Date: Wed, 14 May 1997 10:24:33 +1200
From: Bill Currie <billc AT blackmagic DOT tait DOT co DOT nz>
Subject: Re: Bitwise ASM with DJGPP GCC
To: Keith Hull <keithh AT zetnet DOT co DOT uk>
Cc: djgpp AT delorie DOT com
Reply-to: billc AT blackmagic DOT tait DOT co DOT nz
Message-id: <3378EA21.774B@blackmagic.tait.co.nz>
Organization: Tait Electronics NZ
MIME-version: 1.0
Content-type: text/plain; charset=us-ascii
Content-transfer-encoding: 7bit
References: <1997051321301977565 AT zetnet DOT co DOT uk>
Precedence: bulk

Keith Hull wrote:
> 
> Hi everyone,
> 
> I would like some help with GCC and inline asm (i've looked at
> Brennan's Guide to Inline ASM - it seems to make sense ;) .... heres
> my problem..
> 
> asm ("rorl _val, _num_of_rotates");
> asm ("roll _val, _num_of_rotates");
> asm ("sall _val, _num_of_rotates");
> asm ("sarl _val, _num_of_rotates");
> asm ("shll _val, _num_of_rotates");
> asm ("shrl _val, _num_of_rotates");

Two things: no such instruction (amout to rotate must be either %cl or a
const) and the operands are backwords anyway.

Try:

asm ("rorl %b1,%0":"g"(val):"c"(num_of_rotates))

This will produce something similare to:

  movb _num_of_rotates,%cl
  rorl %cl,_val

depending on where gcc stores val and num_of_rotates.

BTW, gcc will optimise  val=(val>>8)|(val<<24) to rorl $8,_val (or more
likely replace _val with a register).

Bill
-- 
Leave others their otherness.