Re: double max case labels in switch

[Richard Henderson <rth at redhat dot com>]

On Sun, Nov 25, 2001 at 01:38:06PM +1030, Alan Modra wrote:
> >   > This doesn't seem right.  This is used by variable length insn
> >   > processing, isn't it?  Which means that the constants set in
> >   > the md file for branch lengths would break.
> >   > 
> >   > Seems better to just fix the backends in question to represent
> >   > lengths in bytes.
> 
> I'm not quite sure what you're getting at here.

Consider 

  (if_then_else (eq_attr "type" "branch")
                (if_then_else (and (ge (minus (match_dup 0) (pc))
                                       (const_int -32768))
                                   (lt (minus (match_dup 0) (pc))
                                       (const_int 32764)))
                              (const_int 4)
                              (const_int 8))

Which units will be used for this calculation?  I claim that
"length" should be figured consistently in one set of units
lest we get mightly confused.


r~