Compute Library
 19.08
GCLogits1DShiftExpSumKernel Class Reference

Interface for shifting the logits values around the max value and exponentiating the result. More...

#include <GCSoftmaxLayerKernel.h>

Collaboration diagram for GCLogits1DShiftExpSumKernel:
[legend]

Public Member Functions

 GCLogits1DShiftExpSumKernel ()
 Default constructor. More...
 
 GCLogits1DShiftExpSumKernel (const GCLogits1DShiftExpSumKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
GCLogits1DShiftExpSumKerneloperator= (const GCLogits1DShiftExpSumKernel &)=delete
 Prevent instances of this class from being copied (As this class contains pointers) More...
 
 GCLogits1DShiftExpSumKernel (GCLogits1DShiftExpSumKernel &&)=default
 Allow instances of this class to be moved. More...
 
GCLogits1DShiftExpSumKerneloperator= (GCLogits1DShiftExpSumKernel &&)=default
 Allow instances of this class to be moved. More...
 
void configure (const IGCTensor *input, const IGCTensor *max, IGCTensor *output, IGCTensor *sum)
 Set the input and output tensors. More...
 
void run (const Window &window) override
 Enqueue the OpenGL ES shader to process the given window. More...
 
- Public Member Functions inherited from IGCKernel
 IGCKernel ()
 Constructor. More...
 
GCKernelkernel ()
 Returns a reference to the GLES kernel of this object. More...
 
void add_1D_tensor_argument (unsigned int &idx, const IGCTensor *tensor, const unsigned int binding_point, const Window &window)
 Add the passed 1D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_2D_tensor_argument (unsigned int &idx, const IGCTensor *tensor, const unsigned int binding_point, const Window &window)
 Add the passed 2D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_3D_tensor_argument (unsigned int &idx, const IGCTensor *tensor, const unsigned int binding_point, const Window &window)
 Add the passed 3D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
unsigned int num_arguments_per_1D_tensor () const
 Returns the number of arguments enqueued per 1D tensor object. More...
 
unsigned int num_arguments_per_2D_tensor () const
 Returns the number of arguments enqueued per 2D tensor object. More...
 
unsigned int num_arguments_per_3D_tensor () const
 Returns the number of arguments enqueued per 3D tensor object. More...
 
void set_lws_hint (gles::NDRange &lws_hint)
 Set the Local-Workgroup-Size hint. More...
 
void set_target (GPUTarget target)
 Set the targeted GPU architecture. More...
 
GPUTarget get_target () const
 Get the targeted GPU architecture. 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...
 

Detailed Description

Interface for shifting the logits values around the max value and exponentiating the result.

Definition at line 46 of file GCSoftmaxLayerKernel.h.

Constructor & Destructor Documentation

◆ GCLogits1DShiftExpSumKernel() [1/3]

Default constructor.

Definition at line 102 of file GCSoftmaxLayerKernel.cpp.

103  : _input(nullptr), _max(nullptr), _output(nullptr), _sum(nullptr)
104 {
105 }

◆ GCLogits1DShiftExpSumKernel() [2/3]

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

◆ GCLogits1DShiftExpSumKernel() [3/3]

Allow instances of this class to be moved.

Member Function Documentation

◆ configure()

void configure ( const IGCTensor input,
const IGCTensor max,
IGCTensor output,
IGCTensor sum 
)

Set the input and output tensors.

Parameters
[in]inputSource tensor. Data types supported: F16/F32
[in]maxMax values tensor. Data types supported: same as input
[out]outputDestination tensor. Data types supported: same as input
[out]sumSum of 1D logits tensor. Data types supported: same as input

Definition at line 107 of file GCSoftmaxLayerKernel.cpp.

108 {
110  ARM_COMPUTE_ERROR_ON_NULLPTR(max, sum, output);
111 
112  // Output auto initialization if not yet initialized
113  auto_init_if_empty(*sum->info(), max->info()->tensor_shape(), 1, input->info()->data_type());
114  auto_init_if_empty(*output->info(), input->info()->tensor_shape(), 1, input->info()->data_type());
115 
119 
120  _input = input;
121  _max = max;
122  _output = output;
123  _sum = sum;
124 
125  // Set build options
126  std::set<std::string> build_opts;
127  std::string dt_name = (input->info()->data_type() == DataType::F32) ? "DATA_TYPE_FP32" : "DATA_TYPE_FP16";
128  build_opts.insert("#define " + dt_name);
129  build_opts.emplace("#define LOCAL_SIZE_X " + support::cpp11::to_string(1));
130  build_opts.emplace("#define LOCAL_SIZE_Y " + support::cpp11::to_string(1));
131  build_opts.emplace("#define LOCAL_SIZE_Z " + support::cpp11::to_string(1));
132  build_opts.insert("#define SOFTMAX_LAYER_SHIFT_EXP_SUM");
133 
134  // Tell the kernel that the width is not a multiple of 8
135  if((input->info()->dimension(0) % 8) != 0)
136  {
137  build_opts.insert("#define NON_MULTIPLE_OF_8");
138  }
139 
140  // Create kernel
141  _kernel = static_cast<GCKernel>(GCKernelLibrary::get().create_kernel("softmax_layer_shift_exp_sum", build_opts));
142 
143  // Set fixed arguments
144  unsigned int idx = 4 * num_arguments_per_3D_tensor(); //Skip the input and output parameters
145  _kernel.set_argument(idx++, input->info()->dimension(0));
146 
147  // Configure window
148  // The kernel loops over all elements in steps of 8
149  const unsigned int num_elems_processed_per_iteration = ceil_to_multiple(input->info()->dimension(0), 8);
150  unsigned int num_elems_written_per_iteration = 1;
151  if(input->info()->data_type() == DataType::F16)
152  {
153  num_elems_written_per_iteration = 2;
154  }
155 
156  Window win = calculate_max_window(*input->info(), Steps(num_elems_processed_per_iteration));
157 
158  AccessWindowHorizontal input_access(input->info(), 0, num_elems_processed_per_iteration);
159  AccessWindowHorizontal max_access(max->info(), 0, num_elems_written_per_iteration);
160  AccessWindowHorizontal output_access(output->info(), 0, num_elems_processed_per_iteration);
161  AccessWindowHorizontal sum_access(sum->info(), 0, num_elems_written_per_iteration);
162 
163  update_window_and_padding(win, input_access, max_access, output_access, sum_access);
164 
165  output_access.set_valid_region(win, input->info()->valid_region());
166  sum_access.set_valid_region(win, ValidRegion(Coordinates(), sum->info()->tensor_shape()));
167 
168  IGCKernel::configure(win);
169 }
virtual size_t dimension(size_t index) const =0
Return the size of the requested dimension.
#define ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES(...)
Definition: Validate.h:543
DATA_TYPE sum(__global const DATA_TYPE *input)
Calculate sum of a vector.
std::string to_string(T &&value)
Convert integer and float values to string.
virtual DataType data_type() const =0
Data type used for each element of the tensor.
1 channel, 1 F32 per channel
unsigned int num_arguments_per_3D_tensor() const
Returns the number of arguments enqueued per 3D tensor object.
Definition: IGCKernel.cpp:147
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps=Steps(), bool skip_border=false, BorderSize border_size=BorderSize())
Calculate the maximum window for a given tensor shape and border setting.
Definition: Helpers.cpp:28
bool auto_init_if_empty(ITensorInfo &info, const TensorShape &shape, int num_channels, DataType data_type, QuantizationInfo quantization_info=QuantizationInfo())
Auto initialize the tensor info (shape, number of channels and data type) if the current assignment i...
Definition: Helpers.inl:201
#define ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES(...)
Definition: Validate.h:441
1 channel, 1 F16 per channel
virtual ValidRegion valid_region() const =0
Valid region of the tensor.
bool update_window_and_padding(Window &win, Ts &&... patterns)
Update window and padding size for each of the access patterns.
Definition: Helpers.h:402
virtual const TensorShape & tensor_shape() const =0
Size for each dimension of the tensor.
auto ceil_to_multiple(S value, T divisor) -> decltype(((value+divisor - 1)/divisor) *divisor)
Computes the smallest number larger or equal to value that is a multiple of divisor.
Definition: Utils.h:66
Class to describe a number of elements in each dimension.
Definition: Steps.h:40
Coordinates of an item.
Definition: Coordinates.h:37
Implementation of a row access pattern.
virtual ITensorInfo * info() const =0
Interface to be implemented by the child class to return the tensor's metadata.
std::unique_ptr< Kernel > create_kernel()
Helper function to create and return a unique_ptr pointed to a CL/GLES kernel object.
Definition: Helpers.h:86
#define ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN(t, c,...)
Definition: Validate.h:789
static GCKernelLibrary & get()
Get the static instance of GCKernelLibrary.
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:161
Container for valid region of a window.
Definition: Types.h:174
Describe a multidimensional execution window.
Definition: Window.h:39

References ARM_COMPUTE_ERROR_ON_DATA_TYPE_CHANNEL_NOT_IN, ARM_COMPUTE_ERROR_ON_MISMATCHING_DATA_TYPES, ARM_COMPUTE_ERROR_ON_MISMATCHING_SHAPES, ARM_COMPUTE_ERROR_ON_NULLPTR, arm_compute::auto_init_if_empty(), arm_compute::calculate_max_window(), arm_compute::ceil_to_multiple(), arm_compute::create_kernel(), ITensorInfo::data_type(), ITensorInfo::dimension(), arm_compute::F16, arm_compute::F32, GCKernelLibrary::get(), ITensor::info(), IGCKernel::num_arguments_per_3D_tensor(), sum(), ITensorInfo::tensor_shape(), arm_compute::support::cpp11::to_string(), arm_compute::update_window_and_padding(), and ITensorInfo::valid_region().

Referenced by GCSoftmaxLayer::configure().

◆ operator=() [1/2]

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

◆ operator=() [2/2]

Allow instances of this class to be moved.

◆ run()

void run ( const Window window)
overridevirtual

Enqueue the OpenGL ES shader to process the given window.

Parameters
[in]windowRegion on which to execute the kernel. (Must be a valid region of the window returned by window()).

Implements IGCKernel.

Definition at line 171 of file GCSoftmaxLayerKernel.cpp.

172 {
175 
177  Window slice = window_collapsed.first_slice_window_3D();
178 
179  _kernel.use();
180 
181  do
182  {
183  unsigned int idx = 0;
184  unsigned int binding = 1; // SSBO binding starts from 1.
185  // Set inputs
186  add_3D_tensor_argument(idx, _input, binding++, slice);
187  add_3D_tensor_argument(idx, _max, binding++, slice);
188  add_3D_tensor_argument(idx, _output, binding++, slice);
189  add_3D_tensor_argument(idx, _sum, binding++, slice);
190  _kernel.update_shader_params();
191  enqueue(*this, slice);
192  }
193  while(window_collapsed.slide_window_slice_3D(slice));
194 }
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
void add_3D_tensor_argument(unsigned int &idx, const IGCTensor *tensor, const unsigned int binding_point, const Window &window)
Add the passed 3D tensor's parameters to the object's kernel's arguments starting from the index idx.
Definition: IGCKernel.cpp:132
void enqueue(cl::CommandQueue &queue, ICLKernel &kernel, const Window &window, const cl::NDRange &lws_hint=CLKernelLibrary::get().default_ndrange(), bool use_dummy_work_items=false)
Add the kernel to the command queue with the given window.
Definition: ICLKernel.cpp:39
Window collapse_if_possible(const Window &full_window, size_t first, size_t last, bool *has_collapsed=nullptr) const
Collapse the dimensions between first and last if possible.
Definition: Window.inl:54
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:319
static constexpr size_t DimZ
Alias for dimension 2 also known as Z dimension.
Definition: Window.h:47
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:275
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:205
Describe a multidimensional execution window.
Definition: Window.h:39
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:940
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)

References IGCKernel::add_3D_tensor_argument(), ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, Window::collapse_if_possible(), Window::DimZ, arm_compute::enqueue(), Window::first_slice_window_3D(), arm_compute::test::validation::reference::slice(), Window::slide_window_slice_3D(), and IKernel::window().


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