> Am 02.02.2024 um 10:41 schrieb Jakub Jelinek <ja...@redhat.com>:
>
> Hi!
>
> Similar problem to calls with uninitialized large/huge _BitInt SSA_NAME
> arguments, var_to_partition will not work for those, but unlike calls
> where we just create a new uninitialized SSA_NAME here we need to change
> the inline asm input to be an uninitialized VAR_DECL.
>
> Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
Ok
Richard
> 2024-02-02 Jakub Jelinek <ja...@redhat.com>
>
> PR middle-end/113699
> * gimple-lower-bitint.cc (bitint_large_huge::lower_asm): Handle
> uninitialized large/huge _BitInt SSA_NAME inputs.
>
> * gcc.dg/bitint-81.c: New test.
>
> --- gcc/gimple-lower-bitint.cc.jj 2024-01-31 10:55:42.677095630 +0100
> +++ gcc/gimple-lower-bitint.cc 2024-02-01 10:36:08.408867602 +0100
> @@ -5200,9 +5200,18 @@ bitint_large_huge::lower_asm (gimple *st
> && TREE_CODE (TREE_TYPE (s)) == BITINT_TYPE
> && bitint_precision_kind (TREE_TYPE (s)) >= bitint_prec_large)
> {
> - int part = var_to_partition (m_map, s);
> - gcc_assert (m_vars[part] != NULL_TREE);
> - TREE_VALUE (t) = m_vars[part];
> + if (SSA_NAME_IS_DEFAULT_DEF (s)
> + && (!SSA_NAME_VAR (s) || VAR_P (SSA_NAME_VAR (s))))
> + {
> + TREE_VALUE (t) = create_tmp_var (TREE_TYPE (s), "bitint");
> + mark_addressable (TREE_VALUE (t));
> + }
> + else
> + {
> + int part = var_to_partition (m_map, s);
> + gcc_assert (m_vars[part] != NULL_TREE);
> + TREE_VALUE (t) = m_vars[part];
> + }
> }
> }
> update_stmt (stmt);
> --- gcc/testsuite/gcc.dg/bitint-81.c.jj 2024-02-01 10:38:26.132937531 +0100
> +++ gcc/testsuite/gcc.dg/bitint-81.c 2024-02-01 10:38:01.628280939 +0100
> @@ -0,0 +1,12 @@
> +/* PR middle-end/113699 */
> +/* { dg-do compile { target bitint } } */
> +/* { dg-options "-O2 -std=c23" } */
> +
> +void
> +foo (void)
> +{
> +#if __BITINT_MAXWIDTH__ >= 129
> + _BitInt(129) i;
> + __asm__ ("" : : "rm" (i));
> +#endif
> +}
>
> Jakub
>
