> -----Original Message----- > From: Christophe Lyon <christophe.l...@arm.com> > Sent: Wednesday, May 3, 2023 3:37 PM > To: gcc-patches@gcc.gnu.org; Kyrylo Tkachov <kyrylo.tkac...@arm.com>; > Richard Earnshaw <richard.earns...@arm.com>; Richard Sandiford > <richard.sandif...@arm.com> > Cc: Christophe Lyon <christophe.l...@arm.com> > Subject: [PATCH v2 03/22] arm: [MVE intrinsics] Rework vreinterpretq > > This patch implements vreinterpretq using the new MVE intrinsics > framework. > > The old definitions for vreinterpretq are removed as a consequence. > Ok. Thanks, Kyrill > 2022-09-08 Murray Steele <murray.ste...@arm.com> > Christophe Lyon <christophe.l...@arm.com> > > gcc/ > * config/arm/arm-mve-builtins-base.cc (vreinterpretq_impl): New > class. > * config/arm/arm-mve-builtins-base.def: Define vreinterpretq. > * config/arm/arm-mve-builtins-base.h (vreinterpretq): New > declaration. > * config/arm/arm-mve-builtins-shapes.cc (parse_element_type): New > function. > (parse_type): Likewise. > (parse_signature): Likewise. > (build_one): Likewise. > (build_all): Likewise. > (overloaded_base): New struct. > (unary_convert_def): Likewise. > * config/arm/arm-mve-builtins-shapes.h (unary_convert): Declare. > * config/arm/arm-mve-builtins.cc (TYPES_reinterpret_signed1): New > macro. > (TYPES_reinterpret_unsigned1): Likewise. > (TYPES_reinterpret_integer): Likewise. > (TYPES_reinterpret_integer1): Likewise. > (TYPES_reinterpret_float1): Likewise. > (TYPES_reinterpret_float): Likewise. > (reinterpret_integer): New. > (reinterpret_float): New. > (handle_arm_mve_h): Register builtins. > * config/arm/arm_mve.h (vreinterpretq_s16): Remove. > (vreinterpretq_s32): Likewise. > (vreinterpretq_s64): Likewise. > (vreinterpretq_s8): Likewise. > (vreinterpretq_u16): Likewise. > (vreinterpretq_u32): Likewise. > (vreinterpretq_u64): Likewise. > (vreinterpretq_u8): Likewise. > (vreinterpretq_f16): Likewise. > (vreinterpretq_f32): Likewise. > (vreinterpretq_s16_s32): Likewise. > (vreinterpretq_s16_s64): Likewise. > (vreinterpretq_s16_s8): Likewise. > (vreinterpretq_s16_u16): Likewise. > (vreinterpretq_s16_u32): Likewise. > (vreinterpretq_s16_u64): Likewise. > (vreinterpretq_s16_u8): Likewise. > (vreinterpretq_s32_s16): Likewise. > (vreinterpretq_s32_s64): Likewise. > (vreinterpretq_s32_s8): Likewise. > (vreinterpretq_s32_u16): Likewise. > (vreinterpretq_s32_u32): Likewise. > (vreinterpretq_s32_u64): Likewise. > (vreinterpretq_s32_u8): Likewise. > (vreinterpretq_s64_s16): Likewise. > (vreinterpretq_s64_s32): Likewise. > (vreinterpretq_s64_s8): Likewise. > (vreinterpretq_s64_u16): Likewise. > (vreinterpretq_s64_u32): Likewise. > (vreinterpretq_s64_u64): Likewise. > (vreinterpretq_s64_u8): Likewise. > (vreinterpretq_s8_s16): Likewise. > (vreinterpretq_s8_s32): Likewise. > (vreinterpretq_s8_s64): Likewise. > (vreinterpretq_s8_u16): Likewise. > (vreinterpretq_s8_u32): Likewise. > (vreinterpretq_s8_u64): Likewise. > (vreinterpretq_s8_u8): Likewise. > (vreinterpretq_u16_s16): Likewise. > (vreinterpretq_u16_s32): Likewise. > (vreinterpretq_u16_s64): Likewise. > (vreinterpretq_u16_s8): Likewise. > (vreinterpretq_u16_u32): Likewise. > (vreinterpretq_u16_u64): Likewise. > (vreinterpretq_u16_u8): Likewise. > (vreinterpretq_u32_s16): Likewise. > (vreinterpretq_u32_s32): Likewise. > (vreinterpretq_u32_s64): Likewise. > (vreinterpretq_u32_s8): Likewise. > (vreinterpretq_u32_u16): Likewise. > (vreinterpretq_u32_u64): Likewise. > (vreinterpretq_u32_u8): Likewise. > (vreinterpretq_u64_s16): Likewise. > (vreinterpretq_u64_s32): Likewise. > (vreinterpretq_u64_s64): Likewise. > (vreinterpretq_u64_s8): Likewise. > (vreinterpretq_u64_u16): Likewise. > (vreinterpretq_u64_u32): Likewise. > (vreinterpretq_u64_u8): Likewise. > (vreinterpretq_u8_s16): Likewise. > (vreinterpretq_u8_s32): Likewise. > (vreinterpretq_u8_s64): Likewise. > (vreinterpretq_u8_s8): Likewise. > (vreinterpretq_u8_u16): Likewise. > (vreinterpretq_u8_u32): Likewise. > (vreinterpretq_u8_u64): Likewise. > (vreinterpretq_s32_f16): Likewise. > (vreinterpretq_s32_f32): Likewise. > (vreinterpretq_u16_f16): Likewise. > (vreinterpretq_u16_f32): Likewise. > (vreinterpretq_u32_f16): Likewise. > (vreinterpretq_u32_f32): Likewise. > (vreinterpretq_u64_f16): Likewise. > (vreinterpretq_u64_f32): Likewise. > (vreinterpretq_u8_f16): Likewise. > (vreinterpretq_u8_f32): Likewise. > (vreinterpretq_f16_f32): Likewise. > (vreinterpretq_f16_s16): Likewise. > (vreinterpretq_f16_s32): Likewise. > (vreinterpretq_f16_s64): Likewise. > (vreinterpretq_f16_s8): Likewise. > (vreinterpretq_f16_u16): Likewise. > (vreinterpretq_f16_u32): Likewise. > (vreinterpretq_f16_u64): Likewise. > (vreinterpretq_f16_u8): Likewise. > (vreinterpretq_f32_f16): Likewise. > (vreinterpretq_f32_s16): Likewise. > (vreinterpretq_f32_s32): Likewise. > (vreinterpretq_f32_s64): Likewise. > (vreinterpretq_f32_s8): Likewise. > (vreinterpretq_f32_u16): Likewise. > (vreinterpretq_f32_u32): Likewise. > (vreinterpretq_f32_u64): Likewise. > (vreinterpretq_f32_u8): Likewise. > (vreinterpretq_s16_f16): Likewise. > (vreinterpretq_s16_f32): Likewise. > (vreinterpretq_s64_f16): Likewise. > (vreinterpretq_s64_f32): Likewise. > (vreinterpretq_s8_f16): Likewise. > (vreinterpretq_s8_f32): Likewise. > (__arm_vreinterpretq_f16): Likewise. > (__arm_vreinterpretq_f32): Likewise. > (__arm_vreinterpretq_s16): Likewise. > (__arm_vreinterpretq_s32): Likewise. > (__arm_vreinterpretq_s64): Likewise. > (__arm_vreinterpretq_s8): Likewise. > (__arm_vreinterpretq_u16): Likewise. > (__arm_vreinterpretq_u32): Likewise. > (__arm_vreinterpretq_u64): Likewise. > (__arm_vreinterpretq_u8): Likewise. > * config/arm/arm_mve_types.h (__arm_vreinterpretq_s16_s32): > Remove. > (__arm_vreinterpretq_s16_s64): Likewise. > (__arm_vreinterpretq_s16_s8): Likewise. > (__arm_vreinterpretq_s16_u16): Likewise. > (__arm_vreinterpretq_s16_u32): Likewise. > (__arm_vreinterpretq_s16_u64): Likewise. > (__arm_vreinterpretq_s16_u8): Likewise. > (__arm_vreinterpretq_s32_s16): Likewise. > (__arm_vreinterpretq_s32_s64): Likewise. > (__arm_vreinterpretq_s32_s8): Likewise. > (__arm_vreinterpretq_s32_u16): Likewise. > (__arm_vreinterpretq_s32_u32): Likewise. > (__arm_vreinterpretq_s32_u64): Likewise. > (__arm_vreinterpretq_s32_u8): Likewise. > (__arm_vreinterpretq_s64_s16): Likewise. > (__arm_vreinterpretq_s64_s32): Likewise. > (__arm_vreinterpretq_s64_s8): Likewise. > (__arm_vreinterpretq_s64_u16): Likewise. > (__arm_vreinterpretq_s64_u32): Likewise. > (__arm_vreinterpretq_s64_u64): Likewise. > (__arm_vreinterpretq_s64_u8): Likewise. > (__arm_vreinterpretq_s8_s16): Likewise. > (__arm_vreinterpretq_s8_s32): Likewise. > (__arm_vreinterpretq_s8_s64): Likewise. > (__arm_vreinterpretq_s8_u16): Likewise. > (__arm_vreinterpretq_s8_u32): Likewise. > (__arm_vreinterpretq_s8_u64): Likewise. > (__arm_vreinterpretq_s8_u8): Likewise. > (__arm_vreinterpretq_u16_s16): Likewise. > (__arm_vreinterpretq_u16_s32): Likewise. > (__arm_vreinterpretq_u16_s64): Likewise. > (__arm_vreinterpretq_u16_s8): Likewise. > (__arm_vreinterpretq_u16_u32): Likewise. > (__arm_vreinterpretq_u16_u64): Likewise. > (__arm_vreinterpretq_u16_u8): Likewise. > (__arm_vreinterpretq_u32_s16): Likewise. > (__arm_vreinterpretq_u32_s32): Likewise. > (__arm_vreinterpretq_u32_s64): Likewise. > (__arm_vreinterpretq_u32_s8): Likewise. > (__arm_vreinterpretq_u32_u16): Likewise. > (__arm_vreinterpretq_u32_u64): Likewise. > (__arm_vreinterpretq_u32_u8): Likewise. > (__arm_vreinterpretq_u64_s16): Likewise. > (__arm_vreinterpretq_u64_s32): Likewise. > (__arm_vreinterpretq_u64_s64): Likewise. > (__arm_vreinterpretq_u64_s8): Likewise. > (__arm_vreinterpretq_u64_u16): Likewise. > (__arm_vreinterpretq_u64_u32): Likewise. > (__arm_vreinterpretq_u64_u8): Likewise. > (__arm_vreinterpretq_u8_s16): Likewise. > (__arm_vreinterpretq_u8_s32): Likewise. > (__arm_vreinterpretq_u8_s64): Likewise. > (__arm_vreinterpretq_u8_s8): Likewise. > (__arm_vreinterpretq_u8_u16): Likewise. > (__arm_vreinterpretq_u8_u32): Likewise. > (__arm_vreinterpretq_u8_u64): Likewise. > (__arm_vreinterpretq_s32_f16): Likewise. > (__arm_vreinterpretq_s32_f32): Likewise. > (__arm_vreinterpretq_s16_f16): Likewise. > (__arm_vreinterpretq_s16_f32): Likewise. > (__arm_vreinterpretq_s64_f16): Likewise. > (__arm_vreinterpretq_s64_f32): Likewise. > (__arm_vreinterpretq_s8_f16): Likewise. > (__arm_vreinterpretq_s8_f32): Likewise. > (__arm_vreinterpretq_u16_f16): Likewise. > (__arm_vreinterpretq_u16_f32): Likewise. > (__arm_vreinterpretq_u32_f16): Likewise. > (__arm_vreinterpretq_u32_f32): Likewise. > (__arm_vreinterpretq_u64_f16): Likewise. > (__arm_vreinterpretq_u64_f32): Likewise. > (__arm_vreinterpretq_u8_f16): Likewise. > (__arm_vreinterpretq_u8_f32): Likewise. > (__arm_vreinterpretq_f16_f32): Likewise. > (__arm_vreinterpretq_f16_s16): Likewise. > (__arm_vreinterpretq_f16_s32): Likewise. > (__arm_vreinterpretq_f16_s64): Likewise. > (__arm_vreinterpretq_f16_s8): Likewise. > (__arm_vreinterpretq_f16_u16): Likewise. > (__arm_vreinterpretq_f16_u32): Likewise. > (__arm_vreinterpretq_f16_u64): Likewise. > (__arm_vreinterpretq_f16_u8): Likewise. > (__arm_vreinterpretq_f32_f16): Likewise. > (__arm_vreinterpretq_f32_s16): Likewise. > (__arm_vreinterpretq_f32_s32): Likewise. > (__arm_vreinterpretq_f32_s64): Likewise. > (__arm_vreinterpretq_f32_s8): Likewise. > (__arm_vreinterpretq_f32_u16): Likewise. > (__arm_vreinterpretq_f32_u32): Likewise. > (__arm_vreinterpretq_f32_u64): Likewise. > (__arm_vreinterpretq_f32_u8): Likewise. > (__arm_vreinterpretq_s16): Likewise. > (__arm_vreinterpretq_s32): Likewise. > (__arm_vreinterpretq_s64): Likewise. > (__arm_vreinterpretq_s8): Likewise. > (__arm_vreinterpretq_u16): Likewise. > (__arm_vreinterpretq_u32): Likewise. > (__arm_vreinterpretq_u64): Likewise. > (__arm_vreinterpretq_u8): Likewise. > (__arm_vreinterpretq_f16): Likewise. > (__arm_vreinterpretq_f32): Likewise. > * config/arm/mve.md (@arm_mve_reinterpret<mode>): New > pattern. > * config/arm/unspecs.md: (REINTERPRET): New unspec. > > gcc/testsuite/ > * g++.target/arm/mve.exp: Add general-c++ and general directories. > * g++.target/arm/mve/general-c++/nomve_fp_1.c: New test. > * g++.target/arm/mve/general-c++/vreinterpretq_1.C: New test. > * gcc.target/arm/mve/general-c/nomve_fp_1.c: New test. > * gcc.target/arm/mve/general-c/vreinterpretq_1.c: New test. > --- > gcc/config/arm/arm-mve-builtins-base.cc | 33 + > gcc/config/arm/arm-mve-builtins-base.def | 2 + > gcc/config/arm/arm-mve-builtins-base.h | 2 + > gcc/config/arm/arm-mve-builtins-shapes.cc | 28 + > gcc/config/arm/arm-mve-builtins-shapes.h | 8 + > gcc/config/arm/arm-mve-builtins.cc | 60 + > gcc/config/arm/arm_mve.h | 300 ---- > gcc/config/arm/arm_mve_types.h | 1365 +---------------- > gcc/config/arm/mve.md | 18 + > gcc/config/arm/unspecs.md | 1 + > gcc/testsuite/g++.target/arm/mve.exp | 8 +- > .../arm/mve/general-c++/nomve_fp_1.c | 15 + > .../arm/mve/general-c++/vreinterpretq_1.C | 25 + > .../gcc.target/arm/mve/general-c/nomve_fp_1.c | 15 + > .../arm/mve/general-c/vreinterpretq_1.c | 25 + > 15 files changed, 290 insertions(+), 1615 deletions(-) > create mode 100644 gcc/testsuite/g++.target/arm/mve/general- > c++/nomve_fp_1.c > create mode 100644 gcc/testsuite/g++.target/arm/mve/general- > c++/vreinterpretq_1.C > create mode 100644 gcc/testsuite/gcc.target/arm/mve/general- > c/nomve_fp_1.c > create mode 100644 gcc/testsuite/gcc.target/arm/mve/general- > c/vreinterpretq_1.c > > diff --git a/gcc/config/arm/arm-mve-builtins-base.cc b/gcc/config/arm/arm- > mve-builtins-base.cc > index e9f285faf2b..abf6a1e19de 100644 > --- a/gcc/config/arm/arm-mve-builtins-base.cc > +++ b/gcc/config/arm/arm-mve-builtins-base.cc > @@ -38,8 +38,41 @@ using namespace arm_mve; > > namespace { > > +/* Implements vreinterpretq_* intrinsics. */ > +class vreinterpretq_impl : public quiet<function_base> > +{ > + gimple * > + fold (gimple_folder &f) const override > + { > + /* We should punt to rtl if the effect of the reinterpret on > + registers does not conform to GCC's endianness model like we do > + on aarch64, but MVE intrinsics are not currently supported on > + big-endian. For this, we'd need to handle big-endian properly > + in the .md file, like we do on aarch64 with > + define_insn_and_split "*aarch64_sve_reinterpret<mode>". */ > + gcc_assert (targetm.can_change_mode_class (f.vector_mode (0), > + f.vector_mode (1), > + VFP_REGS)); > + > + /* Otherwise vreinterpret corresponds directly to a VIEW_CONVERT_EXPR > + reinterpretation. */ > + tree rhs = build1 (VIEW_CONVERT_EXPR, TREE_TYPE (f.lhs), > + gimple_call_arg (f.call, 0)); > + return gimple_build_assign (f.lhs, VIEW_CONVERT_EXPR, rhs); > + } > + > + rtx > + expand (function_expander &e) const override > + { > + machine_mode mode = e.vector_mode (0); > + return e.use_exact_insn (code_for_arm_mve_reinterpret (mode)); > + } > +}; > + > } /* end anonymous namespace */ > > namespace arm_mve { > > +FUNCTION (vreinterpretq, vreinterpretq_impl,) > + > } /* end namespace arm_mve */ > diff --git a/gcc/config/arm/arm-mve-builtins-base.def b/gcc/config/arm/arm- > mve-builtins-base.def > index d15ba2e23e8..5c0c1b9cee7 100644 > --- a/gcc/config/arm/arm-mve-builtins-base.def > +++ b/gcc/config/arm/arm-mve-builtins-base.def > @@ -18,7 +18,9 @@ > <http://www.gnu.org/licenses/>. */ > > #define REQUIRES_FLOAT false > +DEF_MVE_FUNCTION (vreinterpretq, unary_convert, reinterpret_integer, > none) > #undef REQUIRES_FLOAT > > #define REQUIRES_FLOAT true > +DEF_MVE_FUNCTION (vreinterpretq, unary_convert, reinterpret_float, none) > #undef REQUIRES_FLOAT > diff --git a/gcc/config/arm/arm-mve-builtins-base.h b/gcc/config/arm/arm- > mve-builtins-base.h > index c4d7b750cd5..60e7bd24eda 100644 > --- a/gcc/config/arm/arm-mve-builtins-base.h > +++ b/gcc/config/arm/arm-mve-builtins-base.h > @@ -23,6 +23,8 @@ > namespace arm_mve { > namespace functions { > > +extern const function_base *const vreinterpretq; > + > } /* end namespace arm_mve::functions */ > } /* end namespace arm_mve */ > > diff --git a/gcc/config/arm/arm-mve-builtins-shapes.cc b/gcc/config/arm/arm- > mve-builtins-shapes.cc > index f20660d8319..d0da0ffef91 100644 > --- a/gcc/config/arm/arm-mve-builtins-shapes.cc > +++ b/gcc/config/arm/arm-mve-builtins-shapes.cc > @@ -338,6 +338,34 @@ struct overloaded_base : public function_shape > } > }; > > +/* <T0>_t foo_t0[_t1](<T1>_t) > + > + where the target type <t0> must be specified explicitly but the source > + type <t1> can be inferred. > + > + Example: vreinterpretq. > + int16x8_t [__arm_]vreinterpretq_s16[_s8](int8x16_t a) > + int32x4_t [__arm_]vreinterpretq_s32[_s8](int8x16_t a) > + int8x16_t [__arm_]vreinterpretq_s8[_s16](int16x8_t a) > + int8x16_t [__arm_]vreinterpretq_s8[_s32](int32x4_t a) */ > +struct unary_convert_def : public overloaded_base<1> > +{ > + void > + build (function_builder &b, const function_group_info &group, > + bool preserve_user_namespace) const override > + { > + b.add_overloaded_functions (group, MODE_none, > preserve_user_namespace); > + build_all (b, "v0,v1", group, MODE_none, preserve_user_namespace); > + } > + > + tree > + resolve (function_resolver &r) const override > + { > + return r.resolve_unary (); > + } > +}; > +SHAPE (unary_convert) > + > } /* end namespace arm_mve */ > > #undef SHAPE > diff --git a/gcc/config/arm/arm-mve-builtins-shapes.h b/gcc/config/arm/arm- > mve-builtins-shapes.h > index 9e353b85a76..04d19a02890 100644 > --- a/gcc/config/arm/arm-mve-builtins-shapes.h > +++ b/gcc/config/arm/arm-mve-builtins-shapes.h > @@ -22,8 +22,16 @@ > > namespace arm_mve > { > + /* The naming convention is: > + > + - to use names like "unary" etc. if the rules are somewhat generic, > + especially if there are no ranges involved. */ > + > namespace shapes > { > + > + extern const function_shape *const unary_convert; > + > } /* end namespace arm_mve::shapes */ > } /* end namespace arm_mve */ > > diff --git a/gcc/config/arm/arm-mve-builtins.cc b/gcc/config/arm/arm-mve- > builtins.cc > index b0cceb75ceb..e409a029346 100644 > --- a/gcc/config/arm/arm-mve-builtins.cc > +++ b/gcc/config/arm/arm-mve-builtins.cc > @@ -199,6 +199,52 @@ CONSTEXPR const type_suffix_info > type_suffixes[NUM_TYPE_SUFFIXES + 1] = { > #define TYPES_signed_32(S, D) \ > S (s32) > > +#define TYPES_reinterpret_signed1(D, A) \ > + D (A, s8), D (A, s16), D (A, s32), D (A, s64) > + > +#define TYPES_reinterpret_unsigned1(D, A) \ > + D (A, u8), D (A, u16), D (A, u32), D (A, u64) > + > +#define TYPES_reinterpret_integer(S, D) \ > + TYPES_reinterpret_unsigned1 (D, s8), \ > + D (s8, s16), D (s8, s32), D (s8, s64), \ > + TYPES_reinterpret_unsigned1 (D, s16), \ > + D (s16, s8), D (s16, s32), D (s16, s64), \ > + TYPES_reinterpret_unsigned1 (D, s32), \ > + D (s32, s8), D (s32, s16), D (s32, s64), \ > + TYPES_reinterpret_unsigned1 (D, s64), \ > + D (s64, s8), D (s64, s16), D (s64, s32), \ > + TYPES_reinterpret_signed1 (D, u8), \ > + D (u8, u16), D (u8, u32), D (u8, u64), \ > + TYPES_reinterpret_signed1 (D, u16), \ > + D (u16, u8), D (u16, u32), D (u16, u64), \ > + TYPES_reinterpret_signed1 (D, u32), \ > + D (u32, u8), D (u32, u16), D (u32, u64), \ > + TYPES_reinterpret_signed1 (D, u64), \ > + D (u64, u8), D (u64, u16), D (u64, u32) > + > +/* { _s8 _s16 _s32 _s64 } x { _s8 _s16 _s32 _s64 } > + { _u8 _u16 _u32 _u64 } { _u8 _u16 _u32 _u64 }. */ > +#define TYPES_reinterpret_integer1(D, A) \ > + TYPES_reinterpret_signed1 (D, A), \ > + TYPES_reinterpret_unsigned1 (D, A) > + > +#define TYPES_reinterpret_float1(D, A) \ > + D (A, f16), D (A, f32) > + > +#define TYPES_reinterpret_float(S, D) \ > + TYPES_reinterpret_float1 (D, s8), \ > + TYPES_reinterpret_float1 (D, s16), \ > + TYPES_reinterpret_float1 (D, s32), \ > + TYPES_reinterpret_float1 (D, s64), \ > + TYPES_reinterpret_float1 (D, u8), \ > + TYPES_reinterpret_float1 (D, u16), \ > + TYPES_reinterpret_float1 (D, u32), \ > + TYPES_reinterpret_float1 (D, u64), \ > + TYPES_reinterpret_integer1 (D, f16), \ > + TYPES_reinterpret_integer1 (D, f32), \ > + D (f16, f32), D (f32, f16) > + > /* Describe a pair of type suffixes in which only the first is used. */ > #define DEF_VECTOR_TYPE(X) { TYPE_SUFFIX_ ## X, NUM_TYPE_SUFFIXES } > > @@ -231,6 +277,8 @@ DEF_MVE_TYPES_ARRAY (integer_16_32); > DEF_MVE_TYPES_ARRAY (integer_32); > DEF_MVE_TYPES_ARRAY (signed_16_32); > DEF_MVE_TYPES_ARRAY (signed_32); > +DEF_MVE_TYPES_ARRAY (reinterpret_integer); > +DEF_MVE_TYPES_ARRAY (reinterpret_float); > > /* Used by functions that have no governing predicate. */ > static const predication_index preds_none[] = { PRED_none, NUM_PREDS }; > @@ -253,6 +301,14 @@ static const predication_index preds_p_or_none[] = { > PRED_p, PRED_none, NUM_PREDS > }; > > +/* A list of all MVE ACLE functions. */ > +static CONSTEXPR const function_group_info function_groups[] = { > +#define DEF_MVE_FUNCTION(NAME, SHAPE, TYPES, PREDS) > \ > + { #NAME, &functions::NAME, &shapes::SHAPE, types_##TYPES, > preds_##PREDS, \ > + REQUIRES_FLOAT }, > +#include "arm-mve-builtins.def" > +}; > + > /* The scalar type associated with each vector type. */ > extern GTY(()) tree scalar_types[NUM_VECTOR_TYPES]; > tree scalar_types[NUM_VECTOR_TYPES]; > @@ -431,6 +487,10 @@ handle_arm_mve_h (bool > preserve_user_namespace) > > /* Define MVE functions. */ > function_table = new hash_table<registered_function_hasher> (1023); > + function_builder builder; > + for (unsigned int i = 0; i < ARRAY_SIZE (function_groups); ++i) > + builder.register_function_group (function_groups[i], > + preserve_user_namespace); > } > > /* Return true if CANDIDATE is equivalent to MODEL_TYPE for overloading > diff --git a/gcc/config/arm/arm_mve.h b/gcc/config/arm/arm_mve.h > index 0d2ba968fc0..7688b5a7e53 100644 > --- a/gcc/config/arm/arm_mve.h > +++ b/gcc/config/arm/arm_mve.h > @@ -358,14 +358,6 @@ > #define vstrwq_scatter_shifted_offset_p(__base, __offset, __value, __p) > __arm_vstrwq_scatter_shifted_offset_p(__base, __offset, __value, __p) > #define vstrwq_scatter_shifted_offset(__base, __offset, __value) > __arm_vstrwq_scatter_shifted_offset(__base, __offset, __value) > #define vuninitializedq(__v) __arm_vuninitializedq(__v) > -#define vreinterpretq_s16(__a) __arm_vreinterpretq_s16(__a) > -#define vreinterpretq_s32(__a) __arm_vreinterpretq_s32(__a) > -#define vreinterpretq_s64(__a) __arm_vreinterpretq_s64(__a) > -#define vreinterpretq_s8(__a) __arm_vreinterpretq_s8(__a) > -#define vreinterpretq_u16(__a) __arm_vreinterpretq_u16(__a) > -#define vreinterpretq_u32(__a) __arm_vreinterpretq_u32(__a) > -#define vreinterpretq_u64(__a) __arm_vreinterpretq_u64(__a) > -#define vreinterpretq_u8(__a) __arm_vreinterpretq_u8(__a) > #define vddupq_m(__inactive, __a, __imm, __p) > __arm_vddupq_m(__inactive, __a, __imm, __p) > #define vddupq_u8(__a, __imm) __arm_vddupq_u8(__a, __imm) > #define vddupq_u32(__a, __imm) __arm_vddupq_u32(__a, __imm) > @@ -518,8 +510,6 @@ > #define vfmsq_m(__a, __b, __c, __p) __arm_vfmsq_m(__a, __b, __c, __p) > #define vmaxnmq_m(__inactive, __a, __b, __p) > __arm_vmaxnmq_m(__inactive, __a, __b, __p) > #define vminnmq_m(__inactive, __a, __b, __p) > __arm_vminnmq_m(__inactive, __a, __b, __p) > -#define vreinterpretq_f16(__a) __arm_vreinterpretq_f16(__a) > -#define vreinterpretq_f32(__a) __arm_vreinterpretq_f32(__a) > #define vminnmq_x(__a, __b, __p) __arm_vminnmq_x(__a, __b, __p) > #define vmaxnmq_x(__a, __b, __p) __arm_vmaxnmq_x(__a, __b, __p) > #define vcmulq_x(__a, __b, __p) __arm_vcmulq_x(__a, __b, __p) > @@ -2365,96 +2355,6 @@ > #define vaddq_u32(__a, __b) __arm_vaddq_u32(__a, __b) > #define vaddq_f16(__a, __b) __arm_vaddq_f16(__a, __b) > #define vaddq_f32(__a, __b) __arm_vaddq_f32(__a, __b) > -#define vreinterpretq_s16_s32(__a) __arm_vreinterpretq_s16_s32(__a) > -#define vreinterpretq_s16_s64(__a) __arm_vreinterpretq_s16_s64(__a) > -#define vreinterpretq_s16_s8(__a) __arm_vreinterpretq_s16_s8(__a) > -#define vreinterpretq_s16_u16(__a) __arm_vreinterpretq_s16_u16(__a) > -#define vreinterpretq_s16_u32(__a) __arm_vreinterpretq_s16_u32(__a) > -#define vreinterpretq_s16_u64(__a) __arm_vreinterpretq_s16_u64(__a) > -#define vreinterpretq_s16_u8(__a) __arm_vreinterpretq_s16_u8(__a) > -#define vreinterpretq_s32_s16(__a) __arm_vreinterpretq_s32_s16(__a) > -#define vreinterpretq_s32_s64(__a) __arm_vreinterpretq_s32_s64(__a) > -#define vreinterpretq_s32_s8(__a) __arm_vreinterpretq_s32_s8(__a) > -#define vreinterpretq_s32_u16(__a) __arm_vreinterpretq_s32_u16(__a) > -#define vreinterpretq_s32_u32(__a) __arm_vreinterpretq_s32_u32(__a) > -#define vreinterpretq_s32_u64(__a) __arm_vreinterpretq_s32_u64(__a) > -#define vreinterpretq_s32_u8(__a) __arm_vreinterpretq_s32_u8(__a) > -#define vreinterpretq_s64_s16(__a) __arm_vreinterpretq_s64_s16(__a) > -#define vreinterpretq_s64_s32(__a) __arm_vreinterpretq_s64_s32(__a) > -#define vreinterpretq_s64_s8(__a) __arm_vreinterpretq_s64_s8(__a) > -#define vreinterpretq_s64_u16(__a) __arm_vreinterpretq_s64_u16(__a) > -#define vreinterpretq_s64_u32(__a) __arm_vreinterpretq_s64_u32(__a) > -#define vreinterpretq_s64_u64(__a) __arm_vreinterpretq_s64_u64(__a) > -#define vreinterpretq_s64_u8(__a) __arm_vreinterpretq_s64_u8(__a) > -#define vreinterpretq_s8_s16(__a) __arm_vreinterpretq_s8_s16(__a) > -#define vreinterpretq_s8_s32(__a) __arm_vreinterpretq_s8_s32(__a) > -#define vreinterpretq_s8_s64(__a) __arm_vreinterpretq_s8_s64(__a) > -#define vreinterpretq_s8_u16(__a) __arm_vreinterpretq_s8_u16(__a) > -#define vreinterpretq_s8_u32(__a) __arm_vreinterpretq_s8_u32(__a) > -#define vreinterpretq_s8_u64(__a) __arm_vreinterpretq_s8_u64(__a) > -#define vreinterpretq_s8_u8(__a) __arm_vreinterpretq_s8_u8(__a) > -#define vreinterpretq_u16_s16(__a) __arm_vreinterpretq_u16_s16(__a) > -#define vreinterpretq_u16_s32(__a) __arm_vreinterpretq_u16_s32(__a) > -#define vreinterpretq_u16_s64(__a) __arm_vreinterpretq_u16_s64(__a) > -#define vreinterpretq_u16_s8(__a) __arm_vreinterpretq_u16_s8(__a) > -#define vreinterpretq_u16_u32(__a) __arm_vreinterpretq_u16_u32(__a) > -#define vreinterpretq_u16_u64(__a) __arm_vreinterpretq_u16_u64(__a) > -#define vreinterpretq_u16_u8(__a) __arm_vreinterpretq_u16_u8(__a) > -#define vreinterpretq_u32_s16(__a) __arm_vreinterpretq_u32_s16(__a) > -#define vreinterpretq_u32_s32(__a) __arm_vreinterpretq_u32_s32(__a) > -#define vreinterpretq_u32_s64(__a) __arm_vreinterpretq_u32_s64(__a) > -#define vreinterpretq_u32_s8(__a) __arm_vreinterpretq_u32_s8(__a) > -#define vreinterpretq_u32_u16(__a) __arm_vreinterpretq_u32_u16(__a) > -#define vreinterpretq_u32_u64(__a) __arm_vreinterpretq_u32_u64(__a) > -#define vreinterpretq_u32_u8(__a) __arm_vreinterpretq_u32_u8(__a) > -#define vreinterpretq_u64_s16(__a) __arm_vreinterpretq_u64_s16(__a) > -#define vreinterpretq_u64_s32(__a) __arm_vreinterpretq_u64_s32(__a) > -#define vreinterpretq_u64_s64(__a) __arm_vreinterpretq_u64_s64(__a) > -#define vreinterpretq_u64_s8(__a) __arm_vreinterpretq_u64_s8(__a) > -#define vreinterpretq_u64_u16(__a) __arm_vreinterpretq_u64_u16(__a) > -#define vreinterpretq_u64_u32(__a) __arm_vreinterpretq_u64_u32(__a) > -#define vreinterpretq_u64_u8(__a) __arm_vreinterpretq_u64_u8(__a) > -#define vreinterpretq_u8_s16(__a) __arm_vreinterpretq_u8_s16(__a) > -#define vreinterpretq_u8_s32(__a) __arm_vreinterpretq_u8_s32(__a) > -#define vreinterpretq_u8_s64(__a) __arm_vreinterpretq_u8_s64(__a) > -#define vreinterpretq_u8_s8(__a) __arm_vreinterpretq_u8_s8(__a) > -#define vreinterpretq_u8_u16(__a) __arm_vreinterpretq_u8_u16(__a) > -#define vreinterpretq_u8_u32(__a) __arm_vreinterpretq_u8_u32(__a) > -#define vreinterpretq_u8_u64(__a) __arm_vreinterpretq_u8_u64(__a) > -#define vreinterpretq_s32_f16(__a) __arm_vreinterpretq_s32_f16(__a) > -#define vreinterpretq_s32_f32(__a) __arm_vreinterpretq_s32_f32(__a) > -#define vreinterpretq_u16_f16(__a) __arm_vreinterpretq_u16_f16(__a) > -#define vreinterpretq_u16_f32(__a) __arm_vreinterpretq_u16_f32(__a) > -#define vreinterpretq_u32_f16(__a) __arm_vreinterpretq_u32_f16(__a) > -#define vreinterpretq_u32_f32(__a) __arm_vreinterpretq_u32_f32(__a) > -#define vreinterpretq_u64_f16(__a) __arm_vreinterpretq_u64_f16(__a) > -#define vreinterpretq_u64_f32(__a) __arm_vreinterpretq_u64_f32(__a) > -#define vreinterpretq_u8_f16(__a) __arm_vreinterpretq_u8_f16(__a) > -#define vreinterpretq_u8_f32(__a) __arm_vreinterpretq_u8_f32(__a) > -#define vreinterpretq_f16_f32(__a) __arm_vreinterpretq_f16_f32(__a) > -#define vreinterpretq_f16_s16(__a) __arm_vreinterpretq_f16_s16(__a) > -#define vreinterpretq_f16_s32(__a) __arm_vreinterpretq_f16_s32(__a) > -#define vreinterpretq_f16_s64(__a) __arm_vreinterpretq_f16_s64(__a) > -#define vreinterpretq_f16_s8(__a) __arm_vreinterpretq_f16_s8(__a) > -#define vreinterpretq_f16_u16(__a) __arm_vreinterpretq_f16_u16(__a) > -#define vreinterpretq_f16_u32(__a) __arm_vreinterpretq_f16_u32(__a) > -#define vreinterpretq_f16_u64(__a) __arm_vreinterpretq_f16_u64(__a) > -#define vreinterpretq_f16_u8(__a) __arm_vreinterpretq_f16_u8(__a) > -#define vreinterpretq_f32_f16(__a) __arm_vreinterpretq_f32_f16(__a) > -#define vreinterpretq_f32_s16(__a) __arm_vreinterpretq_f32_s16(__a) > -#define vreinterpretq_f32_s32(__a) __arm_vreinterpretq_f32_s32(__a) > -#define vreinterpretq_f32_s64(__a) __arm_vreinterpretq_f32_s64(__a) > -#define vreinterpretq_f32_s8(__a) __arm_vreinterpretq_f32_s8(__a) > -#define vreinterpretq_f32_u16(__a) __arm_vreinterpretq_f32_u16(__a) > -#define vreinterpretq_f32_u32(__a) __arm_vreinterpretq_f32_u32(__a) > -#define vreinterpretq_f32_u64(__a) __arm_vreinterpretq_f32_u64(__a) > -#define vreinterpretq_f32_u8(__a) __arm_vreinterpretq_f32_u8(__a) > -#define vreinterpretq_s16_f16(__a) __arm_vreinterpretq_s16_f16(__a) > -#define vreinterpretq_s16_f32(__a) __arm_vreinterpretq_s16_f32(__a) > -#define vreinterpretq_s64_f16(__a) __arm_vreinterpretq_s64_f16(__a) > -#define vreinterpretq_s64_f32(__a) __arm_vreinterpretq_s64_f32(__a) > -#define vreinterpretq_s8_f16(__a) __arm_vreinterpretq_s8_f16(__a) > -#define vreinterpretq_s8_f32(__a) __arm_vreinterpretq_s8_f32(__a) > #define vuninitializedq_u8(void) __arm_vuninitializedq_u8(void) > #define vuninitializedq_u16(void) __arm_vuninitializedq_u16(void) > #define vuninitializedq_u32(void) __arm_vuninitializedq_u32(void) > @@ -37874,126 +37774,6 @@ extern void *__ARM_undef; > int (*)[__ARM_mve_type_float16x8_t]: __arm_vuninitializedq_f16 (), \ > int (*)[__ARM_mve_type_float32x4_t]: __arm_vuninitializedq_f32 ());}) > > -#define __arm_vreinterpretq_f16(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_f16_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_f16_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_f16_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_f16_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_f16_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_f16_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_f16_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_f16_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_f16_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_f32(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_f32_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_f32_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_f32_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_f32_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_f32_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_f32_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_f32_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_f32_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_f32_f16 > (__ARM_mve_coerce(__p0, float16x8_t)));}) > - > -#define __arm_vreinterpretq_s16(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_s16_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s16_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s16_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s16_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s16_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s16_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s16_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s16_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_s16_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_s32(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_s32_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s32_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s32_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s32_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s32_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s32_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s32_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s32_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_s32_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_s64(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_s64_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s64_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s64_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s64_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s64_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s64_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s64_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s64_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_s64_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_s8(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_s8_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s8_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s8_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s8_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s8_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s8_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s8_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s8_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_s8_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_u16(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_u16_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u16_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u16_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u16_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u16_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u16_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u16_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u16_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_u16_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_u32(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_u32_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u32_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u32_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u32_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u32_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u32_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u32_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u32_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_u32_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_u64(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_u64_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u64_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u64_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u64_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u64_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u64_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u64_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u64_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_u64_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > -#define __arm_vreinterpretq_u8(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_float16x8_t]: __arm_vreinterpretq_u8_f16 > (__ARM_mve_coerce(__p0, float16x8_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u8_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u8_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u8_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u8_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u8_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u8_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u8_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)), \ > - int (*)[__ARM_mve_type_float32x4_t]: __arm_vreinterpretq_u8_f32 > (__ARM_mve_coerce(__p0, float32x4_t)));}) > - > #define __arm_vstrwq_scatter_base_wb(p0,p1,p2) ({ __typeof(p2) __p2 = > (p2); \ > _Generic( (int (*)[__ARM_mve_typeid(__p2)])0, \ > int (*)[__ARM_mve_type_int32x4_t]: __arm_vstrwq_scatter_base_wb_s32 > (p0, p1, __ARM_mve_coerce(__p2, int32x4_t)), \ > @@ -39931,86 +39711,6 @@ extern void *__ARM_undef; > int (*)[__ARM_mve_type_uint32x4_t]: __arm_vuninitializedq_u32 (), \ > int (*)[__ARM_mve_type_uint64x2_t]: __arm_vuninitializedq_u64 ());}) > > -#define __arm_vreinterpretq_s16(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s16_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s16_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s16_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s16_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s16_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s16_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s16_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_s32(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s32_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s32_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s32_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s32_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s32_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s32_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s32_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_s64(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s64_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s64_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_s64_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s64_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s64_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s64_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s64_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_s8(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_s8_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_s8_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_s8_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_s8_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_s8_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_s8_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_s8_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_u16(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u16_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u16_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u16_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u16_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u16_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u16_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u16_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_u32(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u32_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u32_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u32_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u32_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u32_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u32_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u32_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > -#define __arm_vreinterpretq_u64(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u64_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u64_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u64_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint8x16_t]: __arm_vreinterpretq_u64_u8 > (__ARM_mve_coerce(__p0, uint8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u64_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u64_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u64_s64 > (__ARM_mve_coerce(__p0, int64x2_t)));}) > - > -#define __arm_vreinterpretq_u8(p0) ({ __typeof(p0) __p0 = (p0); \ > - _Generic( (int (*)[__ARM_mve_typeid(__p0)])0, \ > - int (*)[__ARM_mve_type_int16x8_t]: __arm_vreinterpretq_u8_s16 > (__ARM_mve_coerce(__p0, int16x8_t)), \ > - int (*)[__ARM_mve_type_int32x4_t]: __arm_vreinterpretq_u8_s32 > (__ARM_mve_coerce(__p0, int32x4_t)), \ > - int (*)[__ARM_mve_type_int64x2_t]: __arm_vreinterpretq_u8_s64 > (__ARM_mve_coerce(__p0, int64x2_t)), \ > - int (*)[__ARM_mve_type_int8x16_t]: __arm_vreinterpretq_u8_s8 > (__ARM_mve_coerce(__p0, int8x16_t)), \ > - int (*)[__ARM_mve_type_uint16x8_t]: __arm_vreinterpretq_u8_u16 > (__ARM_mve_coerce(__p0, uint16x8_t)), \ > - int (*)[__ARM_mve_type_uint32x4_t]: __arm_vreinterpretq_u8_u32 > (__ARM_mve_coerce(__p0, uint32x4_t)), \ > - int (*)[__ARM_mve_type_uint64x2_t]: __arm_vreinterpretq_u8_u64 > (__ARM_mve_coerce(__p0, uint64x2_t)));}) > - > #define __arm_vabsq_x(p1,p2) ({ __typeof(p1) __p1 = (p1); \ > _Generic( (int (*)[__ARM_mve_typeid(__p1)])0, \ > int (*)[__ARM_mve_type_int8x16_t]: __arm_vabsq_x_s8 > (__ARM_mve_coerce(__p1, int8x16_t), p2), \ > diff --git a/gcc/config/arm/arm_mve_types.h > b/gcc/config/arm/arm_mve_types.h > index 12bb519142f..ae2591faa03 100644 > --- a/gcc/config/arm/arm_mve_types.h > +++ b/gcc/config/arm/arm_mve_types.h > @@ -29,1124 +29,101 @@ typedef float float32_t; > > #pragma GCC arm "arm_mve_types.h" > > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_s32 (int32x4_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_s64 (int64x2_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_s8 (int8x16_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_u16 (uint16x8_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_u32 (uint32x4_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_u64 (uint64x2_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_u8 (uint8x16_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_s16 (int16x8_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_s64 (int64x2_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_s8 (int8x16_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_u16 (uint16x8_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_u32 (uint32x4_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_u64 (uint64x2_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_u8 (uint8x16_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_s16 (int16x8_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_s32 (int32x4_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_s8 (int8x16_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_u16 (uint16x8_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_u32 (uint32x4_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_u64 (uint64x2_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_u8 (uint8x16_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_s16 (int16x8_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_s32 (int32x4_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_s64 (int64x2_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_u16 (uint16x8_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_u32 (uint32x4_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_u64 (uint64x2_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_u8 (uint8x16_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_s16 (int16x8_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_s32 (int32x4_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_s64 (int64x2_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_s8 (int8x16_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_u32 (uint32x4_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_u64 (uint64x2_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_u8 (uint8x16_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_s16 (int16x8_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_s32 (int32x4_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_s64 (int64x2_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_s8 (int8x16_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_u16 (uint16x8_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_u64 (uint64x2_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_u8 (uint8x16_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_s16 (int16x8_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_s32 (int32x4_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_s64 (int64x2_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_s8 (int8x16_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_u16 (uint16x8_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_u32 (uint32x4_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_u8 (uint8x16_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_s16 (int16x8_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_s32 (int32x4_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_s64 (int64x2_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_s8 (int8x16_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_u16 (uint16x8_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_u32 (uint32x4_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_u64 (uint64x2_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_u8 (void) > -{ > - uint8x16_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_u16 (void) > -{ > - uint16x8_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_u32 (void) > -{ > - uint32x4_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_u64 (void) > -{ > - uint64x2_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_s8 (void) > -{ > - int8x16_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_s16 (void) > -{ > - int16x8_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_s32 (void) > -{ > - int32x4_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_s64 (void) > -{ > - int64x2_t __uninit; > - __asm__ ("": "=w"(__uninit)); > - return __uninit; > -} > - > -#if (__ARM_FEATURE_MVE & 2) /* MVE Floating point. */ > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_f16 (float16x8_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32_f32 (float32x4_t __a) > -{ > - return (int32x4_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_f16 (float16x8_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16_f32 (float32x4_t __a) > -{ > - return (int16x8_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_f16 (float16x8_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64_f32 (float32x4_t __a) > -{ > - return (int64x2_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_f16 (float16x8_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8_f32 (float32x4_t __a) > -{ > - return (int8x16_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_f16 (float16x8_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16_f32 (float32x4_t __a) > -{ > - return (uint16x8_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_f16 (float16x8_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32_f32 (float32x4_t __a) > -{ > - return (uint32x4_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_f16 (float16x8_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64_f32 (float32x4_t __a) > -{ > - return (uint64x2_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_f16 (float16x8_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8_f32 (float32x4_t __a) > -{ > - return (uint8x16_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_f32 (float32x4_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_s16 (int16x8_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_s32 (int32x4_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_s64 (int64x2_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_s8 (int8x16_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_u16 (uint16x8_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_u32 (uint32x4_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_u64 (uint64x2_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16_u8 (uint8x16_t __a) > -{ > - return (float16x8_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_f16 (float16x8_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_s16 (int16x8_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_s32 (int32x4_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_s64 (int64x2_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_s8 (int8x16_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_u16 (uint16x8_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_u32 (uint32x4_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_u64 (uint64x2_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32_u8 (uint8x16_t __a) > -{ > - return (float32x4_t) __a; > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_f16 (void) > -{ > - float16x8_t __uninit; > - __asm__ ("": "=w" (__uninit)); > - return __uninit; > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vuninitializedq_f32 (void) > -{ > - float32x4_t __uninit; > - __asm__ ("": "=w" (__uninit)); > - return __uninit; > -} > - > -#endif > - > -#ifdef __cplusplus > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_s16_s32 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_s16_s64 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_s16_s8 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_s16_u16 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_s16_u32 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_s16_u64 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_s16_u8 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_s32_s16 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_s32_s64 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_s32_s8 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_s32_u16 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_s32_u32 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_s32_u64 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_s32_u8 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_s64_s16 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_s64_s32 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_s64_s8 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_s64_u16 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_s64_u32 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_s64_u64 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_s64_u8 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_s8_s16 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_s8_s32 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_s8_s64 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_s8_u16 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_s8_u32 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_s8_u64 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_s8_u8 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_u16_s16 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_u16_s32 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_u16_s64 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_u16_s8 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_u16_u32 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_u16_u64 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_u16_u8 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_u32_s16 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_u32_s32 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_u32_s64 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_u32_s8 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_u32_u16 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_u32_u64 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_u32_u8 (__a); > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_u64_s16 (__a); > -} > - > -__extension__ extern __inline uint64x2_t > +__extension__ extern __inline uint8x16_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (int32x4_t __a) > +__arm_vuninitializedq_u8 (void) > { > - return __arm_vreinterpretq_u64_s32 (__a); > + uint8x16_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint64x2_t > +__extension__ extern __inline uint16x8_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (int64x2_t __a) > +__arm_vuninitializedq_u16 (void) > { > - return __arm_vreinterpretq_u64_s64 (__a); > + uint16x8_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint64x2_t > +__extension__ extern __inline uint32x4_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (int8x16_t __a) > +__arm_vuninitializedq_u32 (void) > { > - return __arm_vreinterpretq_u64_s8 (__a); > + uint32x4_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > __extension__ extern __inline uint64x2_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (uint16x8_t __a) > +__arm_vuninitializedq_u64 (void) > { > - return __arm_vreinterpretq_u64_u16 (__a); > + uint64x2_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint64x2_t > +__extension__ extern __inline int8x16_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (uint32x4_t __a) > +__arm_vuninitializedq_s8 (void) > { > - return __arm_vreinterpretq_u64_u32 (__a); > + int8x16_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint64x2_t > +__extension__ extern __inline int16x8_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (uint8x16_t __a) > +__arm_vuninitializedq_s16 (void) > { > - return __arm_vreinterpretq_u64_u8 (__a); > + int16x8_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint8x16_t > +__extension__ extern __inline int32x4_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (int16x8_t __a) > +__arm_vuninitializedq_s32 (void) > { > - return __arm_vreinterpretq_u8_s16 (__a); > + int32x4_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint8x16_t > +__extension__ extern __inline int64x2_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (int32x4_t __a) > +__arm_vuninitializedq_s64 (void) > { > - return __arm_vreinterpretq_u8_s32 (__a); > + int64x2_t __uninit; > + __asm__ ("": "=w"(__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_u8_s64 (__a); > -} > +#if (__ARM_FEATURE_MVE & 2) /* MVE Floating point. */ > > -__extension__ extern __inline uint8x16_t > +__extension__ extern __inline float16x8_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (int8x16_t __a) > +__arm_vuninitializedq_f16 (void) > { > - return __arm_vreinterpretq_u8_s8 (__a); > + float16x8_t __uninit; > + __asm__ ("": "=w" (__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint8x16_t > +__extension__ extern __inline float32x4_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (uint16x8_t __a) > +__arm_vuninitializedq_f32 (void) > { > - return __arm_vreinterpretq_u8_u16 (__a); > + float32x4_t __uninit; > + __asm__ ("": "=w" (__uninit)); > + return __uninit; > } > > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_u8_u32 (__a); > -} > +#endif > > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_u8_u64 (__a); > -} > +#ifdef __cplusplus > > __extension__ extern __inline uint8x16_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > @@ -1205,244 +182,6 @@ __arm_vuninitializedq (int64x2_t /* __v > ATTRIBUTE UNUSED */) > } > > #if (__ARM_FEATURE_MVE & 2) /* MVE Floating point. */ > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_s32_f16 (__a); > -} > - > -__extension__ extern __inline int32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s32 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_s32_f32 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_s16_f16 (__a); > -} > - > -__extension__ extern __inline int16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s16 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_s16_f32 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_s64_f16 (__a); > -} > - > -__extension__ extern __inline int64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s64 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_s64_f32 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_s8_f16 (__a); > -} > - > -__extension__ extern __inline int8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_s8 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_s8_f32 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_u16_f16 (__a); > -} > - > -__extension__ extern __inline uint16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u16 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_u16_f32 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_u32_f16 (__a); > -} > - > -__extension__ extern __inline uint32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u32 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_u32_f32 (__a); > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_u64_f16 (__a); > -} > - > -__extension__ extern __inline uint64x2_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u64 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_u64_f32 (__a); > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_u8_f16 (__a); > -} > - > -__extension__ extern __inline uint8x16_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_u8 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_u8_f32 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (float32x4_t __a) > -{ > - return __arm_vreinterpretq_f16_f32 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_f16_s16 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_f16_s32 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_f16_s64 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_f16_s8 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_f16_u16 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_f16_u32 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_f16_u64 (__a); > -} > - > -__extension__ extern __inline float16x8_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f16 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_f16_u8 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (float16x8_t __a) > -{ > - return __arm_vreinterpretq_f32_f16 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (int16x8_t __a) > -{ > - return __arm_vreinterpretq_f32_s16 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (int32x4_t __a) > -{ > - return __arm_vreinterpretq_f32_s32 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (int64x2_t __a) > -{ > - return __arm_vreinterpretq_f32_s64 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (int8x16_t __a) > -{ > - return __arm_vreinterpretq_f32_s8 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (uint16x8_t __a) > -{ > - return __arm_vreinterpretq_f32_u16 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (uint32x4_t __a) > -{ > - return __arm_vreinterpretq_f32_u32 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (uint64x2_t __a) > -{ > - return __arm_vreinterpretq_f32_u64 (__a); > -} > - > -__extension__ extern __inline float32x4_t > -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > -__arm_vreinterpretq_f32 (uint8x16_t __a) > -{ > - return __arm_vreinterpretq_f32_u8 (__a); > -} > - > __extension__ extern __inline float16x8_t > __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) > __arm_vuninitializedq (float16x8_t /* __v ATTRIBUTE UNUSED */) > diff --git a/gcc/config/arm/mve.md b/gcc/config/arm/mve.md > index 35eab6c94bf..ab688396f97 100644 > --- a/gcc/config/arm/mve.md > +++ b/gcc/config/arm/mve.md > @@ -10561,3 +10561,21 @@ (define_expand > "vcond_mask_<mode><MVE_vpred>" > } > DONE; > }) > + > +;; Reinterpret operand 1 in operand 0's mode, without changing its contents. > +(define_expand "@arm_mve_reinterpret<mode>" > + [(set (match_operand:MVE_vecs 0 "register_operand") > + (unspec:MVE_vecs > + [(match_operand 1 "arm_any_register_operand")] > + REINTERPRET))] > + "(TARGET_HAVE_MVE && VALID_MVE_SI_MODE (<MODE>mode)) > + || (TARGET_HAVE_MVE_FLOAT && VALID_MVE_SF_MODE > (<MODE>mode))" > + { > + machine_mode src_mode = GET_MODE (operands[1]); > + if (targetm.can_change_mode_class (<MODE>mode, src_mode, > VFP_REGS)) > + { > + emit_move_insn (operands[0], gen_lowpart (<MODE>mode, > operands[1])); > + DONE; > + } > + } > +) > diff --git a/gcc/config/arm/unspecs.md b/gcc/config/arm/unspecs.md > index 84384ee798d..dccda283573 100644 > --- a/gcc/config/arm/unspecs.md > +++ b/gcc/config/arm/unspecs.md > @@ -1255,4 +1255,5 @@ (define_c_enum "unspec" [ > SQRSHRL_64 > SQRSHRL_48 > VSHLCQ_M_ > + REINTERPRET > ]) > diff --git a/gcc/testsuite/g++.target/arm/mve.exp > b/gcc/testsuite/g++.target/arm/mve.exp > index cd824035540..f75ec20ea64 100644 > --- a/gcc/testsuite/g++.target/arm/mve.exp > +++ b/gcc/testsuite/g++.target/arm/mve.exp > @@ -42,8 +42,12 @@ set dg-do-what-default "assemble" > dg-init > > # Main loop. > -dg-runtest [lsort [glob -nocomplain > $srcdir/$subdir/../../gcc.target/arm/mve/intrinsics/*.\[cCS\]]] \ > - "" $DEFAULT_CXXFLAGS > +set gcc_subdir [string replace $subdir 0 2 gcc] > +set files [glob -nocomplain \ > + "$srcdir/$subdir/../../gcc.target/arm/mve/intrinsics/*.\[cCS\]" \ > + "$srcdir/$gcc_subdir/mve/general/*.\[cCS\]" \ > + "$srcdir/$subdir/mve/general-c++/*.\[cCS\]"] > +dg-runtest [lsort $files] "" $DEFAULT_CXXFLAGS > > # All done. > set dg-do-what-default ${save-dg-do-what-default} > diff --git a/gcc/testsuite/g++.target/arm/mve/general-c++/nomve_fp_1.c > b/gcc/testsuite/g++.target/arm/mve/general-c++/nomve_fp_1.c > new file mode 100644 > index 00000000000..e0692ceb8c8 > --- /dev/null > +++ b/gcc/testsuite/g++.target/arm/mve/general-c++/nomve_fp_1.c > @@ -0,0 +1,15 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target arm_v8_1m_mve_ok } */ > +/* Do not use dg-add-options arm_v8_1m_mve, because this might expand > to "", > + which could imply mve+fp depending on the user settings. We want to > make > + sure the '+fp' extension is not enabled. */ > +/* { dg-options "-mfpu=auto -march=armv8.1-m.main+mve" } */ > + > +#include <arm_mve.h> > + > +void > +f1 (uint8x16_t v) > +{ > + vreinterpretq_f16 (v); /* { dg-error {ACLE function 'void > vreinterpretq_f16\(uint8x16_t\)' requires ISA extension 'mve.fp'} } */ > + /* { dg-message {note: you can enable mve.fp by using the command-line > option '-march', or by using the 'target' attribute or pragma} "" {target > *-*-*} .- > 1 } */ > +} > diff --git a/gcc/testsuite/g++.target/arm/mve/general-c++/vreinterpretq_1.C > b/gcc/testsuite/g++.target/arm/mve/general-c++/vreinterpretq_1.C > new file mode 100644 > index 00000000000..8b29ee58163 > --- /dev/null > +++ b/gcc/testsuite/g++.target/arm/mve/general-c++/vreinterpretq_1.C > @@ -0,0 +1,25 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */ > +/* { dg-add-options arm_v8_1m_mve_fp } */ > + > +#include <arm_mve.h> > + > +void > +f1 (int8x16_t s8, uint16x8_t u16, float32x4_t f32) > +{ > + __arm_vreinterpretq_s8 (); /* { dg-error {no matching function for call to > '__arm_vreinterpretq_s8\(\)'} } */ > + __arm_vreinterpretq_s8 (s8, s8); /* { dg-error {no matching function for > call > to '__arm_vreinterpretq_s8\(int8x16_t\&, int8x16_t\&\)'} } */ > + __arm_vreinterpretq_s8 (0); /* { dg-error {no matching function for call to > '__arm_vreinterpretq_s8\(int\)'} } */ > + __arm_vreinterpretq_s8 (s8); /* { dg-error {no matching function for call > to > '__arm_vreinterpretq_s8\(int8x16_t\&\)'} } */ > + __arm_vreinterpretq_s8 (u16); > + __arm_vreinterpretq_u16 (); /* { dg-error {no matching function for call to > '__arm_vreinterpretq_u16\(\)'} } */ > + __arm_vreinterpretq_u16 (u16, u16); /* { dg-error {no matching function > for call to '__arm_vreinterpretq_u16\(uint16x8_t\&, uint16x8_t\&\)'} } */ > + __arm_vreinterpretq_u16 (0); /* { dg-error {no matching function for call > to > '__arm_vreinterpretq_u16\(int\)'} } */ > + __arm_vreinterpretq_u16 (u16); /* { dg-error {no matching function for call > to '__arm_vreinterpretq_u16\(uint16x8_t\&\)'} } */ > + __arm_vreinterpretq_u16 (f32); > + __arm_vreinterpretq_f32 (); /* { dg-error {no matching function for call to > '__arm_vreinterpretq_f32\(\)'} } */ > + __arm_vreinterpretq_f32 (f32, f32); /* { dg-error {no matching function for > call to '__arm_vreinterpretq_f32\(float32x4_t\&, float32x4_t\&\)'} } */ > + __arm_vreinterpretq_f32 (0); /* { dg-error {no matching function for call > to > '__arm_vreinterpretq_f32\(int\)'} } */ > + __arm_vreinterpretq_f32 (f32); /* { dg-error {no matching function for call > to '__arm_vreinterpretq_f32\(float32x4_t\&\)'} } */ > + __arm_vreinterpretq_f32 (s8); > +} > diff --git a/gcc/testsuite/gcc.target/arm/mve/general-c/nomve_fp_1.c > b/gcc/testsuite/gcc.target/arm/mve/general-c/nomve_fp_1.c > new file mode 100644 > index 00000000000..21c2af16a61 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/arm/mve/general-c/nomve_fp_1.c > @@ -0,0 +1,15 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target arm_v8_1m_mve_ok } */ > +/* Do not use dg-add-options arm_v8_1m_mve, because this might expand > to "", > + which could imply mve+fp depending on the user settings. We want to > make > + sure the '+fp' extension is not enabled. */ > +/* { dg-options "-mfpu=auto -march=armv8.1-m.main+mve" } */ > + > +#include <arm_mve.h> > + > +void > +foo (uint8x16_t v) > +{ > + vreinterpretq_f16 (v); /* { dg-error {ACLE function > '__arm_vreinterpretq_f16_u8' requires ISA extension 'mve.fp'} } */ > + /* { dg-message {note: you can enable mve.fp by using the command-line > option '-march', or by using the 'target' attribute or pragma} "" {target > *-*-*} .- > 1 } */ > +} > diff --git a/gcc/testsuite/gcc.target/arm/mve/general-c/vreinterpretq_1.c > b/gcc/testsuite/gcc.target/arm/mve/general-c/vreinterpretq_1.c > new file mode 100644 > index 00000000000..0297bd50198 > --- /dev/null > +++ b/gcc/testsuite/gcc.target/arm/mve/general-c/vreinterpretq_1.c > @@ -0,0 +1,25 @@ > +/* { dg-do compile } */ > +/* { dg-require-effective-target arm_v8_1m_mve_fp_ok } */ > +/* { dg-add-options arm_v8_1m_mve_fp } */ > + > +#include <arm_mve.h> > + > +void > +f1 (int8x16_t s8, uint16x8_t u16, float32x4_t f32) > +{ > + __arm_vreinterpretq_s8 (); /* { dg-error {too few arguments to function > '__arm_vreinterpretq_s8'} } */ > + __arm_vreinterpretq_s8 (s8, s8); /* { dg-error {too many arguments to > function '__arm_vreinterpretq_s8'} } */ > + __arm_vreinterpretq_s8 (0); /* { dg-error {passing 'int' to argument 1 of > '__arm_vreinterpretq_s8', which expects an MVE vector type} } */ > + __arm_vreinterpretq_s8 (s8); /* { dg-error {'__arm_vreinterpretq_s8' has > no form that takes 'int8x16_t' arguments} } */ > + __arm_vreinterpretq_s8 (u16); > + __arm_vreinterpretq_u16 (); /* { dg-error {too few arguments to function > '__arm_vreinterpretq_u16'} } */ > + __arm_vreinterpretq_u16 (u16, u16); /* { dg-error {too many arguments to > function '__arm_vreinterpretq_u16'} } */ > + __arm_vreinterpretq_u16 (0); /* { dg-error {passing 'int' to argument 1 of > '__arm_vreinterpretq_u16', which expects an MVE vector type} } */ > + __arm_vreinterpretq_u16 (u16); /* { dg-error {'__arm_vreinterpretq_u16' > has no form that takes 'uint16x8_t' arguments} } */ > + __arm_vreinterpretq_u16 (f32); > + __arm_vreinterpretq_f32 (); /* { dg-error {too few arguments to function > '__arm_vreinterpretq_f32'} } */ > + __arm_vreinterpretq_f32 (f32, f32); /* { dg-error {too many arguments to > function '__arm_vreinterpretq_f32'} } */ > + __arm_vreinterpretq_f32 (0); /* { dg-error {passing 'int' to argument 1 of > '__arm_vreinterpretq_f32', which expects an MVE vector type} } */ > + __arm_vreinterpretq_f32 (f32); /* { dg-error {'__arm_vreinterpretq_f32' > has no form that takes 'float32x4_t' arguments} } */ > + __arm_vreinterpretq_f32 (s8); > +} > -- > 2.34.1
RE: [PATCH v2 03/22] arm: [MVE intrinsics] Rework vreinterpretq
Kyrylo Tkachov via Gcc-patches Wed, 03 May 2023 07:52:56 -0700
- [PATCH v2 03/22] arm: [MVE intrinsics] Rew... Christophe Lyon via Gcc-patches
- RE: [PATCH v2 03/22] arm: [MVE intrin... Kyrylo Tkachov via Gcc-patches