 mitchellhansen
|
ec630a454d
|
More bugfixing on the negative ray directions. Again having the forgetful while loop
|
7 years ago |
|
mitchellhansen
|
3218a1f35a
|
And fixed a few things that got brought back in that fix
|
7 years ago |
|
mitchellhansen
|
61312b7bc6
|
Somehow fixed a codexl compile error
|
7 years ago |
|
mitchellhansen
|
91e5d1bcd6
|
Getting ever so closer to fully working oct rendering. Fixed a very very weird bug that caused either register corruption or messing with the program counter or whatever. Caused by the parent stack index not being populated for the root index. Weird
|
7 years ago |
|
mitchellhansen
|
c1e18ce17b
|
Fixed small error in scale when doing the downward traversal step. For some reason OpenCL has decided to start completely skipping the downward traversal loop when the jump power is equal to 1/2 dimension. What the hell?
|
7 years ago |
|
mitchellhansen
|
67ec7b7785
|
Traversal is finally able to render coherent images, lots of artifacting. Probably logic error with traversal as opposed to math error in intersection_t's
|
7 years ago |
|
mitchellhansen
|
af7e0bf00b
|
going to need to either add method of getting sub voxel position from the idx mask, or keep track of both vox and sub vox position in the DFS. Converted some logic steps in the DFS over to branchless selects
|
7 years ago |
MitchellHansen
|
3c76a0729a
|
looks like the idx checking got broken somehow
|
7 years ago |
|
MitchellHansen
|
9ab0770bdb
|
tweaking, fixed logic errors in the DFS vox function along with some traversal errors. Hunting down a driver crash that I somehow introduced
|
7 years ago |
|
MitchellHansen
|
20f36d4eb1
|
lots of little tweaks, the cl is still completely broken, need to revert the oct position calc
|
7 years ago |
|
mitchellhansen
|
8f6ecae4cb
|
really bad crashing, need to debug this on a cpu only machine
|
7 years ago |
|
mitchellhansen
|
8f822e14f7
|
Have some real bad lock ups when the oct is semi full
|
7 years ago |
|
mitchellhansen
|
6650eb784d
|
Slightly closer, for some reason the face mask has been -1 and 0 this whole time????
|
7 years ago |
|
mitchellhansen
|
86bcd4f0ae
|
Well I got something coming up, performance is worse than I was hoping, but there's a lot of optimization to go
|
7 years ago |
|
mitchellhansen
|
6464fceece
|
Added enabling and disabling of the octree from the settings buffer
|
7 years ago |
|
mitchellhansen
|
d2bc5e483a
|
The settings buffer is fully operational
|
7 years ago |
|
mitchell hansen
|
d6bdcbdeca
|
Leaving it compiling at least, CL seems to have broken at some point on my thinkpad
|
7 years ago |
|
mitchellhansen
|
176d9f7a54
|
And also made the camera react to gravity, not fall through the floor
|
7 years ago |
|
mitchellhansen
|
c698711fdf
|
Finally converted the camera lens to a proper frustrum, no more fish eye
|
7 years ago |
|
MitchellHansen
|
40634837a9
|
Porting over the traversal algo to do some in situ programming. Running into some pretty bad register pressure bottlenecks. Might pivot to thinking about the multistage kernel for a bit
|
7 years ago |
|
MitchellHansen
|
36bf5697fa
|
Tweaking, fixed a very old off by one bug on voxel gen
|
7 years ago |
|
MitchellHansen
|
dcf355c636
|
Small tweaks while debugging. Will get 1:1 traversal working 100% before tackling jump_power
|
7 years ago |
|
MitchellHansen
|
242aaaa485
|
Tweaking the blinn-phong function
|
7 years ago |
|
MitchellHansen
|
76189ef0b4
|
Optimizing, fixing things in the kernel. More oct work
|
7 years ago |
|
MitchellHansen
|
77e283a4ce
|
256^3 rendering at 15FPS WITHOUT THE TRAVERSAL ALGORITHM! This thing's gonna FLY!
|
7 years ago |
|
MitchellHansen
|
2519532172
|
committing in a broken state, we got problems with the far pointers. Acting as absolute when should be relative
|
7 years ago |
|
MitchellHansen
|
787e308bcb
|
Removed the demo movement, added fog, correctly this time!
|
7 years ago |
|
MitchellHansen
|
f733ca4aec
|
'Demo Mode' release
|
7 years ago |
|
MitchellHansen
|
87b7f60fa0
|
found it, wasn't handling the intersection case
|
7 years ago |
|
MitchellHansen
|
ed99716565
|
More optimizing, removing some dumb casts. Some are needed though when they really shouldn't be? Also somehow broke shadowing in the last few commits and never noticed D=
|
7 years ago |
|
MitchellHansen
|
c5c65474d6
|
~10 FPS from moving some oct stuff to const, ~0.5 fps from adding a few more consts to initializers in the kernel
|
7 years ago |
|
mitchellhansen
|
58ef1da02a
|
Did this work?
|
7 years ago |
|
MitchellHansen
|
5fcf1c0e44
|
working on the frame controller, this might get messy
|
7 years ago |
|
MitchellHansen
|
a6e18bbb54
|
Generalized the way I'm handling reflections. Need to do some soul
searching on how I'm accumulating colors
|
7 years ago |
|
MitchellHansen
|
9ac52aef5e
|
Hah! Was able to get the ray redirection working, very large step in the way towards multi light, refraction, reflection, and indirect lightinggit status
|
7 years ago |
|
MitchellHansen
|
ed250d1291
|
Octree works out to 128x128 as is, and performance seems very promising. Traversal or octree generation fixes will be next
|
7 years ago |
|
MitchellHansen
|
305ef917e0
|
Thankfully it wasn't a problem with the data format, I was passing the
address of a ptr, not the address contained by the ptr. With that,
preliminary GPU octree interaction works perfectly
|
7 years ago |
|
mitchell hansen
|
7c86c60f9f
|
Added EGL linking and useage for the linux cl_khr_gl_sharing
interaction. This still crashes on my machine but at least its getting
closer to working. Fixed some warnings and errors while compiling with
clang++
|
7 years ago |
|
MitchellHansen
|
ebce781eb3
|
Got the octree data to the GPU and it's traversing it, but it having some corruption issues. Endianness???
|
7 years ago |
|
MitchellHansen
|
836af27a3e
|
small tweaks to the kernel code
|
7 years ago |
|
MitchellHansen
|
86f1622090
|
Some machinations on a config structure as well as a restructure on how
I do logging.
|
7 years ago |
|
MitchellHansen
|
9f764f4cbd
|
Lots of little tweaks as I figure out the octree. Fixed bug regarding the selects in the kernel dictating material texturing
|
7 years ago |
|
MitchellHansen
|
4642ab8f0b
|
Fiddling with the traversal algorithm and shoehorning in the DFS algo
|
7 years ago |
|
MitchellHansen
|
c7bde50e0d
|
Saving before breaking changes
|
8 years ago |
|
MitchellHansen
|
ce862feb0b
|
Couple of refactors and tricks in the kernel to speed things up. ~5FPS average improvement
|
8 years ago |
|
MitchellHansen
|
334a375535
|
Commit might have been messed up
|
8 years ago |
|
MitchellHansen
|
8806777dc9
|
Lots of tweaking. Tenative buggy reflection implimented.
|
8 years ago |
|
MitchellHansen
|
5e222a0331
|
Added new demo, tweaked some values to produce better lighting
|
8 years ago |
|
MitchellHansen
|
0d82cd5a20
|
Finally moved screenshots and runtime compilation to GUI elements
|
8 years ago |
|
MitchellHansen
|
7e5d4ef947
|
10 FPS average increase from changing global work size to 2D
|
8 years ago |
