gko::DpcppExecutor

SYCL / DPC++ executor. Runs on Intel GPUs and CPUs (and on any SYCL target a Ginkgo build picks up). The constructor takes a SYCL device type and a host master executor; the runtime selects a concrete device matching the type. Kernels dispatch through SYCL queues; allocations come from SYCL’s USM (Unified Shared Memory) allocators.

class DpcppExecutor

Inherits from

• public gko::detail::ExecutorBase<DpcppExecutor>

• public std::enable_shared_from_this<DpcppExecutor>

This is the Executor subclass which represents a DPC++ enhanced device.

Public Functions

virtual std::shared_ptr<Executor> get_master() noexcept override

Returns the master OmpExecutor of this Executor.

Returns:

the master OmpExecutor of this Executor.

virtual std::shared_ptr<const Executor> get_master(

) const noexcept override

Returns the master OmpExecutor of this Executor.

Returns:

the master OmpExecutor of this Executor.

virtual void synchronize() const override

Synchronize the operations launched on the executor with its master.

virtual std::string get_description() const override

Returns:

a textual representation of the executor and its device.

inline int get_device_id() const noexcept

Get the DPCPP device id of the device associated to this executor.

Returns:

the DPCPP device id of the device associated to this executor

inline const std::vector<int> &get_subgroup_sizes() const noexcept

Get the available subgroup sizes for this device.

Returns:

the available subgroup sizes for this device

inline int get_num_computing_units() const noexcept

Get the number of Computing Units of this executor.

Returns:

the number of Computing Units of this executor

inline int get_num_subgroups() const noexcept

Get the number of subgroups of this executor.

inline const std::vector<int> &get_max_workitem_sizes(

) const noexcept

Get the maximum work item sizes.

Returns:

the maximum work item sizes

inline int get_max_workgroup_size() const noexcept

Get the maximum workgroup size.

Returns:

the maximum workgroup size

inline int get_max_subgroup_size() const noexcept

Get the maximum subgroup size.

Returns:

the maximum subgroup size

inline std::string get_device_type() const noexcept

Get a string representing the device type.

Returns:

a string representing the device type

virtual void run(const Operation &op) const = 0

Runs the specified Operation using this Executor.

Parameters:

op – the operation to run

template<typename ClosureOmp, typename ClosureCuda, typename ClosureHip, typename ClosureDpcpp>
inline void run(

const ClosureOmp &op_omp,

const ClosureCuda &op_cuda,

const ClosureHip &op_hip,

const ClosureDpcpp &op_dpcpp,

) const

Runs one of the passed in functors, depending on the Executor type.

Template Parameters:

• ClosureOmp – type of op_omp

• ClosureCuda – type of op_cuda

• ClosureHip – type of op_hip

• ClosureDpcpp – type of op_dpcpp

Parameters:

• op_omp – functor to run in case of a OmpExecutor or ReferenceExecutor

• op_cuda – functor to run in case of a CudaExecutor

• op_hip – functor to run in case of a HipExecutor

• op_dpcpp – functor to run in case of a DpcppExecutor

template<typename ClosureReference, typename ClosureOmp, typename ClosureCuda, typename ClosureHip, typename ClosureDpcpp>
inline void run(

std::string name,

const ClosureReference &op_ref,

const ClosureOmp &op_omp,

const ClosureCuda &op_cuda,

const ClosureHip &op_hip,

const ClosureDpcpp &op_dpcpp,

) const

Runs one of the passed in functors, depending on the Executor type.

Template Parameters:

• ClosureReference – type of op_ref

• ClosureOmp – type of op_omp

• ClosureCuda – type of op_cuda

• ClosureHip – type of op_hip

• ClosureDpcpp – type of op_dpcpp

Parameters:

• name – the name of the operation

• op_ref – functor to run in case of a ReferenceExecutor

• op_omp – functor to run in case of a OmpExecutor

• op_cuda – functor to run in case of a CudaExecutor

• op_hip – functor to run in case of a HipExecutor

• op_dpcpp – functor to run in case of a DpcppExecutor

Public Static Functions

static std::shared_ptr<DpcppExecutor> create(

int device_id,

std::shared_ptr<Executor> master,

std::string device_type = "all",

dpcpp_queue_property property = dpcpp_queue_property::in_order,

)

Creates a new DpcppExecutor.

Parameters:

• device_id – the DPCPP device id of this device

• master – an executor on the host that is used to invoke the device kernels

• device_type – a string representing the type of device to consider (accelerator, cpu, gpu or all).

static int get_num_devices(std::string device_type)

Get the number of devices present on the system.

Parameters:

device_type – a string representing the device type

Returns:

the number of devices present on the system