The TORCS Racing Board
Username: Password: Remember Me?
Lost Password Register
Author: firechief | Created: 2017-09-18 03:30:19
Subject: Dirt-3 Review & Observations
As is tradition, here's my photo diary:

On a personal note, I do see that some people may (or may not) have installed some robots incorrectly. For example, Replicant failed qualifying for W-D and fa1, USR also failed qualifying for W-D and DanDroid for fa1. Why? We don't know, as the unexpected behaviour was ignored and the races went ahead :\

I'm glad there were enough of us submitting results to balance this out, but I would ask in future that if you see something unexpected like that, please raise it on the forums before proceeding - if you don't, you may not only be needlessly pulling someone else's results down, but you'll also deprive yourself of seeing the best possible race with every robot performing as its author intended.

There have been plenty of occasions this season where we've helped each other out by finding & including bug fixes, making tweaks to config files to prevent pit errors etc, downloading bots from google drive shares ... its created a really nice atmosphere. Lets keep it going :)

I also saw a wide disparity in qualy times for wdbee. For example, a second difference in wdbee's fast lap on Tim/W-D's machines versus mine, fa1's and Danny's. That's considerable - I'd really like to know why it happened that way, as I'm sure W-D would too. Maybe its something to do with you guys using antique Visual Studio versions? Though I didn't see the same thing with Tiger and Mouse.

Hint: I know both Tim & W-D need 2008 for other reasons, but I'm not sure if you're aware that multiple versions can be installed on the one computer - they do co-exist peacefully. I've not yet gotten TORCS building on VS 2015 or 2017 due to the 3rd party libraries in torcs/windows/lib being outdated, but 2013 works fine and can be downloaded from
Last Edited: 2017-09-18 09:32:27 by firechief
    Author: dummy | Created: 2017-09-18 07:07:31
    Subject: Re: Dirt-3 Review & Observations
    Thanks Andrew, very funny indeed and let the panel beaters also clean the dust, when they're at it ;)
    Last Edited: 2017-09-18 07:08:51 by dummy
    Author: wdbee | Created: 2017-09-18 17:48:04
    Subject: Re: Dirt-3 Review & Observations
    Hi Andrew,

    > a second difference in wdbee's fast lap on Tim/W-D's machines versus mine
    I assume, that it is just only second or third lap be accepted while qualifying ;).
    The car was driving at the limit and dirt is an amplifier for small differences.
    BTW: For me we are so close to the limits, that using tdble for TORCS schould be checked against double for all the calculations.

    Fixing bugs:
    (JSR-Code does not compile, Mouse-Code crashed TORCS, Tiger-Code crashed TORCS, Replicant setup with wrong names made TORCS not accept results). In addition three cars qualifying with 0.0 ... and there have been robots blocking the startline while trying to do a pit stop???

    My time is limted, sorry.

    Last Edited: 2017-09-18 17:48:04 by wdbee
      Author: firechief | Created: 2017-09-19 03:13:44
      Subject: Re: Dirt-3 Review & Observations
      > My time is limted, sorry.

      Fair enough - you're not alone with that problem ;) There certainly were a fair few issues this race. I've just (finally) gone through my code and fixed all the double->float warnings.
      Last Edited: 2017-09-19 03:13:44 by firechief
        Author: wdbee | Created: 2017-09-19 07:33:40
        Subject: Re: Dirt-3 Review & Observations
        Well, what I wanted to say is, that doing many of the calculations with that short datatype tdble instead of double in TORCS means that we get much more of theses small differences from computer to computer.

        The hardware is much faster now. We should check wether we can change the TORCS definition of tdble from float to double without lot of rework and wether it does reduce the differences we get.

        Last Edited: 2017-09-19 07:33:40 by wdbee
          Author: firechief | Created: 2017-09-19 08:52:01
          Subject: Re: Dirt-3 Review & Observations
          Yep that makes a huge amount of sense. I'd be very happy if tdble was defined as double & all the float nonsense could be removed.
          Last Edited: 2017-09-19 08:52:01 by firechief