GAS parser

GAS parser

[Ivan Pryanishnikov]

Hi,

I'm going to port binutils to a new target and looking at
the organization of the package.  

I have a question concerning GAS: some ports (eg, bfin) use 
lex/bison extensions (eg, bfin-lex.l, bfin-parse.y) to write their
parsers.

I was wandering about the reasons behind; possibly some of them
are:
 - it's easier (generally) in some sense
 - it's not possible to solve their parsing problems in a standard way
 - they had some non-gnu assembler and just adjust/reuse it for the 
   binutils assembler (because it's easier than to write from scratch)
 - ??

Could you please comment this issue.
Thanks,
Ivan

Re: GAS parser

[Ian Lance Taylor]

Ivan Pryanishnikov <prianich@complang.tuwien.ac.at> writes:

> I have a question concerning GAS: some ports (eg, bfin) use 
> lex/bison extensions (eg, bfin-lex.l, bfin-parse.y) to write their
> parsers.
> 
> I was wandering about the reasons behind; possibly some of them
> are:
>  - it's easier (generally) in some sense
>  - it's not possible to solve their parsing problems in a standard way
>  - they had some non-gnu assembler and just adjust/reuse it for the 
>    binutils assembler (because it's easier than to write from scratch)
>  - ??

I don't know about bfin.  For m68k Ken and I wrote a Bison parser
(though not a flex lexer) because the assembler needed to support two
different syntaxes, and the cases were sufficiently complex that it
seemed worthwhile to express them in a grammar, to minimize the chance
of error.  So I guess that your option 1 applies for the m68k.

Ian

Re: GAS parser

[l l]

Hi,

Little offtopic but, IMHO, self host(self compile assembler like fasm)
make gas more portable.

TIA


2007/8/2, Ivan Pryanishnikov <prianich@complang.tuwien.ac.at>:
> Hi,
>
> I'm going to port binutils to a new target and looking at
> the organization of the package.
>
> I have a question concerning GAS: some ports (eg, bfin) use
> lex/bison extensions (eg, bfin-lex.l, bfin-parse.y) to write their
> parsers.
>
> I was wandering about the reasons behind; possibly some of them
> are:
>  - it's easier (generally) in some sense
>  - it's not possible to solve their parsing problems in a standard way
>  - they had some non-gnu assembler and just adjust/reuse it for the
>    binutils assembler (because it's easier than to write from scratch)
>  - ??
>
> Could you please comment this issue.
> Thanks,
> Ivan
>

RE: GAS parser

[Dave Korn]

On 02 August 2007 14:53, Ivan Pryanishnikov wrote:

> Hi,
> 
> I'm going to port binutils to a new target and looking at
> the organization of the package.
> 
> I have a question concerning GAS: some ports (eg, bfin) use
> lex/bison extensions (eg, bfin-lex.l, bfin-parse.y) to write their
> parsers.
> 
> I was wandering about the reasons behind; possibly some of them
> are:
>  - it's easier (generally) in some sense

  I think that's the main reason for m68k.  There are a couple of syntaxes and
a lot of variations and it's easier than hand-parsing all the operands.

>  - it's not possible to solve their parsing problems in a standard way

  I think that's the main reason for bfin.  It's a DSP and has some very
unusual syntax for specifying parallelism and SIMD.

>  - ??

  I think that's the main reason for MIPS (itbl-parse.y).  It allows the
coprocessor instruction set to be specified at runtime which is useful for
MIPS which has millions of subtypes and custom variants.


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....