The second[1] C program I read in my life was Fractint[2]. It is called that way because it calculated fractals with integer math[3].
At that time integer math was not only faster than floating point, but the only thing available to most of us. Processors like the 80386 did not come with any floating point facilities whatsoever. The FPU was a separate unit (aptly called a coprocessor at the time) which was expensive and hard to obtain where I lived.
[1] The first one was DKBTrace, which lives on as POVRay. Thank you David K. Buck, where ever you are now. I learned a lot from you.
[3] Technically we should call it fixed point I guess, but in the end it was all done with the integer instructions of the CPU. So integer math is not completely wrong.
I still have the entire manual that I printed and bound circa 1996. I saw a program on PBS about fractals and we had just got dial up internet. I found fractint and cut my programming teeth with that and a copy of Borland that I convinced my parents to buy for me. Running on a 75MHz Compaq. Being a teenager at that time was just magical.
Edit: Just noticed you mentioned Povray too. Are you me? Wow, lol.
"Processors like the 80386 did not come with any floating point facilities whatsoever."
The 80286 needed a 80287 - I bought one to run AutoCAD (1MB RAM!)
I can't quite remember the exact order of events but either or both 80386 and 80486 had a co-pro initially. Then Intel re-invented the IBM magic screwdriver. They dropped the 80 prefix and added SX or DX and an i prefix for a laugh. SX had no co-pro or it was disabled and DX was the full fat model.
They really started to screw the punter with the 586 ... Pentium. The trend has continued - Core iBollocks.
I was going to correct you, and say the DX had a coprocessor, but the SX did not.
Apparently, that’s incorrect, and the difference is bus width.
However, at the end of its life, my 386 motherboard was running one of these at 100MHz, with a coprocessor on the daughterboard in the main CPU socket:
If I remember correctly the 486 was just basically, at first, the 386 with the FPU built in. There may have been a couple of new instructions? On that last part I am a bit fuzzy.
The 8086/8088/80186/80188, 80286 & 80386 all required an external chip for floating point.
The 80386DX was the full 32-bit data bus version. The 80386SX was a cost reduced version with a 16-bit data bus (akin to the difference between the 8086 & 8088).
The 80486DX had an FPU; the 80486SX had a disabled or missing FPU (depending on vintage). You could pair the 80486SX with an 80487 FPU, which was just a full 80486DX in a different pinout that took over from the crippled part.
While some Intel doco did start calling them Intel386 & Intel486, they didn't drop the 80 until Pentium. Intel lost a court case trying to prevent clone makers from calling their chips '80486' or whatever because the judge ruled you can't keep your competitor from using a number.
Weitek made floating point units that paired to the 386 and 486. Some motherboards supported this, and a small amount of software did (early Autocad).
Correcting one down thread, the 486 is more than just a 386+387. There a cache, it's pipelined, and there are some changes to the ISA. 4x the transistors.
Yeah, Fractint was great. The VGA color palette cycling was trippy.
In the mid 1980's I typed in a Mandlebrot set program from a magazine. It was for the Apple 2, written in BASIC. It took around 12 hours to generate a 280x192 monochrome Mandlebrot.
At that time integer math was not only faster than floating point, but the only thing available to most of us. Processors like the 80386 did not come with any floating point facilities whatsoever. The FPU was a separate unit (aptly called a coprocessor at the time) which was expensive and hard to obtain where I lived.
[1] The first one was DKBTrace, which lives on as POVRay. Thank you David K. Buck, where ever you are now. I learned a lot from you.
[2] https://fractint.org/
[3] Technically we should call it fixed point I guess, but in the end it was all done with the integer instructions of the CPU. So integer math is not completely wrong.