vacca@burdvax.UUCP (04/18/84)
I am trying to determine the percentage of users (in the following
three environments) who actually use assembler to some degree:
1. UNIX systems programmers
2. UNIX applications programmers
3. students and faculty in a university UNIX environment
The question is whether or not C is a viable alternative to using assembler.
It seems that the SysV manuals are trying to dissuade UNIX programmers from
using assembler. Is this true?rej@cornell.UUCP (04/19/84)
C can replace assembler. The only use for assembler on VAXen is to write the context switching routines and to improve the speed of a few commonly used subroutines. People that I know programming in C on a 68000 never seem to use assembler. The Cornell Program Synthesiser is written in C and runs on many PDP-11s across campus. Its device drivers and overlay routines are written in C, with only the process switching routines being in assembler. Ralph Johnson cornell!rej
willc@iuvax.UUCP (04/20/84)
I agree that there is very little need to program in assembly language under Unix. In our department, the systems programmers generally use C and the faculty generally use languages such as Scheme (a dialect of Lisp). I recently chose to build the kernel of a Scheme processor in Motorola 68000 assembly language rather than in C. The kernel interprets a p-code, so the assembly language code amounts to microcode for a hypothetical machine. For this application, which involves considerable bit-diddling on specific hardware registers, I believe assembly language is easier to work with than C. Such applications are extremely rare. William Clinger willc!iuvax...
laman@sdcsvax.UUCP (04/21/84)
Also used in the C library for the system interface routines, but that's
not the kernel...
Nooooo.... Not ANOTHER set of interface routines.....
Mike Laman
UUCP: {ucbvax,philabs,sdccsu3,sdcsla}!sdcsvax!lamangrunwald@uiuccsb.UUCP (04/23/84)
#R:burdvax:-159300:uiuccsb:13700002:000:443 uiuccsb!grunwald Apr 23 09:47:00 1984 I fall under (3), and I use a small amount of assembler every once in a while. I'm working on a version of the Berkley Pascal compiler which allows multiple processes. There's a small amount of assembler needed to handle context swaps, and that is essentially used as a co-routine call between C/Pascal programs. The only other case is where I hand-code a routine such as "bcopy" to copy blocks of data using the VAX special instructions.
berry@zinfandel.UUCP (04/26/84)
#R:burdvax:-159300:zinfandel:8000007:000:465 zinfandel!berry Apr 23 16:15:00 1984 In the four or so years I have been working with C and UNIX, I have written 2 (count 'em, two) assembler routines. Both were to wring the most out of the innermost loop after profiling. One was in fact, merely a version of strcmp that didn't save any registers, and used some volatile ones instead of the ones the C compiler picked. That got an extra 12% speedup from a cross-loader. Berry Kercheval Zehntel Inc. (ihnp4!zehntel!zinfandel!berry) (415)932-6900