This is the mail archive of the
gcc-regression@gcc.gnu.org
mailing list for the GCC project.
Re: flow.c:verify_local_live_at_start patch (Was: 1 GCC regressions, 1 new, with your patch on 2000-11-02T22:05:00Z)
- To: Joern Rennecke <amylaar at redhat dot com>
- Subject: Re: flow.c:verify_local_live_at_start patch (Was: 1 GCC regressions, 1 new, with your patch on 2000-11-02T22:05:00Z)
- From: Richard Henderson <rth at cygnus dot com>
- Date: Thu, 2 Nov 2000 23:38:01 -0800
- Cc: gcc-patches at gcc dot gnu dot org, gcc-regression at gcc dot gnu dot org, neilb at earthling dot net, GCC regression checker <geoffk at cygnus dot com>
- References: <200011030026.QAA08318@sloth.cygnus.com> <200011030320.eA33Khg18161@tantra.cygnus.co.uk>
On Fri, Nov 03, 2000 at 03:20:43AM +0000, Joern Rennecke wrote:
> This is a actually a flow bug.
This is not a flow bug, this is a combine bug.
> The combiner has combined two insn that use / set hard registers:
Ok, so combine should detect this case and do
SET_BIT (refresh_blocks, this_basic_block);
need_refresh = 1;
which will recompute global register life at
the end of the combine pass.
> * flow.c (verify_local_live_at_start): Allow hard regs to die.
Despite Geoff's ack, this patch is not ok. This severely
weakens our ability to detect bugs elsewhere in the compiler.
r~