It’s no problem writing a 3d game in a month, sunny? What sun?
Pretty much in the month since my last post I've been working on Lagom, which was my project for this semesters Game Design course. Deadline was 30 minutes ago so now I'm finally done, had a little panic towards the end when I found out a game breaking bug that only occurred rarely before now reared it's ugly head very consistently. Thankfully that also allowed me to isolate and find it (or at least I hope so) I'm certain it was related to alignment issues concerning the structures used by bullets for it's SSE operations, which somehow ended up not being aligned. And my initial plan of aligned_storage did not seem to work that well either.
Now with this quarter's courses over (Sweden has a weird system where they split semesters into two quarters) I can start working on my Monte Carlo ray tracer, seems like it will be a lot less work as the requirements are just to write (any) Monte Carlo ray tracer, and even though I'd be interested in writing a much more involved solution; time is limited and there's only 3 month of classes left in my studies.
The other thing I have to force myself to do is write Master Thesis applications, I missed my opportunity at Dice by waiting to long even though the topics they offered would have been extremely interesting. I will also be a Lab Assistant for Matt Cooper in the next quarter, so we'll see if my students will want to jump out of the window as I've told my girlfriend before.
If you want to check out Lagom I made a project page for it, I hope to later make an android version but a few days of dabbling did not result in any quantum leaps of development (as expected) and I was not willing to invest more for the time being. (I didn't even get started converting all the shaders to glsl, though that shouldn't take so long, currently ogre just crashes out on me on android with no error in the log)
First Post +1
I wanted to start this blog over summer, but somehow my mind just ended up stuck in my ideas about how to organize a Graphics Engine using a much more task based approach, I feel like it can save a lot of development time, yet I ended up thinking about a dozen implementations that I was never quite happy with. I still want to get a little demo running, but for now I have to resume my studies, and this semester that means writing a game. Somehow if someone told me I could've written a game for a course project during high school, I would've been overjoyed by the opportunity, but now somehow I'd rather implement a Order Independent Transparency solution. Yet, since I chose to do it in C++ and in 3D so I can at least do some visuals; I'm probably going to end up spending a lot of time on it.
I've thought a lot about Order Independent Transparency recently, and as far as I can tell a stochastic solution seems to offer the most promise to me (Stochastic Transparency) perhaps if we could take some cues from the radiosity guys and lessen the frequency of lesser visible samples, something that would seem to be a possible extension for Stochastic Transparency. I mean in practice having more then a handful of samples per pixel would be prohibitively expensive to shade anyway, so we might be best served as with a hybrid of Stochastic and Adaptive Transparency, pushing samples into our visibility buffer but once it's full instead of merging them (as done in Adaptive Transparency) to discard them semi-stochastically and then after shading merge the samples, recovering the discarded samples from neighbouring pixels.