Thou shalt have no other gods before the ANSI C standard 1360

rpl

There is no simple answer. The basic principle is that computers process data as *aggregations* of bits all at the same time; as an analogy think of a cable with multiple wires all carrying signals in "parallel". The number of such "wires" within the computer hardware is the number of "bits" we attribute to that computer. This is complicated by the fact that in most modern computers, there are *different* "bit widths" used for different parts of the computer; typically at the memory-chip level, the width is 8 or 16 bits, while within the computer's "instruction set" a variety of widths (8, 16, 32, and maybe 64 bits) *for data access* may be supported. Opcodes may be variable width, or perhaps all the same width, anywhere from 8 to 32 bits. When you hear of a *single* bit-width being used to describe a computer, usually it refers to the width of the data paths through the integer Arithmetic-Logic Unit as seen at the Instruction Set Architecture level; this usually is the same as the "integer register" width. (Often, floating-point uses different registers and ALU.)

At the OS level, it is more likely that what is being referred to is the width of such values as *file sizes*, with integer representations that the hardware can use directly with the ALU.

That these distinctions are somewhat arbitrary is shown by the fact that Solaris for the SPARC architecture has for quite a while been available as both "32-bit" and a "64-bit" systems.

Register width has almost nothing to do with vectors.