This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix PR target/5626 (was Re: Infrastructure for long branches?)
- From: Brad Lucier <lucier at math dot purdue dot edu>
- To: jakub at redhat dot com
- Cc: lucier at math dot purdue dot edu, gcc-patches at gcc dot gnu dot org
- Date: Mon, 25 Feb 2002 15:47:35 -0500 (EST)
- Subject: Re: [PATCH] Fix PR target/5626 (was Re: Infrastructure for long branches?)
> Please try the following patch which should increase the +-32kinsn
> limitation for branch on register to +-2Minsns (which I believe is the top
> what can be reasonably supported, because otherwise every branch would
> require a scratch register one could build destination address into
> to issue a jmpl).
Unfortunately, this patch causes a bootstrap comparison failure:
Bootstrap comparison failure!
fold-const.o differs
I'd like to ask another question re:
> - [(set_attr "type" "branch")])
> + [(set_attr "type" "branch")
> + (set (attr "length")
> + (if_then_else (lt (pc) (match_dup 2))
> + (if_then_else (lt (minus (match_dup 2) (pc))
> + (const_int 32000))
> + (const_int 1)
> + (const_int 3))
> + (if_then_else (lt (minus (pc) (match_dup 2))
> + (const_int 32000))
> + (const_int 1)
> + (const_int 3))))])
It seems that you're saying that if the jump is < 32000 instructions,
just put it in, otherwise put in the three-instruction long jump
sequence.
Couldn't this cause a problem if more than 384 jumps between the jump
and the target were lengthened? Then you'd still have a 32768 instruction
jump limit, but a longer jump.
Brad