Re: [glob2-devel] Nicowar and desync

[Bradley Arsenault]

On 7/18/06, Stéphane Magnenat <address@hidden> wrote:
> > There is no float usage in Nicowar, I assure you. And it does use more
> > CPU, but I doubt thats the cause unless its litterally lagging your
> > machine.
>
> grep float AINicowar.h AINicowar.cpp :
>
> AINicowar.h:                            GradientPoll(Gradient::Sources source,
> Gradient::Obstacles obstacle, float weight) : is_null(false), source(source),
> obstacle(obstacle), weight(weight), min_dist(-1), max_dist(-1) {}
> AINicowar.h:                            float weight;
> AINicowar.cpp:          average_unit_strength_level=static_cast<unsigned
> int>(std::floor(static_cast<float>(average_unit_strength_level)/static_cast<float>(total)+0.5))-1;
> AINicowar.cpp:          average_unit_strength_level=static_cast<unsigned
> int>(std::floor(static_cast<float>(average_unit_strength_level)/static_cast<float>(total)+0.5))-1;
> AINicowar.cpp:                  float score=0;
> AINicowar.cpp:                                  const float weight =
> CONSTRUCTION_FACTORS[building_type][factor].weight;
> AINicowar.cpp:
> score+=static_cast<float>(top_left)*weight;
> AINicowar.cpp:
> score+=static_cast<float>(top_right)*weight;
> AINicowar.cpp:
> score+=static_cast<float>(bottom_left)*weight;
> AINicowar.cpp:
> score+=static_cast<float>(bottom_right)*weight;
> AINicowar.cpp:                  tp.percent=static_cast<unsigned
> int>(static_cast<float>(counts[i])/static_cast<float>(num_buildings_wanted[i])*100);
> AINicowar.cpp:  float devisor=1;
> AINicowar.cpp:          devisor=static_cast<float>(max)/16.0;
> AINicowar.cpp:  float
> average_happiness=static_cast<float>(total_happiness)/static_cast<float>(total_units);
> AINicowar.cpp:                          float
> enemy_average_happiness=static_cast<float>(enemy_happiness)/static_cast<float>(enemy_units);
>
> It seems it does. Don't forget floating point format does not constrain
> handling of least significant bits, so different CPU (even both x86) can give
> different results. We had this bug in glob2 earlier in the same type of
> computation Nicowar is currently doing.
>
> I agree for the CPU.
>
> Steph
>
>
> _______________________________________________
> glob2-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/glob2-devel

Yeah, Kai already emailed me with a grep. I, unfortently, am not at my
house, and I can't work on glob2 for atleast three weeks while my
family moves. Worse yet, me might have to get modem-internet in our
next house. Anyway, all of these usages are independant casts, so in
my case, atleast the same CPU will give the same results every time.

I'll fix this as soon as I have time. Sorry everyone, my mistake.

--
Start and finish, Bradley Arsenault

ps: I guess the unstable tag stays on