Yury Khrustalev <[email protected]> writes:
> From: Szabolcs Nagy <[email protected]>
>
> Builtin for chkfeat: the input argument is used to initialize x16 then
> execute chkfeat and return the updated x16.
>
> Note: ACLE __chkfeat(x) plans to flip the bits to be more intuitive
> (xor the input to output), but for the builtin that seems unnecessary
> complication.
Sounds good.
> gcc/ChangeLog:
>
> * config/aarch64/aarch64-builtins.cc (enum aarch64_builtins):
> Define AARCH64_BUILTIN_CHKFEAT.
> (aarch64_general_init_builtins): Handle chkfeat.
> (aarch64_general_expand_builtin): Handle chkfeat.
> ---
> gcc/config/aarch64/aarch64-builtins.cc | 18 ++++++++++++++++++
> 1 file changed, 18 insertions(+)
>
> diff --git a/gcc/config/aarch64/aarch64-builtins.cc
> b/gcc/config/aarch64/aarch64-builtins.cc
> index 7d737877e0b..765f2091504 100644
> --- a/gcc/config/aarch64/aarch64-builtins.cc
> +++ b/gcc/config/aarch64/aarch64-builtins.cc
> @@ -875,6 +875,8 @@ enum aarch64_builtins
> AARCH64_PLDX,
> AARCH64_PLI,
> AARCH64_PLIX,
> + /* Armv8.9-A / Armv9.4-A builtins. */
> + AARCH64_BUILTIN_CHKFEAT,
> AARCH64_BUILTIN_MAX
> };
>
> @@ -2280,6 +2282,12 @@ aarch64_general_init_builtins (void)
> if (!TARGET_ILP32)
> aarch64_init_pauth_hint_builtins ();
>
> + tree ftype_chkfeat
> + = build_function_type_list (uint64_type_node, uint64_type_node, NULL);
> + aarch64_builtin_decls[AARCH64_BUILTIN_CHKFEAT]
> + = aarch64_general_add_builtin ("__builtin_aarch64_chkfeat",
> ftype_chkfeat,
> + AARCH64_BUILTIN_CHKFEAT);
> +
> if (in_lto_p)
> handle_arm_acle_h ();
> }
> @@ -3484,6 +3492,16 @@ aarch64_general_expand_builtin (unsigned int fcode,
> tree exp, rtx target,
> case AARCH64_PLIX:
> aarch64_expand_prefetch_builtin (exp, fcode);
> return target;
> +
> + case AARCH64_BUILTIN_CHKFEAT:
> + {
> + rtx x16_reg = gen_rtx_REG (DImode, R16_REGNUM);
> + op0 = expand_normal (CALL_EXPR_ARG (exp, 0));
> + emit_move_insn (x16_reg, op0);
> + expand_insn (CODE_FOR_aarch64_chkfeat, 0, 0);
> + emit_move_insn (target, x16_reg);
> + return target;
target isn't reuired to be nonnull, so this would be safer as:
return copy_to_reg (x16_reg);
(I don't think it's worth complicating things by trying to reuse target,
since this code isn't going to be performance/memory critical.)
Looks good otherwise.
Thanks,
Richard
> + }
> }
>
> if (fcode >= AARCH64_SIMD_BUILTIN_BASE && fcode <=
> AARCH64_SIMD_BUILTIN_MAX)
