This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: PR opt/10392: Reload bug on SH
- From: Eric Botcazou <ebotcazou at libertysurf dot fr>
- To: <tm_gccmail at mail dot kloo dot net>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Tue, 30 Sep 2003 08:22:51 +0200
- Subject: Re: PR opt/10392: Reload bug on SH
- References: <Pine.LNX.4.21.0309291522130.7602-100000@mail.kloo.net>
> I tried your patch mentioned in:
>
> http://gcc.gnu.org/ml/gcc-patches/2003-09/msg01083.html
Thanks!
> This seems to cause a failure on stress-1.17/layer3.i,
> compiled with -O2 -m4:
>
> layer3.i: In function `III_dequantize_sample':
>
> layer3.i:3515: error: insn does not satisfy its constraints:
> (insn:HI 337 2471 2472 23 (set (mem/s:SI (plus:SI (reg:SI 2 r2)
> (reg:SI 0 r0)) [5 max S4 A32])
> (mem:SI (plus:SI (reg/f:SI 14 r14)
> (const_int 72 [0x48])) [167 cb+0 S4 A32])) 123 {movsi_ie}
> (insn_list 335 (nil))
> (nil))
> layer3.i:3515: internal compiler error: in reload_cse_simplify_operands,
> at postreload.c:378
>
> Evidently, it causes a memory-to-memory insn to be generated, which
> doesn't exist on the SH4...
Interesting. The patch is aimed at only disabling a faulty optimization, so
this is either a bug in the patch or a weird interaction with another part
of reload.
> Testcase is available if you need it.
Of course.
--
Eric Botcazou