[Bug target/51287] New: [4.7 regression] 252.eon compfail with -march=atom

[izamyatin at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51287

             Bug #: 51287
           Summary: [4.7 regression] 252.eon compfail with -march=atom
    Classification: Unclassified
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassig...@gcc.gnu.org
        ReportedBy: izamya...@gmail.com


Compilation string
g++ -c -o ggSolidNoise2.o   -DSPEC_CPU2000_LP64 -DHAS_ERRLIST    -I. -DNDEBUG  
  -O2 -ffast-math -march=atom  ggSolidNoise2.cc

After the fix 

commit 0a55d49743f9fb5b2981037ca7676b818d273073
Author: amodra <amodra@138bc75d-0d04-0410-961f-82ee72b054a4>
Date: Wed Nov 9 01:20:14 2011 +0000
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@181188
138bc75d-0d04-0410-961f-82ee72b054a4


I see following code in rtl dumps since 206r.csa 

(insn 36 10 37 3 (clobber (reg:DI 0 ax)) -1                  (1)
     (nil))

(insn/f 37 36 31 3 (parallel [
            (set (mem:DI (pre_dec:DI (reg/f:DI 7 sp)) [0 S8 A8])
                (reg:DI 0 ax))
            (clobber (mem:BLK (scratch) [0 A8]))
        ]) ./ggRanNum.cc:45 -1
     (expr_list:REG_CFA_ADJUST_CFA (set (reg/f:DI 7 sp)
            (plus:DI (reg/f:DI 7 sp)
                (const_int -8 [0xfffffffffffffff8])))
        (nil)))

(note 31 37 11 3 NOTE_INSN_PROLOGUE_END)

(insn 11 31 12 3 (set (reg:DI 4 si)
        (symbol_ref/f:DI ("*.LC5") [flags 0x2]  <var_decl 0x7f5bc282ae60
*.LC5>)) ./ggRanNum.cc:47 62 {*movdi_internal_rex64}
     (nil))

(insn 12 11 13 3 (set (reg:DI 5 di)
        (symbol_ref:DI ("_ZSt4cerr") [flags 0x40]  <var_decl 0x7f5bc2cb95a0
cerr>)) ./ggRanNum.cc:47 62 {*movdi_internal_rex64}
     (nil))

(call_insn 13 12 15 3 (set (reg:DI 0 ax)
        (call (mem:QI (symbol_ref:DI
("_ZStlsISt11char_traitsIcEERSt13basic_ostreamIcT_ES5_PKc") [flags 0x41] 
<function_decl 0x7f5bc2c9b400 operator<<>) [0 operator<< S1 A8])
            (const_int 0 [0]))) ./ggRanNum.cc:47 636 {*call_value}
     (expr_list:REG_DEAD (reg:DI 5 di)
        (expr_list:REG_DEAD (reg:DI 4 si)
            (nil)))
    (expr_list:REG_FRAME_RELATED_EXPR (use (reg:DI 5 di))
        (expr_list:REG_FRAME_RELATED_EXPR (use (reg:DI 4 si))
            (nil))))

(insn 15 13 16 3 (set (reg:DI 5 di)                      (2)
        (reg/f:DI 0 ax [orig:59 D.23804 ] [59]))
lib/gcc/x86_64-unknown-linux-gnu/4.7.0/../../../../include/c++/4.7.0/ostream:115
62 {*movdi_internal_rex64}
     (expr_list:REG_DEAD (reg/f:DI 0 ax [orig:59 D.23804 ] [59])        (nil)))


  Fail is in i386.c, routine distance_non_agu_define_in_bb (when looking at
(2)). get_attr_type call for (1) raises assert since such instruction cannot be
recognized.

  Fix seems to be obvious but couple of questions left:
   Is rtx (1) legal and if yes what exactly does it mean?
   Is it correct statement that caller of get_attr_type is always responsible
for checking that passing instruction is recognizable?


  Compiler:

Target: x86_64-unknown-linux-gnu
Configured with: ../configure --disable-bootstrap --enable-languages=c,c++
--prefix=....