Newton's method in QBASIC

[nkeck72]

Some talk about the speed of math functions in Commodore BASIC in another thread inspired me to write a quick-and-dirty implementation of the calculus concept of Newton's method in QBASIC as a sort of benchmark for DOS PCs this morning. If you're not familiar, Newton's method is an algorithmic method for estimating zeroes (where a graph crosses the x-axis), and requires a guess as to where this occurs. The definition of this method is:

x[SUB]n+1[/SUB]=x[SUB]n[/SUB]-(f(x[SUB]n[/SUB])/f'(x[SUB]n[/SUB]))

where f(x) is the function you wish to find the zero of, x[SUB]n[/SUB] is the current guess, x[SUB]n+1[/SUB] is the calculated next guess, and f'(x) is the derivative of f(x).

Feel free to try it out and report back. This is by no means a good example of BASIC or QBASIC code and could probably use some optimization and cleanup, I only wrote this in about an hour or so. Everything uses double-precision internally to prevent precision loss so I'd be interested in particular to see the differences with and without an FPU installed.

Full disclosure, I wrote this in Windows 98 so I don't know how well this will run on a pre-286. YMMV. Feel free to direct me to another forum category if QBASIC doesn't run on a < 386.

 

[Trixter]

A couple of random responses:

- This does more to benchmark BASIC implementations than it does to benchmark a system, unless you specify exactly which qbasic interpreter or compiler you wrote it for
- QBASIC runs on every system
- The version of QBASIC that comes with DOS 6 doesn't use floating-point IIRC -- I thought that was only a feature of the compilers, but I could be wrong

 

[krebizfan]

QBASIC supports floating point. It does not support math coprocessors though which was reserved for the compilers.

 

[nkeck72]

[quote Trixter]This does more to benchmark BASIC implementations than it does to benchmark a system, unless you specify exactly which qbasic interpreter or compiler you wrote it for[/quote]

Yeah, I guess that's what I meant. Sorry for the confusion.

[quote Trixter]QBASIC runs on every system
- The version of QBASIC that comes with DOS 6 doesn't use floating-point IIRC -- I thought that was only a feature of the compilers, but I could be wrong[/quote]

I didn't know QBASIC ran on everything, but the QBASIC that comes with DOS 6 definitely does use floating point, that's what I wrote it on.

[quote krebizfan]It does not support math coprocessors though which was reserved for the compilers.[/quote]

That I didn't know, but it does make sense I suppose. I have a copy of Borland C++ that hasn't seen some attention here recently, I might just fire it up and write an implementation in that and compile it specifically for benchmarking FPUs.

 

[Trixter]

I didn't know QBASIC ran on everything, but the QBASIC that comes with DOS 6 definitely does use floating point, that's what I wrote it on.

Apologies to you and krebizfan; I meant that it doesn't support the 8087 math co-processor unless you're using a compiler, as krebizfan confirmed. All versions of Microsoft BASIC in all forms support floating-point routines.