Floating point addiction is the belief, by inexperienced programmers (most notably typed languages, not languages like Perl, PHP, and, in some cases, VB/ASP), that floating point numbers (floats, doubles, etc) are better than integer numbers (intss), because they can do elite things like decimal places, and are therefore more accurate.

It totally ignores the fact that, due to the design of modern processors, with the FPU being the most complex, and, in the case of multimedia, the most stressed part of the CPU, floating point math is usually drastically slower than integer math by design. While floating point numbers are necessary sometimes, the fact that people use signed floating point types when adding 8 and 14 is usually considered blatantly stupid and kludgeworthy.

This is true, mostly. However, the FPU is not the most stressed part of the processor just because you're playing an MPEG and an MP3 in the background. The whole processor is dedicated to running one program at a time, in most cases, and the computers that can make exceptions to that rule (Cray MTAs, etc) don't get programmed in VB.
Also, all floats are signed.

That said, for well-pipelined processors like the PowerPC, doing integer stuff in the FPU can be a 1337 assembler trick. More pipelines create more parallelism.

Log in or register to write something here or to contact authors.