LuxRender v1.0RC1 release

[rendermagic]

I agree with your comments on hardware locks, Dade. Any software development pursuit needs to consider this strategy. Companies that close or lock their technologies make it more difficult for others to progress their technology.

I also think it would be interesting to see a comparison between Cuda running on Nvidia hardware compared to openCL running on Nvidia hardware. Further, running openCL on AMD hardware should be included in the comparison too.

In my understanding, openCL on Nvidia hardware is compiled down to Cuda before it is finally compiled down again. There's a lot less opportunities to fully optimize with this path. So, at the present time, AMD hardware is a great option.

I'm running Nvidia 580GTX hardware and at the time when I purchased it, I was trying to compare some AMD hardware, but I couldn't get the drivers to work correctly on my setup. So, ultimately, Nvidia hardware was faster for me

[J the Ninja]

On the subject of hybrid bidir...Dade, were the changes you committed earlier today supposed to fix the issues with the metal mat? It still doesn't look quite right. Or is that just he lack of MIS? (note, samples or render time are not comparable between these two)

Hybrid:

hybrid-gold copy.jpg (70.05 KiB) Viewed 672 times

Sampler:

sampler copy.jpg (46.7 KiB) Viewed 672 times

[Dade]

On the subject of hybrid bidir...Dade, were the changes you committed earlier today supposed to fix the issues with the metal mat? It still doesn't look quite right. Or is that just he lack of MIS? (note, samples or render time are not comparable between these two)

It should fix all diffuse and all specular materials (i.e. glass, mirror, null, architectural glass, ecc.). I have not yet tested any glossy material but they should work if diffuse materials work. However, missing MIS can be quite "deadly" (i.e. a lot of fireflies) for rendering glossy-nearly-specular materials. If you post the material definition, I can check if it is just the missing MIS or something else.

[J the Ninja]

On the subject of hybrid bidir...Dade, were the changes you committed earlier today supposed to fix the issues with the metal mat? It still doesn't look quite right. Or is that just he lack of MIS? (note, samples or render time are not comparable between these two)

It should fix all diffuse and all specular materials (i.e. glass, mirror, null, architectural glass, ecc.). I have not yet tested any glossy material but they should work if diffuse materials work. However, missing MIS can be quite "deadly" (i.e. a lot of fireflies) for rendering glossy-nearly-specular materials. If you post the material definition, I can check if it is just the missing MIS or something else.

Nothing fancy, just preset gold:

Code: Select all

Texture "Material_nk" "fresnel" "fresnelname"
   "string name" ["gold"]

MakeNamedMaterial "Material"
   "float uroughness" [0.075000002980232]
   "float vroughness" [0.075000002980232]
   "texture fresnel" ["Material_nk"]
   "string type" ["metal2"]

[Dade]

Nothing fancy, just preset gold:

I think it is just the missing MIS. I have done some tests and if you run the rendering long enough, the result is correct, it has just a very high variance (as expected).

If you want to look at a good example of No-MIS BiDir Va MIS BiDir, just check http://www.sjbrown.co.uk/2009/02/05/mul ... mportance/

[Dade]

A small note: it seems to be a typo in Win32/64 SLG2 window title, it says to be 2.0devel4 but I assume it is v2.0devel5 instead.

[jensverwiebe]

Nothing fancy, just preset gold:

I think it is just the missing MIS. I have done some tests and if you run the rendering long enough, the result is correct, it has just a very high variance (as expected).

If you want to look at a good example of No-MIS BiDir Va MIS BiDir, just check http://www.sjbrown.co.uk/2009/02/05/mul ... mportance/

Dade, something here makes me wonder: the bidir/hybrid stats are almost higher that std/bidir, but the resulting spp count does not show it up.
Statistics problem or whats happening here ?

Newest build from source as i write here btw..

Update: this issue has something todo when using volumeintegrator multi vs. single.
One get`s exorbitant high effeciency but the render does not look as expected progresswise, needs more investigation.

Jens

[Lord Crc]

A small note: it seems to be a typo in Win32/64 SLG2 window title, it says to be 2.0devel4 but I assume it is v2.0devel5 instead.

I think it may be down to an incorrect symlink, causing me to build against an older luxrays revision

I've been having some difficulties with the luxrays repository and Win32Text extension in Mercurial. I'm working on converting to the EOLExtension but it's tedious as I need to essentially nuke all my working copies.

[Dade]

A small note: it seems to be a typo in Win32/64 SLG2 window title, it says to be 2.0devel4 but I assume it is v2.0devel5 instead.

I think it may be down to an incorrect symlink, causing me to build against an older luxrays revision

I've been having some difficulties with the luxrays repository and Win32Text extension in Mercurial. I'm working on converting to the EOLExtension but it's tedious as I need to essentially nuke all my working copies.

So do you think it is a real devel4 version, not just a wrong slgcfg.h ?

P.S. Jens, I hope to have time to look into the problem tomorrow.

[Lord Crc]

So do you think it is a real devel4 version, not just a wrong slgcfg.h ?

Unfortunately yes. I can provide new builds today, just not sure how we should do it.