Compute Library
 21.02
NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel Class Reference

Neon kernel used to quantize down the int32 accumulator values of GEMMLowp to QSYMM16. More...

#include <NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.h>

Collaboration diagram for NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel:
[legend]

Public Member Functions

const char * name () const override
 Name of the kernel. More...
 
 NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel ()
 Constructor. More...
 
 NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel (const NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKerneloperator= (const NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
 NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel (NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel &&)=default
 Allow instances of this class to be moved. More...
 
NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKerneloperator= (NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel &&)=default
 Allow instances of this class to be moved. More...
 
 ~NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel ()=default
 Default destructor. More...
 
void configure (const ITensor *input, const ITensor *bias, ITensor *output, int result_fixedpoint_multiplier, int result_shift, int min=0, int max=0)
 Initialise the kernel's input and output. More...
 
void run (const Window &window, const ThreadInfo &info) override
 Execute the kernel on the passed window. More...
 
- Public Member Functions inherited from ICPPKernel
virtual ~ICPPKernel ()=default
 Default destructor. More...
 
virtual void run_nd (const Window &window, const ThreadInfo &info, const Window &thread_locator)
 legacy compatibility layer for implemantions which do not support thread_locator In these cases we simply narrow the interface down the legacy version More...
 
virtual void run_op (ITensorPack &tensors, const Window &window, const ThreadInfo &info)
 Execute the kernel on the passed window. More...
 
- Public Member Functions inherited from IKernel
 IKernel ()
 Constructor. More...
 
virtual ~IKernel ()=default
 Destructor. More...
 
virtual bool is_parallelisable () const
 Indicates whether or not the kernel is parallelisable. More...
 
virtual BorderSize border_size () const
 The size of the border for that kernel. More...
 
const Windowwindow () const
 The maximum window the kernel can be executed on. More...
 

Static Public Member Functions

static Status validate (const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, int min=0, int max=0)
 Static function to check if given info will lead to a valid configuration of NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel. More...
 

Detailed Description

Neon kernel used to quantize down the int32 accumulator values of GEMMLowp to QSYMM16.

This kernel takes a final int32 accumulator value (the output of NEGEMMLowpMatrixMultiplyKernel), and processes it to obtain the final QSYMM16 value. The following computations will be performed by the kernel:

  1. Compute fixed point multiplication between each entry of input by result_fixedpoint_multiplier
  2. Add bias to final result if bias tensor is not a nullptr
  3. Round to nearest division by a power-of-two using result_shift
  4. Clamp the value between the specified min and max bounds
  5. Clamp the resulting int32 values to the [-32768, 32767] range and cast to QSYMM16.

Definition at line 45 of file NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.h.

Constructor & Destructor Documentation

◆ NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel() [1/3]

Constructor.

Definition at line 191 of file NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp.

Referenced by NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::name().

192  : _func(nullptr), _input(nullptr), _bias(nullptr), _output(nullptr), _result_fixedpoint_multiplier(0), _result_shift(0), _min(0), _max(0)
193 {
194 }

◆ NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel() [2/3]

Prevent instances of this class from being copied (As this class contains pointers)

◆ NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel() [3/3]

◆ ~NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel()

Member Function Documentation

◆ configure()

void configure ( const ITensor input,
const ITensor bias,
ITensor output,
int  result_fixedpoint_multiplier,
int  result_shift,
int  min = 0,
int  max = 0 
)

Initialise the kernel's input and output.

Parameters
[in]inputInput tensor. Data type supported: S32
[in]biasBiases tensor. Only shared biases supported and it can be a nullptr if the biases addition is not required. Biases are 1D tensor with dimensions [OFM]. Data type supported: Same as input.
[out]outputOutput tensor. Data type supported: Data type supported: QSYMM16
[in]result_fixedpoint_multiplierFixed point value to be multiplied to each element of the input matrix when once the result_offset has been add
[in]result_shiftInteger value used to round to nearest division by a power-of-two the result after the fixed point multiplication
[in]min(Optional) Min value used to saturate down the output result before converting back to QSYMM16. Defaults to 0.
[in]max(Optional) Max value used to saturate up the output result before converting back to QSYMM16. Along with min, this value can be used to implement "rectified linear unit" activation functions. Defaults to 0.

Definition at line 196 of file NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp.

References ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_THROW_ON, ITensor::info(), arm_compute::test::validation::input, and arm_compute::validate_arguments().

Referenced by NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::name().

198 {
199  // Perform validate step
201  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(input->info(), (bias != nullptr) ? bias->info() : nullptr, output->info(), min, max));
202 
203  _input = input;
204  _bias = bias;
205  _output = output;
206  _result_fixedpoint_multiplier = result_fixedpoint_multiplier;
207  _result_shift = result_shift;
208  _min = min;
209  _max = max;
210 
211  // Configure kernel window
212  auto win_config = validate_and_configure_window(input->info(), output->info());
213  ARM_COMPUTE_ERROR_THROW_ON(win_config.first);
214  INEKernel::configure(win_config.second);
215 
216  // Check if we need to clamp the result using min and max
217  const bool is_bounded_relu = !(min <= -32768 && max >= 32767);
218  _func = is_bounded_relu ? &NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::run<true> : &NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::run<false>;
219 }
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161

◆ name()

◆ operator=() [1/2]

Prevent instances of this class from being copied (As this class contains pointers)

Referenced by NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::name().

◆ operator=() [2/2]

◆ run()

void run ( const Window window,
const ThreadInfo info 
)
overridevirtual

Execute the kernel on the passed window.

Warning
If is_parallelisable() returns false then the passed window must be equal to window()
Note
The window has to be a region within the window returned by the window() method
The width of the window has to be a multiple of num_elems_processed_per_iteration().
Parameters
[in]windowRegion on which to execute the kernel. (Must be a region of the window returned by window())
[in]infoInfo about executing thread and CPU.

Reimplemented from ICPPKernel.

Definition at line 230 of file NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp.

References ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, ARM_COMPUTE_UNUSED, and IKernel::window().

Referenced by NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::name().

231 {
235 
236  (this->*_func)(window);
237 }
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
#define ARM_COMPUTE_UNUSED(...)
To avoid unused variables warnings.
Definition: Error.h:152
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:941
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205

◆ validate()

Status validate ( const ITensorInfo input,
const ITensorInfo bias,
const ITensorInfo output,
int  min = 0,
int  max = 0 
)
static

Static function to check if given info will lead to a valid configuration of NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.

Parameters
[in]inputInput tensor info. Data type supported: S32
[in]biasBiases tensor info. Only shared biases supported and it can be a nullptr if the biases addition is not required. Biases are 1D tensor info with dimensions [OFM]. Data type supported: Same as input.
[in]outputOutput tensor info. Data type supported: Data type supported: QSYMM16
[in]min(Optional) Min value used to saturate down the output result before converting back to QSYMM16. Defaults to 0.
[in]max(Optional) Max value used to saturate up the output result before converting back to QSYMM16, Along with min, this value can be used to implement "rectified linear unit" activation functions. Defaults to 0.
Returns
a status

Definition at line 221 of file NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel.cpp.

References ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_RETURN_ON_ERROR, ICloneable< T >::clone(), and arm_compute::validate_arguments().

Referenced by NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPointKernel::name(), NEGEMMLowpQuantizeDownInt32ToInt16ScaleByFixedPoint::validate(), and NEGEMMLowpOutputStage::validate().

222 {
224  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(input, bias, output, min, max));
225  ARM_COMPUTE_RETURN_ON_ERROR(validate_and_configure_window(input->clone().get(), output->clone().get()).first);
226 
227  return Status{};
228 }
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
Status validate_arguments(const ITensorInfo *input, const ITensorInfo *bias, const ITensorInfo *output, const GEMMLowpOutputStageInfo *output_stage)
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161

The documentation for this class was generated from the following files: