24.02.1
|
Go to the documentation of this file.
24 #ifndef SRC_RUNTIME_HEURISTICS_DIRECT_CONV_ICLDIRECTCONVKERNELCONFIG
25 #define SRC_RUNTIME_HEURISTICS_DIRECT_CONV_ICLDIRECTCONVKERNELCONFIG
35 namespace cl_direct_conv
71 return _configs.at(
DT_F32);
73 return _configs.at(
DT_F16);
84 std::array<T, 3> _configs;
@ QSYMM8_PER_CHANNEL
quantized, symmetric per channel fixed-point 8-bit number
SimpleTensor< float > src
ClDirectConvConfigArray(T func_f32, T func_f16, T func_int8)
Constructor.
static constexpr size_t DT_INT8
Alias for Int8 index.
static constexpr size_t DT_F32
Alias for F32 index.
@ QASYMM8
quantized, asymmetric fixed-point 8-bit number unsigned
Basic interface for the Direct convolution kernel configuration.
static constexpr size_t DT_F16
Alias for F16 index.
IClDirectConvKernelConfig(GPUTarget arch)
Constructor.
virtual ~IClDirectConvKernelConfig()=default
Virtual destructor.
@ QASYMM8_SIGNED
quantized, asymmetric fixed-point 8-bit number signed
Basic container for the OpenCL direct convolution configuration functions.
ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE(IClDirectConvKernelConfig)
virtual DirectConvComputeKernelInfo configure(const ITensorInfo *src, const ITensorInfo *wei, const PadStrideInfo &conv_info)=0
This method returns the DirectConvComputeKernelInfo for the given inputs.
GPUTarget
Available GPU Targets.
Copyright (c) 2017-2024 Arm Limited.
@ F16
16-bit floating-point number
Store the tensor's metadata.
@ F32
32-bit floating-point number
T get_function(DataType data_type)
Method to return the direct convolution configuration function based on data type.
DataType
Available data types.
Compute descriptor used by the direct convolution kernel.