From ae72a46e495742863dba44fcf5fdc673c9d2afbc Mon Sep 17 00:00:00 2001 From: Gunes Bayir Date: Sun, 29 Jan 2023 13:24:24 +0000 Subject: =?UTF-8?q?Add=20new=20operator=20AddMulAdd=20for=20Neon=E2=84=A2?= =?UTF-8?q?=20backend=20for=20Float/Quantized=20types?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit This is a fused operator that merges Add + Mul + Add [+ Relu-based-Activation] layers and have an intermediate output after the first Add. It's supported for FP16/32/QASYMM8/QASYMM8_SIGNED data types. The subsequent Add and Mul are intended for scaling and the coefficients only have one dimension (per channel). The inputs are - input1 : nD tensor [X, Y, Z, W, ..] - input2 : nD tensor [X, Y, Z, W, ..] - add_coef : 1D tensor [X] - mul_coef : 1D tensor [X] The outputs are - out1 : nD tensor (intermediate output) [X, Y, Z, W, ..] - out2 : nD tensor (final output) [X, Y, Z, W, ..] The operation can be summarized as follows: out1 <- input1 + input2 out2 <- Act(out1 * mul_coef + add_coef) The activation function can be Identity, Relu, Bounded Relu or Lower/Upper Bounded Relu. The intermediate output can be skipped by providing a nullptr. The reason of providing this operator is to be able to fuse in case of Residual network patterns and save computations by reducing memory back and forward. Resolves: COMPMID-5463 Signed-off-by: Gunes Bayir Change-Id: I8ef577aa623b036e9a9f655cc088493fd19a6109 Reviewed-on: https://review.mlplatform.org/c/ml/ComputeLibrary/+/9055 Comments-Addressed: Arm Jenkins Reviewed-by: Jakub Sujak Reviewed-by: Viet-Hoa Do Tested-by: Arm Jenkins Benchmark: Arm Jenkins --- src/cpu/kernels/CpuAddMulAddKernel.cpp | 204 +++++++++++++++++++++++++++++++++ 1 file changed, 204 insertions(+) create mode 100644 src/cpu/kernels/CpuAddMulAddKernel.cpp (limited to 'src/cpu/kernels/CpuAddMulAddKernel.cpp') diff --git a/src/cpu/kernels/CpuAddMulAddKernel.cpp b/src/cpu/kernels/CpuAddMulAddKernel.cpp new file mode 100644 index 0000000000..b84bdd54e9 --- /dev/null +++ b/src/cpu/kernels/CpuAddMulAddKernel.cpp @@ -0,0 +1,204 @@ +/* + * Copyright (c) 2023 Arm Limited. + * + * SPDX-License-Identifier: MIT + * + * Permission is hereby granted, free of charge, to any person obtaining a copy + * of this software and associated documentation files (the "Software"), to + * deal in the Software without restriction, including without limitation the + * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or + * sell copies of the Software, and to permit persons to whom the Software is + * furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in all + * copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE + * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE + * SOFTWARE. + */ +#include "src/cpu/kernels/CpuAddMulAddKernel.h" + +#include "arm_compute/core/ITensor.h" +#include "arm_compute/core/TensorInfo.h" +#include "arm_compute/core/Validate.h" + +#include "src/core/CPP/Validate.h" +#include "src/core/common/Registrars.h" +#include "src/core/helpers/AutoConfiguration.h" +#include "src/core/helpers/WindowHelpers.h" +#include "src/cpu/kernels/addmuladd/list.h" + +namespace arm_compute +{ +namespace cpu +{ +namespace kernels +{ +namespace +{ +static const std::vector available_kernels = +{ +#ifdef __aarch64__ + { + "neon_fp32_add_mul_add", + [](const DataTypeISASelectorData & data) { return (data.dt == DataType::F32); }, + REGISTER_FP32_NEON(arm_compute::cpu::add_mul_add_fp32_neon) + }, + { + "neon_fp16_add_mul_add", + [](const DataTypeISASelectorData & data) { return (data.dt == DataType::F16); }, + REGISTER_FP16_NEON(arm_compute::cpu::add_mul_add_fp16_neon) + }, + { + "neon_qasymm8_add_mul_add", + [](const DataTypeISASelectorData & data) { return (data.dt == DataType::QASYMM8); }, + REGISTER_QASYMM8_NEON(arm_compute::cpu::add_mul_add_u8_neon) + }, + { + "neon_qasymm8_signed_add_mul_add", + [](const DataTypeISASelectorData & data) { return (data.dt == DataType::QASYMM8_SIGNED); }, + REGISTER_QASYMM8_SIGNED_NEON(arm_compute::cpu::add_mul_add_s8_neon) + } +#endif // __aarch64__ +}; + +Status validate_arguments(const ITensorInfo *input1, const ITensorInfo *input2, + const ITensorInfo *bn_mul, const ITensorInfo *bn_add, + const ITensorInfo *add_output, const ITensorInfo *final_output, + ConvertPolicy policy, const ActivationLayerInfo &act_info) +{ + ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1, input2, bn_mul, bn_add, final_output); + + ARM_COMPUTE_RETURN_ERROR_ON_MSG(policy != ConvertPolicy::SATURATE, "Only Saturate Policy is supported"); + + using ActFunction = ActivationLayerInfo::ActivationFunction; + const ActFunction act_func = act_info.activation(); + ARM_COMPUTE_RETURN_ERROR_ON_MSG( + (act_func != ActFunction::BOUNDED_RELU && act_func != ActFunction::RELU && act_func != ActFunction::LU_BOUNDED_RELU && act_func != ActFunction::IDENTITY), + "Only RELU Family activations, or no activation, is supported"); + + ARM_COMPUTE_RETURN_ERROR_ON_CPU_F16_UNSUPPORTED(input1); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(input1, 1, DataType::QASYMM8, DataType::QASYMM8_SIGNED, + DataType::F16, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, input2); + + if(is_data_type_quantized(input1->data_type())) + { + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(bn_mul, 1, DataType::F32); + ARM_COMPUTE_RETURN_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(bn_add, 1, DataType::F32); + } + else + { + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, bn_mul); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, bn_add); + } + + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, input2); // No broadcasting + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(bn_mul, bn_add); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(bn_mul->num_dimensions() != 1, "BatchNorm coefficients should be 1D array"); + ARM_COMPUTE_RETURN_ERROR_ON_MSG(bn_mul->tensor_shape()[0] != input1->tensor_shape()[0], "First dimensions of inputs and batchNorm coefs should match"); + + // Validate in case we have add layer's output (intermediate) initialized + if(add_output != nullptr && add_output->total_size() > 0) + { + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, add_output); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, add_output); + } + + // Validate in case final output has been initialized + if(final_output->total_size() > 0) + { + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_DATA_TYPES(input1, final_output); + ARM_COMPUTE_RETURN_ERROR_ON_MISMATCHING_SHAPES(input1, final_output); + } + + const auto uk = CpuAddMulAddKernel::get_implementation(DataTypeISASelectorData{ input1->data_type(), CPUInfo::get().get_isa() }); + ARM_COMPUTE_RETURN_ERROR_ON(uk == nullptr || uk->ukernel == nullptr); + + return Status{}; +} +} // namespace + +void CpuAddMulAddKernel::configure(const ITensorInfo *input1, const ITensorInfo *input2, + const ITensorInfo *bn_mul, const ITensorInfo *bn_add, + ITensorInfo *add_output, ITensorInfo *final_output, + ConvertPolicy policy, const ActivationLayerInfo &act_info) +{ + ARM_COMPUTE_UNUSED(bn_mul, bn_add, input2); + ARM_COMPUTE_ERROR_ON_NULLPTR(input1, input2, bn_add, bn_mul, final_output); + ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input1, input2, bn_mul, bn_add, add_output, final_output, policy, act_info)); + + const auto uk = CpuAddMulAddKernel::get_implementation(DataTypeISASelectorData{ input1->data_type(), CPUInfo::get().get_isa() }); + ARM_COMPUTE_ERROR_ON_NULLPTR(uk); + ARM_COMPUTE_ERROR_ON(uk->ukernel == nullptr); + + _policy = policy; + _act_info = act_info; + _run_method = uk->ukernel; + _name = std::string("CpuAddMulAddKernel/").append(uk->name); + + // Auto initialize outputs if not initialized + set_shape_if_empty(*final_output, input1->tensor_shape()); + set_data_type_if_unknown(*final_output, input1->data_type()); + + if(add_output != nullptr) + { + set_shape_if_empty(*add_output, input1->tensor_shape()); + set_data_type_if_unknown(*add_output, input1->data_type()); + } + + // Configure kernel window + Window win; + win = calculate_max_window(*final_output, Steps()); + ICpuKernel::configure(win); +} + +Status CpuAddMulAddKernel::validate(const ITensorInfo *input1, const ITensorInfo *input2, + const ITensorInfo *bn_mul, const ITensorInfo *bn_add, + const ITensorInfo *add_output, const ITensorInfo *final_output, + ConvertPolicy policy, const ActivationLayerInfo &act_info) +{ + ARM_COMPUTE_RETURN_ERROR_ON_NULLPTR(input1, input2, bn_mul, bn_add, final_output); + + ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input1, input2, bn_mul, bn_add, add_output, final_output, policy, act_info)); + + return Status{}; +} + +void CpuAddMulAddKernel::run_op(ITensorPack &tensors, const Window &window, const ThreadInfo &info) +{ + ARM_COMPUTE_UNUSED(info); + + ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this); + ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(ICpuKernel::window(), window); + + ARM_COMPUTE_ERROR_ON(tensors.empty()); + ARM_COMPUTE_ERROR_ON(_run_method == nullptr); + + const ITensor *input1 = tensors.get_const_tensor(TensorType::ACL_SRC_0); + const ITensor *input2 = tensors.get_const_tensor(TensorType::ACL_SRC_1); + const ITensor *bn_mul = tensors.get_const_tensor(TensorType::ACL_SRC_2); + const ITensor *bn_add = tensors.get_const_tensor(TensorType::ACL_SRC_3); + ITensor *add_output = tensors.get_tensor(TensorType::ACL_DST_0); + ITensor *final_output = tensors.get_tensor(TensorType::ACL_DST_1); + + _run_method(input1, input2, bn_mul, bn_add, add_output, final_output, _policy, _act_info, window); +} + +const char *CpuAddMulAddKernel::name() const +{ + return _name.c_str(); +} + +const std::vector &CpuAddMulAddKernel::get_available_kernels() +{ + return available_kernels; +} +} // namespace kernels +} // namespace cpu +} // namespace arm_compute -- cgit v1.2.1