Compute Library
 22.11
ClGemmMatrixMultiplyReshapedOnlyRhsKernel Class Reference

OpenCL kernel to multiply matrices when only the input matrix RHS (src1) has been reshaped. More...

#include <ClGemmMatrixMultiplyReshapedOnlyRhsKernel.h>

Collaboration diagram for ClGemmMatrixMultiplyReshapedOnlyRhsKernel:
[legend]

Public Member Functions

 ClGemmMatrixMultiplyReshapedOnlyRhsKernel ()
 
 ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE (ClGemmMatrixMultiplyReshapedOnlyRhsKernel)
 
void configure (const ClCompileContext &compile_context, const ITensorInfo *src0, const ITensorInfo *src1, const ITensorInfo *src2, ITensorInfo *dst, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info, const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
 Initialise the kernel's input and output. More...
 
void run_op (ITensorPack &tensors, const Window &window, cl::CommandQueue &queue) override
 Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue. More...
 
- 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_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 *src0, const ITensorInfo *src1, const ITensorInfo *src2, const ITensorInfo *dst, float alpha, float beta, const GEMMLHSMatrixInfo &lhs_info, const GEMMRHSMatrixInfo &rhs_info, const GEMMKernelInfo &gemm_info)
 Static function to check if given info will lead to a valid configuration. 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

OpenCL kernel to multiply matrices when only the input matrix RHS (src1) has been reshaped.

Note
The input matrix src1 must be reshaped through ClGemmReshapeRhsMatrixKernel

Definition at line 43 of file ClGemmMatrixMultiplyReshapedOnlyRhsKernel.h.

Constructor & Destructor Documentation

◆ ClGemmMatrixMultiplyReshapedOnlyRhsKernel()

Member Function Documentation

◆ ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE()

ARM_COMPUTE_DISALLOW_COPY_ALLOW_MOVE ( ClGemmMatrixMultiplyReshapedOnlyRhsKernel  )

◆ configure()

void configure ( const ClCompileContext compile_context,
const ITensorInfo src0,
const ITensorInfo src1,
const ITensorInfo src2,
ITensorInfo dst,
float  alpha,
float  beta,
const GEMMLHSMatrixInfo lhs_info,
const GEMMRHSMatrixInfo rhs_info,
const GEMMKernelInfo gemm_info 
)

Initialise the kernel's input and output.

Note
If rhs_info.export_to_cl_image = true, this OpenCL kernel will fetch the RHS data using the OpenCL read_image built-in function. Reading from the OpenCL image object can increase the performance. However, since the OpenCL image object is created importing the OpenCL buffer, the following conditions are required:
  1. rhs_info.n0 can only be 4, 8 and 16
  2. rhs_info.k0 can only be 4, 8 and 16
  3. Data type can only be F32
  4. The platform should support the OpenCL cl_khr_image2d_from_buffer extension
  5. The stride Y for the src1 should satisfy the OpenCL pitch alignment requirement
  6. src1 width should be less or equal to (CL_DEVICE_IMAGE2D_MAX_WIDTH * 4)
  7. src1 (height * depth) should be less or equal to CL_DEVICE_IMAGE2D_MAX_HEIGHT
Parameters
[in]compile_contextThe compile context to be used.
[in]src0Input tensor containing the LHS matrix. Data type supported: F16/F32 (only F32 if rhs_info.export_to_cl_image = true). The number of dimensions for the LHS matrix must be less or equal than 4.
[in]src1Input tensor containing the RHS reshaped matrix. Data type supported: same as src0. The number of dimensions for the RHS matrix must be less or equal than 3.
[in]src2Input tensor containing the bias matrix. Data type supported: same as src0.
[out]dstOutput tensor to store the result of matrix multiplication. Data type supported: same as src0
[in]alphaWeight of the matrix product
[in]betaWeight of the matrix bias
[in]lhs_infoLHS matrix information used to retrieve the number of rows to be processed by each thread. Only the following values are supported: lhs_info.m0: 1,2,3,4,5,6,7,8
[in]rhs_infoRHS matrix information used for reshaping the src1 tensor. Only the following values are supported: rhs_info.k0: 2,3,4,8,16 rhs_info.n0: 2,3,4,8,16 rhs_info.transpose: true,false
[in]gemm_infoGEMM information used to retrieve the original dimensions of the input matrices

Definition at line 187 of file ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp.

References CLBuildOptions::add_option(), CLBuildOptions::add_option_if(), ARM_COMPUTE_ERROR_ON, ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_THROW_ON, arm_compute::auto_init_if_empty(), ICloneable< T >::clone(), arm_compute::misc::shape_calculator::compute_mm_shape(), arm_compute::create_kernel(), ITensorInfo::data_type(), GEMMKernelInfo::depth_output_gemm3d, ITensorInfo::dimension(), GEMMRHSMatrixInfo::export_to_cl_image, arm_compute::float_to_string_with_full_precision(), arm_compute::test::validation::gemm_info, CLKernelLibrary::get(), arm_compute::get_cl_type_from_data_type(), arm_compute::get_padding_info(), GEMMKernelInfo::has_pad_y, arm_compute::has_padding_changed(), arm_compute::helpers::float_ops::is_one(), kernel_name, arm_compute::test::validation::lhs_info, arm_compute::lower_string(), GEMMLHSMatrixInfo::m0, ITensorInfo::num_dimensions(), CLBuildOptions::options(), GEMMKernelInfo::post_ops, arm_compute::preferred_dummy_work_items_support(), GEMMKernelInfo::reinterpret_input_as_3d, arm_compute::test::validation::rhs_info, arm_compute::string_from_activation_func(), arm_compute::string_from_data_type(), arm_compute::support::cpp11::to_string(), arm_compute::upper_string(), arm_compute::cpu::kernels::validate_and_configure_window(), and arm_compute::cpu::kernels::validate_arguments().

190 {
191  ARM_COMPUTE_ERROR_ON_NULLPTR(src0, src1, dst);
192 
193  // dst tensor auto initialization if not yet initialized
194  auto_init_if_empty(*dst, src0->clone()->set_tensor_shape(misc::shape_calculator::compute_mm_shape(*src0, *src1, gemm_info)));
195 
196  ARM_COMPUTE_ERROR_THROW_ON(validate_arguments(src0, src1, src2, dst, alpha, beta, lhs_info, rhs_info, gemm_info));
197 
198  _reinterpret_input_as_3d = gemm_info.reinterpret_input_as_3d;
199  _reinterpret_output_as_3d = gemm_info.depth_output_gemm3d != 0;
200  _use_dummy_work_items = preferred_dummy_work_items_support(CLKernelLibrary::get().get_device());
201  _add_bias = src2 != nullptr;
202  _export_to_cl_image = rhs_info.export_to_cl_image;
203  _has_pad_y = gemm_info.has_pad_y;
204  _num_post_op_args = gemm_info.post_ops.total_num_arguments();
205 
206  auto padding_info = get_padding_info({ src0, src1, src2, dst });
207 
208  // In case both input and dst have to be reinterpreted as 3D tensors,
209  // force reinterpret_input_as_3d and reinterpret_output_as_3d to be false.
210  if((_reinterpret_input_as_3d == _reinterpret_output_as_3d) && _has_pad_y)
211  {
212  _reinterpret_input_as_3d = false;
213  _reinterpret_output_as_3d = false;
214  }
215 
216  // Check if we need to slide the matrix B
217  const unsigned int num_dimensions_src0 = src0->num_dimensions();
218  _slide_matrix_b = (src1->num_dimensions() >= num_dimensions_src0);
219 
220  ElementsProcessed num_elements_processed{};
221 
222  // Configure kernel window
223  Window win = validate_and_configure_window(src0->clone().get(), src1->clone().get(), (src2 != nullptr) ? src2->clone().get() : nullptr, dst->clone().get(), lhs_info, rhs_info, gemm_info,
224  num_elements_processed);
225  ICLKernel::configure_internal(win);
226 
227  // If _reinterpret_input_as_3d = reinterpret_output_as_3d = true,
228  // we will dispatch a batched-GEMM to reduce the complexity of the address calculation within the OpenCL kernel.
229  // This means that the actual m used by the kernel is given by dst->dimension(1) and not by gemm_info.m
230  const unsigned int internal_m = _reinterpret_output_as_3d ? gemm_info.m : dst->dimension(1);
231 
232  // These variables are used only if gemm_info.has_pad_y == true
233  const unsigned int h_gemm_3d = _reinterpret_output_as_3d ? dst->dimension(1) : src0->dimension(1);
234  const unsigned int d_gemm_3d = _reinterpret_output_as_3d ? dst->dimension(2) : src0->dimension(2);
235 
236  // Shrink M0 to be always <= M (internal_m) to prevent out-of-bounds reads.
237  // NOTE: This might have implications on heuristics and performance
238  const unsigned int internal_m0 = std::min(internal_m, lhs_info.m0);
239 
240  // Calculate partial (store instead of load) M0 and partial N0 for the partial blocks at the end of a row/column if any. This is to avoid padding.
241  const unsigned int partial_store_m0 = internal_m % internal_m0;
242  const unsigned int partial_store_n0 = gemm_info.n % rhs_info.n0;
243  _m = internal_m;
244  _n = gemm_info.n;
245  _k = gemm_info.k;
246  // Create build options
247  CLBuildOptions build_opts;
248  build_opts.add_option("-DDATA_TYPE=" + get_cl_type_from_data_type(src0->data_type()));
249  build_opts.add_option_if(!(helpers::float_ops::is_one(alpha)), "-DALPHA=" + float_to_string_with_full_precision(alpha));
250  build_opts.add_option_if(src2 != nullptr, "-DBETA=" + float_to_string_with_full_precision(beta));
251  build_opts.add_option_if(helpers::float_ops::is_one(beta), "-DUNIT_BETA");
252  build_opts.add_option_if(gemm_info.broadcast_bias, "-DBROADCAST_BIAS");
253  build_opts.add_option_if(!_slide_matrix_b, "-DMATRIX_B_DEPTH=" + support::cpp11::to_string(src1->dimension(2)));
254  build_opts.add_option_if(rhs_info.interleave, "-DRHS_INTERLEAVE");
255  build_opts.add_option_if(_use_dummy_work_items, "-DDUMMY_WORK_ITEMS");
256  build_opts.add_option_if(rhs_info.export_to_cl_image, "-DOPENCL_IMAGE_SUPPORT");
257  build_opts.add_option("-DRHS_HEIGHT=" + support::cpp11::to_string(src1->dimension(1)));
258  build_opts.add_option("-DM0=" + support::cpp11::to_string(internal_m0));
259  build_opts.add_option("-DN0=" + support::cpp11::to_string(rhs_info.n0));
260  build_opts.add_option("-DK0=" + support::cpp11::to_string(rhs_info.k0));
261  build_opts.add_option("-DH0=" + support::cpp11::to_string(rhs_info.h0));
262  build_opts.add_option("-DPARTIAL_STORE_M0=" + support::cpp11::to_string(partial_store_m0));
263  build_opts.add_option("-DPARTIAL_STORE_N0=" + support::cpp11::to_string(partial_store_n0));
264  if(_has_pad_y)
265  {
266  build_opts.add_option_if(_reinterpret_input_as_3d, "-DREINTERPRET_INPUT_AS_3D");
267  build_opts.add_option_if(_reinterpret_output_as_3d, "-DREINTERPRET_OUTPUT_AS_3D");
268  build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DHEIGHT_GEMM3D=" + support::cpp11::to_string(h_gemm_3d));
269  build_opts.add_option_if(_reinterpret_input_as_3d || _reinterpret_output_as_3d, "-DDEPTH_GEMM3D=" + support::cpp11::to_string(d_gemm_3d));
270  }
271  // If post_ops are used, then we disable the use of gemm_info.activation_info
272  if(gemm_info.post_ops.size() > 0)
273  {
274  post_op_utils.set_post_ops_cl_build_options(build_opts, gemm_info.post_ops);
275  }
276  else
277  {
278  build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DACTIVATION_TYPE=" + lower_string(string_from_activation_func(gemm_info.activation_info.activation())));
279  build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DA_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.a()));
280  build_opts.add_option_if(gemm_info.activation_info.enabled(), "-DB_VAL=" + float_to_string_with_full_precision(gemm_info.activation_info.b()));
281  }
282 
283  std::string kernel_name("gemm_mm_reshaped_only_rhs_");
284  kernel_name += rhs_info.transpose ? "t" : "nt";
285  kernel_name += rhs_info.export_to_cl_image ? "_texture" : "";
286  post_op_utils.set_post_ops_cl_kernel_name(kernel_name, gemm_info.post_ops);
287 
288  // A macro guard to compile ONLY the kernel of interest
289  build_opts.add_option("-D" + upper_string(kernel_name));
290 
291  // Create kernel
292  _kernel = create_kernel(compile_context, kernel_name, build_opts.options());
293 
294  // Set config_id for enabling LWS tuning
295  _config_id = kernel_name;
296  _config_id += "_";
297  _config_id += (_has_pad_y ? "" : "no_pad_y_");
298  _config_id += (_add_bias ? "add_bias_" : "");
299  _config_id += (gemm_info.broadcast_bias ? "broadcast_bias_" : "");
300  _config_id += (_reinterpret_input_as_3d ? "3di_" : "");
301  _config_id += (_reinterpret_output_as_3d ? "3do_" : "");
302  _config_id += (gemm_info.activation_info.enabled() ? "fused_activation_" : "");
303  _config_id += lower_string(string_from_data_type(src0->data_type()));
304  _config_id += "_";
305  _config_id += support::cpp11::to_string(dst->dimension(1));
306  _config_id += "_";
307  _config_id += support::cpp11::to_string(dst->dimension(0));
308  _config_id += "_";
309  _config_id += support::cpp11::to_string(gemm_info.k);
310  _config_id += "_";
311  _config_id += support::cpp11::to_string(dst->dimension(2));
312  _config_id += "_";
313  _config_id += support::cpp11::to_string(lhs_info.m0);
314  _config_id += "_";
315  _config_id += support::cpp11::to_string(rhs_info.n0);
316  _config_id += "_";
317  _config_id += support::cpp11::to_string(rhs_info.k0);
318  _config_id += "_";
319  _config_id += support::cpp11::to_string(rhs_info.h0);
320  _config_id += "_";
321  _config_id += support::cpp11::to_string(rhs_info.interleave);
322 
324 }
bool is_one(float a, float epsilon=0.00001f)
Checks if the input floating point number is 1.0f checking if the difference is within a range define...
Definition: float_ops.h:97
bool preferred_dummy_work_items_support(const cl::Device &device)
Helper function to check if "dummy work-items" are preferred to have a power of two NDRange In case d...
Definition: CLHelpers.cpp:367
std::string to_string(T &&value)
Convert integer and float values to string.
TensorShape compute_mm_shape(const ITensorInfo &input0, const ITensorInfo &input1, bool is_interleaved_transposed, const GEMMReshapeInfo &reshape_info)
Calculate the matrix multiplication output shape of two tensors.
#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
const std::string & string_from_activation_func(ActivationLayerInfo::ActivationFunction act)
Translates a given activation function to a string.
Definition: Utils.cpp:163
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
#define ARM_COMPUTE_ERROR_THROW_ON(status)
Definition: Error.h:455
std::string lower_string(const std::string &val)
Lower a given string.
Definition: Utils.cpp:353
Status validate_arguments(const ITensorInfo *src, const ITensorInfo *weights, const ITensorInfo *dst, const PadStrideInfo &conv_info)
std::string upper_string(const std::string &val)
Raise a given string to upper case.
Definition: Utils.cpp:360
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:404
const std::string & string_from_data_type(DataType dt)
Convert a data type identity into a string.
Definition: Utils.cpp:135
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 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:603
std::pair< Status, Window > validate_and_configure_window(ITensorInfo *src, ITensorInfo *dst)
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:588
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
std::string kernel_name

◆ run_op()

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

Enqueue the OpenCL kernel to process the given window on the passed OpenCL command queue.

Note
The queue is not flushed by this method, and therefore the kernel will not have been executed by the time this method returns.
Parameters
[in]tensorsA vector containing the tensors to operato on.
[in]windowRegion on which to execute the kernel. (Must be a valid region of the window returned by window()).
[in,out]queueCommand queue on which to enqueue the kernel.

Reimplemented from ICLKernel.

Definition at line 334 of file ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp.

References arm_compute::ACL_DST, arm_compute::ACL_SRC_0, arm_compute::ACL_SRC_1, arm_compute::ACL_SRC_2, ICLKernel::add_2D_tensor_argument(), ICLKernel::add_2D_tensor_argument_if(), ARM_COMPUTE_ERROR_ON, ARM_COMPUTE_ERROR_ON_INVALID_SUBWINDOW, ARM_COMPUTE_ERROR_ON_NULLPTR, ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL, arm_compute::create_image2d_from_buffer(), Window::DimX, Window::DimY, arm_compute::enqueue(), Window::first_slice_window_3D(), CLKernelLibrary::get(), ITensorPack::get_const_tensor(), arm_compute::experimental::get_post_op_arg_type(), ITensorPack::get_tensor(), ICLKernel::lws_hint(), Window::set(), arm_compute::test::validation::reference::slice(), Window::slide_window_slice_3D(), and IKernel::window().

335 {
338 
339  const auto src0 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_0));
340  const auto src1 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_1));
341  const auto src2 = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(TensorType::ACL_SRC_2));
342  auto dst = utils::cast::polymorphic_downcast<ICLTensor *>(tensors.get_tensor(TensorType::ACL_DST));
343 
344  ARM_COMPUTE_ERROR_ON_NULLPTR(src0, src1, dst);
345  ARM_COMPUTE_ERROR_ON(_add_bias && src2 == nullptr);
346 
347  if(src1->info()->num_dimensions() < 3)
348  {
349  // The stride_z for matrix B must be zero if we do not slice
350  ARM_COMPUTE_ERROR_ON(src1->info()->strides_in_bytes()[3] != 0);
351  }
352 
353  const size_t lhs_idx_batch_size = _reinterpret_input_as_3d && !_has_pad_y ? 3u : 2u;
354  const size_t rhs_idx_batch_size = 2u;
355  const size_t bia_idx_batch_size = 2u;
356  const size_t out_idx_batch_size = _reinterpret_output_as_3d && !_has_pad_y ? 3u : 2u;
357 
359  Window slice_matrix_b = slice;
360 
361  slice_matrix_b.set(Window::DimX, Window::Dimension(0, 1, 1));
362  slice_matrix_b.set(Window::DimY, Window::Dimension(0, 1, 1));
363 
364  // Get cross plane pads
365  const unsigned int total_cross_plane_pad_lhs = src0->info()->padding().top + src0->info()->padding().bottom;
366  const unsigned int total_cross_plane_pad_out = dst->info()->padding().top + dst->info()->padding().bottom;
367 
368  // The execution should fail if we try to run with has_pad_y = false but we have padding in either the LHS or DST tensor
369  ARM_COMPUTE_ERROR_ON(!_has_pad_y && ((total_cross_plane_pad_lhs != 0) || (total_cross_plane_pad_out != 0)));
370 
371  cl::Image2D src1_image2d;
372 
373  if(_export_to_cl_image)
374  {
375  const TensorShape shape2d(src1->info()->dimension(0) / 4, src1->info()->dimension(1) * src1->info()->dimension(2));
376  const size_t image_row_pitch = src1->info()->strides_in_bytes()[1];
377 
378  src1_image2d = create_image2d_from_buffer(CLKernelLibrary::get().context(), src1->cl_buffer(), shape2d, src1->info()->data_type(), image_row_pitch);
379  }
380 
381  do
382  {
383  Window slice_b = slice;
384  // Don't slice matrix B along the z dimension if matrix B has just 2 dimensions and matrix A more than 2
385  // This scenario can happen when the matrix multiplication is used to perform a convolution operation
386  if(!_slide_matrix_b)
387  {
388  slice_b = slice_matrix_b;
389  }
390 
391  unsigned int idx = 0;
392 
393  // LHS buffer
394  add_2D_tensor_argument(idx, src0, slice);
395 
396  // RHS buffer or RHS OpenCL image (_export_to_cl_image == true)
397  if(_export_to_cl_image)
398  {
399  _kernel.setArg(idx++, src1_image2d);
400  }
401  else
402  {
403  add_2D_tensor_argument(idx, src1, slice_b);
404  }
405 
406  // Bias buffer (_add_bias == true)
407  add_2D_tensor_argument_if(_add_bias, idx, src2, slice);
408 
409  // dst buffer
410  add_2D_tensor_argument(idx, dst, slice);
411 
412  // post op argument buffers
413  for(size_t i = 0; i < _num_post_op_args; ++i)
414  {
415  const auto post_op_arg = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(experimental::get_post_op_arg_type(i)));
416  add_2D_tensor_argument(idx, post_op_arg, slice);
417  }
418 
419  // LHS stride_z
420  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(src0->info()->strides_in_bytes()[lhs_idx_batch_size]));
421 
422  // RHS stride_z (not used if _export_to_cl_image == true)
423  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(src1->info()->strides_in_bytes()[rhs_idx_batch_size]));
424 
425  // Bias stride_z (if _add_bias == true)
426  if(_add_bias)
427  {
428  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(src2->info()->strides_in_bytes()[bia_idx_batch_size]));
429  }
430 
431  // dst stride_z
432  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(dst->info()->strides_in_bytes()[out_idx_batch_size]));
433  // post op argument stride_z
434  for(size_t i = 0; i < _num_post_op_args; ++i)
435  {
436  const auto post_op_arg = utils::cast::polymorphic_downcast<const ICLTensor *>(tensors.get_const_tensor(experimental::get_post_op_arg_type(i)));
437  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(post_op_arg->info()->strides_in_bytes()[2]));
438  }
439 
440  // Cross-plan padding (if _reinterpret_input_as_3d = true)
441  if(_reinterpret_input_as_3d && _has_pad_y)
442  {
443  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(total_cross_plane_pad_lhs));
444  }
445 
446  // Cross-plan padding (if reinterpret_output_as_3d = true)
447  if(_reinterpret_output_as_3d && _has_pad_y)
448  {
449  _kernel.setArg<cl_uint>(idx++, static_cast<unsigned int>(total_cross_plane_pad_out));
450  }
451 
452  // Pass m, n and k at runtime as signed ints, to ensure results of any subractions they could be operand in, would still be signed.
453  _kernel.setArg<cl_int>(idx++, _m);
454  _kernel.setArg<cl_int>(idx++, _n);
455  _kernel.setArg<cl_int>(idx++, _k);
456 
457  enqueue(queue, *this, slice, lws_hint(), _use_dummy_work_items);
458  }
459  while(window.slide_window_slice_3D(slice));
460 }
const Window & window() const
The maximum window the kernel can be executed on.
Definition: IKernel.cpp:28
void add_2D_tensor_argument_if(bool cond, unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 2D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx ...
Definition: ICLKernel.h:213
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:383
#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
static CLKernelLibrary & get()
Access the KernelLibrary singleton.
TensorType get_post_op_arg_type(size_t index)
Get post op argument TensorType from post op argument index in a flattened, ordered post op argument ...
Definition: PostOpUtils.h:79
static constexpr size_t DimX
Alias for dimension 0 also known as X dimension.
Definition: Window.h:43
bool slide_window_slice_3D(Window &slice) const
Slide the passed 3D window slice.
Definition: Window.h:349
#define ARM_COMPUTE_ERROR_ON_UNCONFIGURED_KERNEL(k)
Definition: Validate.h:915
static constexpr size_t DimY
Alias for dimension 1 also known as Y dimension.
Definition: Window.h:45
void add_2D_tensor_argument(unsigned int &idx, const ICLTensor *tensor, const Window &window)
Add the passed 2D tensor&#39;s parameters to the object&#39;s kernel&#39;s arguments starting from the index idx...
Definition: ICLKernel.h:202
#define ARM_COMPUTE_ERROR_ON_NULLPTR(...)
Definition: Validate.h:157
cl::Image2D create_image2d_from_buffer(const cl::Context &ctx, const cl::Buffer &buffer, const TensorShape &shape2d, DataType data_type, size_t image_row_pitch)
Create a cl::Image2D object from an OpenCL buffer.
Definition: CLUtils.cpp:35
Window first_slice_window_3D() const
First 3D slice of the window.
Definition: Window.h:305
#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 src0,
const ITensorInfo src1,
const ITensorInfo src2,
const ITensorInfo dst,
float  alpha,
float  beta,
const GEMMLHSMatrixInfo lhs_info,
const GEMMRHSMatrixInfo rhs_info,
const GEMMKernelInfo gemm_info 
)
static

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

Similar to ClGemmMatrixMultiplyReshapedOnlyRhsKernel::configure()

Returns
a status

Definition at line 326 of file ClGemmMatrixMultiplyReshapedOnlyRhsKernel.cpp.

References ARM_COMPUTE_RETURN_ON_ERROR, and arm_compute::cpu::kernels::validate_arguments().

329 {
330  ARM_COMPUTE_RETURN_ON_ERROR(validate_arguments(src0, src1, src2, dst, alpha, beta, lhs_info, rhs_info, gemm_info));
331  return Status{};
332 }
#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 *src, const ITensorInfo *weights, const ITensorInfo *dst, const PadStrideInfo &conv_info)

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