re PR rtl-optimization/59915 (LRA ICE - Repeated looping over subreg reloads…

re PR rtl-optimization/59915 (LRA ICE - Repeated looping over subreg reloads (gcc.c-torture/compile/simd-3.c))

2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>

	PR regression/59915
	* lra-constraints.c (simplify_operand_subreg): Spill pseudo if
	there is a danger of looping.

From-SVN: r207007

gcc/ChangeLog

+2014-01-23  Vladimir Makarov  <vmakarov@redhat.com>
+
+	PR regression/59915
+	* lra-constraints.c (simplify_operand_subreg): Spill pseudo if
+	there is a danger of looping.
+
 2014-01-23  Pat Haugen  <pthaugen@us.ibm.com>
 
 	* config/rs6000/rs6000.c (rs6000_option_override_internal): Don't

gcc/lra-constraints.c

@@ -1291,9 +1291,20 @@ simplify_operand_subreg (int nop, enum machine_mode reg_mode)
        && ! LRA_SUBREG_P (operand))
       || CONSTANT_P (reg) || GET_CODE (reg) == PLUS || MEM_P (reg))
     {
-      /* The class will be defined later in curr_insn_transform.  */
-      enum reg_class rclass
-	= (enum reg_class) targetm.preferred_reload_class (reg, ALL_REGS);
+      enum reg_class rclass;
+
+      if (REG_P (reg)
+	  && curr_insn_set != NULL_RTX
+	  && (REG_P (SET_SRC (curr_insn_set))
+	      || GET_CODE (SET_SRC (curr_insn_set)) == SUBREG))
+	/* There is big probability that we will get the same class
+	   for the new pseudo and we will get the same insn which
+	   means infinite looping.  So spill the new pseudo.  */
+	rclass = NO_REGS;
+      else
+	/* The class will be defined later in curr_insn_transform.  */
+	rclass
+	  = (enum reg_class) targetm.preferred_reload_class (reg, ALL_REGS);
 
       if (get_reload_reg (curr_static_id->operand[nop].type, reg_mode, reg,
 			  rclass, "subreg reg", &new_reg))