Visible to Intel only — GUID: GUID-50A92F3D-B202-4D0E-BC29-42701044CF2E
Visible to Intel only — GUID: GUID-50A92F3D-B202-4D0E-BC29-42701044CF2E
namespace dnnl::graph::sycl_interop
Overview
SYCL interoperability namespace. More…
namespace sycl_interop { // global functions allocator make_allocator( dnnl_graph_sycl_allocate_f sycl_malloc, dnnl_graph_sycl_deallocate_f sycl_free ); engine make_engine_with_allocator( const sycl::device& adevice, const sycl::context& acontext, const allocator& alloc ); sycl::event execute( compiled_partition& c_partition, stream& astream, const std::vector<tensor>& inputs, std::vector<tensor>& outputs, const std::vector<sycl::event>& deps = {} ); } // namespace sycl_interop
Detailed Documentation
SYCL interoperability namespace.
Global Functions
allocator make_allocator( dnnl_graph_sycl_allocate_f sycl_malloc, dnnl_graph_sycl_deallocate_f sycl_free )
Constructs an allocator from SYCL malloc and free function pointer.
SYCL allocator should be used for SYCL runtime and host allocator should be used for non-SYCL. Currently, only device USM allocator is supported.
Parameters:
sycl_malloc |
The pointer to SYCL malloc function |
sycl_free |
The pointer to SYCL free function |
Returns:
Created allocator
sycl::event execute( compiled_partition& c_partition, stream& astream, const std::vector<tensor>& inputs, std::vector<tensor>& outputs, const std::vector<sycl::event>& deps = {} )
Executes a compiled partition in a specified stream and returns a SYCL event.
Parameters:
c_partition |
Compiled partition to execute. |
astream |
Stream object to run over |
inputs |
Arguments map. |
outputs |
Arguments map. |
deps |
Optional vector with sycl::event dependencies. |
Returns:
Output event.