Racer v0.8.27 released!

[Ruud]

Hopefully quite close to v0.9RC1.

Get it at http://www.racer.nl/download/racer0.8.27.zip (75Mb)

The changes:
- move_cg=1 didn't offset correctly: cameras, wipers, generic models, brake calipers, exhausts, dials, helmet, carlights Z/far.
- 'rewind' commands for live rewinding (if RTD logging is on)
- Default font made a little bigger again
- onspeedlimiter.rex gets executed when the speed limiter is turned on.
- Shadowmapping tweaked for director and other cameras.
- Fuel consumption reset to 0 when standing still (was 99.9 L/100km).
- A missing controller (defined in a controls file) would crash Racer
- Using something other than default.ini in racer.ini's ini.controls now works (can be saved)
- 3 profiles available in the controller setup screen.
- The names above the cars were black - fixed with floatingname_*.cg shaders.
- The wheel blur models would show in the car selection garage.
- Added grass smoke particles
- Select car/track screens sped up by loading only part of the car.ini/track.ini files. Assumes 'car' is at the top though (!).
- Selecting a car, then exiting would crash.
- When cars stood still, too much was damped away and the differential wheels would rotate a bit.

 

[Mr Whippy]

Perfect case of needing to read the manual

That said, we have all missed things... I know many have only just started getting the best from the FF due to control panel settings!


One other thing to ask about. DOF is a nice effect, and we had it working at one stage, and it was rather nice. As was the scene motion blur.

Neither are essential, but, they make replays look really damn good

Are these relatively cheap via CG and full screen shaders, or costly?

Thanks

Dave

 

[Ruud]

Motion blur is quite expensive. It's there with fs_filter1=bloom_shadows_blur_f.cg and velocity_map=1 in racer.ini. But there's some kind of aliasing/moire going on here.
Also, ghosting appears, so motion vs non-motion should be detected (to avoid moving areas to blend into non-moving areas or vice versa). I won't polish that for v0.9. The possibilities are there though for people to jump in and create motion blur postprocess shaders. It will be costly.

I'm looking into the splitdist2=500 thing. Blur=1 also hides some aliasing, but blur=0 looks good already compared to existing games. Perhaps only turning blur on in cockpit views might work (although not that elegant to implement it seems). I'll try and make it fade. Strangely, I get no performance increase with blur=0 vs blur=1. Probably too much other bandwidth that is the bottleneck.
BTW Are you sure the 4 splits are working? ;-) In shadowmapping.cg, there is in fact a hardcoded splits=3 method. A bit faster.

 

[Mr Whippy]

Profiles work ok, but the menu buttons didn't seem to work sometimes, you'd press them but nothing happened. Trying to set live envmap stuff, fullscreen, etc, no joy. Buttons just sat there dead.

When you get to the friction sliders etc, they don't seem to be interactive anymore, the result of changes isn't seen till reboot of Racer... from what I can tell anyway. These also become non-responsive often.

G25 profile is way out. Steering axis and pedal axis seem wrong. 0-1000 rather than -1000 > 1000 etc... so I can't steer right, but can left etc, or throttle/brakes have big dead-zones at each end (sure my wheel is set up right in control panel)
Also frictions/inertias etc are very low default. Are you checking these across wheels so all wheels 'feel' about the same?



I'm not sure if you have ever seen a shadow, but they don't look like Racer draws them

They are sharp, they don't go blurry 25m away, and generally don't look all nasty.

My settings on the other release thread (26?) looked tons better.

Right now a car 50m away from the camera literally has a shadow so diffused from blur on a sunny bright day, that it is more towards diffused than it is an actual shadow. It looks like bad fake AO, rather than a shadow.

4 splits, 10 40 160 640 (add a fade on that last one and it'll look nice), blur off, and they look(ed) ace.

If you could reduce blur amount it would help, it simply destroys the shadows on bigger splits. The blurring is nice up close when details are covering lots of pixels, it removes the jaggies, great... but at a distance that bunch of pixels can be a whole car and it's simply blurred out of existence!



Hmmm, looks like a fade has been added to the last split, but it's set between the 2nd/3rd split distances? Any chance this can auto-detect the split count and fade between the last 2 linearly? I can't get a 1000m 4th split to work with splits set to 4!


GT5 is bad, but these are worse

Is there any easy way to remove the fade between the 2/3 splits, and move it to the 3/4 split? Or 6/7 split if I so wish? (users with high end gear might want this so hard coding that fade seems bad)

I don't think we'll get nice shadows with just 3 splits, and a 4th doesn't seem costly here... as per 0.8.26 release example settings I posted, it looks so much better.

Dave

 

[Some1]

I've noticed something very strange going on with lights, see the video attached.
In the video, the time is set to 00:00, and the car lights are turned on, but they don't appear at first. As I start to accelerate and drive, suddenly the lights appear on the track surface, then disappear again. And this is repeating itself over and over again. I've also noticed, that either the fog or something is also blending in-out occasionally, as I drive.
I'm not sure, if it is the particular track setup causing it or it is a shader bug somewhere...

Also, here's the link to the track for testing: http://www.mediafire.com/file/u5s280kx9mfm1rk/desert.zip

 

[Mr Whippy]

I noticed that last night too, but with track shadows vs car shadows.

Drive alongside a row of buildings with gaps between them in your car, and the buildings are casting shadows over the track and shadowing the car, but the car isn't shadowing the track between the building shadows.
Ie, you are driving in and out of the light, but the car isn't casting shadows when it's in the light... well not until the 1st or 2nd split anyway.


This is probably to do with that split fade thing...?!

 

[Ruud]

Profiles work ok, but the menu buttons didn't seem to work sometimes, you'd press them but nothing happened. Trying to set live envmap stuff, fullscreen, etc, no joy. Buttons just sat there dead.

I'll check out the menu at home (if I ever get there; so much snow here now).
As for the shadows, I guess blur is a bit personal. I find blur=0 to be quite aliasing. I've modified the split distances and set splits to 4.
Every flexibility there (fading and such) is nice but is costly, so I'd rather keep performance rather than keeping a lot of variables being used for every pixel onscreen. In the future we'll need to generate our own shaders, so semi-constant vars (like split count) can perhaps be baked into generated shaders to get maximum performance. Anyway, after v0.9.

The splits=4 on your side may trigger a fade effect, but there is none in the code. There's only fades inbetween the splits themselves (like the outer 10% of the shadow maps get cross-faded). The last split fade is in now, which looks gentle.

Things like blurring only split0 seem a bad idea; all things that change are more noticable than things that look a bit worse, but consistently.
I've taken your 4 split distances, hope to get through the snow now.

 

[Ruud]

I've noticed something very strange going on with lights, see the video attached.
In the video, the time is set to 00:00, and the car lights are turned on, but they don't appear at first. As I start to accelerate and drive, suddenly the lights appear on the track surface, then disappear again. And this is repeating itself over and over again. I've also noticed, that either the fog or something is also blending in-out occasionally, as I drive.

That track looks VERY dark. Exposure here goes up to 999.0. That may trigger floating point errors (HDR can't store values >65000 for example). Try giving it some light.

 

[Some1]

That track looks VERY dark. Exposure here goes up to 999.0. That may trigger floating point errors (HDR can't store values >65000 for example). Try giving it some light.

I don't think it is related to that. I disabled auto_exposure, changed the time to 0600 (to still see the headlights) and they still act the same way.

It looks more like some value is either overflowing or "spinning"...

Try out the track yourself, you should get the same effect.

 

[Mr Whippy]

I don't think it is related to that. I disabled auto_exposure, changed the time to 0600 (to still see the headlights) and they still act the same way.

It looks more like some value is either overflowing or "spinning"...

Try out the track yourself, you should get the same effect.

Ruud, this looks very similar to what I have seen here... but with shadows... this is probably what I am seeing as a fade in the shadow splits but isn't!

Get a track where there is a row of buildings maybe 20m apart, so they cast shadows over the road in blocks. Then drive along that road into and out of shadow, but watch from a track camera from about 100m away.
The buildings cast shadows, but when the car comes into the light it isn't casting shadows until it's very close. When it does cast shadows, it fades in from nothing to 100% over about 25m or so.


Looks just like Some1's issue there, but with shadows.


Will post a video later.


Something fishy going on.


As per hard coding splits etc, that makes sense if it's cheaper on the hardware when doing shading. BUT, while we are still tweaking and tinkering it makes sense to not hard code anything until we are all happy? That way we can tinker around with the settings to try improve it BEFORE it's too late

Dave

 

[Ruud]

I don't think it is related to that. I disabled auto_exposure, changed the time to 0600 (to still see the headlights) and they still act the same way.

It looks more like some value is either overflowing or "spinning"...

Try out the track yourself, you should get the same effect.

It's the vertex spacing. Projected lights calculate light falloff per vertex. And the vertices are very wide apart; tesselate the track and it'll be fine.

 

[Mr Whippy]

Here is my issue Ruud.

Note the trees are shadowing at a distance of about 100m from the camera, but the car stops shadowing (it fades out) well before the last split, about 40m from the camera.



Dave

 

[krsman3d]

There is something I wanted to know that has been bothering me for years.

When a car is FWD, I set up its brakes but they never feel as they should because the rear wheels keep getting locked up while braking and ABS is not really able to solve this. As soon as I set the car to be RWD (braking forces, weight distribution... are not changed) it feels much better. The car brakes better and there is no locking up, as if something is changed.

Can this perhaps be solved?

 

[Stereo]

Things like blurring only split0 seem a bad idea; all things that change are more noticable than things that look a bit worse, but consistently.
I've taken your 4 split distances, hope to get through the snow now.

Didn't completely trace through the code but it seemed like this was happening - something like

iSqrtSamples = max(iSqrtSamples-index, 2);

So as the index of the shadow split increases, it blurs less? (iSqrtSample = 0 is no blur)

 

[Mr Whippy]

There is something I wanted to know that has been bothering me for years.

When a car is FWD, I set up its brakes but they never feel as they should because the rear wheels keep getting locked up while braking and ABS is not really able to solve this. As soon as I set the car to be RWD (braking forces, weight distribution... are not changed) it feels much better. The car brakes better and there is no locking up, as if something is changed.

Can this perhaps be solved?

The rear brake force could be too strong.

Or, there is a weakness in the logic of the ABS right now. If the rear wheels lock up first, then the front wheels can lock up afterwards.

The best thing to do would be to reduce rear braking force to half what it is now, then slowly work back up until the car starts to feel bad again, then back off a little bit.


Ideally, ABS would use each wheels slip ratio, or pairs of wheels on each side of the car, which would be multi-channel EBD, or old two channel ABS respectively.
What we have simulated now isn't really ideal at all when compared to any road car implementation of ABS.

Dave

 

[krsman3d]

With braking_factor=1, 60/40% weight distribution I have to set the front wheels braking forces to 1900 and the rear wheels to 300. Only this stops the rear wheels from locking up, but I don't think that a real car uses such small amounts of rear wheel braking.

 

[Mr Whippy]

That does sound very low.

What is the CTRL 9 debug screen readout for the front/rear weights under braking? It sounds like a lot of weight is moving to the front under braking perhaps?

Best to put a thread up under the physics problems section, then we can discuss at length and not clog up this thread

Dave

 

[DavidI]

Sorry, Ruud you only half-nailed the old standing still on a hill issues. The tyres don't rotate by themselves anymore but....
I think a few pictures describe it better. two shots below, Lambo was stopped a few metres from the start line on Mugello with brakes full on (using mouse for controller, note the cursor bottom of screenie), took a screenshot, waited about two minutes, took another one. Nothing else was touched or moved, except you will notice the car slid down the (tiny) incline a foot in those two minutes. Happens on any track with any car I've tested.

[edit] More info, ctrl-1 info shows that grip on all wheels seems to oscillate in cycles back to almost zero very quickly when standing still, smallest amount of movement seems to settle the oscillations. The cycling is too fast for me to see exactly what values it goes to & from, but the grip value seems to get very low in fast cycles. [/edit]

The tar-tyre-smoke IS definetely working heaps better, whatever you didn't do worked wonders.

 

[Mr Whippy]

That standing still thing is a funny problem.

I'm easy about it so for all I mind it's just a zero oscillation problem and it's not damped perfectly.


However, purely as an observation, the previous behaviour was that on a hill brakes off, the wheels would turn and spin against the ground and quickly be followed by the car moving forward with them.
Now the behaviour is that the wheels turn and slip against the ground at maybe 5cm/s until the car begins moving...

The improvement is that a car is more steady when stood still, but the result is that it can look more weird now as the wheels have to get going faster before the car will move.


I'll take a video of it if needed.

As said though, I'm easy on this. I imagine when forces/frictions get very small, and the forces are or need to be equalling out and balanced, things can get a bit hard to get perfect.

Dave

 

[ammods]

It has been there from the very beginning. Just select an external car camera (keys 1-9) and then use the numpad to rotate: 4 and 5 - left and right, 2 and 8 - up and down, 1 and 3 move closer or away and with 'plus' or 'minus' you can increase or decrease the FOV.

cheers bud, i didnt know that, that will come in handy cheers for the advice

 

[Alex Forbin]

cheers bud, i didnt know that, that will come in handy cheers for the advice

I still really want some sort of mouse-look. Most of us use a wheel for driving which leaves the mouse free.

Alex Forbin