diff options
Diffstat (limited to 'arm_compute/runtime/NEON/functions/NEGEMMConv2d.h')
| -rw-r--r-- | arm_compute/runtime/NEON/functions/NEGEMMConv2d.h | 40 |
1 files changed, 25 insertions, 15 deletions
diff --git a/arm_compute/runtime/NEON/functions/NEGEMMConv2d.h b/arm_compute/runtime/NEON/functions/NEGEMMConv2d.h index 2bd233f520..d1c5a1c9b3 100644 --- a/arm_compute/runtime/NEON/functions/NEGEMMConv2d.h +++ b/arm_compute/runtime/NEON/functions/NEGEMMConv2d.h @@ -27,22 +27,20 @@ #include "arm_compute/runtime/FunctionDescriptors.h" #include "arm_compute/runtime/IFunction.h" #include "arm_compute/runtime/IMemoryManager.h" -#include "arm_compute/runtime/NEON/functions/NEActivationLayer.h" -#include "arm_compute/runtime/NEON/functions/NEPermute.h" -#include "arm_compute/runtime/Tensor.h" #include <memory> + namespace arm_compute { // Forward declarations class ITensor; -class NEGEMMAssemblyDispatch; +class ITensorInfo; -/** Basic function to compute the convolution layer. This function calls the following Neon kernels/functions: +/** Basic function to compute the convolution layer. This function calls the following kernels/functions: * * Supports only NHWC data layout * - * -# @ref NEGEMMAssemblyDispatch + * -# @ref cpu::CpuGemmAssemblyDispatch * -# @ref NEActivationLayer, in case activation cannot be fused in the assembly dispatch * * Weights are transformed from OHWI to HWIO format using the following kernels: @@ -65,6 +63,18 @@ public: ~NEGEMMConv2d(); /** Set the input and output tensors. * + * Valid data layouts: + * - All + * + * Valid data type configurations: + * |src0 |src1 |src2 |dst | + * |:--------------|:--------------|:--------------|:--------------| + * |QASYMM8 |QASYMM8 |S32 |QASYMM8 | + * |QASYMM8_SIGNED |QASYMM8_SIGNED |S32 |QASYMM8_SIGNED | + * |F16 |F16 |F16 |F16 | + * |F32 |F32 |F32 |F32 | + * |BFLOAT16 |BFLOAT16 |BFLOAT16 |BFLOAT16 | + * * @param[in] input Source tensor. 3 lower dimensions represent a single input [width, height, IFM], * while every optional dimension from 4 and above represent a batch of inputs. * Data types supported: QASYMM8/QASYMM8_SIGNED/BFLOAT16/F16/F32. @@ -76,7 +86,8 @@ public: * Data types supported: Same as @p input. * @param[in] info Convolution layer descriptor */ - void configure(ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const Conv2dInfo &info); + void + configure(ITensor *input, const ITensor *weights, const ITensor *biases, ITensor *output, const Conv2dInfo &info); /** Static function to check if given info will lead to a valid configuration of @ref NEGEMMConv2d * * @param[in] input Source tensor info. 3 lower dimensions represent a single input [width, height, IFM], @@ -92,20 +103,19 @@ public: * * @return a status */ - static Status validate(const ITensorInfo *input, const ITensorInfo *weights, const ITensorInfo *biases, const ITensorInfo *output, const Conv2dInfo &info); + static Status validate(const ITensorInfo *input, + const ITensorInfo *weights, + const ITensorInfo *biases, + const ITensorInfo *output, + const Conv2dInfo &info); // Inherited methods overridden: void run() override; void prepare() override; private: - std::unique_ptr<NEGEMMAssemblyDispatch> _gemm_asm_func; - NEActivationLayer _activation_func; - NEPermute _weights_permute_func; - const ITensor *_original_weights; - Tensor _permuted_weights; - bool _is_prepared; - bool _run_activation; + struct Impl; + std::unique_ptr<Impl> _impl; }; } // namespace arm_compute #endif /* ARM_COMPUTE_NEGEMMCONV2D_H */ |