On 05/01/18 22:14, Steve Ellcey wrote:
> This is a fix for PR target/83335. We are asserting in
> aarch64_print_address_internal because we have a non Pmode
> address coming from an asm instruction. My fix is to
> just allow this by checking this_is_asm_operands. This is
> what it was doing before the assert was added that caused
> the ICE.
>
> Verified that it fixed gcc.target/aarch64/asm-2.c in ILP32
> mode and that it caused no regressions.
>
> Steve Ellcey
> sell...@cavium.com
>
>
> 2018-01-05 Steve Ellcey <sell...@cavium.com>
>
> PR target/83335
> * config/aarch64/aarch64.c (aarch64_print_address_internal):
> Allow non Pmode address in asm statements.
>
>
> diff --git a/gcc/config/aarch64/aarch64.c b/gcc/config/aarch64/aarch64.c
> index a189605..af74212 100644
> --- a/gcc/config/aarch64/aarch64.c
> +++ b/gcc/config/aarch64/aarch64.c
> @@ -5670,8 +5670,9 @@ aarch64_print_address_internal (FILE *f, machine_mode
> mode, rtx x,
> {
> struct aarch64_address_info addr;
>
> - /* Check all addresses are Pmode - including ILP32. */
> - gcc_assert (GET_MODE (x) == Pmode);
> + /* Check all addresses are Pmode - including ILP32,
> + unless this is coming from an asm statement. */
> + gcc_assert (GET_MODE (x) == Pmode || this_is_asm_operands);
>
> if (aarch64_classify_address (&addr, x, mode, true, type))
> switch (addr.type)
>
Wouldn't it be better to call output_operand_lossage() with a suitable
diagnostic message? If the operand isn't in Pmode assembly will
(should) fail anyway.
R.
