24.04
|
Go to the documentation of this file.
24 #ifndef ACL_SRC_RUNTIME_HEURISTICS_MATMUL_NATIVE_ICLMATMULNATIVEKERNELVARIANT_H
25 #define ACL_SRC_RUNTIME_HEURISTICS_MATMUL_NATIVE_ICLMATMULNATIVEKERNELVARIANT_H
99 std::array<T, 2> _configs;
135 #endif // ACL_SRC_RUNTIME_HEURISTICS_MATMUL_NATIVE_ICLMATMULNATIVEKERNELVARIANT_H
@ QSYMM8_PER_CHANNEL
quantized, symmetric per channel fixed-point 8-bit number
@ NATIVE_QUANTIZED
Native matrix multiplication for quantized types.
ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(IClMatMulNativeKernelVariant)
@ QASYMM8
quantized, asymmetric fixed-point 8-bit number unsigned
@ NATIVE_MMUL_QUANTIZED
Native matrix multiplication using MMUL extension for Quantized types.
T get_function(DataType data_type)
Method to return the matmul native variant function based on data type.
static constexpr size_t DT_FLOAT
Alias for Float index.
static constexpr size_t DT_QUANTIZED
Alias for Quantized type index.
Activation Layer Information class.
@ QASYMM8_SIGNED
quantized, asymmetric fixed-point 8-bit number signed
virtual ~IClMatMulNativeKernelVariant()=default
Virtual destructor.
@ NATIVE_FP
Native matrix multiplication for FP types.
@ NATIVE_MMUL_FP
Native matrix multiplication using MMUL extension for FP types.
Basic interface for the matmul native kernel variant This is the base class that chooses architecture...
GPUTarget
Available GPU Targets.
Copyright (c) 2017-2024 Arm Limited.
ClMatMulNativeVariantArray(T func_float, T func_quantized)
Constructor.
@ F16
16-bit floating-point number
virtual MatMulKernelType select_kernel(const ITensorInfo *lhs, const ITensorInfo *rhs, const MatMulInfo &info, const ActivationLayerInfo &act_info)=0
This method returns the MatMulKernelType for the given inputs.
Store the tensor's metadata.
@ F32
32-bit floating-point number
IClMatMulNativeKernelVariant(GPUTarget arch)
Constructor.
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
Basic container for the OpenCL MatMul Native variant functions.
DataType
Available data types.
Class for holding information related to matrix multiplication function.