This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] DCE: Remove LIBCALL/RETVAL notes if they are moved to the same insn


On Fri, 2007-10-26 at 16:54 +0200, Andreas Krebbel wrote:
> Hello,
> 
> mainline GCC currently doesn't bootstrap on s390 and s390x.
> 
> The problem is that the recent dce changes might produce insns like:
> 
> 144r.ce1:
> 
> (insn 14 11 15 2 /build2/gcc-4.3/libgomp/iter.c:45 (set (reg:DI 73)
>         (reg:DI 71)) 50 {*movdi_64} (insn_list:REG_LIBCALL 14 (expr_list:REG_DEAD (reg:DI 71)
>             (insn_list:REG_RETVAL 14 (expr_list:REG_EQUAL (unspec:DI [
>                             (symbol_ref:DI ("gomp_tls_data") [flags 0x4a] <var_decl 0x200003741f8 gomp_tls_data>)
>                         ] 502)
>                     (nil))))))
> 
> Which can't be handled by cse.  Note that both notes - LIBCALL and
> RETVAL - point to this insn.
> 
> The rest of the libcall is dead and therefore correctly thrown away.
> 
> I'm currently testing the attached patch.  It fixes the testcase which
> made the s390x GCC fail bootstrapping.
> 
> OK for mainline?

This also fixes powerpc64-linux bootstrap, which died with an ICE
building libjava.

Janis


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]