Hi!
The vla15.C testcase ICEs with
-mcpu=cortex-m1 -mpure-code -fstack-protector -mthumb
as what force_const_mem returns (a SYMBOL_REF) is not a valid
memory address.
Previously the code was moving the address of the force_const_mem
into a register rather than the content of that MEM, so that instruction
must have been supported and loading from a MEM with a single REG base ought
to be valid too.
Bootstrapped/regtested on armv7hl-linux-gnueabi, ok for trunk?
No testcase, as I haven't figured out my way through gcc.target/arm, sorry.
2021-02-12 Jakub Jelinek <ja...@redhat.com>
PR target/98998
* config/arm/arm.md (*stack_protect_combined_set_insn,
*stack_protect_combined_test_insn): If force_const_mem result
is not valid general operand, force its address into the destination
register first.
--- gcc/config/arm/arm.md.jj 2021-01-04 10:25:44.404170742 +0100
+++ gcc/config/arm/arm.md 2021-02-11 12:50:26.049604711 +0100
@@ -9216,6 +9216,11 @@ (define_insn_and_split "*stack_protect_c
else
{
rtx mem = force_const_mem (SImode, operands[1]);
+ if (!general_operand (mem, SImode))
+ {
+ emit_move_insn (operands[2], XEXP (mem, 0));
+ mem = replace_equiv_address (mem, operands[2], false);
+ }
emit_move_insn (operands[2], mem);
}
}
@@ -9299,6 +9304,11 @@ (define_insn_and_split "*stack_protect_c
else
{
rtx mem = force_const_mem (SImode, operands[1]);
+ if (!general_operand (mem, SImode))
+ {
+ emit_move_insn (operands[3], XEXP (mem, 0));
+ mem = replace_equiv_address (mem, operands[3], false);
+ }
emit_move_insn (operands[3], mem);
}
}
Jakub
