TCR Passive Damage Calculator

posted on June 10th, 2011, 7:43 pm

Floats can still be accurate to within about 7 digits in accuracy.  Which makes it a very precise approximation.

By the way, due to some technical difficulties with the database structure, I'm going to have to start over.  This time, I'm going to start with the passives, because there are fewer of them to deal with.

posted on June 10th, 2011, 8:01 pm

floats (aka a single precission floating point number as used by armada) are precise at around 2 digits. The 3 digit rounding we use in FOScript is already more precise than the game engine will handle reliably.

The formulas given in the guide are actually the reversed formulas as used in FOScript, the scripting language used in the Fleet Operations tools (yea i know, the name is very imaginative). But other subsystems of the balancing system are rounded as well for the GUI representation or filled with off numbers as they are gained from prediction. That's why your calculations are a bit off. But the difference is realy irrelevant. Wont appear in the game engine either way

posted on June 10th, 2011, 8:16 pm

The seven digits accuracy accounts for every digit used.  Not just the ones to the right of the decimal point.  The data range is "+/-3.8x10[sup]+/-38[/sup] with 7 digits of accuracy."  That's what my C++ book says about floats.

posted on June 10th, 2011, 8:47 pm

accuracy precision

a float can express numbers with 7 digits. but it cant express all values with 7 digits. thats why its precision is around 2 digits as a rule of thumb.

Thats the cause if z-buffer-flickering, energy-creation in physics engines and many many more its actually one of the common challanges game engines have to deal with

posted on June 10th, 2011, 9:10 pm

That's what my C++ book says about floats.

Maybe you should also recheck that book called bible! There are some things written down that many ppl misinterpret.

Regarding the passive calc...would be nice to have that tool!

posted on June 10th, 2011, 9:35 pm

accuracy precision

a float can express numbers with 7 digits. but it cant express all values with 7 digits. thats why its precision is around 2 digits as a rule of thumb.

Thats the cause if z-buffer-flickering, energy-creation in physics engines and many many more its actually one of the common challanges game engines have to deal with

Computer Science kitteh agrees - the dev is right.  Might want to look up the IEEE-754 definition for clarification.

posted on June 10th, 2011, 10:01 pm

accuracy precision

a float can express numbers with 7 digits. but it cant express all values with 7 digits. thats why its precision is around 2 digits as a rule of thumb.

Thats the cause if z-buffer-flickering, energy-creation in physics engines and many many more its actually one of the common challanges game engines have to deal with

So that's why my fence texture is flickering through the wall texture at my little quarter-mile oval.  I hate these computer limitations.  However, I'm still going to go for the maximum potential accuracy.

posted on June 11th, 2011, 2:15 pm

Thats why I wish they would add a Decimal data type to Java because i dont like floating point approximation. btw im not suggesting that games need to be that precise because that would quadruple the memory usage (decimals are 16 bit right?)

posted on June 11th, 2011, 2:28 pm

Floats are 4-bytes while doubles are 8-bytes.  I think there should be a decimal type as well in C++.

posted on June 11th, 2011, 2:48 pm

ehm, just for the small difference that they are incredibly slow

posted on June 11th, 2011, 2:49 pm

Decimals are slow?

posted on June 11th, 2011, 3:19 pm

yes. floating point numbers are processed in a single CPU instruction (there are floating point units on CPUs or GPUs), while decimals have to be processed programatically

posted on June 11th, 2011, 3:29 pm

Then the solution to that is simple, quadruple the size of the float and double data types.  It'll cause some incompatibility issues, but it'll greatly increase the maximum potential precision.  Otherwise, build faster processors that can handle repeated use of the decimal data type.

posted on June 11th, 2011, 3:41 pm

Float and double are fine, they are appropriate for the tasks they are designed for. If you really care about precision that much just used fixed point.