• Please review our updated Terms and Rules here

8087 speedup?

J

JoJo_ReloadeD

Experienced Member
Joined
May 12, 2007
Messages
120
Hi, i've just ported wolf3d to work with v20 / v30 / 80186 and with a 10mhz machine it's obviously slow :) so, do you think adding a 8087 would speedup the overall speed?

I think at least it would speedup the bottleneck of float calculation so... what do you think ?
 
How much of Wolf3d uses floating point?

The 8087 is far faster than using emulation (software) on the 8088. If only a small amount of the total time on the CPU involves floating point, then it won't make a difference. If a good deal of time (10% or more), then it will make a difference. Your mileage will vary.

If you don't want to bother with the 8087 (which many machines don't have), you could convert that code to use integer, or use FP emulation code that is 'sloppier' but faster. For a game both approaches might be good enough.
 
mbbrutman said:
How much of Wolf3d uses floating point?

The 8087 is far faster than using emulation (software) on the 8088. If only a small amount of the total time on the CPU involves floating point, then it won't make a difference. If a good deal of time (10% or more), then it will make a difference. Your mileage will vary.

If you don't want to bother with the 8087 (which many machines don't have), you could convert that code to use integer, or use FP emulation code that is 'sloppier' but faster. For a game both approaches might be good enough.
Click to expand...

Thank you for answering.

I'll see how much of float code is used, but i think it's done every frame, to calculate the player's relative position :)
 
I really didn't know wolf3D would run on an XT, I have to try that.

But it does need VGA, doesn't it? I've got EGA installed now.
 
If it's only a handlful of calculations per frame then it's obviously not a big deal. If it is a handful of calculations for each pixel in the frame, then obviously it is a big deal.

You have the code .. you should be able to analyze it and see how dependent on floating point it really is, or if those floating point operations can be substituted with integer operations instead.
 
You must log in or register to reply here.
Back
Top