What was the first publication to document what is (now) known as AT&T syntax assembly language?
As sampablokuper (OP) has noted, the UNIX Programmer's Manual cites PAL-11R as the source of the syntax, with only minor changes. So I've found a published book from May 1971, the PAL-11R Assembler Programmer's Manual that describes the syntax. Given that the syntax was probably around in 1969, this is still almost definitely not the earliest.
The earliest I can find with a casual web search is the UNIX Programmers Manual of November 1971 (better version). If you scroll to page 195 (page 5), you'll see this:
173700: mov $177472,r0 12700;177472 mov $3,—(r0) 12740;3 mov $140000,—(r0) 12740;140000 mov $54000,—(r0) 12740;54000 mov $—2000,—(r0) 12740;176000 mov $5,—(r0) 12740;5 tstb (r0) 105710 bge .—2 2376 jmp *$5400Q 137;54000 173740: mov $177350,r0 12700;177350 clr —(r0) 5040 mov r0,—(r0) 10040 mov $3,—(r0) 12740;3 tstb (r0) 105710 bge .—2 2376 tst *$177350 5737;177350 bne . 1377 movb $5,(r0) 112710;5 tstb (r0) 105710 bge .—2 2376 clr pc 5007
That, if I'm not mistaken, is AT&T syntax!
The earliest paper on UNIX, according to this source, is this paper. This document is from July 1974, so it's no good as a first, and does not include any AT&T syntax anyway. It does, however, state that the first version of UNIX was made circa. 1969–1970. This isn't necessarily a limit on the earliest that the syntax could've been described in a publication, since it might predate UNIX; I don't know.
Thanks! I had not previously seen
movused by itself in AT&T assembly code (always
movl, etc.), but that does not mean
movwas (is?) not valid AT&T syntax. (I am new to the syntax.) I will upvote this and provisionally mark it as correct. If any other answers come along, with earlier examples, I will move the "correct" mark accordingly.– user11467Dec 30, 2018 at 23:27
1On page 21 of the first PDF in your answer, the man page for
asis based on the DEC-provided assembler PAL-11R, although it was coded locally. Therefore, only the differences will be recorded [in this man page]." The differences seem fairly minor. So, it seems that what we now call "AT&T syntax" is largely due to DEC rather than AT&T. Interesting.– user11467Dec 31, 2018 at 0:23
#foofor an immediate operand in -11 assemblers; the Unix guys apparently preferred
$foo, which to my eyes lacks any mnemonic significance. Similarly
*. Does anyone know the rationale or is this just personal taste? Dec 31, 2018 at 15:19
2@another-dave Ask a new question and more people will see this to be able to answer it.– wizzwizz4 ♦Dec 31, 2018 at 15:47
1@another-dave: # and @ were apparently hard to type (Unix Assembler Reference Manual §8.1 - Dennis M. Ritchie, referenced in my answer of stackoverflow.com/questions/4193827/…), fuz says "As the UNIX tty driver used @ and # as control characters, $ was substituted for # and * for @. at stackoverflow.com/questions/42244028/…– ninjaljJan 1, 2019 at 13:06