Compute Library
 22.05
ClLogits1DMaxShiftExpSumKernel Class Reference

Interface for max, shifting, exponentiating and summing the logits. More...

#include <ClSoftmaxKernel.h>

Collaboration diagram for ClLogits1DMaxShiftExpSumKernel:
[legend]

Public Types

using ParallelReductionInfo = std::tuple< bool, unsigned int >
 Info for whether a parallel reduction will be run and the vector size of the execution. More...
 

Public Member Functions

 ClLogits1DMaxShiftExpSumKernel ()
 
 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE (ClLogits1DMaxShiftExpSumKernel)
 
void configure (const CLCompileContext &compile_context, const ITensorInfo &src, ITensorInfo &max, ITensorInfo &dst, ITensorInfo &sum, const SoftmaxKernelInfo &info)
 Configure the kernel using the given information about tensors. More...
 
void run_op (ITensorPack &tensors, const Window &window, ::cl::CommandQueue &queue) override
 
- Public Member Functions inherited from ICLKernel
 ICLKernel ()
 Constructor. More...
 
cl::Kernel & kernel ()
 Returns a reference to the OpenCL kernel of this object. More...
 
CLKernelType type () const
 Returns the CL kernel type. More...
 
template<typename T >
void add_1D_array_argument (unsigned int &idx, const ICLArray< T > *array, const Strides &strides, unsigned int num_dimensions, const Window &window)
 Add the passed 1D array's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_1D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, 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_1D_tensor_argument_if (bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
 Add the passed 1D tensor's parameters to the object's kernel's arguments starting from the index idx if the condition is true. More...
 
void add_2D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, 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_2D_tensor_argument_if (bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
 Add the passed 2D tensor's parameters to the object's kernel's arguments starting from the index idx if the condition is true. More...
 
void add_3D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
 Add the passed 3D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_4D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
 Add the passed 4D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_5D_tensor_argument (unsigned int &idx, const ICLTensor *tensor, const Window &window)
 Add the passed 5D tensor's parameters to the object's kernel's arguments starting from the index idx. More...
 
void add_3d_tensor_nhw_argument (unsigned int &idx, const ICLTensor *tensor)
 Add the passed NHW 3D tensor's parameters to the object's kernel's arguments by passing strides, dimensions and the offset to the first valid element in bytes. More...
 
void add_4d_tensor_nhwc_argument (unsigned int &idx, const ICLTensor *tensor)
 Add the passed NHWC 4D tensor's parameters to the object's kernel's arguments by passing strides, dimensions and the offset to the first valid element in bytes. More...
 
virtual void run (const Window &window, cl::CommandQueue &queue)
 Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue. More...
 
virtual void run_op (ITensorPack &tensors, const Window &window, cl::CommandQueue &queue)
 Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue. More...
 
virtual void run_composite_op (ITensorPack &tensors, const Window &window, cl::CommandQueue &queue, const experimental::dynamic_fusion::ClExecutionDescriptor &exec_desc)
 The execution is carried out through run_op method. But the run_op method needs to be extended to include ClExecutionDescriptor as now LWS GWS tuning will be separated from the IKernel. More...
 
template<typename T >
void add_argument (unsigned int &idx, T value)
 Add the passed parameters to the object's kernel's arguments starting from the index idx. More...
 
void set_lws_hint (const cl::NDRange &lws_hint)
 Set the Local-Workgroup-Size hint. More...
 
cl::NDRange lws_hint () const
 Return the Local-Workgroup-Size hint. More...
 
void set_wbsm_hint (const cl_int &wbsm_hint)
 Set the workgroup batch size modifier hint. More...
 
cl_int wbsm_hint () const
 Return the workgroup batch size modifier hint. More...
 
const std::string & config_id () const
 Get the configuration ID. More...
 
void set_target (GPUTarget target)
 Set the targeted GPU architecture. More...
 
void set_target (cl::Device &device)
 Set the targeted GPU architecture according to the CL device. More...
 
GPUTarget get_target () const
 Get the targeted GPU architecture. More...
 
size_t get_max_workgroup_size ()
 Get the maximum workgroup size for the device the CLKernelLibrary uses. More...
 
template<unsigned int dimension_size>
void add_tensor_argument (unsigned &idx, const ICLTensor *tensor, const Window &window)
 
template<typename T , unsigned int dimension_size>
void add_array_argument (unsigned &idx, const ICLArray< T > *array, const Strides &strides, unsigned int num_dimensions, const Window &window)
 Add the passed array's parameters to the object's kernel's arguments starting from the index idx. 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...
 
bool is_window_configured () const
 Function to check if the embedded window of this kernel has been configured. More...
 

Static Public Member Functions

static Status validate (const ITensorInfo &src, const ITensorInfo &max, const ITensorInfo &dst, const ITensorInfo &sum)
 Static function to check if given info will lead to a valid configuration. More...
 
static ParallelReductionInfo is_parallel_reduction (size_t size)
 Checks if the given size is eligible for parallel reduction. More...
 
- Static Public Member Functions inherited from ICLKernel
static constexpr unsigned int num_arguments_per_3d_tensor_nhw ()
 Returns the number of arguments enqueued per NHW 3D Tensor object. More...
 
static constexpr unsigned int num_arguments_per_4d_tensor_nhwc ()
 Returns the number of arguments enqueued per NHWC 4D Tensor object. More...
 
static constexpr unsigned int num_arguments_per_1D_array ()
 Returns the number of arguments enqueued per 1D array object. More...
 
static constexpr unsigned int num_arguments_per_1D_tensor ()
 Returns the number of arguments enqueued per 1D tensor object. More...
 
static constexpr unsigned int num_arguments_per_2D_tensor ()
 Returns the number of arguments enqueued per 2D tensor object. More...
 
static constexpr unsigned int num_arguments_per_3D_tensor ()
 Returns the number of arguments enqueued per 3D tensor object. More...
 
static constexpr unsigned int num_arguments_per_4D_tensor ()
 Returns the number of arguments enqueued per 4D tensor object. More...
 
static cl::NDRange gws_from_window (const Window &window)
 Get the global work size given an execution window. More...
 

Detailed Description

Interface for max, shifting, exponentiating and summing the logits.

Definition at line 40 of file ClSoftmaxKernel.h.

Member Typedef Documentation

◆ ParallelReductionInfo

using ParallelReductionInfo = std::tuple<bool, unsigned int>

Info for whether a parallel reduction will be run and the vector size of the execution.

Definition at line 51 of file ClSoftmaxKernel.h.

Constructor & Destructor Documentation

◆ ClLogits1DMaxShiftExpSumKernel()

Definition at line 157 of file ClSoftmaxKernel.cpp.

References arm_compute::ELEMENTWISE.

158 {
160 }
Elementeise CL kernel type.
Definition: CLTypes.h:84

Member Function Documentation

◆ ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE()

ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE ( ClLogits1DMaxShiftExpSumKernel  )

◆ configure()

void configure ( const CLCompileContext compile_context,
const ITensorInfo src,
ITensorInfo max,
ITensorInfo dst,
ITensorInfo sum,
const SoftmaxKernelInfo info 
)

Configure the kernel using the given information about tensors.

Parameters
[in]compile_contextThe compile context to be used.
[in]srcSource tensor. Data types supported: QASYMM8/QASYMM8_SIGNED/F16/F32
[in,out]maxMax values tensor. Data types supported: same as src
[out]dstDestination tensor. Data types supported: same as src
[out]sumSum of 1D logits tensor. Data types supported: same as src
[in]infoContains information consumed by kernels for softmax described in SoftmaxKernelInfo.

Definition at line 162 of file ClSoftmaxKernel.cpp.

References CLBuildOptions::add_option(), CLBuildOptions::add_option_if(), CLBuildOptions::add_options_if(), arm_compute::adjust_vec_size(), ARM_COMPUTE_ERROR_ON, ARM_COMPUTE_ERROR_THROW_ON, arm_compute::auto_init_if_empty(), SoftmaxKernelInfo::beta, arm_compute::calculate_max_window(), ICloneable< T >::clone(), arm_compute::create_kernel(), ITensorInfo::data_type(), ITensorInfo::dimension(), arm_compute::test::validation::dst, dt, arm_compute::F16, arm_compute::float_to_string_with_full_precision(), arm_compute::get_cl_type_from_data_type(), arm_compute::get_padding_info(), arm_compute::has_padding_changed(), SoftmaxKernelInfo::input_data_type, arm_compute::is_data_type_float(), arm_compute::is_data_type_quantized_asymmetric(), arm_compute::is_data_type_quantized_asymmetric_signed(), SoftmaxKernelInfo::is_log, kernel_name, arm_compute::support::cpp11::lround(), ICLKernel::lws_hint(), CLBuildOptions::options(), arm_compute::test::validation::qinfo, ITensorInfo::quantization_info(), UniformQuantizationInfo::scale, arm_compute::test::validation::src, ITensorInfo::tensor_shape(), arm_compute::support::cpp11::to_string(), and QuantizationInfo::uniform().

163 {
164  auto padding_info = get_padding_info({ &src, &max, &dst, &sum });
165 
166  // Output auto initialization if not yet initialized
167  auto_init_if_empty(sum, src.clone()->set_tensor_shape(max.tensor_shape()));
168  auto_init_if_empty(dst, *src.clone());
169 
170  // Perform validation step
171  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments_1DMaxShiftExpSum(src, max, dst, sum));
172 
173  const DataType dt = src.data_type();
174  const UniformQuantizationInfo qinfo = src.quantization_info().uniform();
175  const size_t reduction_dim_size = src.dimension(0);
176  const float beta = info.beta;
177  const auto is_signed_qasymm8 = is_data_type_quantized_asymmetric_signed(info.input_data_type);
178  const int min_value = is_signed_qasymm8 ? CL_SCHAR_MIN : 0;
179 
180  const unsigned int vector_size = adjust_vec_size(_serial_vector_size, reduction_dim_size);
181 
182  // Set build options
183  CLBuildOptions build_opts;
184  build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(dt));
185  build_opts.add_option("-DMIN_VALUE=" + support::cpp11::to_string(min_value));
186  build_opts.add_option("-DVECTOR_SIZE=" + support::cpp11::to_string(vector_size));
187  build_opts.add_option("-DSRC_WIDTH=" + support::cpp11::to_string(reduction_dim_size));
188  build_opts.add_option("-DVECTOR_SIZE_LEFTOVER=" + support::cpp11::to_string(reduction_dim_size % vector_size));
189  build_opts.add_option("-DLOG_VECTOR_SIZE=" + support::cpp11::to_string(lround(log2(vector_size))));
190  build_opts.add_option_if((reduction_dim_size % vector_size) != 0, "-DNON_MULTIPLE_OF_VECTOR_SIZE");
191  build_opts.add_option_if(is_signed_qasymm8, "-DQASYMM8_SIGNED");
192  build_opts.add_option_if(is_data_type_float(dt) && (beta != 1.0f), "-DBETA=" + float_to_string_with_full_precision(beta));
193  build_opts.add_option_if(is_data_type_float(dt) && info.is_log, "-DLOG_SOFTMAX");
194  build_opts.add_option_if(is_data_type_float(dt), "-DMINVAL=" + ((dt == DataType::F16) ? std::string("-HALF_MAX") : std::string("-FLT_MAX")));
195  build_opts.add_option_if(is_data_type_quantized_asymmetric(dt), "-DSCALE=" + float_to_string_with_full_precision(qinfo.scale));
196  build_opts.add_option_if(is_data_type_quantized_asymmetric(dt), "-DBETA=" + float_to_string_with_full_precision(beta));
197  build_opts.add_options_if(is_data_type_quantized_asymmetric(dt), prepare_quantized_softmax_build_options(qinfo.scale, beta).options());
198 
199  cl::NDRange lws_hint(cl::NullRange);
200  std::string kernel_name = std::string("softmax_layer_max_shift_exp_sum_") + (is_data_type_quantized_asymmetric(dt) ? "quantized_" : "") + "serial";
201 
202  // Create kernel.
203  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
204 
205  // Configure window
206  Window win = calculate_max_window(src, Steps(reduction_dim_size));
207  IClKernel::configure_internal(win, lws_hint);
208 
210 }
Window calculate_max_window(const ValidRegion &valid_region, const Steps &steps, bool skip_border, BorderSize border_size)
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:384
std::string to_string(T &&value)
Convert integer and float values to string.
#define ARM_COMPUTE_ERROR_ON(cond)
If the condition is true then an error message is printed and an exception thrown.
Definition: Error.h:466
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
SimpleTensor< float > src
Definition: DFT.cpp:155
1 channel, 1 F16 per channel
cl::Kernel create_kernel(const CLCompileContext &ctx, const std::string &kernel_name, const std::set< std::string > &build_opts=std::set< std::string >())
Creates an opencl kernel using a compile context.
Definition: CLHelpers.cpp:391
std::string float_to_string_with_full_precision(float val)
Create a string with the float in full precision.
Definition: Utils.h:1124
std::string get_cl_type_from_data_type(const DataType &dt)
Translates a tensor data type to the appropriate OpenCL type.
Definition: CLHelpers.cpp:39
bool is_data_type_quantized_asymmetric_signed(DataType dt)
Check if a given data type is of asymmetric quantized signed type.
Definition: Utils.h:1071
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...
bool has_padding_changed(const std::unordered_map< const ITensorInfo *, PaddingSize > &padding_map)
Check if the previously stored padding info has changed after configuring a kernel.
Definition: Utils.cpp:601
bool is_data_type_quantized_asymmetric(DataType dt)
Check if a given data type is of asymmetric quantized type.
Definition: Utils.h:1052
ScaleKernelInfo info(interpolation_policy, default_border_mode, PixelValue(), sampling_policy, false)
const QuantizationInfo qinfo
Definition: Im2Col.cpp:155
std::unordered_map< const ITensorInfo *, PaddingSize > get_padding_info(std::initializer_list< const ITensorInfo *> infos)
Stores padding information before configuring a kernel.
Definition: Utils.cpp:586
long lround(T value)
Round floating-point value with half value rounding away from zero and cast to long.
unsigned int adjust_vec_size(unsigned int vec_size, size_t dim0)
Returns the adjusted vector size in case it is less than the input&#39;s first dimension, getting rounded down to its closest valid vector size.
Definition: Utils.h:1222
std::string kernel_name
DataType
Available data types.
Definition: Types.h:79
bool is_data_type_float(DataType dt)
Check if a given data type is of floating point type.
Definition: Utils.h:1010

◆ is_parallel_reduction()

ClLogits1DMaxShiftExpSumKernel::ParallelReductionInfo is_parallel_reduction ( size_t  size)
static

Checks if the given size is eligible for parallel reduction.

Note
Serial reduction is launched for width < (_grid_size * _serial_vector_size).
Parallel reduction is launched for width >= (_grid_size * _serial_vector_size) and vector_size is forced to 4.
Parameters
[in]sizeSize to check
Returns
A two-element tuple where the first element is a boolean specifying if a parallel reduction will be run, while the second element is the vector size of the execution.

Definition at line 218 of file ClSoftmaxKernel.cpp.

Referenced by ClLogits1DMaxShiftExpSumKernel::run_op().

219 {
220  bool is_parallel_reduction = (size >= (_grid_size * _serial_vector_size)) && (_grid_size > 1);
221  unsigned int vector_size = is_parallel_reduction ? _parallel_vector_size : _serial_vector_size;
222  return std::make_tuple(is_parallel_reduction, vector_size);
223 }
static ParallelReductionInfo is_parallel_reduction(size_t size)
Checks if the given size is eligible for parallel reduction.

◆ run_op()

void run_op ( ITensorPack tensors,
const Window window,
::cl::CommandQueue &  queue 
)
override

Definition at line 225 of file ClSoftmaxKernel.cpp.

References arm_compute::ACL_DST, arm_compute::ACL_INT_0, arm_compute::ACL_INT_1, arm_compute::ACL_SRC, ICLKernel::add_3D_tensor_argument(), ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, Window::collapse_if_possible(), Window::DimX, Window::DimZ, arm_compute::enqueue(), Window::first_slice_window_3D(), ITensorPack::get_const_tensor(), ITensorPack::get_tensor(), ClLogits1DMaxShiftExpSumKernel::is_parallel_reduction(), ICLKernel::lws_hint(), Window::set(), arm_compute::test::validation::reference::slice(), Window::slide_window_slice_3D(), and IKernel::window().

226 {
229 
230  auto src = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC));
231  auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
232  auto max = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_INT_0));
233  auto sum = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_INT_1));
234 
235  ARM_COMPUTE_ERROR_ON_NULLPTR(src, dst, max, sum);
236 
237  // Collapse window in Z dimension
238  Window window_collapsed = window.collapse_if_possible(IClKernel::window(), Window::DimZ);
239 
240  // Reconfigure window in case of parallel reduction
241  ParallelReductionInfo parallel_reduction_info = is_parallel_reduction(src->info()->dimension(0));
242  if(std::get<0>(parallel_reduction_info))
243  {
244  // Launch grid_size parallel work items
245  window_collapsed.set(Window::DimX, Window::Dimension(0, _grid_size, 1));
246  }
247 
248  // Get slices
249  Window slice = window_collapsed.first_slice_window_3D();
250  do
251  {
252  unsigned int idx = 0;
253  // Set inputs
254  add_3D_tensor_argument(idx, src, slice);
255  add_3D_tensor_argument(idx, max, slice);
256  add_3D_tensor_argument(idx, dst, slice);
257  add_3D_tensor_argument(idx, sum, slice);
258  enqueue(queue, *this, slice, lws_hint());
259  }
260  while(window_collapsed.slide_window_slice_3D(slice));
261 }
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
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:32
cl::NDRange lws_hint() const
Return the Local-Workgroup-Size hint.
Definition: ICLKernel.h:384
static ParallelReductionInfo is_parallel_reduction(size_t size)
Checks if the given size is eligible for parallel reduction.
std::tuple< bool, unsigned int > ParallelReductionInfo
Info for whether a parallel reduction will be run and the vector size of the execution.
void add_3D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 3D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:227
SimpleTensor< float > src
Definition: DFT.cpp:155
static constexpr size_t DimX
Alias for dimension 0 also known as X dimension.
Definition: Window.h:43
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:68
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:915
static constexpr size_t DimZ
Alias for dimension 2 also known as Z dimension.
Definition: Window.h:47
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
#define ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW(f, s)
Definition: Validate.h:201
SimpleTensor< T > slice(const SimpleTensor< T > &src, Coordinates starts, Coordinates ends)

◆ validate()

Status validate ( const ITensorInfo src,
const ITensorInfo max,
const ITensorInfo dst,
const ITensorInfo sum 
)
static

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

Similar to ClLogits1DMaxShiftExpSumKernel::configure()

Returns
a status

Definition at line 212 of file ClSoftmaxKernel.cpp.

References ARM_COMPUTE_RETURN_ON_ERROR.

Referenced by ClSoftmax::validate().

213 {
214  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments_1DMaxShiftExpSum(src, max, dst, sum));
215  return Status{};
216 }
#define ARM_COMPUTE_RETURN_ON_ERROR(status)
Checks if a status contains an error and returns it.
Definition: Error.h:204
SimpleTensor< float > src
Definition: DFT.cpp:155

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