|Re: 8051 assembler hgw@rht32.PCS.COM (h.-g. willers) (1991-03-12)|
|Re: 8051 assembler email@example.com (1991-03-14)|
|Re: 8051 assembler mzenier@polari.UUCP (1991-03-15)|
|Re: 8051 assembler firstname.lastname@example.org (1991-03-19)|
|From:||email@example.com (Barton Christopher Massey)|
|Keywords:||8051, assembler, design, lex|
|Organization:||Department of Computer Science, University of Oregon|
|Date:||Tue, 19 Mar 91 22:18:27 GMT|
In article <3502@polari.UUCP> mzenier@polari.UUCP (Mark Zenier) writes:
> But not Lex or Flex. The implementation of include files at the
> yacc grammar level caused a such a messy conflict with the lookahead
> in both Flex and Decus Lex, that I ended up rewriting the the lexical
> analyzer by hand. Lex source isn't portable, especially if you
> play with the input() macro.
Hmm. The Motorola 56000 family assembly language has the interesting
property that label addresses may be used in preprocessor expressions(!).
In spite of this, I recently completed a 56K assembler written using LEX and
YACC -- it's actually compatible with LEX or FLEX, and YACC, BISON, or BYACC
(obviously, I tend to use FLEX and BYACC, but it needed to be portable to
some weird environments, such as Stanford's V OS).
It required some ifdefs, but other than that, the lexer was actually quite
straightforward. I had no serious trouble modifying the input routines
to switch between raw input and input from preprocessor output buffers.
It meant that the lexer only looked ahead a line at a time, but that's
about the best one can do anyhow, I think...
Pushing information back through the *parser*, on the other hand, was a
real nightmare, and still isn't quite right. Unfortunately, 56K expression
syntax is non-trivial, and having a machine-generated operand parser was
a real win.
Anyhow, although I sympathise with the above-quoted paragraph, I strongly
recommend using a FLEX lexer for an assembler -- my experience was much more
Return to the
Search the comp.compilers archives again.