Subject: sizes of bytes [was: Re: IO in R6RS]
From: (Rob Warnock)
Date: 1998/06/11
Newsgroups: comp.lang.scheme
Message-ID: <6lni1a$>

Kumar Balachandran <kumar*xspam*@*xspam*> wrote:
| (Was it IBM where "byte" was never supposed to be used without the number
| of bits).

You're probably thinking of the DEC PDP-10, where a "byte" could be
(dynamically) *any* width between 1 & 36 bits, and writing programs
required using (at least) two different character sets, 7-bit ASCII
(packed 5 to a word) for text files & terminal I/O and a 6-bit character
set (packed 6 to a word) for communicating filenames to the operating

Look at Common Lisp's "LDB" and "DPB" library routines. Nearly *exactly*
those functions existed as single PDP-10 instructions of the same name.
E.g., here's some code to convert a filename in null-terminated 7-bit
ASCII (a.k.a. "ASCIZ") to SIXBIT [leaving out all the error checking
for name-too-long and illegal-SIXBIT-char]. My apologies if this isn't
exactly correct MACRO-10 syntax; it's been quite a while...

	; Call with a0 := either a conventionally pre-decremented byte-pointer
	; to an ASCIZ string or a word address; returns v0 := SIXBIT equiv.
	; Caveats: No error checking. Modifies a0.

	a2six:	movei	v0, 0		; initialize result
		move	t0, [xwd 440600,v0] ;and a pointer to it
		tlne	a0,-1		; pointer bits already set in arg?
		hrli	a0, 440700	; no, point to non-existent prev. byte
	a2six0:	ildb	t1, a0		; get next (or first) byte
		skipne	t1		; null?
		popj	p, 0		; yup, return.
		trc	t1, 40		; complement (what will be) the MSB
		idpb	t1, t0		; insert into result [in *v0*!]
		jrst	a2six0		; and loop



Rob Warnock, 7L-551
Silicon Graphics, Inc.		Phone: 650-933-1673
2011 N. Shoreline Blvd.		FAX: 650-933-4392
Mountain View, CA  94043	PP-ASEL-IA