Change of Linux Distribution

At home my PC has 3 OS’s installed, Windows 7, Windows XP and Ubuntu. I don’t really use a bootloader, i use the motherboard boot device selector to switch between the two drives (7 has a drive to itself, and XP and Ubuntu shares)

I decided that I would switch over to just having Windows 7 and Linux, as I haven’t used XP in a long time. I would also change distributions of Linux away from Ubuntu to… Scientific Linux.

What? Why? What the Hell is that?

Scientific Linux was the first linux I was introduced to during my PhD, back in those days (despite it being only 5 years ago) SL 3 was on the systems at Uni, and, while functional it was pretty terrible and getting anything to work was extremely hard work, getting the correct versions of everything was a big task. SL is maintained mainly by the particle physics community, and such it is loaded with useful things as a development environment pre-shipped, however like I said above, it was a little behind the pace back in those days.

Things are a bit better now with SL6 and in the whole is more up to date and compatible
with hardware.

I have collected together all the helpful and handy hints I used after SL was installed in order to get LuxRender built, along with useful things like… graphics drivers.

Graphics Drivers (Nvidia) The original is instructions are listed here but I present a shortened version

  1. Get the required dependancies : yum install kernel-devel gcc
  2. Grab the Nvidia Driver of your choice
  3. black list the nouveau driver
    edit : /etc/modprobe.d/blacklist.conf
    Add : blacklist nouveau at the end
    black list in grub by editing, : /boot/grub/menu.lst
    Add : rdblacklist=nouveau at the end of the line starting with kernal, of the one you want to install the drivers
  4. Set the default init level to 3 to stop the graphics drivers being loaded during start up
    Edit : /etc/inittab
    Replace id:5:initdefault: with id:3:initdefault:
  5. Reboot
  6. log in as normal and run the nVidia installer
  7. Reboot
  8. log in again, and type init 5
    If everything went well, the GUI will load and everything will work perfectly 😉
  9. If all is good, reverse point number 4.
  10. If all is bad… follow all this backwards

My experience with this isnt great, so i take no responsibility if you brick your linux… I certainly have tried to cut corners in the past and ended up in a total mess. These instructions i followed to the letter and successfully installed a v275 series driver in Scientific Linux 6_x86_64.

Building LuxRender

I used a guide posted on the LuxForum  If you follow it to the word, everything should work fine. You can use yum or apt-get in general to install all the dependancies you need and in the case of boost and python, follow things to the letter of the guide and you should not get any problems.

Notes :

If you have installed the nVidia drivers, you will usually have OpenCL and Cuda libraries installed in your system, and thus not require the ATI Stream SDK as listed in the guide. HOWEVER you will need to make a /usr/include/CL folder and copy the OpenCL headers in there. This can be taken from the ATI Stream SDK or, from the khronos group page.

The only things i changed was that i didnt use the ATI stream SDK as explained. I also copied the compiled boost libraies to my system and removed version 1.41 which shipped with my system.  Other than those  you might need to get hold of is libGLEW and compile and install which isn’t really listed as an instruction.

With a little discussion and following that guide i was able to get LuxRays and LuxRender built and running on Scientific linux… which was a nice moment.

Advertisements
| 1 Comment

[WIP] Downed Starship – Cleanup, Mountains and stones.

Working on the Downed Starship in blender 2.57 has brought a few challenges, one of them being placement of debris. Previously I used the game engine to simulate physics. however ,with some of the simulation occuring using the older 2.49 engine, parts of the scene seemed to be suffering from caching issues and for other parts were suffering from being, ‘un-modifiable’ that is, if i move an object to lay it better or put it somewhere a little different, when i hit the render button, blender resets the position of the object back to were it was in the animation.

I cant really complain since in this respect blender isnt really performing an animation, it is the game engine. I finally found a way to remove animation information, but it is somewhat tedeaous and im sure there is  a better way. All you do is go to the Dropesheet and select the object, and delete the position and rotation channels. For the most part this esensially clears all the information. For my scene it didnt reset the position or rotation but left the object where it was on that frame. I should experiment more, but for my cleanup it worked out just fine.

In order to make my background a little more interesting i added a second set of mountains/hills by instancing the original background and pushing it back a little, and rotating. I experimented adding a few rocks, this looks reasonable, however I think i need to remove the texture from them as they look like sandy blobs rather than rocks.

I also used arch glass rather than the full glass refraction, this was an experiemnt to speed up the render time. I also changed the volume integrator to single over multi, and in effect boosted my render speed, from about 500S/s to 10kS/s which is nice. I had to increase the scattering scale however to get a scene similar to what i had before.

Here is the result.

I have some kind of grid artefact in the sky, likely caused by numerical accuracy, but i will be checking with the people at Lux to see if anything obvious can cause this.

| Leave a comment

[WIP] Downed Starship, in helmet render

I wanted to do some kind of real view, as though looking through the eyes of someone standing at the scene of the downed starship. This presents a few issues, scale, focal length and getting the materials and volumetrics to behave. All this within blender 2.57!

I spent some time modelling a helmet, this was a challenge for me but rewarding none the less,

One helmet, as you can see i placed the camera inside it, the focal length of an eye is somewhere between 17 and 24mm however for a camera this doesnt appear to give the same sense of peripheral vision. So i made it shorter to get more of the scene in.

The next part is working out the reflections for the inside. Now i did a number of test renders with glass and the number of reflections where minimal, this is good you might say, as a helmet would hope to reduce the number of reflections as not to confuse the user. And you would be right… however, to have zero reflections is also not in my opinion realistic and takes away from the fact it is an inside helmet view.

I thus added a mix material of glass2 and a 1%mirror and tested that. I also placed a mesh face inside the helmet. As you can see the face is placed far far far behind and not exactly where the real viewing place would be. again, this will need some moving and adjusting.

The last thing i tried was to make a HUD, in order to have some kind of computer overlay to the scene, im not sure i like this but will at least keep it there for the time being.

The last point to make is the volumetrics,

I initially had problems that the volumetrics would be reset when the light passes through the glass of the helmet… that is, the scattering simply would not be there and it would be perfect daylight outside. It appears that if a volume is planar and thin, then this can happen for a scene like this because of rounding errors and the step size of the volume. In total i had to use a thickness of 0.04 using the blender solidify tool, which is somewhat unphysical, but i will look at that later.

Here is the resulting 10hr render

As you can see i dint play around with all the materials this time, because im in blender 2.57 i have to rework all the materials. I think i need to raise the scattering a little bit, but other all, it isnt too bad a progression. The soil needs better detailing i think, probably a displacement map or higher bump map.

| Leave a comment

LuxRender 0.8 Final

As the tile says, LuxRender 0.8 is finally released as a stable version. Barring any major bugfixes it is good to use. I suspect a 0.8.1 version will be released as appears traditional.

To get hold of it, along with the exporter of your choice, click this link

A number of things have been changed since 0.7.1 a few of which are listed below

  • New GUI options for saving tonemapped images
  • render queues
  • new lightgroup handling, and GUI options
  • Glossy Translucent material
  • Band texture for complex gradients
  • Volume scattering (homogeneous)
  • ply mesh loader
  • on the fly microdisplacement subdivision

The full list is rather long, but these are the ones that spring to mind as the major features as a user.

 

 

 

 

| Leave a comment

Large Scale Test

To continue my previous post, a little late however, here is a test i did of a blender generated terrain. I scaled it up so it is approximately 10km wide (from memory) and put it in the background of the scene. Here the scattering seems to give the sense of scale much clearer.

I am moving over to Blender2.57 and such the scene has slowed down a little once more as i try and match the luxblend settings.

| Leave a comment

[WIP] Downed starship revisited with LuxRender 0.8

With the inclusion or at least with the possibility of using scattering to create a dusty environment, I have revisited the previous project spoken about in a few posts – Example – .

One thing that I always found difficult to create was a truly different or strange environment and make it look like i didnt just do something strange with the tone mapping.

I loaded up an old version of the scene and applied minimum settings for volumetric scattering. Here is the result.

The number of samples here is fairly low so it is very noisy, but despite that, the scattering gives a impressive glow from the sun and the sunsky. It also gives a slight perception of depth as the hill fades to the left of the scene.

Since then I decided the following

1) Make my own Environment exactly how i want it and to the scale i want it. The specs of this are approximately the following;

Scale -> about 2km square
Environement -> A long trench drawn out be the ship/object, surface should be sandy with rocks and debris scattered.

I spent a little time trying to create this world, and imported the starship part. Once more i applied minimal settings for the scattering volume and hit go.

One thing that always bothered me in my previous work with this type of scene is that ultimately the scene will be from someone looking out. So I used a cube as a setting device. The camera is located 1.7m above the ground, with the cube being 1m in dimensions set a short distance ahead so it is visible.

It is a start

Posted in Blender, LuxRender | Leave a comment

Splash Screen entry and result

Alas! my entry didnt win, however I did come second place with about 17% of the votes, the winner was moure with 21% of the votes. His scene is great, the first version of it drew an instant “WOW” when I first saw it.

The entry can be viewed here

My entry in its final form can is as follows

Since the competition I have been extremely busy with work and been unable to put any effort into testing or the next competition which is a test example scene competition. I have had a small play around with a scene but Im not extremely happy with it.

| Leave a comment