Racer v0.8.12 is out

[Ruud]

Yep, it's going a bit quick, but this is all for targeting a non-moving target, v0.9.

v0.8.12 is at http://www.racer.nl/download/racer0.8.12.zip for download

The changes are:

• Added camera.auto_switch_director_track_cam setting to avoid automatic track/director camera mode switches. By default it's off.
• Lighting/shadowing fixes in shaders (dyn_standard_bump*.cg, dyn_standard_bump_speca*.cg)
• The Baja had some bad shaders (metal parts in car.shd)
• Modeler gave a lot of QLOG errors.
• Modeler ASE imports now show a dialog if too many indices are used in 1 geob.
• Accepting new graphics options would crash Racer
• Added auto-calibration for controls (not too useful for consumer controllers though).See http://www.racer.nl/tutorial/ctrlset.htm#values
• Shadowmapping now done using a texture atlas instead of a texture array. This for ATI cards.
• On ATI, shadowmaps could not only be a Z-channel. ATI seems to need color buffers.
• More tweaks to get ATI to work better. Our experiments show that alpha_to_coverage is very poorly supported, but at least shadowmapping seems to work better. nVidia cards recommended.
• sky_daynight*.cg fixed to do the same as sky_*.cg, only with an added night texture.
• Halo effects in shadow mapping improved through the use of Z information
• Added dyn_standard_speca_f.cg shader for 1 texture with specular control using alpha.
• Cg parameter passing in the render engine was cached. Lots faster.

 

[Mr Whippy]

There is certainly something wrong with the fresnel mixing using the ambient/diffuse colour (why chrome is now broken?)

A black car at a glancing angle in midday sunlight is like a mirror, yet mine is very dark in it's reflection.

I'll do more testing, but it's probably in the blend for the envmap I am guessing...!?

Dave

 

[JagerMan]

Thanks ruud, ATI issues have almost totally gone for me. Although, shadow mapping seems to half fps.

 

[n'is]

thanks for ati version

 

[boomer541]

Got increase of 10 fps straight out of the box!

Testing skies, compared with 0.0811 looks good but just marginally brighter. Have to try the carlswood gfx section next.

Have a few questions about the "empty" cg in the Qlog that will have to wait until I check futher.

Thanks Ruud and company!

 

[Mr Whippy]

The brightness is the bloom multiplier in 0.8.12, it is now 1 where it was 0.5 before...

I'd recommend reverting to 0.5 for now. I think this has been boosted to blow out the sky more so reflections look right, but the reflections are broken right now being rather dull (somehow), which means this 'fix' is actually a bodge for a different problem (not that I think Ruud did it realising the reflections are wrong)

Ie, chrome looks dull, and reflections in black cars at glancing angles are not mirror like, so the ref map mixing is not right right now, so skies look dull, so I can see the logic for pumping up the bloom and making skies white out more, making reflections look more intense, but it's the wrong way to do it!


Cheers

Dave

 

[Stereo]

Just a little thing, but worth doing right - movables, like the cones on Carlswood, need a dynamic shader.

eg.

shader_cone~vf_standard
{
  ; Cones (used to have the material 'shader_c_orange) in v0.8.2-
  ; Change to dyn shader
  vertex_shader
  {
    file=dyn_standard_v.cg
  }
  layer0
  {
    map=c_orange.tga
  }
}

Or if there were more than one type of movable material, it would be worth giving it a more general one (vf_dyn_standard I guess would be a reasonable name) instead of vf_standard.

 

[boomer541]

Qlog errors:
Sat Jul 10 02:31:21 (INFO): [racer/2956] Your gfxcard supports 32 units - render pipe only supports 8; capped
Sat Jul 10 02:38:35 (INFO): [racer/2956] Your gfxcard supports 32 units - render pipe only supports 8; capped
Sat Jul 10 02:38:39 (WARN): [racer/2956] QObjMgr: object 'data/renderer/shaders/empty_f.cg' still around with 2 reference(s)

Why the gfxcard info twice? Do we really need it at all?

What is the problem with the empty_f.cg?

 

[Some1]

Boomer (you Qlog nazi! ), Qlog doesn't only contain error messages, but also informative and warning messages. It is nice to have information messages logged for whatever troubleshooting reasons.

So, if a Qlog line contains *WARN* or *INFO* or *DEBUG*, it is not really an error.

The WARN message basically just says that something is still using the empty_f.cg shader while some system wants to remove it perhaps.

Never mind these Qlog messages!

 

[Mr Whippy]

Just a little thing, but worth doing right - movables, like the cones on Carlswood, need a dynamic shader.

eg.

shader_cone~vf_standard
{
  ; Cones (used to have the material 'shader_c_orange) in v0.8.2-
  ; Change to dyn shader
  vertex_shader
  {
    file=dyn_standard_v.cg
  }
  layer0
  {
    map=c_orange.tga
  }
}

Or if there were more than one type of movable material, it would be worth giving it a more general one (vf_dyn_standard I guess would be a reasonable name) instead of vf_standard.

I thought that too, but the the cones seem to tumble around and be lit appropriately (iirc)...

Hmmm

Dave

 

[Lipton]

What is the error?

I`ve attached a screenshot.

http://img683.imageshack.us/img683/7465/theerror.jpg

 

[Mr Whippy]

I`ve attached a screenshot.

http://img683.imageshack.us/img683/7465/theerror.jpg

The shader you have referenced in there doesn't exist in the same place in the v0.8.12 release, so it's trying to include a file that has moved... the shader you are using there isn't one in the general release, so you really need to either use one that is, or maintain your custom shader.

Dave

 

[Bumper]

I`ve attached a screenshot.

Just open up the .cg -files in the cars cg-folder, and replace the "shaders_hdr" -path with just "shaders" instead. Guess you'll have to do the same with Tib's X-Type as well since the shaders were taken from it.

That did the trick to me, though I still can't figure out a good replacement for the metallic shader and the window shader acts a bit weird as well...
Oh and the whole interior appears a bit too dark now.. Could be my textures, but is there any other way to brighten it up a little bit?

 

[boomer541]

The attached images show the same sky dome and textures on Ver. 0.0881 and 0.0812. I think we need to have Ruud and company fix the way skies are done so that track makers can have adjustabiity to allow them to make good skies.

Notice the curved sky using flags=0 and the other problems shown vs flags=73 on 0.0881.

I tried the gfx section from carlswood with the following settings TURNED OFF (commented out):
exposure, flare, and intensity.
Difuse, ambient and specular were all set to 0, 0, 0 and only rayleigh, mie and clouds values adjusted and they were the only ones that had an effect on the sky image!

Please fix so that we can have good skies, thank you!

Noted that headlights don't cast shadows and they are too dim as they were in an earlier version.
data/renderer/shaders/projected_texture_f.cg being = 0.6 makes headlights too dim, I set at 0.9.

 

[Mr Whippy]

Noted that headlights don't cast shadows and they are too dim as they were in an earlier version.
data/renderer/shaders/projected_texture_f.cg being = 0.6 makes headlights too dim, I set at 0.9.

There is a fundamental issue with lights, since they seem to 'add' to the specular of a material, rather than pump up the diffuse... they are also massively bright in the day time, and TOD night time settings are too light.

The whole mix is just bad. Headlights need to be set to 0.05 in power or something, and night times should be very dark, and auto exposure boost the nights up (say exposure 10)

Right now the dynamic range of day and night are about 50% different, rather than maybe 1000x as in real life!

The whole headlights thing right now is just unuseable, as are flares. They are too bright in the day, and too dull at night. Either they need to scale inverse to scene brightness, or something... but as they are they just never look anything other than wrong.


Twill be good to get auto-exposure working nicely and have more realistic intensity values for day/night so these things work more elegantly. Ie, thunder during the day can be too bright, while on a night it is really nice. Flares look good at dusk/dawn, but are crazy in the middle of the day

All just need those detail tweaks to get working perfectly!

Dave

 

[Lipton]

Thank you very much. I`ll try your suggestions ASAP.

 

[Bram Hengeveld]

Article promoted to frontpage news: http://www.racedepartment.com/content/512-racer-free-car-simulation-v0-8-12-out.html

Hope it helps Racer grow a little more

 

[camsinny]

Gah, Ruud, any chance of getting a few more light flares? ...I need 50... The limit seems to be about 40 currently?
While you're messing with them...any chance of fixing bug ref. #G5? All these flares would look fantastic if they rotated properly.

Email me and I'll give you a preview

 

[boomer541]

Working with skies from the console gives some troublesome results.

Don't use number pad to enter numbers, it'll move the car.

fog mie <v> only changes size of sun, fog stays the same.

fog rayleigh <v> 0.01 - all white, 0.001 - too dark, this was with clouds 0.2.

extinctionfactor <v> had no effect.

sun ambient and diffuse had no change.

This was done with my Surfaces and sounds track which you can try to get the sky to work good on by downloading it as tested in the download thread.

 

[Alex Forbin]

Has anyone tried GPLex lately, I can get it to convert to .dof but when tracked imports it nothing shows up. The shader template
is blank after the standard cg shaders are written as well.

Alex Forbin

 

[camsinny]

With CG is there a better way to do some reflections?
I.e. It'd be nice to have the car self-reflect and have things like wing mirrors in the reflection whereas things like the $mirror could use the FBO and envmap as it currently does?
Especially with interiors it'd be nice to have some proper reflections...

Hmmm...