WWHC-Diaz + (NEW!) The VR Missions (help wanted!)

wildweasel · Post by **wildweasel** » Fri Sep 28, 2007 4:05 pm

Okay, finally got it fixed. Thanks!

Zippy · Post by **Zippy** » Fri Sep 28, 2007 4:15 pm

DoomRater wrote:Making it an int and multiplying it by 65536 so that it is correctly offsetted like the fixed point is.

'fixed point' is not really fixed point per se, it is actually an int that is divided by 65536 and then displayed with a decimal point. The script entry I posted above I tested this time so it should work- input 100 for normal speed and 35 for the shouldered speed.

Almost. ACS only supports integers. "Fixed point" is thus represented by using the high 16 bits of the integer as the integer part of the number and the low 16 bits as the decimal part of the number. Hence, 1.0 is 65536. That's why multiplying by 65536 works. It would be the same thing to multiply by 1.0. It would also be the same thing to bit shift it left by 16. In this sense, because you are limiting the digits on each side of the radix point, fixed point really is fixed point.

Code: Select all

1.0 == 65536 == 1 << 16

EDIT: minor typo.

DoomRater · Post by **DoomRater** » Fri Sep 28, 2007 4:35 pm

...isn't that what I just said? The only difference is when they're displayed- fixed point would display the result as 1.0 whereas int will display as 65536.

Unless you are saying how fixed point is being handled here really IS how fixed point is handled correctly, in which I can't contest that.

Zippy · Post by **Zippy** » Fri Sep 28, 2007 6:59 pm

It is basically fixed point. There is no real focus on specific number of digits on either side of the radix point like there might be in other languages, but I've never used a language that has fixed point as a primitive type so I don't actually know how common that is/was. Anyway, the fact still remains that you are just representing a decimal number as int16.int16 and so you have no floating point.

I apologize for any confusion. The way your last post read (or more specifically, Gez's post) made it sound like there was some kind of significant difference with integer and fixed point in ACS. Of course, there isn't any difference to ACS, and I believe when working with fixed point in ACS it is actually very important to understand that fact.

DoomRater · Post by **DoomRater** » Fri Sep 28, 2007 8:08 pm

Anyway, I looked around for some way to accomplish the same thing without ACS, but there is no such animal. Not until SetActorProperty becomes available to DECORATE anyway, which a ways off.

Matt · Post by **Matt** » Fri Sep 28, 2007 9:10 pm

@WW: One thing you can do is, instead of sending the information directly, create an array or (if the array doesn't work either) a set of if/else statements, that check the value of a variable and set the correct number accordingly, e.g., if 0 then SetActorProperty(...1.0...), if 1 then SAP (...0.35...), if 2 then SAP (...0.62...), etc.

DoomRater · Post by **DoomRater** » Fri Sep 28, 2007 9:23 pm

Actually, that wouldn't be a bad solution. I just happen to like being able to plug in the value faster as it means I wouldn't have to recompile should I want to change a number.

Matt · Post by **Matt** » Fri Sep 28, 2007 9:58 pm

Frankly, I suggested that method because all this "fixed point" crap confuses and scares the shit out of me. I've been staring at the wiki entry for about five minutes now and I still can't wrap my head around just why anything would be done like this except some extremely vague, assumed notion of "computer go faster with fix point lol"...

DoomRater · Post by **DoomRater** » Fri Sep 28, 2007 10:00 pm

It was done that way because there was no other way to add fractional numbers to ACS otherwise. ACS only understands the int type.

Matt · Post by **Matt** » Fri Sep 28, 2007 10:06 pm

...would 65536 happen to be represented in base-16 as 1000? Because if it does, I think I get it now.

EDITsage because I don't want to spam this thread with my own ignorance: Right, 10000 with four zeroes, thanks.

DoomRater · Post by **DoomRater** » Fri Sep 28, 2007 10:07 pm

10000, but yeah.

Zippy · Post by **Zippy** » Sat Sep 29, 2007 9:57 am

Another way to think of it is on the binary level. 65536 is 2^16, so as a 32 bit binary number:

Code: Select all

65536 = 00000000000000010000000000000000

Now, let's keep the data exactly the same, but change the way we look at it. Let's consider the high 16 bits separately from the low 16 bits.

Code: Select all

   high bits           low bits
0000000000000001   0000000000000000

Now in one sense what we have a two 16 bit integers. The first one has the value 1, and the second one has the value 0. So, if we stick a decimal inbetween them when we display the value... it would be 1.0. The binary representation hasn't changed at all. It is still 65536, if you look at it in that way. With the fixed point number representation, we're just looking at the 16 high bits as the integer part of the number (to the left of the decimal) and the 16 low bits as the fractional part of the number (to the right of the decimal.)

The reason one may choose to use fixed point is because it is MUCH faster than floating point, at the cost of precision. Fixed point works with regular integer math, which is quite fast. Also, because it works with regular integer math, you don't need to have a floating point unit on your CPU to use it. If you recall, Doom was coded in 1993. Getting as much out of the computers of the day as possible was paramount, and so fixed point was probably one of the choices they made to help retain engine speed and possibility compatibility, but I'm not one to be aware of exactly how frequent floating point units were at the time so that particular point may be moot.

Post by **Gez** » Sat Sep 29, 2007 1:55 pm

1993 was the time of "mathematic coprocessors" which allowed to perform floating point fast.

wildweasel · Post by **wildweasel** » Sat Sep 29, 2007 2:29 pm

Yeah, this is all really scary...

But enough of fixed/floating point stuff because the problem's been fixed. I'm going to make my next task importing the recoil system from Hideous Destructor.

Matt · Post by **Matt** » Sat Sep 29, 2007 3:32 pm

Know what might be a neat addition? An altfire that shoots a gun gangsta-style so the recoil pushes you sideways...

Also, I should note that the highest pitch (lowest number) that works properly for ZDoom is -5825, not whatever the number is on the version of HD I currently have uploaded.