The following patch fixes
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67954
The patch was bootstrapped on x86/x86-64. The test is too big and was
not included to the patch.
Committed as rev. 230893 to gcc-5 branch and as rev. 230894 to trunk.
Index: ChangeLog
===================================================================
--- ChangeLog (revision 230887)
+++ ChangeLog (working copy)
@@ -1,3 +1,9 @@
+2015-11-25 Vladimir Makarov <vmaka...@redhat.com>
+
+ PR rtl-optimization/67954
+ * lra-constraints.c (curr_insn_transform): Add check on scratch
+ pseudo when change class to NO_REGS. Add an assert.
+
2015-11-24 Michael Meissner <meiss...@linux.vnet.ibm.com>
Backport form mainline
Index: lra-constraints.c
===================================================================
--- lra-constraints.c (revision 230887)
+++ lra-constraints.c (working copy)
@@ -3738,7 +3738,8 @@ curr_insn_transform (bool check_only_p)
assigment pass and the scratch pseudo will be
spilled. Spilled scratch pseudos are transformed
back to scratches at the LRA end. */
- && lra_former_scratch_operand_p (curr_insn, i))
+ && lra_former_scratch_operand_p (curr_insn, i)
+ && lra_former_scratch_p (REGNO (op)))
{
int regno = REGNO (op);
lra_change_class (regno, NO_REGS, " Change to", true);
@@ -3747,6 +3748,8 @@ curr_insn_transform (bool check_only_p)
spilled pseudo as there is only one such insn, the
current one. */
reg_renumber[regno] = -1;
+ lra_assert (bitmap_single_bit_set_p
+ (&lra_reg_info[REGNO (op)].insn_bitmap));
}
/* We can do an optional reload. If the pseudo got a hard
reg, we might improve the code through inheritance. If
