separate subscenes. I think I see a few visual artefacts on the BSP level which I’ll look at, even though BSP is not our favoured format.
I finally got around to ordering some upgrades for my machine here, something I’ve been meaning to do for ages but just never seem to find the time. I haven’t upgraded the core of my main machine for over 2 years now, mainly because these days I spend far, far more time coding than playing games and thus not only do I not have time to tinker with the hardware much anymore, but having a graphics card that’s not the latest and greatest is actually a bonus - I’m always scared I’ll get sloppy if I’m working with the best all the time.
I’m quite pleased with the amount of work I’ve got done on OGRE over the past couple of weeks. As well as the things I’ve already mentioned in this blog, I’ve added a new ParticleSystem optimisation feature that lets you tell particle systems to stop updating if they’ve been out of camera shot for a defined period of time. Since particle systems can be CPU heavy, and you might want to distribute them around a bit, this should save some cycles and allow more particle systems to be used practically.
God, MS really need to get SP1 for VS 2005 out, now. I’ve been mildy irritated by the appearance of a little bar at the bottom of the VS window labelled ‘Updating Intellisense…’ which kicks in and generally screws up my performance measurement after I do a large build. I have to sit there waiting for it to finish before I take any performance stats because it hogs the CPU badly.
I’ve bitten the bullet, and have taken the opportunity raised by writing my version of zeroskills patch on enumerating SceneManagers to set right a few things in SceneManager in general. One of the goals is to make it safe to use more than one SceneManager at once, particularly for rendering subscenes using different techniques. In theory this has always been possible, but in practice some optimisations and assumptions put a few hurdles in the way.
Well, it’s been a seriously crazy week, I’ve been flat-out the whole time. The good news is that plenty of OGRE work got done, including fixing a few relatively obscure bugs in the stable version (Azathoth), which will thus make it in to version 1.0.7, which I’m building right as we speak. I believe this will be the last Azathoth release before Dagon goes planet-wide. I also managed to clear a good few of the remaining TODO items on Dagon this week, leaving only compositor scripting and a couple of platform & rendersystem-specific bugs to resolve before release, which with any luck we’ll be done with by the end of the month.
Well, Borland have finally accepted defeat and are looking to sell off their IDE business, including Delphi, JBuilder and C++ Builder. It’s a long way from the heady early 90’s when Turbo C and Turbo Assembler were my two favorite development tools and the Borland name was synonymous with software development. Despite it being a little sad, I’m not surprised in the least. The fact is that since those heady days, Borland have consistently gotten it wrong.