will do, sir will let you know when i’ve uploaded it. i actually may need to repatch and recompile first, because i’m not sure i applied everything correctly.
guys, i’m getting a weird effect when trying to use the strand info node as an input to a color ramp, which is the color input to a diffuse shader. that is to say, i’m getting a weird effect when trying to use the strand node at ALL. the material renders solid black, no shading. it doesn’t compute shading or shadowing at all (ie, it renders so fast in such a case that it’s not really “rendering” the hair, just its outline in the image). did i miss a file when patching?
The BMW one right? I didn’t change any of the other render settings, just the tiles. I was trying to get it to render in one tile like the older build, but it would crash when I did that.
As I noted I used a 256 tile - “The hair build did one 256x256 tile in 1:48.” So by that math it did almost 1/8 of the image but took 17 seconds more than the older build. (1920x1080 @ 50% = 920x540 - a 256x256 tile would require 3.75x2.11 tiles to complete the image or just under 8 tiles.)
I’ll try a bot build like you ask, give me a half an hour or so. I’m sure there will be 20 other post for you to reply to in the meantime.
I’ve updated the patch with three simple changes. A correction to the encasing system, fixed a clear bug for emissive materials and changed the procedure when no strands or triangles exist. I would be very interested to see if this allows the basic scene to render on a gpu. I wouldn’t recommend testing on objects that don’t have both until this is solved, even with a cpu.
As for this odd circular effect. I need to look at it in detail but believe it may actually stem from having incredibly short strand segments (ones that are shorter than the strand radius).
I scoff at feeling like crap tomorrow
I actually tried to set my avatar to be a guy kicking his self in the butt but the board wouldn’t allow it
twiddling thumbs waiting for cuda to compile
@drobbins
I laughed uncontrollably at the thought of that avatar. Perhaps I’m one of those old ladies in need of sleep?
Here I plead ignorance at how things work, but I think this is need of it’s own build branch on Graphicall. It would certainly be easier for me to involved in the testing as I have neither the tools, nay even a clue as to how to build myself.
do you still have both Mingw64 builds on your computer?
If so… go inside the Hair build
\2.64\scripts\addons\cycles(rename the cycles to cycles_old or even delete it if you have the .zip still
Now go to the buildbot folder the same location and copy the cycles folder
and paste it into the other location. (1. see if it works. 2. see if it has speed increases)
probably wont work, im working on compiling the “new patch” maybe it will be better.
It didn’t like that at all. But I’m going to try a couple other things first.
On first try, I just got a “Blender Stopped Responding” Windows message and it shuts down. Second try, I opened the terminal window before doing anything else, and it was printing out some trace/callback errors related to the QuickPref addon - over and over without stopping. I’m next going to setup the hair build as a portable and disable all addons except for cycles.
EDIT:
Disabling all add-ons didn’t help, renaming folder and removing folder are no different. Looking in the folders of each version there are a number of extra and different files in the BotBuild versus the hair build.
EDIT2:
I let the hair build finish this time using the 256 tiles and it finished @ 8:06 min.
you could try just swapping out just the CUDA kernels in the
\2.64\scripts\addons\cycles\lib
folder with the buildbot one… see that works… (leave all the other files alone)
(you will need to delete the stuff in the “pycache” folder)
then uncheck cycles in preferences and re-check it probably ro re-enable it.
Ok, out CUDA machine finally got open i tested this with a
Geforce 550 CUDA is VERY slow in this build
Will wait for some VC builds to popup(or more MinGW builds even, see if that fixes it)
till then, CPU renders pretty quickly.
the new patch does fix the problem of being able to render scene without particles and CUDA,
(its just not even worth it, at least with this build since its so slow LOL,
better off just rendering with CPU if using this build.
**just to save some people some time i did try swapping out MinGW kernels with buildbot
ones CUDA and just got errors. so… needs to be done at compile time.
i dont know why the MikePan scene should render so slow with this build on GPU, it
makes no sense. I somehow compiled the slowest GPU render of all time.
@ohsnapitsjoel: here's the node setup I used for the furry character, as I said it allows the hair to be transparent in the tip, but it increases the render times A LOT!!
So instead you should lower the Width Multiplier to something like 0.02 and render more children particles to avoid that.
EDIT:
Here’s another one, took an old scene and set all the materials for cycles to do the comparison, and it is much better and much easier to setup in cycles than in BI.
Last edited by julperado; 09-Dec-12 at 23:35.
I hate to read such posts; julperado, your Internal render looked bad because you made a horrible shader and particle setup, in my opinion BI strand rendering was solid and I saw some incredible works using it.
Nevertheless - I’m really happy to see that Cycles will soon support strands.
Agreed, I love this patch and do not mistake my thoughts here, but old hair, to me, was really very good…aside from the render times it was actually very very good imho…I have not tested this patch yet…it seems to look good with less tweaking and fast render times(relative)…hope it gets reviewed, ironed out, and into trunk by next release.
I totally agree. BI could give you decent result if you took a little time to figure it out. Strand settings, Material setting and how you light it(deep shadow maps) is important.
This does seem a lot easier though. I am very excited about this feature and it seems to work quite well at such an early stage.
Cycles is my renderer of choice and hair was the missing feature that was holding me back the most.