Re: Csillámtrace - my new graphics engine back
Board:
Home
Board index
Raytracing
General Development
(L) [2012/04/12] [tby jbikker] [Re: Csillámtrace - my new graphics engine] Wayback!Arauna? Ow, Arauna. Long time ago. I did some measurements for you, see attachement. Three scenes, three cameras, 12 cores (24 when counting HT), 3.4Ghz, 1280x704 (because 720 is not a multiple of 32), CPU-only. First scene has 12 lights that all illuminate the full scene (hence the huge average number of rays per pixel). Second scene has more lights, but they are smaller. Third scene is many polygons (600k), and 4 lights, all affecting all scene surfaces. Up to 550M rays/s. [SMILEY 8-)]
[IMG #1 arauna_performance.png]
[IMG #1]:Not scraped:
/web/20161005155601im_/http://ompf2.com/download/file.php?id=10&sid=c8847134741648110dc3ee7fdf367714
(L) [2012/04/12] [ost
by jbikker] [Re: Csillámtrace - my new graphics engine] Wayback!Arauna? Ow, Arauna. Long time ago. I did some measurements for you, see attachement. Three scenes, three cameras, 12 cores (24 when counting HT), 3.4Ghz, 1280x704 (because 720 is not a multiple of 32), CPU-only. First scene has 12 lights that all illuminate the full scene (hence the huge average number of rays per pixel). Second scene has more lights, but they are smaller. Third scene is many polygons (600k), and 4 lights, all affecting all scene surfaces. Up to 550M rays/s. [SMILEY 8-)]
arauna_performance.png
(L) [2012/04/12] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!thank you for sharing these results.
based on your raw ray/sec rate, i can compare it with my own rate too.
the performance of my code is the folowing:
on a 6 core AMD machine, i got 75 fps in res 256x128, with a simple shadow (1 light).
this means:
fps * x * y * lights
75*256*128*2 = ~ 5 million ray per second. (this means a pixel costs 2400 clock)
so your code is actually ~100x faster [SMILEY :D] (so your code eats around 24(!) clock per pixel)
btw, i hope the speed of my code will also be improved soon (the speed of effects mainly), but first i must fix the bugs in the rendering.
(L) [2012/04/12] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!the songoku scene in 512x256:
~38 fps with 6 core on the same cpu ~ 10 million ray per second
(L) [2012/04/12] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i disabled the light in my scene
5% speed up
so i now laugh. something is very fucked in my code, somewhere. but not in the scene tree *sigh*
(L) [2012/04/13] [tby jbikker] [Re: Csillámtrace - my new graphics engine] Wayback!Your calculations are wrong.
Let's start with the 550M = 24cycles:
At 3.4Ghz, 12 cores do 12 * 3.4 * 10^9 cycles = 4.8e10. Divided by 5.5e7 is 872.7, *per ray*, which seems reasonable.
Then, your code:
256 * 128 with one shadow ray, at 75fps = 4.9M rays/s.
Assuming your AMD machine can do 3.6Ghz (I did a quick google), it does 6 * 3.6e10 = 2.16e10 cycles/s.
Each of your rays thus eats 4408 cycles.
Your code is currently roughly 5 times slower, which is reasonable, considering that I spent basically the past decade optimizing mine.
(L) [2012/04/13] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!jbikker:
that AMD machine isnt mine, and i dont remember, how many mhz it was. but i am happy to hear that my code is ,,just'' 5x slower, since i develop ecsactly this algorythms since 4 months. thankfully i can forge around 5% performance up in my code in every day. this means that i have luck, i maybe can reach arauna performance after a month (wich will never happen i think [SMILEY :D]). but the image quality of your code is *mutch* better.
(L) [2012/04/13] [tby Vilem Otte] [Re: Csillámtrace - my new graphics engine] Wayback!Just a little note...
Arauna is still open-source and you can learn a lot from going through it [SMILEY ;)] (I did, thanks jbikker - but still Arauna is faster (in Sponza a lot faster) ... "What kind of sorcery is this" [SMILEY :mrgreen:] ).
The image quality is heavily based on models/textures quality [SMILEY ;)] You can try some Sponza or such models, or I have some usable models on my HDD (even some scenes will be ready soon) created for really small "semestral" game project - so if someone would be interested in ... I could post model or two (and the scenes when ready).
(L) [2012/04/14] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!Vilem Otte:
>> Arauna is still open-source and you can learn a lot from going through it
well, i have downloaded its source code days before, and investigated it. The principles behind the code are not compatible. Its very different from inside. So i cant stole jbikker's tricks from arauna.
and i also dont want to use/rely on any thirdparty code/codesnippets/solutions/principles, becouse it would be a pain in the butt to fix if somethings gets broken in the middle of the code, somewhere, somehow...
>> Sponza
i didnt knewd whats a sponza is. [LINK http://hdri.cgtechniques.com/~sponza/files/] you meant this? ^^
(L) [2012/04/14] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!(i got some requests from people to share this minibenchmark in standalone version. if somebody wants, can get it here: [LINK http://gamegirls.no-ip.org/forum/viewthread.php?thread_id=207&cim=Minibenchmark] )
(L) [2012/04/14] [tby Vilem Otte] [Re: Csillámtrace - my new graphics engine] Wayback!>> Geri wrote:and i also dont want to use/rely on any thirdparty code/codesnippets/solutions/principles
I understand this, I also have my code all written from scratch [SMILEY ;)] It gives better feeling about it, not mentioning when a bug is occured (then it can be a bit nightmarish when heavily using third party code).
>> Geri wrote:i didnt knewd whats a sponza is. [LINK http://hdri.cgtechniques.com/~sponza/files/] you meant this? ^^
Yup, thats the Sponza. There are some other well-known models to work with (Sibenik cathedral, Happy buddha, Power plant, etc.). Of course none of them will actually give the best info you need - as your scenes will definitely differ (and different geometry layout in scenes *can* make quite difference).
I also ran the benchmark on Core i7 (system Linux, Wine 1.4 ... why it doesn't write out the correct system [SMILEY :mrgreen:])
Code: [LINK # Select all] MMX: OK 3Dnow: x SSE: OK SSE2: OK SSE3: OK SSE4: OK
SYS_init: Windows NT/XP/2k/Vista/7/8
CPU raw performance:
Cache16k read: 1525 MByte/sec
Cache128k read: 4746 MByte/sec
Cache256k read: 4502 MByte/sec
Cache512k read: 4808 MByte/sec
Cache16k write: 2533 MByte/sec
Cache128k write: 5808 MByte/sec
Cache256k write: 6571 MByte/sec
Cache512k write: 6774 MByte/sec
MIPS (simple) SMP: 4963
MIPS (simple): 4994
MIPS SMP: 1203
MIPS: 1237
CMP: 1099 million ops/sec
FPU raw performance:
FLOPS (simple) SMP: 1211 (megaflops)
FLOPS (simple): 1181 (megaflops)
FLOPS SMP: 353 (megaflops)
FLOPS: 337 (megaflops)
FCMP: 1162 million ops/sec
FDIV: 112 million ops/sec
CPU algorithmic performance:
PI: 56 million iteration/sec
VEC3 normalize: 37 million/sec
VEC3 normalize (SSE): 45 million/sec
Ray-Sphere collision: 225 million/sec
Ray-Tri collision: 9 million/sec
TMU (+bilinear filter):11.2 million sample/sec
Square Root (formula): 8.3 million res/sec
Square Root (x87): 111.2 million res/sec
Shell sort 1M: 6.2 result/sec
8 queen: 3.7 million queen test/sec
RSA gen/en/decrypt: 0.72 million/sec
SYSTEM performance:
Memory Read SMP: 2560 MByte/sec
Memory Read: 3800 MByte/sec
Memory Write SMP: 1655 MByte/sec
Memory Write: 2999 MByte/sec
Memory Copy SMP: 1050 MByte/sec
Memory Copy: Press any key to continue... 1050 MByte/sec
Address translation: 825 million result/sec
Library calls: 86 million call/sec
malloc/free: 4.8 million pairs/sec
Thread creation: 11249 thread/sec
Minibenchmark done.
I think I would get better numbers, but I'm currently running just Linux systems [SMILEY :ugeek:] , so this ran under Wine 1.4
(L) [2012/04/14] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!ty for watching my benchmark.
8 queen: 3.7 million queen test/sec
RSA gen/en/decrypt: 0.72 million/sec
this two is broken in that version, there was some bugs but i fixed it. (also maybe ray-tri is also broken in that.) strange that it not dumps the cpu type, becouse if its not able to request it from the system, it should be get it with cpuid. and yes, wine runied the results, it wasnt able to get the number of the cpu, so SMP tests also ran without threads.
(L) [2012/04/17] [tby Vilem Otte] [Re: Csillámtrace - my new graphics engine] Wayback!The sky support is handled through sky-dome (testing ray against sphere), sky box (testing ray agains aabb), or just through "hack"?
(L) [2012/04/17] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!if(result==sky){
float u, float v;dual_paraboloid_mapping(&u, &v, vector);
TMU(result_rgba, u,v,sky_texid);
return result;
}
(L) [2012/04/24] [tby Vilem Otte] [Re: Csillámtrace - my new graphics engine] Wayback!Maybe I'll spam a little bit here, but I think jbikker did a little mistake during computation of MRays/s and cycles...
>> Let's start with the 550M = 24cycles:
At 3.4Ghz, 12 cores do 12 * 3.4 * 10^9 cycles = 4.8e10. Divided by 5.5e7 is 872.7, *per ray*, which seems reasonable.
We've got 12 * 3.4 * 10^9 = 40.8 * 10^9 = 4.08 * 10^10 cycles.
Also we've got 550 MRays/s - thats 5.5 * 10^8 MRays/s.
And according to wolframalpha - 4.08 * 10^10 / (5.5 * 10^8) is approx. 74.2 cycles *per ray*.
So actually his code is 10 times faster than he states - 74.2 cycles *per ray* (which is amazingly fast, because from what I've got now - Core i3 @ 2.27, on single core, I can get roughly 2.7 MRays/s -> with KD-Tree -> So I'm at some 840 cycles *per ray* now -> 10 times slower than Arauna [SMILEY :o] ).
For single triangle it's quite different... I can get some 89 MRay-tri tests per second - for packets (on 1920x1080 screen) - thats some 25 cycles for ray-tri test [SMILEY :twisted:] (100 cycles for 2x2 ray packet - tri test)! (the assembly of code (not hand written! - generated by compiler) has about 130 lines) - A good optimization Arauna is using (and I'm not) is that it first tests ray packet frustum against triangle, whether it can hit - which indeed gives huge boost (sometimes). I get some 14 ray-triangle tests on average (interior "ball-in-a-stadium" like scene) - so it means, 350 cycles *per ray* are ray-triangle tests, so another 490 cycles I search in KD-tree.
Getting to code as fast (or should I rather say slow) as mine is quite easy. I just have compact classes for everything and try-and-fail optimized intrinsics functions. Getting to Arauna from here is a lot harder - In my opinion (from what I've gathered when reading Arauna code):
1.) Trade-off speed for precision - I'm still using (right now) quite precise ray-triangle test (barycentric one), Arauna uses Pluecker (imprecise, but a lot faster).
2.) Use ray packets and frustums -> Frustums will quickly tell whether the whole packet hits the node (or the object), and SoA layout of ray packets will give you some 2x - 2.5x speedup over standard AoS code (I'm currently using just ray packets, not frustums).
3.) Do everything in SoA layout, it'll speed things up a bit.
4.) You still won't be there, but you'll be just 2 times or so slower, that'd be the time to use magic wand! [SMILEY :mrgreen:] Or you can buy 4 16-core Opterons [SMILEY :ugeek:]
(L) [2012/05/01] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i am currently writing key/mouse/joystic input system, becouse replace that too. so no strictly ray tracing related thing in the project in these days. after i finished with that, i will expand the scene rebuilding algorithm to be able to process up to 64 threads instead of the current maximum 8, then i will refactor the the camera and character handling from my old engine. I also should do a linux port from the platform api, wich will be easy, according to the platform parts are strictly separated away from the algorithmical implementations. if i finish this, i will bugfix the 2d handling and implements some missing effects to the ray tracing pipeline, and so i will have the base engine in fully functional state, wich i am very waiting, since i want to test it on some complex situations.
(L) [2012/05/28] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i optimized another 25% in speed
(L) [2012/05/29] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback![IMG #1 Image]
the engine now can use hibrid rendering: immediate swap to polygon based ray tracing, with real time texture mapping. purple color indicates where the new algorithm activates, green color indicates that where the algorithm was actually able to switch to polygon based ray tracing, and bilinearry filtered texture addressing. The algorithm only affects the locations that have greater difference in the color, and on near voxels, depends on the texturing, intelligently, automaticly selects those areas, where the hibrid rendering can be accessed. as on the picture, it can result mutch better graphics output without incrasing the voxel number or resolution to 10x or so, only pushing the memory requietment with 35% more, and requires 30% speed (wich will hopefully domesticated back to soon 20% i hope)
[IMG #1]:Not scraped:
https://web.archive.org/web/20161005155519im_/http://yscik.com/jf/datas/users/122-hibridvox.png
(L) [2012/06/01] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!now i have again some mentionable bottlenecks in the pre or postprocess.
maybe i will reach 800x400 on a 8-12 core computer in playable speeds.
i also fixed a problem that caused the far voxels to flicker. the x86 fpu caused the problem, a *=1.000001f in the required place solved the issue.
(L) [2012/06/02] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i made an acceleration structure for the lights also. From the advice from my friend, if i want once to manage 10000 lights simultanously, or so.
(L) [2012/06/02] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!20000 light without notable speed drop - done.
(L) [2012/06/04] [tby toxie] [Re: Csillámtrace - my new graphics engine] Wayback!will there be a downloadable demo at some point? curious to see all that in action.. [SMILEY :)]
(L) [2012/06/04] [tby Serendipity] [Re: Csillámtrace - my new graphics engine] Wayback!>> Geri wrote:20000 light without notable speed drop - done.
Sounds interesting. May I ask how you do that? How do you figure out which lights illuminate a pixel? I always had troubles getting glossy brdfs to work correctly when I tried this.
(L) [2012/06/05] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!toxie: its very fugly at the moment, and nothing to see - i havent even made the effects yet, just sucking with the scene management. but i can give you executables through skype or msn, but i dont take responsibility if cthulhu walks out from your computer = ^ ^ =
Serendipity: basically:
int * lights[10][10][10]
int lights_num[10][10][10]
(of course not *literally* this crap, just logically makes this. also detects the dimensions automatically, minimum 8x8x8, maximum 34x34x34)
but above 20k light, the code starts to slow down - the acc. structure suddenly eats the cache memory. but not a problem, why da fuck would i use such many lights aniway ... .. .. maybe with this, for example i can render a lighter in every soldier's arm of an army from 20k soldier.
(L) [2012/06/10] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!static lighting:
ligting+shadows now can be statically precached to the static objects, so they dont need to trace the whole scene for shadows on those voxels.
i will add an updater that updates this staticness only if its necessary (for example if the light changes position on it, or an object that casts shadow on the specified vox, is moving) so it may will be able to turn it on on the whole scene. this would mean that the global illumination will be complitely free on these objects
(L) [2012/06/12] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!threw windows default memory pool/malloc handlers to garbage - 10x speed up in the scene update. another 2x speed up is necessary to reach *somehow*.
(L) [2012/06/16] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!-i modifyed the scene update. now it runs on max 64 thread instead of the previous 8.
-after fixing a tons of pointer problems, the scene management succesfully performed the stress-test:
dinamic scene with multiple scene reset, static and moving object, automatic vox level recalculation turned on, running on up to 64 threads. runned for 3 hours without bug.
(L) [2012/07/14] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i also finished the normal recalculation with welding mechanism that uses the beginned phisics engine, and its properly multithreaded.
(L) [2012/07/16] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!ability to reflect colorless or semi-colorless, indicated by a variable beethwen 0...1. this results more enjoyable matt surfaces.
(L) [2012/07/20] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!now the reflectioning geometry also antialiased. i was able to do it without a new zbuffer.
(L) [2012/07/20] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i made some optimisations, they speed up the calculation of the reflected surfaces in some cases. the optimisations will also required by other optimisations.
(L) [2012/07/21] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!today feature:
if you go too close to low voxel resolution geometry, it smooths it with dof.
(L) [2012/07/22] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!\lgfx_RayTracer\ujraytracer4.cpp|2933|internal compiler error: Segmentation fault|
[IMG #1 Image]
[IMG #1]:
![[IMG:#0]](images/5daa30c68cb9a3dea579386bcc1749031da5e603d325940a5444a23fb90fffcf.jpg)
(L) [2012/08/27] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i switched to linux, i ported the code within 1 days, so it renders (no input yet)
(L) [2012/08/29] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!-input finished for linux
-a titanic-like bug fixed in the renderer
(L) [2012/10/04] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!particle system almost finished (99%)
i will continue with the anim system
(L) [2012/10/17] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!animation system is almost finished, my character animates properly. however, its leg does not visible for some reason in some cases.
(L) [2012/10/25] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!animation system seems to be done.
develop suspended: memory died in my computer.
since this is a voxel based renderer, i cant continue the code without proper quantity of memory. i alreday ordered the new memory modules. it will arrive within a week.
(L) [2012/10/25] [ost
by paulusul] [Re: Csillámtrace - my new graphics engine] Wayback!Sometimes it's good to program blind without being able to test.
refactor, clean up and comment all your code [SMILEY :)]
(L) [2012/10/25] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!originally, i wanted to finish before christmas. but now this conception is in danger [SMILEY :|]
(L) [2012/10/27] [ost
by mpeterson] [Re: Csillámtrace - my new graphics engine] Wayback!shit, i like this thread...
(L) [2013/08/17] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!thanks!
(L) [2013/08/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!2 billion polygon now loads within 1-2 mins, renders over 17 fps on overclocked 4 core lowend amd athlon2 x4 635. (512x256)
since i have beat arauna (i dont know at what point is it actually happend, since i was not avare), i need another target to competite. is there a good place to find brigade fps results?
(L) [2013/09/03] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!fixed some bugs with the phisic engine, becouse i am now wiring a game for a competition, and had to fix some non-graphics related issues. also there was design flaw in the sound engine for linux: it used fork, i had to eliminate that from the code.
also i started to prepare the code for the arrival of intels upcoming 8 core / 16 thread ddr4 haswell monster desktop cpu with ddr4.
that will give us brutal speed incrase in ray tracing too, and i want my software to be prepared when it arrives in next year.
(L) [2013/09/03] [ost
by mpeterson] [Re: Csillámtrace - my new graphics engine] Wayback!go ahead, this will become the longest ompf thread ever !
(L) [2013/09/03] [ost
by davepermen] [Re: Csillámtrace - my new graphics engine] Wayback!and is not really a thread, but simply his personal blog
(L) [2013/09/03] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!yes, i use this topic as some kind of multiplayer raytracer blog about my ray tracer software.
also
-10% speed improvement in scene loading/refreshing (i meditated at that at night, and i implemented it within 5 minutes (+20 code line))
(L) [2013/09/03] [ost
by jbikker] [Re: Csillámtrace - my new graphics engine] Wayback!>> Geri wrote:2 billion polygon now loads within 1-2 mins, renders over 17 fps on overclocked 4 core lowend amd athlon2 x4 635. (512x256)
since i have beat arauna (i dont know at what point is it actually happend, since i was not avare), i need another target to competite. is there a good place to find brigade fps results?
Beaten Arauna? [SMILEY :)] That one is doing nearly 500M rays in Whitted style ray tracing for Sibenik. It doesn't do instancing so I can't tell you how it would perform on 2 billion triangles.
EDIT: Just checked the chess game. I don't think you can compare your engine to Arauna; you seem to have cut quite some corners in terms of image quality. There's hardly a straight line in the image. Do you know what is causing that?
(L) [2013/09/03] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!its caused by the fpu. not considering to fix it, since i will use smaller resolution in future.
(L) [2013/09/21] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!new game released, for a game creating competition
Dark Tower
Pannika, the animeslut has been kidnapped by ufos, and forced to climb on a dangerous tower. Traps, enemies, and survival. Maybe the ufos let you escape, if you reach the top of the tower. linux and windows version available.
[LINK http://geri.comeze.com/darktower/]
[IMG #1 Image]
4 core i7 recommended.
[IMG #1]:
![[IMG:#0]](images/77b6fac4f6d1abb82944fb8c111d9c080af09a462b6387aef52d2d3500ce21fd.jpg)
(L) [2013/09/24] [tby toxie] [Re: Csillámtrace - my new graphics engine] Wayback!it's very glitchy on my machine (especially when using the two highest resolution modes), and also seems to have some kind of memory leak, as windows has to swap out every now and then until it crashes at some point.
also jumping up the stairs is very difficult as one keeps getting stuck in the blocks often. and please reduce the amount of DOF as often one cannot even see excatly where one has to jump next.
(L) [2013/09/24] [tby sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!It runs fine, when I use the lowest resolution on a mobile I7 CPU, but it doesn't work on Centos 6.4, the usual problem. Can you target an older glibc?
./darktower_x86-64.lnx: /lib64/libm.so.6: version `GLIBC_2.15' not found (required by ./darktower_x86-64.lnx)
./darktower_x86-64.lnx: /lib64/libc.so.6: version `GLIBC_2.14' not found (required by ./darktower_x86-64.lnx)
My issues with the game / engine.
- Everything is blocky, and lots of small black points are jumping around. Quite annoying.
- Slow, compared to aruana (sorry I always mistype the name). At 800x400 it's simply unusable.
- Controls are like from the 90s. Either use the mouse, or introduce something like in croc, that worked fine for situations like this. Or just simply support the mouse, and instead of turning, do sidestepping when pressing a or d.
- Camera is locked to the character, that's usually a very bad idea. Try to add some extra movement there.
- Jumping is slow and buggy.
- DoF is too much.
(L) [2013/09/25] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!toxie: yes, the 32 bit windows version have some kind of memory fragmentation problems, and after a few deaths, it crashes for some unknown reason. i have discovered this bug a few days ago, after releasing the game. but it does not happens on my computer for some reason, so first i will have to find a computer that produces this problem.
sirpalee: hi, long time no speak. i would need an older linux installation with older glibc and gcc to compile a binary for older gcc-s. maybe glibc and gcc update available for centos instead?
-dotty, blocky: yes. block reducing techniques are in my mind, maybe implemented in next year.
-aruana vs csillám: nope. dont forget that this have a complete, real gameplay with full phisics, controlled character, interactive scene and enemies, music, sound. besides, aruana is maybe ~20-25% faster on scenes with small polygon number like this, above, my code becomes significant faster.
-controls - i wanted to add mouse to look around too, but it not worked. i didnt wanted to spend much time on this game competiton, to find out, why.
-camera, jumping --> will not fixed, compo over after 4 days, i have now other things to do.
-dof <- necessary. it makes the things a bit faster. on doffy points, i can bypass some work.
(L) [2013/09/25] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!update: now, due to the fragmentation issuges, the windows version is limited to 320x160. who want to get full experience, please run the 64 bit linux version. i dont support windows any more aniway.
(L) [2013/09/25] [tby sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!Ok, let me rephrase [SMILEY :)], aruana feels faster. From my experiences, the rest of the stuff should be fine on "less" one core. As I said, it feels slower, but maybe I'm not taking some background stuff into account.
GLIBC - sadly I cannot update that. Is there any chance for using an older compiler / glibc? (maybe a small VM?) Or a 64bit OSX version would do as well.
Anyway continue making games! It would be good to see more raytraced games in the future. (even though I'm not a believer... yet)
(L) [2013/09/25] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!yes, i have a compilation environment with a fully working legacy 64 bit linux installation with a distrib from 2007.
somewhere.
that can produce more widely usable binaries, however, sadly somewhat slower binaries.
(L) [2013/09/26] [tby sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!Ok, I'm fine with that, but maybe it would worth setting up a centos 6, that's more or less the common lowest linux these days. If it runs there, it runs everywhere.
One more question, would you mind telling me, how many rays are you using per pixel? (camera, shadow, gi all summed) 2-3-4?
(L) [2013/09/26] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!depends. at default, 2-3 around the character, maximum around 8 becouse of recursion of shiney evil spheres. the engine works automaticly, so i have no control, how large precisely the ray counter is. also in this scene, no gi, just ssao.
(L) [2013/09/26] [tby Geri] [Re: Csillámtrace - my new graphics engine] Wayback!okay i tryed to compile with that old linux, however, the result is broken (no geometry + crash). i am afraid i cant give release for older glibc.
(L) [2013/09/26] [tby sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!What about intel compiler? It generates really good quality code (better than gcc), and you are still able to build binaries compatible with older glibcs.
(you can easily get a 30 day trial)
(L) [2013/09/26] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i have amd, i am not sure if i cant test that properly.
(L) [2013/09/26] [ost
by sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!It works perfectly on amds, you just have to properly set up the minimum instruction set, and you are good to go.
(L) [2013/09/27] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i dont see any clear license informations about the conditions of using it on the webpage.
serial number validation, drm, 700 usd crap. i would not give one cent for it.
also, clang: old versions of clang is capable to compile it, but its tooo slow, binary is significantly slower compared to gcc.
(L) [2013/10/02] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!dark tower is updated
a joystick problem fixed
voxel update got +15% speedup
(L) [2013/11/05] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!32-bit linux binary was accidentally compiled to my current cpu arch. i added a new binary that only uses pure 6x86 instruction set.
(L) [2013/11/19] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!planning to create a technology that decrases blocking, creating the feeling that the software runs on a lit bigger resolution. i plan this as a postprocess technique, to solve the uglyness of pixelcache, relying on a special kind of edge detection, to decrase the blockieness of non horisontal/vertical surfaces with some kind of special interpolation. it will not feelably incrase the machine demand, due to it will be postprocess only, but it will maybe incrase the visual detial greatly.
(L) [2013/11/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!-
(L) [2013/11/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback![quote="Geri"]320x160, start location, without moving, pressing F
Intel Xeon E3-1230 v3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 fps
i5-3470 @ 3.2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 fps
i7-4770 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 60 fps
i7 2600 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .54 fps
i7 2600K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 fps
i7 2700K . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53 fps
Core i7 3630Q . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51 fps
Core i7 4700MQ. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 fps
Core i7-2670QM 2.20GHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 fps
Core i7 2670QM @ 2.2ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 fps
Core i7 2630QM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 fps
Athlon2 X4 635 @ 3.4 ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 fps
Intel Core 2 Duo E7500 @ 2.93GHz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 fps
Phenom2 X4 840 . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 fps
Intel Core 2 Duo T6600 2.2Ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 fps
AMD Athlon 64 5200+ 2x2.7 ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ..16 fps
Athlon64 x2 2.1 ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .15 fps
Athlon 64 X2 4000+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 fps
Athlon 64 X2 3800+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .13 fps
Intel atom z2760 1.8GHz 2c/4t . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .9,0 fps
Pentium4 2.2 ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .7,0 fps
Athlon64 3200+ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .6,9 fps
E1-2100 1 ghz dual core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5,5 fps
Celeron (p3) 1.1 ghz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . .3,8 fps
(L) [2013/12/05] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!first heavyweight game started to be designed with this engine. its not yet public. it will need the brutal vertex and object count offered by the engine. as cpus being faster and faster, resolution 600x300 is nearly playable.
(L) [2013/12/06] [ost
by alidade] [Re: Csillámtrace - my new graphics engine] Wayback!Hi Geri,
You load up polygon models and then voxelize them, or are they voxel models to begin with? Not sure what you're talking about with the vertex count. Any reason you don't use marching cubes instead of the post-process smoothing?
(L) [2013/12/07] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!hi, i use polygon models
(L) [2014/01/04] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!two serious issues fixed in the network code:
-connecting to ip adress was flawed, copying the ip adress to the function copyed less byte than necessary
-under linux, entering to nonblocking mode was flawed
(L) [2014/01/09] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!-no-threaded path crashed due to a bug. fixed.
(L) [2014/01/09] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!when profiling large resolutions - like 800x400 - i have found out that in a generic scene the ssao almost eats the 35% of the performance. i have designed the SSAO in ways to be aware of the low memory bandwidth, but it seems it eats the cpu.
its not alowed for the ssao to have such brutal demand (14-> 9 fps on my computer just becouse of it) so i decided to quickly optimize ssao - or if its not possible, the quality of ssao will be decrased more.
(L) [2014/01/09] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!14->11
however, made the ssao noticably uglyer, but i cant allow ssao to eat so lot resources.
i cant made it even more uglyer, it seems for this feature the 20 gbyte/sec memory bandwidth of a middleclass computer is not really enough.
(L) [2014/01/10] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!+ another 5%
(L) [2014/01/19] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i made some expansions to have larger maps. originally, the engine was only able to carry in enjoyable graphics a 3km*3km*3km sized scene.
first i enhanced this to 20km*20km*20km, later i added a trick to calculate the voxels eternally in double precisity, but stay at float at the other parts of the engine. this needed only to modify a 30-40 line, and allowed me to up to 120km*120km*120km, or even above. thats should be enough for my new game.
i also try to optimize the pre-setup parts of the engine even more, there is some 4 threaded party in the presetup of the cycle, i will soon put all of them to 16 thread (currently, one or two pass left).
(L) [2014/01/19] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!512x256 ~ 10 billion polygon -> 14 fps (idle)
Athlon2 x4 635 @ 3.3 ghz
(L) [2014/01/21] [ost
by lion] [Re: Csillámtrace - my new graphics engine] Wayback!Is there public demo available?
(L) [2014/01/21] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!you can try the two game i posted before.
(however, those are compiled with older versions)
(L) [2014/01/21] [ost
by lion] [Re: Csillámtrace - my new graphics engine] Wayback!I mean lib+headers+sample or viewer which can load some popular format (.obj for example) to try own scene.
PS: have you try use sse\avx to process packet of rays at once? Also you can use Intel\AMD OpenCL (CPU) and achieve sse\avx packet processing using serial C code.
(L) [2014/01/21] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!no, the engine itself not available separately.
i am not interested in inline assembly, and i dont want opencl or any other gpgpu related craps around my code at all. thats why i made this engine, to avoid using graphics cards.
(L) [2014/01/22] [ost
by lion] [Re: Csillámtrace - my new graphics engine] Wayback!I've read about GPU, I mean CPU OpenCL, it's easy way to see what you can get if you use C (not C++) code, so you can use same serial C code for your initial CPU path and OpenCL CPU path. Also take a look at this [LINK http://embree.github.io/] it's CPU only, have serial C++ code as well as optimized sse\avx.
(L) [2014/01/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!well this is a very very interesting question, and here is my opinion about this:
-i dont want such dependencies in my code
-i dont really think that opencl would compile better code than g++. of course, maybe in special cases, where opencl is meant to, it would, but in real-life softwares, i think its impossible
-i dont really have any vectorizable code, except some very few lines in non-so-speed-critical positions that would mean nothing to speed-up
-i dont understand, why they force vectorization, in real-world programs, only very very few can benefit from them, since we alreday have superscalar cpus. vector instructions just waste of the transistors
it would maybe an interesting topic, to see, how cpu-s run these gpu-style algorythms, only for scientifical reasons, but i dont really see any point using it in real-world programs.
judging from the ray/sec rate on your link, on an i7 they only have 7 million triangle/sec render time, so very very time-waste conception - jbikker's arauna alreday can do 3x more in real-world in real environment. so this opencl conception for cpu-s is alreday a dead idea.
i would not waste my time on such things, also, i dont really have opportunity to start sculpting my almost 40.000 code long engine to use opencl in the middle of the renderer for such tests, escpecially when the whole point of this engine was to avoid using such things at all. so its not just so easy. i dont see point in this at all.
(L) [2014/01/22] [ost
by lion] [Re: Csillámtrace - my new graphics engine] Wayback!There no opencl on link.
7 million triangle/sec - I think you look at wrong place, we look for rays\s, you need to calculate and show it in one of tutorialXX_ispc samples.
Also note that arauna have sse optimizations and process packet of rays at once, so you need compare it with _ispc version of embree binary. In opposite embree is serial C++ code (like your engine) which turned to packet using ispc compiler.
PS: darktower game is fun, keep that way =)
PS2: better add game links to first post, so all can easily find it.
(L) [2014/01/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!yes, sorry, i meant ray/sec.
dark tower and the chess game was just basically public beta-tests of the technology, i have learnt a lot of them how NOT to make games in this engine [SMILEY :D]
so i not really linking them. the first heavyweight software maybe arrive with this engine in 2015 Q1, that will be properly advertised on the necessary places.
(L) [2014/01/26] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!please, fill this survey. results will be public
[LINK http://esurv.org/online-survey.php?surveyID=OCDDHH_73dddd5a]
(L) [2014/01/28] [ost
by sirpalee] [Re: Csillámtrace - my new graphics engine] Wayback!The link brings me to the main page. Are you redoing the survey?
(L) [2014/01/30] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!sorry, bad link.
i have fixed it.
[LINK http://esurv.org/online-survey.php?surveyID=OCDDHH_73dddd5a]
(L) [2014/06/08] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!fixed some bugs in the ssao. two component in a vector was bad.
also, i made some notable speed improvements in the scene update, but more speed-ups coming soon.
(L) [2014/10/13] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!i started to optimize 2d sprite rasterization, becouse it was too slow. the goal is to reach 60% performance gain in typical cases. this is nearly reached.
(L) [2014/10/13] [ost
by ultimatemau] [Re: Csillámtrace - my new graphics engine] Wayback!First: What the fudge? Quote: "i dont understand, why they force vectorization, in real-world programs, only very very few can benefit from them, since we alreday have superscalar cpus. vector instructions just waste of the transistors" [SMILEY :shock:] . Very few cannot benefit . [SMILEY 8-)]
Second: Love you're survey (especially the last question [SMILEY :D] over 9000)!
Keep up the good work and discussion.
Cheers
(L) [2015/01/05] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!thankyou. also, i compiled the thing on my android cell-phone. however, i not yet have tested the 3d perfornance, the 2d only. but the 3d will be tested soon too.
(L) [2015/02/01] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!with 2015 01 01, the develop ended.
i dont need more features, so i stopped developing the whole engine, including of course the 3d renderer.
i achieved every goal with this engine.
of course, in the future, i will fix all the bugs still left in the code. also there is an exception on the sprite renderer, which i will use in different projects too - that need some features i havent implemented before.
there is also exception on the platform file, which is needed to compile the project on new platforms (that part are well separated from the engine itself). (i really need to finish the input for android sometimes.)
-i will possibly update the two games using this engine soon (the chess and the dark tower).
-somebody requested the engine as a standalone dll for his project, i will write a minimal documentation for him
-i will update my music software too
-my 3d modeler does not need any updates, that have a relatively new version. i forgot to show this:
[IMG #1 Image]
[LINK http://www.cute3dmodeler.tk/]
[IMG #1]:
![[IMG:#0]](images/947daafbc90bb1cd51391bcbfe15c1828db2f213d49e92c9f02b904b4d2fe4af.png)
(L) [2015/04/22] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!I released a simple benchmark program called GokuBench.
It runs on Linux, Android, and Windows.
[LINK http://gokubench.tk/]
[IMG #1 Image]
[IMG #1]:
![[IMG:#0]](images/bd73ce0002c54030cb8646f4aa5928782dbecf8538e7e13f7194c2e488aff505.jpg)
(L) [2015/04/22] [ost
by zsolnai] [Re: Csillámtrace - my new graphics engine] Wayback!That is a fair bit of dedication right there. [SMILEY :)]
(L) [2015/04/23] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!just realist [SMILEY :D]
(L) [2015/09/20] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!AnimeGirl3D - a tool to convert 2D graphics art to 3D models (obj files)
[IMG #1 Image]
[IMG #2 Image]
[LINK http://animegirl3d.tk/]
I discovered several bugs in my ray tracer, while i created this software.
-the memory management was a bit broken: the optimiztaion, that supposed to choose, wich memory size from the voxels should alloc the memory, was broken, and always choosed the largest. fixing this, resulted memory allocation drop by 70%
-a quick voxel adder optimization was broken, resulted slow scene updates, and larger memory sizes. again, another 70% drop in memory usage.
so when i started to write this, it ate 1-2 gbte of ram, now typically it eats a few 100 mbyte.
i use one ray tracer effect in this besides the light/shadow: a simply reflection on the ground.
[IMG #1]:
![[IMG:#0]](images/abf3b09890d7432e66167f7f78d810e6226b021ca8f72850e41591362d985cb0.jpg)
[IMG #2]:
![[IMG:#1]](images/3157de4108dbd3f31ff7c91820cc85eb611a2b37f57c2ce9c812101debaa0d9f.jpg)
(L) [2016/03/15] [ost
by Geri] [Re: Csillámtrace - my new graphics engine] Wayback!Sorry for not talking for a while, but as i alreday mentioned, the ray tracer is finished, and no future developments will be made on it - except fixing the bugs affecting the softwares i am mading with the game engine.
nowdays i am not into visualization or gaming, i am focusing into a very different, and more important part of informatics, which is needing all of my time. it is an another holy grail of the industry, but nobody knows about it yet. after i properly finished that project (that will be probably happen in this year), i will slowly refocus on ray tracing, becouse i have a new idea. i accidentally invented a whole new ray tracing method from a logic error, which was never even discovered or disucces by anybody else.
that method is 10-1000x faster, than my current ray tracer. i am not aware if i still having the worlds only working-in-reality-and-hosting-games realtime ray tracer or not, maybe somebody beat me in the last year, becouse i not folow the news about this industry any more, but i am a kinda shocked from the new method i discovered. it totally distracts me from working on my currelty more important project.
we was so blind.
i will reveal the method later, after i finished my current project, and implemented the new method, for now, in short, it seems, it fixes every current issues with the real time ray tracers even existed, but it creates some new issues that wasnt yet known, and it will probably run on a cell phone on 720p resolution above 25 fps with real global illumination.
the method is not relying on any current algorythms or mathematical trics, therefore, literally everything must be implemented from zero, but a fully funcioning and optimized code is much simplyer than a current triangle/box/voxel/whatever based ray tracer/caster.
i am not yet sure if i were wrong or correct, so thats why i not reveal anything about it currently, but if my method is okay, that will be like throwing out the piston engine (rasterer) and the jet engine (ray tracers), and the new one is like black hole drive in a romulan destroyer.
i am not even sure if i invented a new way of ray tracing, or is it a new form of rendering.
so after i finished my current work properly, i will begin to experiment with that. its shocking how simple the algorythm is.
(L) [2016/04/05] [ost
by toxie] [Re: Csillámtrace - my new graphics engine] Wayback!keep us updated.. and maybe some pictures incl. framerates?
back
