aboutsummaryrefslogtreecommitdiff
path: root/src/core/NEON/kernels/NERangeKernel.cpp
diff options
context:
space:
mode:
Diffstat (limited to 'src/core/NEON/kernels/NERangeKernel.cpp')
-rw-r--r--src/core/NEON/kernels/NERangeKernel.cpp90
1 files changed, 35 insertions, 55 deletions
diff --git a/src/core/NEON/kernels/NERangeKernel.cpp b/src/core/NEON/kernels/NERangeKernel.cpp
index ec63a35de9..87b7b76b72 100644
--- a/src/core/NEON/kernels/NERangeKernel.cpp
+++ b/src/core/NEON/kernels/NERangeKernel.cpp
@@ -29,11 +29,12 @@
#include "arm_compute/core/TensorInfo.h"
#include "arm_compute/core/Utils.h"
#include "arm_compute/core/Validate.h"
-#include "src/core/NEON/NEAsymm.h"
-#include "src/core/NEON/wrapper/wrapper.h"
+
#include "src/core/common/Registrars.h"
#include "src/core/helpers/AutoConfiguration.h"
#include "src/core/helpers/WindowHelpers.h"
+#include "src/core/NEON/NEAsymm.h"
+#include "src/core/NEON/wrapper/wrapper.h"
#include "src/cpu/kernels/range/list.h"
namespace arm_compute
@@ -55,48 +56,23 @@ struct RangeUKernel
RangeUKernelPtr ukernel;
};
-static const RangeUKernel available_kernels[] =
-{
- {
- "fp16_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::F16; },
- REGISTER_FP16_NEON(arm_compute::cpu::fp16_neon_range_function)
- },
- {
- "f32_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::F32; },
- REGISTER_FP32_NEON(arm_compute::cpu::fp32_neon_range_function)
- },
- {
- "u8_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::U8; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::u8_neon_range_function)
- },
- {
- "u16_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::U16; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::u16_neon_range_function)
- },
- {
- "u32_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::U32; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::u32_neon_range_function)
- },
- {
- "s8_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::S8; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::s8_neon_range_function)
- },
- {
- "s16_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::S16; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::s16_neon_range_function)
- },
- {
- "s32_neon_range",
- [](const RangeSelectorData & data) { return data.dt == DataType::S32; },
- REGISTER_INTEGER_NEON(arm_compute::cpu::s32_neon_range_function)
- },
+static const RangeUKernel available_kernels[] = {
+ {"fp16_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::F16; },
+ REGISTER_FP16_NEON(arm_compute::cpu::fp16_neon_range_function)},
+ {"f32_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::F32; },
+ REGISTER_FP32_NEON(arm_compute::cpu::fp32_neon_range_function)},
+ {"u8_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::U8; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::u8_neon_range_function)},
+ {"u16_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::U16; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::u16_neon_range_function)},
+ {"u32_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::U32; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::u32_neon_range_function)},
+ {"s8_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::S8; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::s8_neon_range_function)},
+ {"s16_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::S16; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::s16_neon_range_function)},
+ {"s32_neon_range", [](const RangeSelectorData &data) { return data.dt == DataType::S32; },
+ REGISTER_INTEGER_NEON(arm_compute::cpu::s32_neon_range_function)},
};
/** Micro-kernel selector
@@ -107,9 +83,9 @@ static const RangeUKernel available_kernels[] =
*/
const RangeUKernel *get_implementation(const RangeSelectorData &data)
{
- for(const auto &uk : available_kernels)
+ for (const auto &uk : available_kernels)
{
- if(uk.is_selected(data))
+ if (uk.is_selected(data))
{
return &uk;
}
@@ -119,28 +95,31 @@ const RangeUKernel *get_implementation(const RangeSelectorData &data)
Status validate_arguments(const ITensorInfo &output, const float start, const float end, const float step)
{
- const auto *uk = get_implementation(RangeSelectorData{ output.data_type() });
+ const auto *uk = get_implementation(RangeSelectorData{output.data_type()});
ARM_COMPUTE_RETURN_ERROR_ON(uk == nullptr || uk->ukernel == nullptr);
ARM_COMPUTE_RETURN_ERROR_ON_MSG((start == end), "start of the requested sequence must not be equal to the end");
ARM_COMPUTE_RETURN_ERROR_ON_MSG(((start < end) && (step <= 0)), "step must be greater than 0 when start < end");
ARM_COMPUTE_RETURN_ERROR_ON_MSG(((start > end) && (step >= 0)), "step must be less than 0 when start > end");
- ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(start, output.data_type(), output.quantization_info()), "start value is outside the range of the data type");
- ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(end, output.data_type(), output.quantization_info()), "end value is outside the range of the data type");
- ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(step, output.data_type(), output.quantization_info()), "step value is outside the range of the data type");
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(start, output.data_type(), output.quantization_info()),
+ "start value is outside the range of the data type");
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(end, output.data_type(), output.quantization_info()),
+ "end value is outside the range of the data type");
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(!check_value_range(step, output.data_type(), output.quantization_info()),
+ "step value is outside the range of the data type");
ARM_COMPUTE_RETURN_ERROR_ON_MSG((start == end), "start of the requested sequence must not be equal to the end");
ARM_COMPUTE_RETURN_ERROR_ON_MSG(output.num_dimensions() != 1, "Output has to be a 1-D tensor");
- ARM_COMPUTE_RETURN_ERROR_ON_MSG(output.tensor_shape().total_size() < num_of_elements_in_range(start, end, step), "Output tensor size is incorrect");
+ ARM_COMPUTE_RETURN_ERROR_ON_MSG(output.tensor_shape().total_size() < num_of_elements_in_range(start, end, step),
+ "Output tensor size is incorrect");
return Status{};
}
} // namespace
-NERangeKernel::NERangeKernel()
- : _start(0), _end(1), _step(1), _output(nullptr)
+NERangeKernel::NERangeKernel() : _start(0), _end(1), _step(1), _output(nullptr)
{
}
@@ -151,7 +130,8 @@ void NERangeKernel::configure(ITensor *output, float start, float end, float ste
ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(*(output->info()), start, end, step));
// Auto initialize output if not initialized
- auto_init_if_empty(*output->info(), TensorShape(num_of_elements_in_range(start, end, step)), 1, output->info()->data_type(), output->info()->quantization_info());
+ auto_init_if_empty(*output->info(), TensorShape(num_of_elements_in_range(start, end, step)), 1,
+ output->info()->data_type(), output->info()->quantization_info());
// Configure kernel window
Window win = calculate_max_window(*output->info(), Steps());
@@ -178,7 +158,7 @@ void NERangeKernel::run(const Window &window, const ThreadInfo &info)
ARM_COMPUTE_UNUSED(info);
ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(this);
ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(INEKernel::window(), window);
- const auto *uk = get_implementation(RangeSelectorData{ _output->info()->data_type() });
+ const auto *uk = get_implementation(RangeSelectorData{_output->info()->data_type()});
uk->ukernel(_output, _start, _step, window);
}
generated by cgit v1.2.1 (git 2.23.0) at 2024-05-10 04:55:06 +0000