https://gcc.gnu.org/g:4290fbf37252244d3b5bfe722401b2beb09ec225
commit r17-6217-g4290fbf37252244d3b5bfe722401b2beb09ec225 Author: Karl Meakin <[email protected]> Date: Thu Sep 17 11:09:00 2026 +0000 aarch64: Port NEON multiply-accumulate intrinsics to pragma-based framework Ports the NEON multiply-accumulate intrinsics to the pragma-based framework, and add corresponding assembly tests for each variant. Also deletes builtin functions that were only used by `arm_neon.h`: * __builtin_aarch64_mla * __builtin_aarch64_mla_n * __builtin_aarch64_mls * __builtin_aarch64_mls_n * __builtin_aarch64_float_mla * __builtin_aarch64_float_mla_n * __builtin_aarch64_float_mls * __builtin_aarch64_float_mls_n * __builtin_aarch64_float_mla_lane * __builtin_aarch64_float_mla_laneq * __builtin_aarch64_float_mls_lane * __builtin_aarch64_float_mls_laneq and their corresponding expanders in `aarch64-simd.md`. gcc/ChangeLog: * config/aarch64/aarch64-acle-builtins.h (TYPES_neon_mla, TYPES_neon_mla_lane): New type arrays. * config/aarch64/aarch64-neon-builtins-base.cc (struct multiply_accumulate): New struct. * config/aarch64/aarch64-neon-builtins-base.def (vmla, vmlaq, vmla_n, vmlaq_n, vmla_lane, vmla_laneq, vmlaq_lane, vmlaq_laneq, vmls, vmlsq, vmls_n, vmlsq_n, vmls_lane, vmls_laneq, vmlsq_lane, vmlsq_laneq): New function declarations. * config/aarch64/aarch64-simd-builtins.def: Delete `__builtin_aarch64_*` functions. * config/aarch64/aarch64-simd.md (aarch64_float_mla<mode>, aarch64_float_mls<mode>, aarch64_float_mla_n<mode>, aarch64_float_mls_n<mode>, aarch64_float_mla_lane<mode>, aarch64_float_mls_lane<mode>, aarch64_float_mla_laneq<mode>, aarch64_float_mls_laneq<mode>): Delete now-obsolete insns. * config/aarch64/arm_neon.h: (vmla_n_f32, vmla_n_s16, vmla_n_s32, vmla_n_u16, vmla_n_u32, vmla_s8, vmla_s16, vmla_s32, vmla_u8, vmla_u16, vmla_u32, vmlaq_n_f32, vmlaq_n_s16, vmlaq_n_s32, vmlaq_n_u16, vmlaq_n_u32, vmlaq_s8, vmlaq_s16, vmlaq_s32, vmlaq_u8, vmlaq_u16, vmlaq_u32, vmls_n_f32, vmls_n_s16, vmls_n_s32, vmls_n_u16, vmls_n_u32, vmls_s8, vmls_s16, vmls_s32, vmls_u8, vmls_u16, vmls_u32, vmlsq_n_f32, vmlsq_n_s16, vmlsq_n_s32, vmlsq_n_u16, vmlsq_n_u32, vmlsq_s8, vmlsq_s16, vmlsq_s32, vmlsq_u8, vmlsq_u16, vmlsq_u32, vmla_f32, vmla_f64, vmlaq_f32, vmlaq_f64, vmla_lane_f32, vmla_lane_s16, vmla_lane_s32, vmla_lane_u16, vmla_lane_u32, vmla_laneq_f32, vmla_laneq_s16, vmla_laneq_s32, vmla_laneq_u16, vmla_laneq_u32, vmlaq_lane_f32, vmlaq_lane_s16, vmlaq_lane_s32, vmlaq_lane_u16, vmlaq_lane_u32, vmlaq_laneq_f32, vmlaq_laneq_s16, vmlaq_laneq_s32, vmlaq_laneq_u16, vmlaq_laneq_u32, vmls_f32, vmls_f64, vmlsq_f32, vmlsq_f64, vmls_lane_f32, vmls_lane_s16, vmls_lane_s32, vmls_lane_u16, vmls_lane_u32, vmls_laneq_f32, vmls_laneq_s16, vmls_laneq_s32, vmls_laneq_u16, vmls_laneq_u32, vmlsq_lane_f32, vmlsq_lane_s16, vmlsq_lane_s32, vmlsq_lane_u16, vmlsq_lane_u32, vmlsq_laneq_f32, vmlsq_laneq_s16, vmlsq_laneq_s32, vmlsq_laneq_u16, vmlsq_laneq_u32): Delete function definitions. gcc/testsuite/ChangeLog: * gcc.target/aarch64/neon/vmla.c: New test. * gcc.target/aarch64/neon/vmls.c: New test. Diff: --- gcc/config/aarch64/aarch64-acle-builtins.h | 15 + gcc/config/aarch64/aarch64-neon-builtins-base.cc | 74 +++ gcc/config/aarch64/aarch64-neon-builtins-base.def | 21 + gcc/config/aarch64/aarch64-simd-builtins.def | 22 - gcc/config/aarch64/aarch64-simd.md | 146 ----- gcc/config/aarch64/arm_neon.h | 703 ---------------------- gcc/testsuite/gcc.target/aarch64/neon/vmla.c | 350 +++++++++++ gcc/testsuite/gcc.target/aarch64/neon/vmls.c | 350 +++++++++++ 8 files changed, 810 insertions(+), 871 deletions(-) diff --git a/gcc/config/aarch64/aarch64-acle-builtins.h b/gcc/config/aarch64/aarch64-acle-builtins.h index bd07225f6c1d..91c922fdae6f 100644 --- a/gcc/config/aarch64/aarch64-acle-builtins.h +++ b/gcc/config/aarch64/aarch64-acle-builtins.h @@ -1981,6 +1981,19 @@ function_expander::result_mode () const #define TYPES_b_neon(S, D, T) \ S (p8), S (s8), S (u8) +/* _s8 _s16 _s32 + _u8 _u16 _u32 + _f32 _f64. */ +#define TYPES_neon_mla(S, D, T) \ + TYPES_bhs_integer (S, D, T), \ + TYPES_sd_float (S, D, T) + +/* _s16 _u16 + _s32 _u32 _f32. */ +#define TYPES_neon_mla_lane(S, D, T) \ + TYPES_hs_integer (S, D, T), \ + TYPES_s_float (S, D, T) + /* _p8 _s8 _u8 _mf8. */ #define TYPES_neon_rev16(S, D, T) \ S (p8), S (s8), S (u8), S (mf8) @@ -2246,6 +2259,8 @@ DEF_SVE_TYPES_ARRAY (bhdq_poly); DEF_SVE_TYPES_ARRAY (bhs_neon); DEF_SVE_TYPES_ARRAY (bhsd_neon); DEF_SVE_TYPES_ARRAY (neon_copy_lane); +DEF_SVE_TYPES_ARRAY (neon_mla_lane); +DEF_SVE_TYPES_ARRAY (neon_mla); DEF_SVE_TYPES_ARRAY (neon_rev16); DEF_SVE_TYPES_ARRAY (neon_rev32); DEF_SVE_TYPES_ARRAY (neon_rev64); diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.cc b/gcc/config/aarch64/aarch64-neon-builtins-base.cc index 4b66a93d0f01..4be73c54f0a2 100644 --- a/gcc/config/aarch64/aarch64-neon-builtins-base.cc +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.cc @@ -1140,6 +1140,69 @@ public: } }; +// vmla (acc, lhs, rhs) = acc + lhs * rhs +// vmls (acc, lhs, rhs) = acc - lhs * rhs +template <arith_kind kind, bool negate> +struct multiply_accumulate : public gimple_function_base +{ + gimple *fold (gimple_folder &f) const override + { + tree acc = gimple_call_arg (f.call, 0); + tree lhs = gimple_call_arg (f.call, 1); + tree rhs = gimple_call_arg (f.call, 2); + tree lane_index; + tree ret_type = TREE_TYPE (f.lhs); + + switch (kind) + { + case arith_kind::normal: + break; + + case arith_kind::dup: + rhs = f.force_val (build_vec_dup (TREE_TYPE (lhs), rhs)); + break; + + case arith_kind::lane: + lane_index = gimple_call_arg (f.call, 3); + rhs = f.force_val (build_lane_get (rhs, lane_index)); + if (VECTOR_TYPE_P (ret_type)) + rhs = f.force_val (build_vec_dup (TREE_TYPE (lhs), rhs)); + break; + } + + tree_code code = negate ? MINUS_EXPR : PLUS_EXPR; + if (FLOAT_TYPE_P (ret_type)) + { + /* Use a PAREN_EXPR (equivalent to C's `__builtin_assoc_barrier`) to + prevent floating-point add and mul being transformed to an FMA. */ + auto mul = f.force_val (fold_build2 (MULT_EXPR, ret_type, lhs, rhs)); + auto paren = f.force_val (fold_build1 (PAREN_EXPR, ret_type, mul)); + auto addsub = f.force_val (fold_build2 (code, ret_type, acc, paren)); + return this->maybe_optimization_barrier (f, addsub); + } + else if (!TYPE_UNSIGNED (ret_type)) + { + /* Convert signed inputs to unsigned first, to avoid undefined behaviour + on overflow. */ + auto signed_type = ret_type; + auto unsigned_type = unsigned_type_for (signed_type); + lhs = f.force_val (build_cast (unsigned_type, lhs)); + rhs = f.force_val (build_cast (unsigned_type, rhs)); + acc = f.force_val (build_cast (unsigned_type, acc)); + auto mul = f.force_val (fold_build2 (MULT_EXPR, unsigned_type, lhs, rhs)); + auto addsub = f.force_val (fold_build2 (code, unsigned_type, acc, mul)); + return gimple_build_assign (f.lhs, build_cast (signed_type, addsub)); + } + else + { + auto mul = f.force_val (fold_build2 (MULT_EXPR, ret_type, lhs, rhs)); + return gimple_build_assign (f.lhs, code, acc, mul); + } + } +}; +template <arith_kind kind> using gimple_mla = multiply_accumulate<kind, false>; +template <arith_kind kind> using gimple_mls = multiply_accumulate<kind, true>; + // Reinterpret NEON_FUNCTION (vreinterpret, vreinterpretq, gimple_reinterpret,) @@ -1170,6 +1233,17 @@ NEON_FUNCTION (vmul_lane, vmul_laneq, vmuls_lane, vmuls_laneq, vmuld_lane, vmuld_laneq, gimple_arith_lane, (MULT_EXPR)) +// Multiply-accumulate +NEON_FUNCTION (vmla, vmlaq, gimple_mla<arith_kind::normal>,) +NEON_FUNCTION (vmla_n, vmlaq_n, gimple_mla<arith_kind::dup>,) +NEON_FUNCTION (vmla_lane, vmla_laneq, vmlaq_lane, vmlaq_laneq, + gimple_mla<arith_kind::lane>,) + +NEON_FUNCTION (vmls, vmlsq, gimple_mls<arith_kind::normal>,) +NEON_FUNCTION (vmls_n, vmlsq_n, gimple_mls<arith_kind::dup>,) +NEON_FUNCTION (vmls_lane, vmls_laneq, vmlsq_lane, vmlsq_laneq, + gimple_mls<arith_kind::lane>,) + // Saturating arithmetic NEON_FUNCTION (vqaddb, vqaddh, vqadds, vqaddd, vqadd, vqaddq, gimple_ifn, (IFN_SAT_ADD)) NEON_FUNCTION (vqsubb, vqsubh, vqsubs, vqsubd, vqsub, vqsubq, gimple_ifn, (IFN_SAT_SUB)) diff --git a/gcc/config/aarch64/aarch64-neon-builtins-base.def b/gcc/config/aarch64/aarch64-neon-builtins-base.def index db9defc454c9..47510bc39b11 100644 --- a/gcc/config/aarch64/aarch64-neon-builtins-base.def +++ b/gcc/config/aarch64/aarch64-neon-builtins-base.def @@ -276,6 +276,26 @@ DEF_NEON_FUNCTION (vrndx, h_float, ("D0,D0")) DEF_NEON_FUNCTION (vrndxq, h_float, ("Q0,Q0")) #undef REQUIRED_EXTENSIONS +// Multiply-accumulate +#define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD) +DEF_NEON_FUNCTION (vmla, neon_mla, ("D0,D0,D0,D0")) +DEF_NEON_FUNCTION (vmlaq, neon_mla, ("Q0,Q0,Q0,Q0")) +DEF_NEON_FUNCTION (vmla_n, neon_mla_lane, ("D0,D0,D0,s0")) +DEF_NEON_FUNCTION (vmlaq_n, neon_mla_lane, ("Q0,Q0,Q0,s0")) +DEF_NEON_FUNCTION (vmla_lane, neon_mla_lane, ("D0,D0,D0,D0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmla_laneq, neon_mla_lane, ("D0,D0,D0,Q0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmlaq_lane, neon_mla_lane, ("Q0,Q0,Q0,D0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmlaq_laneq, neon_mla_lane, ("Q0,Q0,Q0,Q0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmls, neon_mla, ("D0,D0,D0,D0")) +DEF_NEON_FUNCTION (vmlsq, neon_mla, ("Q0,Q0,Q0,Q0")) +DEF_NEON_FUNCTION (vmls_n, neon_mla_lane, ("D0,D0,D0,s0")) +DEF_NEON_FUNCTION (vmlsq_n, neon_mla_lane, ("Q0,Q0,Q0,s0")) +DEF_NEON_FUNCTION (vmls_lane, neon_mla_lane, ("D0,D0,D0,D0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmls_laneq, neon_mla_lane, ("D0,D0,D0,Q0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmlsq_lane, neon_mla_lane, ("Q0,Q0,Q0,D0,ss32", lane<3>)) +DEF_NEON_FUNCTION (vmlsq_laneq, neon_mla_lane, ("Q0,Q0,Q0,Q0,ss32", lane<3>)) +#undef REQUIRED_EXTENSIONS + // Bitwise operations #define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_SIMD) DEF_NEON_FUNCTION (vand, all_integer, ("D0,D0,D0")) @@ -455,6 +475,7 @@ DEF_NEON_FUNCTION (vsm3partw2q, s_unsigned, ("Q0,Q0,Q0,Q0")) DEF_NEON_FUNCTION (vsm4eq, s_unsigned, ("Q0,Q0,Q0")) DEF_NEON_FUNCTION (vsm4ekeyq, s_unsigned, ("Q0,Q0,Q0")) #undef REQUIRED_EXTENSIONS + // FP16 matrix-multiply accumulate (widening) #define REQUIRED_EXTENSIONS nonstreaming_only (AARCH64_FL_F16F32MM) DEF_NEON_FUNCTION (vmmlaq, cvt_f32_f16, ("Q0,Q0,Q1,Q1")) diff --git a/gcc/config/aarch64/aarch64-simd-builtins.def b/gcc/config/aarch64/aarch64-simd-builtins.def index 0a84dfd459d1..1fb7c7639907 100644 --- a/gcc/config/aarch64/aarch64-simd-builtins.def +++ b/gcc/config/aarch64/aarch64-simd-builtins.def @@ -233,20 +233,6 @@ BUILTIN_VQN (UNOP, xtn, 0, DEFAULT) BUILTIN_VQN (UNOPU, xtn, 0, DEFAULT) - /* Implemented by aarch64_mla<mode>. */ - BUILTIN_VDQ_BHSI (TERNOP, mla, 0, DEFAULT) - BUILTIN_VDQ_BHSI (TERNOPU, mla, 0, DEFAULT) - /* Implemented by aarch64_mla_n<mode>. */ - BUILTIN_VDQHS (TERNOP, mla_n, 0, DEFAULT) - BUILTIN_VDQHS (TERNOPU, mla_n, 0, DEFAULT) - - /* Implemented by aarch64_mls<mode>. */ - BUILTIN_VDQ_BHSI (TERNOP, mls, 0, DEFAULT) - BUILTIN_VDQ_BHSI (TERNOPU, mls, 0, DEFAULT) - /* Implemented by aarch64_mls_n<mode>. */ - BUILTIN_VDQHS (TERNOP, mls_n, 0, DEFAULT) - BUILTIN_VDQHS (TERNOPU, mls_n, 0, DEFAULT) - BUILTIN_VQN (SHIFTIMM, shrn_n, 0, DEFAULT) BUILTIN_VQN (USHIFTIMM, shrn_n, 0, DEFAULT) @@ -689,14 +675,6 @@ BUILTIN_VHSDF (TERNOP, fnma, 4, FP) VAR1 (TERNOP, fnma, 4, FP, hf) - BUILTIN_VDQF_DF (TERNOP, float_mla, 0, FP) - BUILTIN_VDQF_DF (TERNOP, float_mls, 0, FP) - BUILTIN_VDQSF (TERNOP, float_mla_n, 0, FP) - BUILTIN_VDQSF (TERNOP, float_mls_n, 0, FP) - BUILTIN_VDQSF (QUADOP_LANE, float_mla_lane, 0, FP) - BUILTIN_VDQSF (QUADOP_LANE, float_mls_lane, 0, FP) - BUILTIN_VDQSF (QUADOP_LANE, float_mla_laneq, 0, FP) - BUILTIN_VDQSF (QUADOP_LANE, float_mls_laneq, 0, FP) /* Implemented by aarch64_crypto_pmull<mode>. */ VAR1 (BINOPP, crypto_pmull, 0, DEFAULT, di) VAR1 (BINOPP, crypto_pmull, 0, DEFAULT, v2di) diff --git a/gcc/config/aarch64/aarch64-simd.md b/gcc/config/aarch64/aarch64-simd.md index 7691fe2fdcd5..696d229d7d33 100644 --- a/gcc/config/aarch64/aarch64-simd.md +++ b/gcc/config/aarch64/aarch64-simd.md @@ -3059,152 +3059,6 @@ [(set_attr "type" "neon_fp_abs_<stype><q>")] ) -(define_expand "aarch64_float_mla<mode>" - [(set (match_operand:VDQF_DF 0 "register_operand") - (plus:VDQF_DF - (mult:VDQF_DF - (match_operand:VDQF_DF 2 "register_operand") - (match_operand:VDQF_DF 3 "register_operand")) - (match_operand:VDQF_DF 1 "register_operand")))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul<mode>3 (scratch, operands[2], operands[3])); - emit_insn (gen_add<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mls<mode>" - [(set (match_operand:VDQF_DF 0 "register_operand") - (minus:VDQF_DF - (match_operand:VDQF_DF 1 "register_operand") - (mult:VDQF_DF - (match_operand:VDQF_DF 2 "register_operand") - (match_operand:VDQF_DF 3 "register_operand"))))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul<mode>3 (scratch, operands[2], operands[3])); - emit_insn (gen_sub<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mla_n<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (plus:VDQSF - (mult:VDQSF - (vec_duplicate:VDQSF - (match_operand:<VEL> 3 "register_operand")) - (match_operand:VDQSF 2 "register_operand")) - (match_operand:VDQSF 1 "register_operand")))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_n<mode>3 (scratch, operands[2], operands[3])); - emit_insn (gen_add<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mls_n<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (minus:VDQSF - (match_operand:VDQSF 1 "register_operand") - (mult:VDQSF - (vec_duplicate:VDQSF - (match_operand:<VEL> 3 "register_operand")) - (match_operand:VDQSF 2 "register_operand"))))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_n<mode>3 (scratch, operands[2], operands[3])); - emit_insn (gen_sub<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mla_lane<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (plus:VDQSF - (mult:VDQSF - (vec_duplicate:VDQSF - (vec_select:<VEL> - (match_operand:V2SF 3 "register_operand") - (parallel [(match_operand:SI 4 "immediate_operand")]))) - (match_operand:VDQSF 2 "register_operand")) - (match_operand:VDQSF 1 "register_operand")))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_lane<mode>3 (scratch, operands[2], - operands[3], operands[4])); - emit_insn (gen_add<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mls_lane<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (minus:VDQSF - (match_operand:VDQSF 1 "register_operand") - (mult:VDQSF - (vec_duplicate:VDQSF - (vec_select:<VEL> - (match_operand:V2SF 3 "register_operand") - (parallel [(match_operand:SI 4 "immediate_operand")]))) - (match_operand:VDQSF 2 "register_operand"))))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_lane<mode>3 (scratch, operands[2], - operands[3], operands[4])); - emit_insn (gen_sub<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mla_laneq<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (plus:VDQSF - (mult:VDQSF - (vec_duplicate:VDQSF - (vec_select:<VEL> - (match_operand:V4SF 3 "register_operand") - (parallel [(match_operand:SI 4 "immediate_operand")]))) - (match_operand:VDQSF 2 "register_operand")) - (match_operand:VDQSF 1 "register_operand")))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_laneq<mode>3 (scratch, operands[2], - operands[3], operands[4])); - emit_insn (gen_add<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - -(define_expand "aarch64_float_mls_laneq<mode>" - [(set (match_operand:VDQSF 0 "register_operand") - (minus:VDQSF - (match_operand:VDQSF 1 "register_operand") - (mult:VDQSF - (vec_duplicate:VDQSF - (vec_select:<VEL> - (match_operand:V4SF 3 "register_operand") - (parallel [(match_operand:SI 4 "immediate_operand")]))) - (match_operand:VDQSF 2 "register_operand"))))] - "TARGET_SIMD" - { - rtx scratch = gen_reg_rtx (<MODE>mode); - emit_insn (gen_mul_laneq<mode>3 (scratch, operands[2], - operands[3], operands[4])); - emit_insn (gen_sub<mode>3 (operands[0], operands[1], scratch)); - DONE; - } -) - (define_insn "fma<mode>4<vczle><vczbe>" [(set (match_operand:VHSDF 0 "register_operand" "=w") (fma:VHSDF (match_operand:VHSDF 1 "register_operand" "w") diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h index a0dca5e8da9e..47a0b541398c 100644 --- a/gcc/config/aarch64/arm_neon.h +++ b/gcc/config/aarch64/arm_neon.h @@ -1290,83 +1290,6 @@ vcvtxd_f32_f64 (float64_t __a) return __builtin_aarch64_float_trunc_rodd_df (__a); } -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_n_f32 (float32x2_t __a, float32x2_t __b, float32_t __c) -{ - return __builtin_aarch64_float_mla_nv2sf (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_n_s16 (int16x4_t __a, int16x4_t __b, int16_t __c) -{ - return __builtin_aarch64_mla_nv4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_n_s32 (int32x2_t __a, int32x2_t __b, int32_t __c) -{ - return __builtin_aarch64_mla_nv2si (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_n_u16 (uint16x4_t __a, uint16x4_t __b, uint16_t __c) -{ - return __builtin_aarch64_mla_nv4hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_n_u32 (uint32x2_t __a, uint32x2_t __b, uint32_t __c) -{ - return __builtin_aarch64_mla_nv2si_uuuu (__a, __b, __c); -} - -__extension__ extern __inline int8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_s8 (int8x8_t __a, int8x8_t __b, int8x8_t __c) -{ - return __builtin_aarch64_mlav8qi (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_s16 (int16x4_t __a, int16x4_t __b, int16x4_t __c) -{ - return __builtin_aarch64_mlav4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_s32 (int32x2_t __a, int32x2_t __b, int32x2_t __c) -{ - return __builtin_aarch64_mlav2si (__a, __b, __c); -} - -__extension__ extern __inline uint8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_u8 (uint8x8_t __a, uint8x8_t __b, uint8x8_t __c) -{ - return __builtin_aarch64_mlav8qi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_u16 (uint16x4_t __a, uint16x4_t __b, uint16x4_t __c) -{ - return __builtin_aarch64_mlav4hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_u32 (uint32x2_t __a, uint32x2_t __b, uint32x2_t __c) -{ - return __builtin_aarch64_mlav2si_uuuu (__a, __b, __c); -} - __extension__ extern __inline int32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlal_high_lane_s16(int32x4_t __a, int16x8_t __b, int16x4_t __v, @@ -1627,160 +1550,6 @@ vmlal_u32 (uint64x2_t __a, uint32x2_t __b, uint32x2_t __c) return __builtin_aarch64_umlalv2si_uuuu (__a, __b, __c); } -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_n_f32 (float32x4_t __a, float32x4_t __b, float32_t __c) -{ - return __builtin_aarch64_float_mla_nv4sf (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_n_s16 (int16x8_t __a, int16x8_t __b, int16_t __c) -{ - return __builtin_aarch64_mla_nv8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_n_s32 (int32x4_t __a, int32x4_t __b, int32_t __c) -{ - return __builtin_aarch64_mla_nv4si (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_n_u16 (uint16x8_t __a, uint16x8_t __b, uint16_t __c) -{ - return __builtin_aarch64_mla_nv8hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_n_u32 (uint32x4_t __a, uint32x4_t __b, uint32_t __c) -{ - return __builtin_aarch64_mla_nv4si_uuuu (__a, __b, __c); -} - -__extension__ extern __inline int8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_s8 (int8x16_t __a, int8x16_t __b, int8x16_t __c) -{ - return __builtin_aarch64_mlav16qi (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_s16 (int16x8_t __a, int16x8_t __b, int16x8_t __c) -{ - return __builtin_aarch64_mlav8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_s32 (int32x4_t __a, int32x4_t __b, int32x4_t __c) -{ - return __builtin_aarch64_mlav4si (__a, __b, __c); -} - -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_u8 (uint8x16_t __a, uint8x16_t __b, uint8x16_t __c) -{ - return __builtin_aarch64_mlav16qi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_u16 (uint16x8_t __a, uint16x8_t __b, uint16x8_t __c) -{ - return __builtin_aarch64_mlav8hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_u32 (uint32x4_t __a, uint32x4_t __b, uint32x4_t __c) -{ - return __builtin_aarch64_mlav4si_uuuu (__a, __b, __c); -} - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_n_f32 (float32x2_t __a, float32x2_t __b, float32_t __c) -{ - return __builtin_aarch64_float_mls_nv2sf (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_n_s16 (int16x4_t __a, int16x4_t __b, int16_t __c) -{ - return __builtin_aarch64_mls_nv4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_n_s32 (int32x2_t __a, int32x2_t __b, int32_t __c) -{ - return __builtin_aarch64_mls_nv2si (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_n_u16 (uint16x4_t __a, uint16x4_t __b, uint16_t __c) -{ - return __builtin_aarch64_mls_nv4hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_n_u32 (uint32x2_t __a, uint32x2_t __b, uint32_t __c) -{ - return __builtin_aarch64_mls_nv2si_uuuu (__a, __b, __c); -} - -__extension__ extern __inline int8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_s8 (int8x8_t __a, int8x8_t __b, int8x8_t __c) -{ - return __builtin_aarch64_mlsv8qi (__a, __b, __c); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_s16 (int16x4_t __a, int16x4_t __b, int16x4_t __c) -{ - return __builtin_aarch64_mlsv4hi (__a, __b, __c); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_s32 (int32x2_t __a, int32x2_t __b, int32x2_t __c) -{ - return __builtin_aarch64_mlsv2si (__a, __b, __c); -} - -__extension__ extern __inline uint8x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_u8 (uint8x8_t __a, uint8x8_t __b, uint8x8_t __c) -{ - return __builtin_aarch64_mlsv8qi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_u16 (uint16x4_t __a, uint16x4_t __b, uint16x4_t __c) -{ - return __builtin_aarch64_mlsv4hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_u32 (uint32x2_t __a, uint32x2_t __b, uint32x2_t __c) -{ - return __builtin_aarch64_mlsv2si_uuuu (__a, __b, __c); -} - __extension__ extern __inline int32x4_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmlsl_high_lane_s16(int32x4_t __a, int16x8_t __b, int16x4_t __v, @@ -2045,83 +1814,6 @@ vmlsl_u32 (uint64x2_t __a, uint32x2_t __b, uint32x2_t __c) return __builtin_aarch64_umlslv2si_uuuu (__a, __b, __c); } -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_n_f32 (float32x4_t __a, float32x4_t __b, float32_t __c) -{ - return __builtin_aarch64_float_mls_nv4sf (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_n_s16 (int16x8_t __a, int16x8_t __b, int16_t __c) -{ - return __builtin_aarch64_mls_nv8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_n_s32 (int32x4_t __a, int32x4_t __b, int32_t __c) -{ - return __builtin_aarch64_mls_nv4si (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_n_u16 (uint16x8_t __a, uint16x8_t __b, uint16_t __c) -{ - return __builtin_aarch64_mls_nv8hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_n_u32 (uint32x4_t __a, uint32x4_t __b, uint32_t __c) -{ - return __builtin_aarch64_mls_nv4si_uuuu (__a, __b, __c); -} - -__extension__ extern __inline int8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_s8 (int8x16_t __a, int8x16_t __b, int8x16_t __c) -{ - return __builtin_aarch64_mlsv16qi (__a, __b, __c); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_s16 (int16x8_t __a, int16x8_t __b, int16x8_t __c) -{ - return __builtin_aarch64_mlsv8hi (__a, __b, __c); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_s32 (int32x4_t __a, int32x4_t __b, int32x4_t __c) -{ - return __builtin_aarch64_mlsv4si (__a, __b, __c); -} - -__extension__ extern __inline uint8x16_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_u8 (uint8x16_t __a, uint8x16_t __b, uint8x16_t __c) -{ - return __builtin_aarch64_mlsv16qi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_u16 (uint16x8_t __a, uint16x8_t __b, uint16x8_t __c) -{ - return __builtin_aarch64_mlsv8hi_uuuu (__a, __b, __c); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_u32 (uint32x4_t __a, uint32x4_t __b, uint32x4_t __c) -{ - return __builtin_aarch64_mlsv4si_uuuu (__a, __b, __c); -} - __extension__ extern __inline int16x8_t __attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) vmovl_high_s8 (int8x16_t __a) @@ -11179,401 +10871,6 @@ vminvq_f64 (float64x2_t __a) return __builtin_aarch64_reduc_smin_nan_scal_v2df (__a); } -/* vmla */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_f32 (float32x2_t __a, float32x2_t __b, float32x2_t __c) -{ - return __builtin_aarch64_float_mlav2sf (__a, __b, __c); -} - -__extension__ extern __inline float64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c) -{ - return (float64x1_t) {__builtin_aarch64_float_mladf (__a[0], __b[0], __c[0])}; -} - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_f32 (float32x4_t __a, float32x4_t __b, float32x4_t __c) -{ - return __builtin_aarch64_float_mlav4sf (__a, __b, __c); -} - -__extension__ extern __inline float64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_f64 (float64x2_t __a, float64x2_t __b, float64x2_t __c) -{ - return __builtin_aarch64_float_mlav2df (__a, __b, __c); -} - -/* vmla_lane */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_lane_f32 (float32x2_t __a, float32x2_t __b, - float32x2_t __c, const int __lane) -{ - return __builtin_aarch64_float_mla_lanev2sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_lane_s16 (int16x4_t __a, int16x4_t __b, - int16x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_lane_s32 (int32x2_t __a, int32x2_t __b, - int32x2_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_lane_u16 (uint16x4_t __a, uint16x4_t __b, - uint16x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_lane_u32 (uint32x2_t __a, uint32x2_t __b, - uint32x2_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -/* vmla_laneq */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_laneq_f32 (float32x2_t __a, float32x2_t __b, - float32x4_t __c, const int __lane) -{ - return __builtin_aarch64_float_mla_laneqv2sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_laneq_s16 (int16x4_t __a, int16x4_t __b, - int16x8_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_laneq_s32 (int32x2_t __a, int32x2_t __b, - int32x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_laneq_u16 (uint16x4_t __a, uint16x4_t __b, - uint16x8_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmla_laneq_u32 (uint32x2_t __a, uint32x2_t __b, - uint32x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -/* vmlaq_lane */ - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_lane_f32 (float32x4_t __a, float32x4_t __b, - float32x2_t __c, const int __lane) -{ - return __builtin_aarch64_float_mla_lanev4sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_lane_s16 (int16x8_t __a, int16x8_t __b, - int16x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_lane_s32 (int32x4_t __a, int32x4_t __b, - int32x2_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_lane_u16 (uint16x8_t __a, uint16x8_t __b, - uint16x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_lane_u32 (uint32x4_t __a, uint32x4_t __b, - uint32x2_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - - /* vmlaq_laneq */ - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_laneq_f32 (float32x4_t __a, float32x4_t __b, - float32x4_t __c, const int __lane) -{ - return __builtin_aarch64_float_mla_laneqv4sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_laneq_s16 (int16x8_t __a, int16x8_t __b, - int16x8_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_laneq_s32 (int32x4_t __a, int32x4_t __b, - int32x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_laneq_u16 (uint16x8_t __a, uint16x8_t __b, - uint16x8_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlaq_laneq_u32 (uint32x4_t __a, uint32x4_t __b, - uint32x4_t __c, const int __lane) -{ - return (__a + (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -/* vmls */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_f32 (float32x2_t __a, float32x2_t __b, float32x2_t __c) -{ - return __builtin_aarch64_float_mlsv2sf (__a, __b, __c); -} - -__extension__ extern __inline float64x1_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c) -{ - return (float64x1_t) {__builtin_aarch64_float_mlsdf (__a[0], __b[0], __c[0])}; -} - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_f32 (float32x4_t __a, float32x4_t __b, float32x4_t __c) -{ - return __builtin_aarch64_float_mlsv4sf (__a, __b, __c); -} - -__extension__ extern __inline float64x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_f64 (float64x2_t __a, float64x2_t __b, float64x2_t __c) -{ - return __builtin_aarch64_float_mlsv2df (__a, __b, __c); -} - -/* vmls_lane */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_lane_f32 (float32x2_t __a, float32x2_t __b, - float32x2_t __c, const int __lane) -{ - return __builtin_aarch64_float_mls_lanev2sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_lane_s16 (int16x4_t __a, int16x4_t __b, - int16x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_lane_s32 (int32x2_t __a, int32x2_t __b, - int32x2_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_lane_u16 (uint16x4_t __a, uint16x4_t __b, - uint16x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_lane_u32 (uint32x2_t __a, uint32x2_t __b, - uint32x2_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -/* vmls_laneq */ - -__extension__ extern __inline float32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_laneq_f32 (float32x2_t __a, float32x2_t __b, - float32x4_t __c, const int __lane) -{ - return __builtin_aarch64_float_mls_laneqv2sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_laneq_s16 (int16x4_t __a, int16x4_t __b, - int16x8_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_laneq_s32 (int32x2_t __a, int32x2_t __b, - int32x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_laneq_u16 (uint16x4_t __a, uint16x4_t __b, - uint16x8_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x2_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmls_laneq_u32 (uint32x2_t __a, uint32x2_t __b, - uint32x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -/* vmlsq_lane */ - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_lane_f32 (float32x4_t __a, float32x4_t __b, - float32x2_t __c, const int __lane) -{ - return __builtin_aarch64_float_mls_lanev4sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_lane_s16 (int16x8_t __a, int16x8_t __b, - int16x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_lane_s32 (int32x4_t __a, int32x4_t __b, - int32x2_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_lane_u16 (uint16x8_t __a, uint16x8_t __b, - uint16x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_lane_u32 (uint32x4_t __a, uint32x4_t __b, - uint32x2_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - - /* vmlsq_laneq */ - -__extension__ extern __inline float32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_laneq_f32 (float32x4_t __a, float32x4_t __b, - float32x4_t __c, const int __lane) -{ - return __builtin_aarch64_float_mls_laneqv4sf (__a, __b, __c, __lane); -} - -__extension__ extern __inline int16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_laneq_s16 (int16x8_t __a, int16x8_t __b, - int16x8_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline int32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_laneq_s32 (int32x4_t __a, int32x4_t __b, - int32x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} -__extension__ extern __inline uint16x8_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_laneq_u16 (uint16x8_t __a, uint16x8_t __b, - uint16x8_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - -__extension__ extern __inline uint32x4_t -__attribute__ ((__always_inline__, __gnu_inline__, __artificial__)) -vmlsq_laneq_u32 (uint32x4_t __a, uint32x4_t __b, - uint32x4_t __c, const int __lane) -{ - return (__a - (__b * __aarch64_vget_lane_any (__c, __lane))); -} - /* vpadd */ __extension__ extern __inline float32x2_t diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vmla.c b/gcc/testsuite/gcc.target/aarch64/neon/vmla.c new file mode 100644 index 000000000000..a247987b1fa8 --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/neon/vmla.c @@ -0,0 +1,350 @@ +/* { dg-do compile } */ +/* { dg-final { check-function-bodies "**" "" } } */ + +#include "arm_neon_test.h" + +#define TEST_LANE(NAME, RET_TYPE, TYPE1, TYPE2, TYPE3) \ + RET_TYPE test_##NAME (TYPE1 a, TYPE2 b, TYPE3 c) \ + { \ + return NAME (a, b, c, VEC_LEN (c) - 1); \ + } + +/* +** test_vmla_s8: +** mla v0\.8b, (v1\.8b, v2\.8b|v2\.8b, v1\.8b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_s8, int8x8_t) + +/* +** test_vmla_u8: +** mla v0\.8b, (v1\.8b, v2\.8b|v2\.8b, v1\.8b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_u8, uint8x8_t) + +/* +** test_vmlaq_s8: +** mla v0\.16b, (v1\.16b, v2\.16b|v2\.16b, v1\.16b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_s8, int8x16_t) + +/* +** test_vmlaq_u8: +** mla v0\.16b, (v1\.16b, v2\.16b|v2\.16b, v1\.16b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_u8, uint8x16_t) + +/* +** test_vmla_s16: +** mla v0\.4h, (v1\.4h, v2\.4h|v2\.4h, v1\.4h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_s16, int16x4_t) + +/* +** test_vmla_u16: +** mla v0\.4h, (v1\.4h, v2\.4h|v2\.4h, v1\.4h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_u16, uint16x4_t) + +/* +** test_vmlaq_s16: +** mla v0\.8h, (v1\.8h, v2\.8h|v2\.8h, v1\.8h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_s16, int16x8_t) + +/* +** test_vmlaq_u16: +** mla v0\.8h, (v1\.8h, v2\.8h|v2\.8h, v1\.8h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_u16, uint16x8_t) + +/* +** test_vmla_s32: +** mla v0\.2s, (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_s32, int32x2_t) + +/* +** test_vmla_u32: +** mla v0\.2s, (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_u32, uint32x2_t) + +/* +** test_vmlaq_s32: +** mla v0\.4s, (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_s32, int32x4_t) + +/* +** test_vmlaq_u32: +** mla v0\.4s, (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_u32, uint32x4_t) + +/* +** test_vmla_f32: +** fmul (v[0-9]+\.2s), (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** fadd v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_f32, float32x2_t) + +/* +** test_vmlaq_f32: +** fmul (v[0-9]+\.4s), (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** fadd v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_f32, float32x4_t) + +/* +** test_vmla_f64: +** fmul (d[0-9]+), (d1, d2|d2, d1) +** fadd d0, (\1, d0|d0, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmla_f64, float64x1_t) + +/* +** test_vmlaq_f64: +** fmul (v[0-9]+\.2d), (v1\.2d, v2\.2d|v2\.2d, v1\.2d) +** fadd v0\.2d, (\1, v0\.2d|v0\.2d, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlaq_f64, float64x2_t) + +/* +** test_vmla_n_s16: +** dup v([0-9]+)\.4h, w0 +** mla v0\.4h, v1\.4h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmla_n_s16, int16x4_t, int16x4_t, int16x4_t, int16_t) + +/* +** test_vmla_n_u16: +** dup v([0-9]+)\.4h, w0 +** mla v0\.4h, v1\.4h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmla_n_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16_t) + +/* +** test_vmlaq_n_s16: +** dup v([0-9]+)\.4h, w0 +** mla v0\.8h, v1\.8h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmlaq_n_s16, int16x8_t, int16x8_t, int16x8_t, int16_t) + +/* +** test_vmlaq_n_u16: +** dup v([0-9]+)\.4h, w0 +** mla v0\.8h, v1\.8h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmlaq_n_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16_t) + +/* +** test_vmla_n_s32: +** fmov s([0-9]+), w0 +** mla v0\.2s, v1\.2s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmla_n_s32, int32x2_t, int32x2_t, int32x2_t, int32_t) + +/* +** test_vmla_n_u32: +** fmov s([0-9]+), w0 +** mla v0\.2s, v1\.2s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmla_n_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32_t) + +/* +** test_vmlaq_n_s32: +** fmov s([0-9]+), w0 +** mla v0\.4s, v1\.4s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmlaq_n_s32, int32x4_t, int32x4_t, int32x4_t, int32_t) + +/* +** test_vmlaq_n_u32: +** fmov s([0-9]+), w0 +** mla v0\.4s, v1\.4s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmlaq_n_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32_t) + +/* +** test_vmla_n_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[0\] +** fadd v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_TERNARY (vmla_n_f32, float32x2_t, float32x2_t, float32x2_t, float32_t) + +/* +** test_vmlaq_n_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[0\] +** fadd v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_TERNARY (vmlaq_n_f32, float32x4_t, float32x4_t, float32x4_t, float32_t) + +/* +** test_vmla_lane_s16: +** mla v0\.4h, v1\.4h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmla_lane_s16, int16x4_t, int16x4_t, int16x4_t, int16x4_t) + +/* +** test_vmla_lane_u16: +** mla v0\.4h, v1\.4h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmla_lane_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16x4_t) + +/* +** test_vmla_lane_s32: +** mla v0\.2s, v1\.2s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmla_lane_s32, int32x2_t, int32x2_t, int32x2_t, int32x2_t) + +/* +** test_vmla_lane_u32: +** mla v0\.2s, v1\.2s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmla_lane_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32x2_t) + +/* +** test_vmla_lane_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[1\] +** fadd v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_LANE (vmla_lane_f32, float32x2_t, float32x2_t, float32x2_t, float32x2_t) + +/* +** test_vmla_laneq_s16: +** mla v0\.4h, v1\.4h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmla_laneq_s16, int16x4_t, int16x4_t, int16x4_t, int16x8_t) + +/* +** test_vmla_laneq_u16: +** mla v0\.4h, v1\.4h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmla_laneq_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16x8_t) + +/* +** test_vmla_laneq_s32: +** mla v0\.2s, v1\.2s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmla_laneq_s32, int32x2_t, int32x2_t, int32x2_t, int32x4_t) + +/* +** test_vmla_laneq_u32: +** mla v0\.2s, v1\.2s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmla_laneq_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32x4_t) + +/* +** test_vmla_laneq_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[3\] +** fadd v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_LANE (vmla_laneq_f32, float32x2_t, float32x2_t, float32x2_t, float32x4_t) + +/* +** test_vmlaq_lane_s16: +** mla v0\.8h, v1\.8h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmlaq_lane_s16, int16x8_t, int16x8_t, int16x8_t, int16x4_t) + +/* +** test_vmlaq_lane_u16: +** mla v0\.8h, v1\.8h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmlaq_lane_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16x4_t) + +/* +** test_vmlaq_lane_s32: +** mla v0\.4s, v1\.4s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmlaq_lane_s32, int32x4_t, int32x4_t, int32x4_t, int32x2_t) + +/* +** test_vmlaq_lane_u32: +** mla v0\.4s, v1\.4s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmlaq_lane_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32x2_t) + +/* +** test_vmlaq_lane_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[1\] +** fadd v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_LANE (vmlaq_lane_f32, float32x4_t, float32x4_t, float32x4_t, float32x2_t) + +/* +** test_vmlaq_laneq_s16: +** mla v0\.8h, v1\.8h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmlaq_laneq_s16, int16x8_t, int16x8_t, int16x8_t, int16x8_t) + +/* +** test_vmlaq_laneq_u16: +** mla v0\.8h, v1\.8h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmlaq_laneq_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16x8_t) + +/* +** test_vmlaq_laneq_s32: +** mla v0\.4s, v1\.4s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmlaq_laneq_s32, int32x4_t, int32x4_t, int32x4_t, int32x4_t) + +/* +** test_vmlaq_laneq_u32: +** mla v0\.4s, v1\.4s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmlaq_laneq_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32x4_t) + +/* +** test_vmlaq_laneq_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[3\] +** fadd v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_LANE (vmlaq_laneq_f32, float32x4_t, float32x4_t, float32x4_t, float32x4_t) diff --git a/gcc/testsuite/gcc.target/aarch64/neon/vmls.c b/gcc/testsuite/gcc.target/aarch64/neon/vmls.c new file mode 100644 index 000000000000..cb851681b181 --- /dev/null +++ b/gcc/testsuite/gcc.target/aarch64/neon/vmls.c @@ -0,0 +1,350 @@ +/* { dg-do compile } */ +/* { dg-final { check-function-bodies "**" "" } } */ + +#include "arm_neon_test.h" + +#define TEST_LANE(NAME, RET_TYPE, TYPE1, TYPE2, TYPE3) \ + RET_TYPE test_##NAME (TYPE1 a, TYPE2 b, TYPE3 c) \ + { \ + return NAME (a, b, c, VEC_LEN (c) - 1); \ + } + +/* +** test_vmls_s8: +** mls v0\.8b, (v1\.8b, v2\.8b|v2\.8b, v1\.8b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_s8, int8x8_t) + +/* +** test_vmls_u8: +** mls v0\.8b, (v1\.8b, v2\.8b|v2\.8b, v1\.8b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_u8, uint8x8_t) + +/* +** test_vmlsq_s8: +** mls v0\.16b, (v1\.16b, v2\.16b|v2\.16b, v1\.16b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_s8, int8x16_t) + +/* +** test_vmlsq_u8: +** mls v0\.16b, (v1\.16b, v2\.16b|v2\.16b, v1\.16b) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_u8, uint8x16_t) + +/* +** test_vmls_s16: +** mls v0\.4h, (v1\.4h, v2\.4h|v2\.4h, v1\.4h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_s16, int16x4_t) + +/* +** test_vmls_u16: +** mls v0\.4h, (v1\.4h, v2\.4h|v2\.4h, v1\.4h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_u16, uint16x4_t) + +/* +** test_vmlsq_s16: +** mls v0\.8h, (v1\.8h, v2\.8h|v2\.8h, v1\.8h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_s16, int16x8_t) + +/* +** test_vmlsq_u16: +** mls v0\.8h, (v1\.8h, v2\.8h|v2\.8h, v1\.8h) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_u16, uint16x8_t) + +/* +** test_vmls_s32: +** mls v0\.2s, (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_s32, int32x2_t) + +/* +** test_vmls_u32: +** mls v0\.2s, (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_u32, uint32x2_t) + +/* +** test_vmlsq_s32: +** mls v0\.4s, (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_s32, int32x4_t) + +/* +** test_vmlsq_u32: +** mls v0\.4s, (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_u32, uint32x4_t) + +/* +** test_vmls_f32: +** fmul (v[0-9]+\.2s), (v1\.2s, v2\.2s|v2\.2s, v1\.2s) +** fsub v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_f32, float32x2_t) + +/* +** test_vmlsq_f32: +** fmul (v[0-9]+\.4s), (v1\.4s, v2\.4s|v2\.4s, v1\.4s) +** fsub v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_f32, float32x4_t) + +/* +** test_vmls_f64: +** fmul (d[0-9]+), (d1, d2|d2, d1) +** fsub d0, (\1, d0|d0, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmls_f64, float64x1_t) + +/* +** test_vmlsq_f64: +** fmul (v[0-9]+\.2d), (v1\.2d, v2\.2d|v2\.2d, v1\.2d) +** fsub v0\.2d, (\1, v0\.2d|v0\.2d, \1) +** ret +*/ +TEST_UNIFORM_TERNARY (vmlsq_f64, float64x2_t) + +/* +** test_vmls_n_s16: +** dup v([0-9]+)\.4h, w0 +** mls v0\.4h, v1\.4h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmls_n_s16, int16x4_t, int16x4_t, int16x4_t, int16_t) + +/* +** test_vmls_n_u16: +** dup v([0-9]+)\.4h, w0 +** mls v0\.4h, v1\.4h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmls_n_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16_t) + +/* +** test_vmlsq_n_s16: +** dup v([0-9]+)\.4h, w0 +** mls v0\.8h, v1\.8h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmlsq_n_s16, int16x8_t, int16x8_t, int16x8_t, int16_t) + +/* +** test_vmlsq_n_u16: +** dup v([0-9]+)\.4h, w0 +** mls v0\.8h, v1\.8h, v\1\.h\[0\] +** ret +*/ +TEST_TERNARY (vmlsq_n_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16_t) + +/* +** test_vmls_n_s32: +** fmov s([0-9]+), w0 +** mls v0\.2s, v1\.2s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmls_n_s32, int32x2_t, int32x2_t, int32x2_t, int32_t) + +/* +** test_vmls_n_u32: +** fmov s([0-9]+), w0 +** mls v0\.2s, v1\.2s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmls_n_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32_t) + +/* +** test_vmlsq_n_s32: +** fmov s([0-9]+), w0 +** mls v0\.4s, v1\.4s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmlsq_n_s32, int32x4_t, int32x4_t, int32x4_t, int32_t) + +/* +** test_vmlsq_n_u32: +** fmov s([0-9]+), w0 +** mls v0\.4s, v1\.4s, v\1\.s\[0\] +** ret +*/ +TEST_TERNARY (vmlsq_n_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32_t) + +/* +** test_vmls_n_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[0\] +** fsub v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_TERNARY (vmls_n_f32, float32x2_t, float32x2_t, float32x2_t, float32_t) + +/* +** test_vmlsq_n_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[0\] +** fsub v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_TERNARY (vmlsq_n_f32, float32x4_t, float32x4_t, float32x4_t, float32_t) + +/* +** test_vmls_lane_s16: +** mls v0\.4h, v1\.4h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmls_lane_s16, int16x4_t, int16x4_t, int16x4_t, int16x4_t) + +/* +** test_vmls_lane_u16: +** mls v0\.4h, v1\.4h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmls_lane_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16x4_t) + +/* +** test_vmls_lane_s32: +** mls v0\.2s, v1\.2s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmls_lane_s32, int32x2_t, int32x2_t, int32x2_t, int32x2_t) + +/* +** test_vmls_lane_u32: +** mls v0\.2s, v1\.2s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmls_lane_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32x2_t) + +/* +** test_vmls_lane_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[1\] +** fsub v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_LANE (vmls_lane_f32, float32x2_t, float32x2_t, float32x2_t, float32x2_t) + +/* +** test_vmls_laneq_s16: +** mls v0\.4h, v1\.4h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmls_laneq_s16, int16x4_t, int16x4_t, int16x4_t, int16x8_t) + +/* +** test_vmls_laneq_u16: +** mls v0\.4h, v1\.4h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmls_laneq_u16, uint16x4_t, uint16x4_t, uint16x4_t, uint16x8_t) + +/* +** test_vmls_laneq_s32: +** mls v0\.2s, v1\.2s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmls_laneq_s32, int32x2_t, int32x2_t, int32x2_t, int32x4_t) + +/* +** test_vmls_laneq_u32: +** mls v0\.2s, v1\.2s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmls_laneq_u32, uint32x2_t, uint32x2_t, uint32x2_t, uint32x4_t) + +/* +** test_vmls_laneq_f32: +** fmul (v[0-9]+\.2s), v1\.2s, v2\.s\[3\] +** fsub v0\.2s, (\1, v0\.2s|v0\.2s, \1) +** ret +*/ +TEST_LANE (vmls_laneq_f32, float32x2_t, float32x2_t, float32x2_t, float32x4_t) + +/* +** test_vmlsq_lane_s16: +** mls v0\.8h, v1\.8h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmlsq_lane_s16, int16x8_t, int16x8_t, int16x8_t, int16x4_t) + +/* +** test_vmlsq_lane_u16: +** mls v0\.8h, v1\.8h, v2\.h\[3\] +** ret +*/ +TEST_LANE (vmlsq_lane_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16x4_t) + +/* +** test_vmlsq_lane_s32: +** mls v0\.4s, v1\.4s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmlsq_lane_s32, int32x4_t, int32x4_t, int32x4_t, int32x2_t) + +/* +** test_vmlsq_lane_u32: +** mls v0\.4s, v1\.4s, v2\.s\[1\] +** ret +*/ +TEST_LANE (vmlsq_lane_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32x2_t) + +/* +** test_vmlsq_lane_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[1\] +** fsub v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_LANE (vmlsq_lane_f32, float32x4_t, float32x4_t, float32x4_t, float32x2_t) + +/* +** test_vmlsq_laneq_s16: +** mls v0\.8h, v1\.8h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmlsq_laneq_s16, int16x8_t, int16x8_t, int16x8_t, int16x8_t) + +/* +** test_vmlsq_laneq_u16: +** mls v0\.8h, v1\.8h, v2\.h\[7\] +** ret +*/ +TEST_LANE (vmlsq_laneq_u16, uint16x8_t, uint16x8_t, uint16x8_t, uint16x8_t) + +/* +** test_vmlsq_laneq_s32: +** mls v0\.4s, v1\.4s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmlsq_laneq_s32, int32x4_t, int32x4_t, int32x4_t, int32x4_t) + +/* +** test_vmlsq_laneq_u32: +** mls v0\.4s, v1\.4s, v2\.s\[3\] +** ret +*/ +TEST_LANE (vmlsq_laneq_u32, uint32x4_t, uint32x4_t, uint32x4_t, uint32x4_t) + +/* +** test_vmlsq_laneq_f32: +** fmul (v[0-9]+\.4s), v1\.4s, v2\.s\[3\] +** fsub v0\.4s, (\1, v0\.4s|v0\.4s, \1) +** ret +*/ +TEST_LANE (vmlsq_laneq_f32, float32x4_t, float32x4_t, float32x4_t, float32x4_t)
