------- Comment #10 from jakub at gcc dot gnu dot org 2006-02-05 09:48 -------
This patch introduces a regression
FAIL: g++.dg/opt/complex5.C (test for excess errors)
ICE on at least {i686,s390,s390x}-linux on gcc-4_1-branch in dom2 pass (with
-O2).
At copyrename2 the function is:
double __complex__ elt_zero() ()
{
doubleD.28 CR.13D.1769;
doubleD.28 D.1768;
doubleD.28 CI.12D.1767;
doubleD.28 CR.11D.1766;
doubleD.28 rD.1754;
cdoubleD.1733 zD.1755;
cdoubleD.1733 aD.1741;
complex doubleD.32 D.1745;
complex doubleD.32 a.0D.1744;
complex doubleD.32 D.1743;
# BLOCK 0 freq:10000
# PRED: ENTRY [100.0%] (fallthru,exec)
REALPART_EXPR <D.1743_2> = 0.0;
CR.13D.1769_8 = CR.11D.1766_5 + 1.0e+0;
D.1745_6 = COMPLEX_EXPR <CR.13D.1769_8, CI.12D.1767_4>;
CR.13D.1769_1 = CR.13D.1769_8;
# <retval>_9 = V_MAY_DEF <<retval>_3>;
REALPART_EXPR <<retval>> = CR.13D.1769_1;
# <retval>_10 = V_MAY_DEF <<retval>_9>;
IMAGPART_EXPR <<retval>> = 0.0;
# VUSE <<retval>_10>;
return <retval>;
# SUCC: EXIT [100.0%]
}
In simple_iv_increment_p
1738 phi = SSA_NAME_DEF_STMT (preinc);
1739 if (TREE_CODE (phi) != PHI_NODE)
1740 return false;
phi is NULL. preinc is:
CR.11D.1761_5
--
jakub at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |jakub at gcc dot gnu dot
| |org, dnovillo at gcc dot gnu
| |dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25977
