[Bug rtl-optimization/55093] [4.8 Regression] [x32] -maddress-mode=long failed

["hjl.tools at gmail dot com" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55093

--- Comment #7 from H.J. Lu <hjl.tools at gmail dot com> 2012-10-29 02:31:44 UTC ---
(In reply to comment #6)
> I am testing this patch:
> 
> diff --git a/gcc/lra-eliminations.c b/gcc/lra-eliminations.c
> index 2222d80..cbfbe7a 100644
> --- a/gcc/lra-eliminations.c
> +++ b/gcc/lra-eliminations.c
> @@ -272,7 +272,7 @@ get_elimination (rtx reg)
>    if ((hard_regno = REGNO (reg)) < 0 || hard_regno >= FIRST_PSEUDO_REGISTER)
>      return NULL;
>    if ((ep = elimination_map[hard_regno]) != NULL)
> -    return ep->from_rtx != reg ? NULL : ep;
> +    return ep->from != hard_regno ? NULL : ep;
>    if ((offset = self_elim_offsets[hard_regno]) == 0)
>      return NULL;
>    /* This is an iteration to restore offsets just after HARD_REGNO

It doesn't solve all problems. I got

Starting program:
/export/build/gnu/gcc-x32-mx32-native-long/build-x86_64-linux/gcc/cc1
-fpreprocessed /tmp/bad.i -march=corei7-avx -mcx16 -msahf -mno-movbe -maes
-mpclmul -mpopcnt -mno-abm -mno-lwp -mno-fma -mno-fma4 -mno-xop -mno-bmi
-mno-bmi2 -mno-tbm -mavx -mno-avx2 -msse4.2 -msse4.1 -mno-lzcnt -mno-rtm
-mno-hle -mrdrnd -mf16c -mfsgsbase -mno-rdseed -mno-prfchw -mno-adx -mfxsr
-mxsave -mxsaveopt --param l1-cache-size=32 --param l1-cache-line-size=64
--param l2-cache-size=8192 -mtune=generic -quiet -dumpbase bad.i -msse2 -mx32
-auxbase-strip O3-pr41881.s -O2 -O3 -version -fno-diagnostics-show-caret
-ftree-vectorize -fno-vect-cost-model -fno-common -fdump-tree-vect-details
-fno-ipa-cp-clone -o O3-pr41881.s
GNU C (GCC) version 4.8.0 20121029 (experimental) (x86_64-unknown-linux-gnu)
    compiled by GNU C version 4.8.0 20121029 (experimental), GMP version 5.0.2,
MPFR version 3.1.0, MPC version 0.9
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
GNU C (GCC) version 4.8.0 20121029 (experimental) (x86_64-unknown-linux-gnu)
    compiled by GNU C version 4.8.0 20121029 (experimental), GMP version 5.0.2,
MPFR version 3.1.0, MPC version 0.9
GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: 8b8bc795ba054ad7d30608d02345517d

Program received signal SIGSEGV, Segmentation fault.
0x0073eddb in dump_gimple_bb_header (outf=outf@entry=0x136dab0, 
    bb=bb@entry=0xf6943380, indent=indent@entry=0, flags=flags@entry=469762568)
    at /export/gnu/import/git/gcc-misc/gcc/gimple-pretty-print.c:2097
2097          memset (s_indent, ' ', (size_t) indent);
(gdb) p/x $rsp
$6 = 0xffffffffffffc8b0
(gdb) 

When -maddress-mode=long is used to compile x32 GCC, Pmode
is DImode and ptr_mode is SImode.  LRA fails to properly handle
ptr_mode != Pmode:

   0x0073edb6 <+278>:    lea    -0x1(%rbx),%eax
   0x0073edb9 <+281>:    mov    %ebx,%edx
   0x0073edbb <+283>:    mov    $0x20,%esi
   0x0073edc0 <+288>:    add    $0x2e,%rax
   0x0073edc4 <+292>:    shr    $0x4,%rax
   0x0073edc8 <+296>:    shl    $0x4,%rax
   0x0073edcc <+300>:    sub    %rax,%rsp
   0x0073edcf <+303>:    lea    0x1f(%rsp),%r8
   0x0073edd4 <+308>:    and    $0xffffffe0,%r8d
   0x0073edd8 <+312>:    mov    %r8,%rdi
=> 0x0073eddb <+315>:    callq  0x4a5970 <memset@plt>