Re: [TUHS] Array index history

On 6/9/2017 5:21 PM, Steve Johnson wrote: Being an assembler snob for a number of years when I first started programming, having started with MACRO-10 on a PDP-10 in high school, I was reluctant to consider C as an alternative. At one point early on, I decided I would see what the difference was in actual performance. So, on an IBM-AT, I wrote two programs. Both did a wire-frame rotation of a 3D cube, rotating around both the X and Y axis. They both used my own graphics library written in assembler that used the basic BIOS graphics calls to draw lines in 3D. One program was written in straight Intel 80286 assembler. The other, in C, C86 from Computer Innovations, Inc. The straight assembler version, I wrote using double-word integers. One word for everything above the decimal point, one word for everything below the decimal point. Now, because of the loss of precision, and that I defined the matrices for rotation as static integers, after a few hundred iterations, the cube would start to deform but it was a proof of concept. The C version, I used real sin(), cos(), but - the interesting part is that I didn't have a floating point coprocessor - so the floating point was emulated in the supplied math library. The assembler version ran at a ratio of 3:4 compared to the C version.Meaning if the assembler version did X rotations in 3 minutes, the C version did it in 4 minutes. BUT - the C version cube didn't deform over time, because the accuracy was intact. After that, I gave up on assembler except where critical. C as a concept did what I needed it to do ;) On topic for this conversation is the fact that arrays in C are basically a perfect mapping to assembler. if you moved from assembler to C like I did, there was no concern over "arrays should start at 1". It was just intuitive anyway. On the other hand, if you went from some other language to C, well, I can understand the alien-ness. art k.