Racer v0.8.13 is out

[Ruud]

Nice for the weekend, v0.8.13
Get it at http://www.racer.nl/download/racer0813.zip (60Mb)

Have a good weekend!
Ruud

Changes:
- Updated FMOD to v4.30.5 (needed in a first attempt to create a 64-bit Racer, for what it's worth)
- standard_blur_f.cg had a syntax error; fixed.
- You couldn't select shadowmapping (CSM) but no bloom. Now you can with fs_filter1=shadows_f.cg (automatically done when selected from the menu)
- Bugfix: first moving menu image was faulty
- Added 'bloomFactor' to data/renderer/common/constants.cg (0.5) to adjust bloom amount
- Special.ini's gfx defaults were not too handy, giving division by zero and slow framerates.
- Number of available flares raised to 60 (was 40)
- Added physics.tracks.use_road_noise option to optionally ignore road_noise from track special.ini files
(not really useful for the internet version)
- 'show carpoints' now also shows flare normals
- Flare aspect was changed with viewing angle, making it appear as if the normal was bad.
- Flare texture rotation now a bit nicer (dot product with camera X instead of Z)
- Updated CSM filtering (+bloom) to use GPTDBiz's PCF filtering (nice work!)
It does have some aliasing/Z artifacts currently though, which need fixing.
- dyn_standard_bump_reflect*.cg had a bug in normal passing, which generated too much shadow areas
(all areas where the normals had negative components)
- Added 'track load <name>' console command to load an entire track (not too useful for general use)
- empty_f.cg shader was not released, giving QLOG warnings
- Rain could disappear sometimes (bad renderstate)

 

[boomer541]

With the note about sky image alpha being used and my sky dome is textured with a .jpg image I'll try using a .tga and check it out, Thanks for the input.

Now to a shader problem on my Surfaces and sounds track as well as carlswood. See the images below. if I can get permission to post them which it does say I have below, damn MS IE8. The car shadow disappears when driving on some surfaces, like mud for example.




We still need particles for mud and stones.

Can some one tell me how to do sparks, etc. thank you.

 

[Ruud]

Sounds good Ruud

Is there any way we could have a shader that makes a plane move around it's pivot point in the manner above...

You could try and create your own, not sure if it's possible, but there are matrices available for vertex shaders to do billboarding (and billboarding can be done in many ways). 'modelView' and 'modelViewIT' are parameters, which are what I use in the flare code to billboard things. Also, the 'center' uniform variable will contain the center location (not sure whether it's world or object coordinates, I think world) of the object. You also need 'modelMatrix' to move along with the car. Hmmm. Still, then you just get 3D objects per flare and you're even farther away I think.

 

[Mr Whippy]

Gtpdzbiz, showed me that the problem was the fact that I had render once set in racer.ini and the sun was showing due to the fact that it was baked into the envmap and not using the live envmap (which works fine). I guess you could render the envmap once without the sun effect
and then just use the shadow mapping to test when to block it thereafter.

I look foreward to your tests, maybe we can give GT5 a run for it's money.

Alex Forbin

Yeah, your problem was an easy fix in a way because the sun became occluded in the envmap for the car from an object a good distance off the car, however, it doesn't stop the problem as per the Lamborghini as pointed out... or any car with upward pointing (ish) polygons on the shaded side of the car

Only solution is as you say to have two envmaps somehow, one with the sun spot applied, one without, and then have it applied appropriately in the deferred rendering pipeline.



There are plenty of areas where the current envmapping techniques fall down, GT5 Prologue and even GT5 proper from screenies, has issues with cars reflecting things they shouldn't. Ie, the lower lip of an air dam at the front of the car reflecting clear blue sky despite it only really being able to reflect the radiator of the car...
I guess there are some things we just won't be able to do really nicely for a few more years until some kind of new method of realtime reflections comes in

Dave

 

[Some1]

Yeah, your problem was an easy fix in a way because the sun became occluded in the envmap for the car from an object a good distance off the car, however, it doesn't stop the problem as per the Lamborghini as pointed out... or any car with upward pointing (ish) polygons on the shaded side of the car

Only solution is as you say to have two envmaps somehow, one with the sun spot applied, one without, and then have it applied appropriately in the deferred rendering pipeline.



There are plenty of areas where the current envmapping techniques fall down, GT5 Prologue and even GT5 proper from screenies, has issues with cars reflecting things they shouldn't. Ie, the lower lip of an air dam at the front of the car reflecting clear blue sky despite it only really being able to reflect the radiator of the car...
I guess there are some things we just won't be able to do really nicely for a few more years until some kind of new method of realtime reflections comes in

Dave

I really think this is currently quite a low priority problem. Just use live envmap, it has performance hit, but you can specify the update interval etc. We should really focus on fixing bugs in the shadow system and lighting setups rather than think whether something should be reflected on the envmap or not.

Does anybody already have good values for the splitdist's and smCors? While the aliasing is not very noticeable in proper car-track setup, it is noticeable in my test track where I only have the standard shader with a white texture everywhere (a scrteen on previous page).

EDIT: Don't know, if it has been mentioned before, but I noticed, that the frustum or whatever used for shadow splitting doesn't properly follow the camera. Sometimes, the smallest split shadow map is in front of the car, sometimes behind, depending on the direction in world.

 

[Mr Whippy]

I really think this is currently quite a low priority problem. Just use live envmap, it has performance hit, but you can specify the update interval etc. We should really focus on fixing bugs in the shadow system and lighting setups rather than think whether something should be reflected on the envmap or not.

Does anybody already have good values for the splitdist's and smCors? While the aliasing is not very noticeable in proper car-track setup, it is noticeable in my test track where I only have the standard shader with a white texture everywhere (a scrteen on previous page).

I agree, it's not a big problem.

But just to clarify, the leaking off the envmap specular through to the shaded side of objects is NOT a problem that goes away by making the envmap live... the sun-spot is burnt into the envmap if the car is not in the shade, and then applied to the reflective objects.

In the example Alex showed, the whole car was in shadow thus the sun spot was occluded, so with live envmap on, the live envmap would occlude the sun spot, and problem solved.

When the car is just sat with one side to the sun and the other in shade, the shaded side upward pointing surfaces reflect the sun spot, despite them being well in the shade!


There are two problems really. Neither needs a fix right now, but there is a fundamental quality issue with having an object that is shading itself only having one envmap to choose from

The simplest example I can give is a car with a spoiler. The top of the spoiler will show the sun spot, but the shadow it casts onto the surface just below it, will also cast that sun spot from the envmap... there needs to be a second envmap for shadowed areas generally. Live envmap doesn't sort out that example

Dave

 

[Some1]

I noticed that the sun specular is drawn on the static car shadow (the occluded old style shadow) when it should really be occluded by the cars dynamic shadow.

Also, how hard it is to implement blending between cascade splits? See Blend between Cascades

My current CSM settings are:

racer.ini:

splits=3
splitdist0=5
splitdist1=25
splitdist2=100
splitdist3=1000

constants.cg:

const float smCor[SM_MAX_SPLITS]={0.0001, 0.0002, 0.0005, 0.0008};

It gives quite good results but the shadows are rendered quite close... maybe adding another split would solve this, but it certainly would hurt performance.

Right now, on big unoptimized tracks, my mid-end system gets about 20-30 fps...

 

[boomer541]

Tried sky with .tga texture in sky shader, several tries later reversing the alpha made it work BUT...





As you can see I'll use what works until something better come along!

tried Some1's CSM settings!

fps=125 fps=122 fps=100
splitdist0=5 splitdist0=10 splitdist0=30
splitdist1=25 splitdist1=80 splitdist1=200
splitdist2=100 splitdist2=200 splitdist2=800

dark square was 55, 115 and 470 meters from car as measured on Surfaces and sounds brake test area.

Then I tried changing the graphics with config.exe and the dark square dissappeared with shadowmapping=0 or covered the entire track with shadowmapping=1 and the water was all black.

I'll have to reset the racer.ini to the settings before I did the above to determine whats going on.

Interesting though!

 

[QuadCoreMax]

In the example Alex showed, the whole car was in shadow thus the sun spot was occluded, so with live envmap on, the live envmap would occlude the sun spot, and problem solved.

When the car is just sat with one side to the sun and the other in shade, the shaded side upward pointing surfaces reflect the sun spot, despite them being well in the shade!

Noticed the same...

Here's some other bugs :

- smoke inside cockpit
- flickering using CSM/Bloom
- Replays - I never managed to record & replay my races.