This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: extv, extzv and insv patterns
- To: Richard Henderson <rth at redhat dot com>
- Subject: Re: extv, extzv and insv patterns
- From: Roman Lechtchinsky <rl at cs dot tu-berlin dot de>
- Date: Thu, 19 Jul 2001 14:16:45 +0200 (MET DST)
- cc: David Edelsohn <dje at watson dot ibm dot com>, gcc at gcc dot gnu dot org
On Wed, 18 Jul 2001, Richard Henderson wrote:
> > I suggest that the bits be relative to the mode of operand0.
Are you talking about extv/extzv or insv? Anyway, right now it is relative
to the mode of the zero_extract operand (i.e. operand1 for extv and
extzv). I think this is what one would expect, the only problem is that
operand1 isn't always big enough for the bit field - it is QImode for MEM
operands, for instance. Is this actually correct? In fact, shouldn't
expmed call those patterns with operands in the correct modes? It won't
even look at the mode in the pattern for MEM operands right now.
> Which would, I see, make the current Alpha definition incorrect,
Why? AFAICS, the bit number is the same regardless of the mode unless
BITS_BIG_ENDIAN != BYTES_BIG_ENDIAN. What am I missing?
Bye
Roman