Searched refs:FP16 (Results 1 – 20 of 20) sorted by relevance
1 //===--- arm_fp16.td - ARM FP16 compiler interface ------------------------===//9 // This file defines the TableGen definitions from which the ARM FP16 header16 // ARMv8.2-A FP16 intrinsics.
141 // AMX_FP16 FP16
1651 // ARMv8.2-A FP16 vector intrinsics for A32/A64.1654 // ARMv8.2-A FP16 one-operand vector intrinsics.1692 // ARMv8.2-A FP16 two-operands vector intrinsics.1745 // ARMv8.2-A FP16 three-operands vector intrinsics.1751 // ARMv8.2-A FP16 lane vector intrinsics.1775 // ARMv8.2-A FP16 vector intrinsics for A64 only.1799 // ARMv8.2-A FP16 lane vector intrinsics.1850 // ARMv8.2-A FP16 reduction vector intrinsics.1882 // v8.2-A FP16 fused multiply-add long instructions.
189 //AMX-FP16
182 // FIXME: FP16 scalar intrinsics use the type v8f16, which is supposed to be184 // FIXME: FP16 conversion between f16 and i64 customize type v8i64, which is
4031 // Always select FP16 instructions if available.
11751 // Always select FP16 instructions if available.
1675 let isUnpredicable = 1; // FP16 instructions cannot in general be conditional2024 let isUnpredicable = 1; // FP16 instructions cannot in general be conditional2054 let isUnpredicable = 1; // FP16 instructions cannot in general be conditional2082 let isUnpredicable = 1; // FP16 instructions cannot in general be conditional2113 let isUnpredicable = 1; // FP16 instructions cannot in general be conditional
136 bool IsAddressingMode5(SDValue N, SDValue &Base, SDValue &Offset, bool FP16);987 bool FP16) { in IsAddressingMode5() argument1007 const int Scale = FP16 ? 2 : 4; in IsAddressingMode5()1023 if (FP16) in IsAddressingMode5()1035 if (FP16) in IsAddressingMode5()
43 // 16-bit FP loads/stores/moves, shared between VFP (with the v8.2A FP16
107 "Full FP16 (FEAT_FP16)", [FeatureFPARMv8]>;110 "Enable FP16 FML instructions (FEAT_FHM)", [FeatureFullFP16]>;
1135 // ARMv8.2-A FP16 Fused Multiply-Add Long1263 // inside the multiclass as the FP16 versions need different predicates.4222 // Pattern for FP16 immediates6760 // Patterns for FP16 Intrinsics - requires reg copy to/from as i16s not supported.
310 * Support ISA of ``AMX-FP16`` which contains ``tdpfp16ps`` instruction.
532 // v8.2-A FP16 Fused Multiply-Add Long
5350 // AMX-FP16 - Intel FP16 AMX extensions
224 // FP16 ops are identified by accumulator & result type.
967 - Add ISA of ``AMX-FP16`` which support ``_tile_dpfp16ps``.
786 AVX512-FP16, ``_Float16`` arithmetic is performed using that native support.
1777 * ``_Float16`` on X86 targets without ``AVX512-FP16``.
1018 // Loads FP16 constant into a register.