Ok, a lot has happened recently, so let me update you on things and thoughts I had.
First, I finally got a real job, I am now a professional web programmer, not exactly my first choice but hey you can’t get it all, well at least it allows me to develop some ideas I have while I’m getting paid for doing things, and the upside is that it’s a pretty flexible job, being deadline based and all, I can take an extra long lunch if I need to as long as I get things done on time.
Fortunately I do a lot of back end work so deadlines and jobs are reasonably evenly spaced and phased, most of the time.
Anyway moving on to anime, my current anime viewing list is
Zero no tsukaima 3
Nogizaka Haruka no Himitsu
ToLoveRu – Trouble
See at least some of them or forfeit you license to be of the ruling class when I take over the world.
OpenGL 3.0 came out in a hailstorm of controversy, they basically removed all the stuff that made 3.0 something better than 2.x, this is basically openGL2.5 and that’s not to great.
I have however not done so much work in openGL as I should, I still have to do the deferred Rendering and stuff like SSAO.
No this summer has been all about ray tracing, I figured that if I was going to continue advocating it I better do some more testing and development of my own ray tracer code, it’s pretty cool now, much better than before.
But I decided to rewrite it all for the future, possibly also using Cuda or openCL, I have a pretty neat software design targeted at multiprocessor systems, so all I am waiting for is the overwhelming urge to take the plunge and start coding, I am thinking of maybe couple it with a deferred rendering engine, just like nvidia did.
Speaking of them, if I take what I know about the mechanics of ray tracing and apply it to what Intel is saying about ray tracing I can safely say that they are full of shit, they are claiming that in a few years cpu:s or possibly the mythical larabee + a 8 core intel will be able to muscle through the rigors of ray tracing, and that’s bull, sure they have the processing power to do the shading and maybe even the ray testing part.
Theoretically I might add, you see it’s not so much about the processing power as we have some ray tracing chip prototypes and they can do that stuff at decently high speeds compared to cpu:s or even gpu:s, no most of the speed sink is related to the raytesting, I know this because when I implemented a system to disregard some ray tests depending on various factors I increased the speed by 3-5 times, and that algorithm is not that aggressive or perfect.
And this was on a scene with relatively few objects to test, basically for each ray you have to test it against all objects.
So lets talk numbers, lets say we have a scene of 1 million polygons renders at 1080p 30Hrz.
There are 1080*1920*30=124416000 pixels rendered each second, that means that in the worst case without optimization we would need 124.416.000.000.000 ray tests each second for only one light source thats right 124,416 trillion tests.
So if you have a system with 1ghz sub processors that can do one ray test per clock cycle you would need 124416 processors, the latest Nvidia card has 280.
So to get around this you need to do some impressive optimizations, and that will require specialized hardware, and neither nvidia or Intel has this, yet.
Now you may have seen some examples like q4 raytraced or the recent nvidia demo, they are both bullshows, the results are significantly fudged, so lat’s just say I am not going to hold my breath.
Outside af all this I am also working on a draft on a new html replacement, basically I am pretty fed up with the frustrations of html 4 and how all browser renders it differently, css is much nicer as I can pretty much depend on certain things be rendered the same thing.
So why not just strip out most the tags and let the programmer make his own.
I would much sooner have a more powerful css than html 5 or even 6 cause I don’t think they will add anything useful, you know like they did with html 5.
Well it’s 03:42 and I’m pooped, so see ya later