YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT

(L) [2006/12/06] [bouliiii] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

Hello there.
I am "proud" to announce that we have done the first release (well the second release) of YACORT (Yet Another COherent Ray Tracer).
This is GPL licensed implementation of OpenRT. Some details:

1/ Everything can be found in my professional homepage:
[LINK http://bat710.univ-lyon1.fr/~bsegovia/yacort/]

2/ This is an *early* release so the code is incomplete and may be unstable

3/ We did *not* implement RTS as presented in the non commercial API since we found it a bit limited
We use a new shading interface RP entirely written with SSE / SSE 2 and so one. It is not complete
but you can find some useful information in the seven tutorials. Furthermore, everything is SSE optimized

4/ Texture accesses are not really optimized and there is no mip-map right now

5/ This is Linux only since we did encapsulte thread / mutex functions (todo)

6/ Unfortunately, there is no MLRTA techniques. It was a bit hard to implement it with a generic shading interface.
But, it will come !

7/ The memory handler is not in paranoid mode and this is an coding design error (malloc are just asserted and it is unsufficient for
a run time API). But, it is currently changing !

8/ GCC 4.* is REQUIRED !!!

Please, all your remarks are welcome.
Enjoy

Ben

(L) [2006/12/06] [tbp] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

Salut ben [SMILEY ;)]

I've given a cursory look at YACORT when you registered and a few things made me wince, above all the use of C in these days and age. I can't remember details to whine about, i should have written a memo right on the spot.

Other than that it nice to see a convincing try at implementing/cloning OpenRT. Now where is the CVS/SVN server again? [SMILEY :)]

(L) [2006/12/06] [bouliiii] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

Unfortunately, there is no public svn server right now. It may be a bit long to make it since French public services are sometimes sloooowwww. A better thing to do
would be to use an extern server instead our currently one. Right now, we can see if it is possible to register developers for our private svn server.
Would you be interested ?
Ben

(L) [2006/12/06] [tbp] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

I would have contributed my GPLed code right away if only you were using C++. But that's not going to happen. I was just suggesting that GPL code and no repository was a strange combination.

Btw kludging away with -fno-strict-aliasing is a sure sign there's something fishy with your codebase and it can only get worse with gcc 4.3.

(L) [2006/12/06] [bouliiii] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

tbp you'reright: aliasing is BAD and my laziness is worst.
So, I union typed what must be union typed and strict aliasing rule is not broken any more.
See you soon

Ben

(L) [2006/12/06] [tbp] [YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT] Wayback!

[crouching tiger stance]I've looked into mm_state.c and found the usual suspects, that is robustness issues, admitedly a pet peeve of mine, with intersection & traversal the culprit being, mostly, [jump at throat] liberal use of reciprocal SSE ops. And lack of corner cases attention.
See [LINK http://ompf.org/forum/viewtopic.php?t=281 this thread] for some pointers.

I can't cook a test case as i'm currently operating within the Evil Empire.

YACORT: Yet Another COherent Ray Tracer, GPL licensed OpenRT back

Board: Board index ‹ Ray tracing ‹ General Development