The following patch has been committed to gcc 5 branch as rev. 228256.
The patch was bootstrapped and tested on x86/x86-64.
Index: ChangeLog
===================================================================
--- ChangeLog (revision 228250)
+++ ChangeLog (working copy)
@@ -1,3 +1,12 @@
+2015-09-29 Vladimir Makarov <vmaka...@redhat.com>
+
+ Backport from mainline
+ 2015-07-21 Vladimir Makarov <vmaka...@redhat.com>
+
+ PR ipa/66424.
+ * lra-remat.c (operand_to_remat): Prevent using insns with input
+ subregs processed separately by IRA.
+
2015-09-29 Andreas Krebbel <kreb...@linux.vnet.ibm.com>
Backport from mainline
@@ -31,7 +40,7 @@
("vec_scatter_element<V_HW_64:mode>_SI"): Replace gf mode
attribute with bhfgq.
-2015-09-29 Andrew Pinski <apin...@cavium.com>
+2015-09-29 Andrew Pinski <apin...@cavium.com>
* config/aarch64/aarch64.md (prefetch):
Change the predicate of operand 0 to register_operand.
Index: lra-remat.c
===================================================================
--- lra-remat.c (revision 228250)
+++ lra-remat.c (working copy)
@@ -432,6 +432,16 @@ operand_to_remat (rtx_insn *insn)
return -1;
found_reg = reg;
}
+ /* IRA calculates conflicts separately for subregs of two words
+ pseudo. Even if the pseudo lives, e.g. one its subreg can be
+ used lately, another subreg hard register can be already used
+ for something else. In such case, it is not safe to
+ rematerialize the insn. */
+ else if (reg->type == OP_IN && reg->subreg_p
+ && reg->regno >= FIRST_PSEUDO_REGISTER
+ && (GET_MODE_SIZE (PSEUDO_REGNO_MODE (reg->regno))
+ == 2 * UNITS_PER_WORD))
+ return -1;
if (found_reg == NULL)
return -1;
if (found_reg->regno < FIRST_PSEUDO_REGISTER)
Index: testsuite/ChangeLog
===================================================================
--- testsuite/ChangeLog (revision 228250)
+++ testsuite/ChangeLog (working copy)
@@ -1,3 +1,11 @@
+2015-09-29 Vladimir Makarov <vmaka...@redhat.com>
+
+ Backport from mainline
+ 2015-07-21 Vladimir Makarov <vmaka...@redhat.com>
+
+ PR ipa/66424.
+ * gcc.target/i386/pr66424.c: New.
+
2015-09-29 Andreas Krebbel <kreb...@linux.vnet.ibm.com>
Backport from mainline
