#include <sycl/detail/export.hpp>
#include <cstddef>
#include <cstdint>
#include <variant>
#include <sycl/detail/pi_error.def>
#include <sycl/detail/pi.def>
Go to the source code of this file.
Classes | |
struct | _pi_offload_entry_struct |
struct | _pi_device_binary_property_struct |
struct | _pi_device_binary_property_set_struct |
struct | pi_device_binary_struct |
This struct is a record of the device binary information. More... | |
struct | pi_buffer_region_struct |
struct | pi_buff_rect_offset_struct |
struct | pi_buff_rect_region_struct |
struct | pi_image_offset_struct |
struct | pi_image_region_struct |
struct | pi_device_binaries_struct |
This struct is a record of all the device code that may be offloaded. More... | |
struct | _pi_image_format |
struct | _pi_image_desc |
struct | _pi_mem_obj_property |
struct | pi_ext_command_buffer_desc |
struct | _pi_plugin |
struct | _pi_plugin::FunctionPointers |
Macros | |
#define | _PI_H_VERSION_MAJOR 14 |
#define | _PI_H_VERSION_MINOR 39 |
#define | _PI_STRING_HELPER(a) #a |
#define | _PI_CONCAT(a, b) _PI_STRING_HELPER(a.b) |
#define | _PI_TRIPLE_CONCAT(a, b, c) _PI_STRING_HELPER(a.b.c) |
#define | _PI_PLUGIN_VERSION_STRING(PluginVersion) _PI_TRIPLE_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR, PluginVersion) |
#define | _PI_H_VERSION_STRING _PI_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR) |
#define | _PI_STRING_SUBSTITUTE(X) _PI_STRING_HELPER(X) |
#define | _PI_PLUGIN_VERSION_CHECK(PI_API_VERSION, PI_PLUGIN_VERSION) |
#define | _PI_ERRC(NAME, VAL) NAME = VAL, |
#define | _PI_ERRC_WITH_MSG(NAME, VAL, MSG) NAME = VAL, |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_UNKNOWN "<unknown>" |
Target identification strings for pi_device_binary_struct.DeviceTargetSpec. More... | |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV32 "spir" |
SPIR-V 32-bit image <-> "spir", 32-bit OpenCL device. More... | |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64 "spir64" |
SPIR-V 64-bit image <-> "spir64", 64-bit OpenCL device. More... | |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_X86_64 "spir64_x86_64" |
Device-specific binary images produced from SPIR-V 64-bit <-> various "spir64_*" triples for specific 64-bit OpenCL devices. More... | |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_GEN "spir64_gen" |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_FPGA "spir64_fpga" |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_NVPTX64 "nvptx64" |
PTX 64-bit image <-> "nvptx64", 64-bit NVIDIA PTX device. More... | |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_AMDGCN "amdgcn" |
#define | __SYCL_PI_DEVICE_BINARY_TARGET_NATIVE_CPU "native_cpu" |
#define | PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT "pi_ext_intel_devicelib_assert" |
Extension to denote native support of assert feature by an arbitrary device piDeviceGetInfo call should return this extension when the device supports native asserts if supported extensions' names are requested. More... | |
#define | __SYCL_PI_PROPERTY_SET_SPEC_CONST_MAP "SYCL/specialization constants" |
Device binary image property set names recognized by the SYCL runtime. More... | |
#define | __SYCL_PI_PROPERTY_SET_SPEC_CONST_DEFAULT_VALUES_MAP "SYCL/specialization constants default values" |
PropertySetRegistry::SYCL_SPEC_CONSTANTS_DEFAULT_VALUES defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_DEVICELIB_REQ_MASK "SYCL/devicelib req mask" |
PropertySetRegistry::SYCL_DEVICELIB_REQ_MASK defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_KERNEL_PARAM_OPT_INFO "SYCL/kernel param opt" |
PropertySetRegistry::SYCL_KERNEL_PARAM_OPT_INFO defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_PROGRAM_METADATA "SYCL/program metadata" |
PropertySetRegistry::SYCL_KERNEL_PROGRAM_METADATA defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_MISC_PROP "SYCL/misc properties" |
PropertySetRegistry::SYCL_MISC_PROP defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_ASSERT_USED "SYCL/assert used" |
PropertySetRegistry::SYCL_ASSERT_USED defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_EXPORTED_SYMBOLS "SYCL/exported symbols" |
PropertySetRegistry::SYCL_EXPORTED_SYMBOLS defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_GLOBALS "SYCL/device globals" |
PropertySetRegistry::SYCL_DEVICE_GLOBALS defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_REQUIREMENTS "SYCL/device requirements" |
PropertySetRegistry::SYCL_DEVICE_REQUIREMENTS defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROPERTY_SET_SYCL_HOST_PIPES "SYCL/host pipes" |
PropertySetRegistry::SYCL_HOST_PIPES defined in PropertySetIO.h. More... | |
#define | __SYCL_PI_PROGRAM_METADATA_TAG_REQD_WORK_GROUP_SIZE "@reqd_work_group_size" |
Program metadata tags recognized by the PI backends. More... | |
#define | __SYCL_PI_PROGRAM_METADATA_GLOBAL_ID_MAPPING "@global_id_mapping" |
#define | PI_BIT(_i) (1 << _i) |
#define | _PI_API(api) decltype(::api) *api; |
Functions | |
pi_result | piPluginInit (pi_plugin *plugin_info) |
pi_result | piPlatformsGet (pi_uint32 num_entries, pi_platform *platforms, pi_uint32 *num_platforms) |
pi_result | piPlatformGetInfo (pi_platform platform, pi_platform_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piextPlatformGetNativeHandle (pi_platform platform, pi_native_handle *nativeHandle) |
Gets the native handle of a PI platform object. More... | |
pi_result | piextPlatformCreateWithNativeHandle (pi_native_handle nativeHandle, pi_platform *platform) |
Creates PI platform object from a native handle. More... | |
pi_result | piDevicesGet (pi_platform platform, pi_device_type device_type, pi_uint32 num_entries, pi_device *devices, pi_uint32 *num_devices) |
pi_result | piextEnablePeerAccess (pi_device command_device, pi_device peer_device) |
pi_result | piextDisablePeerAccess (pi_device command_device, pi_device peer_device) |
pi_result | piextPeerAccessGetInfo (pi_device command_device, pi_device peer_device, pi_peer_attr attr, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piDeviceGetInfo (pi_device device, pi_device_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
Returns requested info for provided native device Return PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT for PI_DEVICE_INFO_EXTENSIONS query when the device supports native asserts. More... | |
pi_result | piDeviceRetain (pi_device device) |
pi_result | piDeviceRelease (pi_device device) |
pi_result | piDevicePartition (pi_device device, const pi_device_partition_property *properties, pi_uint32 num_devices, pi_device *out_devices, pi_uint32 *out_num_devices) |
pi_result | piextDeviceGetNativeHandle (pi_device device, pi_native_handle *nativeHandle) |
Gets the native handle of a PI device object. More... | |
pi_result | piextDeviceCreateWithNativeHandle (pi_native_handle nativeHandle, pi_platform platform, pi_device *device) |
Creates PI device object from a native handle. More... | |
pi_result | piextDeviceSelectBinary (pi_device device, pi_device_binary *binaries, pi_uint32 num_binaries, pi_uint32 *selected_binary_ind) |
Selects the most appropriate device binary based on runtime information and the IR characteristics. More... | |
pi_result | piextGetDeviceFunctionPointer (pi_device device, pi_program program, const char *function_name, pi_uint64 *function_pointer_ret) |
Retrieves a device function pointer to a user-defined function. More... | |
pi_result | piContextCreate (const pi_context_properties *properties, pi_uint32 num_devices, const pi_device *devices, void(*pfn_notify)(const char *errinfo, const void *private_info, size_t cb, void *user_data), void *user_data, pi_context *ret_context) |
pi_result | piContextGetInfo (pi_context context, pi_context_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piContextRetain (pi_context context) |
pi_result | piContextRelease (pi_context context) |
pi_result | piextContextSetExtendedDeleter (pi_context context, pi_context_extended_deleter func, void *user_data) |
pi_result | piextContextGetNativeHandle (pi_context context, pi_native_handle *nativeHandle) |
Gets the native handle of a PI context object. More... | |
pi_result | piextContextCreateWithNativeHandle (pi_native_handle nativeHandle, pi_uint32 numDevices, const pi_device *devices, bool pluginOwnsNativeHandle, pi_context *context) |
Creates PI context object from a native handle. More... | |
pi_result | piQueueCreate (pi_context context, pi_device device, pi_queue_properties properties, pi_queue *queue) |
pi_result | piextQueueCreate (pi_context context, pi_device device, pi_queue_properties *properties, pi_queue *queue) |
pi_result | piQueueGetInfo (pi_queue command_queue, pi_queue_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piQueueRetain (pi_queue command_queue) |
pi_result | piQueueRelease (pi_queue command_queue) |
pi_result | piQueueFinish (pi_queue command_queue) |
pi_result | piQueueFlush (pi_queue command_queue) |
pi_result | piextQueueGetNativeHandle (pi_queue queue, pi_native_handle *nativeHandle, int32_t *nativeHandleDesc) |
Gets the native handle of a PI queue object. More... | |
pi_result | piextQueueCreateWithNativeHandle (pi_native_handle nativeHandle, int32_t nativeHandleDesc, pi_context context, pi_device device, bool pluginOwnsNativeHandle, pi_queue_properties *Properties, pi_queue *queue) |
Creates PI queue object from a native handle. More... | |
pi_result | piMemBufferCreate (pi_context context, pi_mem_flags flags, size_t size, void *host_ptr, pi_mem *ret_mem, const pi_mem_properties *properties=nullptr) |
pi_result | piMemImageCreate (pi_context context, pi_mem_flags flags, const pi_image_format *image_format, const pi_image_desc *image_desc, void *host_ptr, pi_mem *ret_mem) |
pi_result | piMemGetInfo (pi_mem mem, pi_mem_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piMemImageGetInfo (pi_mem image, pi_image_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piMemRetain (pi_mem mem) |
pi_result | piMemRelease (pi_mem mem) |
pi_result | piMemBufferPartition (pi_mem buffer, pi_mem_flags flags, pi_buffer_create_type buffer_create_type, void *buffer_create_info, pi_mem *ret_mem) |
pi_result | piextMemGetNativeHandle (pi_mem mem, pi_native_handle *nativeHandle) |
Gets the native handle of a PI mem object. More... | |
pi_result | piextMemCreateWithNativeHandle (pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, pi_mem *mem) |
Creates PI mem object from a native handle. More... | |
pi_result | piextMemImageCreateWithNativeHandle (pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, const pi_image_format *ImageFormat, const pi_image_desc *ImageDesc, pi_mem *img) |
Creates PI image object from a native handle. More... | |
pi_result | piProgramCreate (pi_context context, const void *il, size_t length, pi_program *res_program) |
pi_result | piProgramCreateWithBinary (pi_context context, pi_uint32 num_devices, const pi_device *device_list, const size_t *lengths, const unsigned char **binaries, size_t num_metadata_entries, const pi_device_binary_property *metadata, pi_int32 *binary_status, pi_program *ret_program) |
Creates a PI program for a context and loads the given binary into it. More... | |
pi_result | piProgramGetInfo (pi_program program, pi_program_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piProgramLink (pi_context context, pi_uint32 num_devices, const pi_device *device_list, const char *options, pi_uint32 num_input_programs, const pi_program *input_programs, void(*pfn_notify)(pi_program program, void *user_data), void *user_data, pi_program *ret_program) |
pi_result | piProgramCompile (pi_program program, pi_uint32 num_devices, const pi_device *device_list, const char *options, pi_uint32 num_input_headers, const pi_program *input_headers, const char **header_include_names, void(*pfn_notify)(pi_program program, void *user_data), void *user_data) |
pi_result | piProgramBuild (pi_program program, pi_uint32 num_devices, const pi_device *device_list, const char *options, void(*pfn_notify)(pi_program program, void *user_data), void *user_data) |
pi_result | piProgramGetBuildInfo (pi_program program, pi_device device, _pi_program_build_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piProgramRetain (pi_program program) |
pi_result | piProgramRelease (pi_program program) |
pi_result | piextProgramSetSpecializationConstant (pi_program prog, pi_uint32 spec_id, size_t spec_size, const void *spec_value) |
Sets a specialization constant to a specific value. More... | |
pi_result | piextProgramGetNativeHandle (pi_program program, pi_native_handle *nativeHandle) |
Gets the native handle of a PI program object. More... | |
pi_result | piextProgramCreateWithNativeHandle (pi_native_handle nativeHandle, pi_context context, bool pluginOwnsNativeHandle, pi_program *program) |
Creates PI program object from a native handle. More... | |
pi_result | piKernelCreate (pi_program program, const char *kernel_name, pi_kernel *ret_kernel) |
pi_result | piKernelSetArg (pi_kernel kernel, pi_uint32 arg_index, size_t arg_size, const void *arg_value) |
pi_result | piKernelGetInfo (pi_kernel kernel, pi_kernel_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piKernelGetGroupInfo (pi_kernel kernel, pi_device device, pi_kernel_group_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piKernelGetSubGroupInfo (pi_kernel kernel, pi_device device, pi_kernel_sub_group_info param_name, size_t input_value_size, const void *input_value, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
API to query information from the sub-group from a kernel. More... | |
pi_result | piKernelRetain (pi_kernel kernel) |
pi_result | piKernelRelease (pi_kernel kernel) |
pi_result | piextKernelSetArgPointer (pi_kernel kernel, pi_uint32 arg_index, size_t arg_size, const void *arg_value) |
Sets up pointer arguments for CL kernels. More... | |
pi_result | piKernelSetExecInfo (pi_kernel kernel, pi_kernel_exec_info value_name, size_t param_value_size, const void *param_value) |
API to set attributes controlling kernel execution. More... | |
pi_result | piextKernelCreateWithNativeHandle (pi_native_handle nativeHandle, pi_context context, pi_program program, bool pluginOwnsNativeHandle, pi_kernel *kernel) |
Creates PI kernel object from a native handle. More... | |
pi_result | piextKernelGetNativeHandle (pi_kernel kernel, pi_native_handle *nativeHandle) |
Gets the native handle of a PI kernel object. More... | |
pi_result | piEventCreate (pi_context context, pi_event *ret_event) |
Create PI event object in a signalled/completed state. More... | |
pi_result | piEventGetInfo (pi_event event, pi_event_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piEventGetProfilingInfo (pi_event event, pi_profiling_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piEventsWait (pi_uint32 num_events, const pi_event *event_list) |
pi_result | piEventSetCallback (pi_event event, pi_int32 command_exec_callback_type, void(*pfn_notify)(pi_event event, pi_int32 event_command_status, void *user_data), void *user_data) |
pi_result | piEventSetStatus (pi_event event, pi_int32 execution_status) |
pi_result | piEventRetain (pi_event event) |
pi_result | piEventRelease (pi_event event) |
pi_result | piextEventGetNativeHandle (pi_event event, pi_native_handle *nativeHandle) |
Gets the native handle of a PI event object. More... | |
pi_result | piextEventCreateWithNativeHandle (pi_native_handle nativeHandle, pi_context context, bool ownNativeHandle, pi_event *event) |
Creates PI event object from a native handle. More... | |
pi_result | piSamplerCreate (pi_context context, const pi_sampler_properties *sampler_properties, pi_sampler *result_sampler) |
pi_result | piSamplerGetInfo (pi_sampler sampler, pi_sampler_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
pi_result | piSamplerRetain (pi_sampler sampler) |
pi_result | piSamplerRelease (pi_sampler sampler) |
pi_result | piEnqueueKernelLaunch (pi_queue queue, pi_kernel kernel, pi_uint32 work_dim, const size_t *global_work_offset, const size_t *global_work_size, const size_t *local_work_size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueEventsWait (pi_queue command_queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueEventsWaitWithBarrier (pi_queue command_queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferRead (pi_queue queue, pi_mem buffer, pi_bool blocking_read, size_t offset, size_t size, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferReadRect (pi_queue command_queue, pi_mem buffer, pi_bool blocking_read, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferWrite (pi_queue command_queue, pi_mem buffer, pi_bool blocking_write, size_t offset, size_t size, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferWriteRect (pi_queue command_queue, pi_mem buffer, pi_bool blocking_write, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferCopy (pi_queue command_queue, pi_mem src_buffer, pi_mem dst_buffer, size_t src_offset, size_t dst_offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferCopyRect (pi_queue command_queue, pi_mem src_buffer, pi_mem dst_buffer, pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin, pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferFill (pi_queue command_queue, pi_mem buffer, const void *pattern, size_t pattern_size, size_t offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemImageRead (pi_queue command_queue, pi_mem image, pi_bool blocking_read, pi_image_offset origin, pi_image_region region, size_t row_pitch, size_t slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemImageWrite (pi_queue command_queue, pi_mem image, pi_bool blocking_write, pi_image_offset origin, pi_image_region region, size_t input_row_pitch, size_t input_slice_pitch, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemImageCopy (pi_queue command_queue, pi_mem src_image, pi_mem dst_image, pi_image_offset src_origin, pi_image_offset dst_origin, pi_image_region region, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemImageFill (pi_queue command_queue, pi_mem image, const void *fill_color, const size_t *origin, const size_t *region, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piEnqueueMemBufferMap (pi_queue command_queue, pi_mem buffer, pi_bool blocking_map, pi_map_flags map_flags, size_t offset, size_t size, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event, void **ret_map) |
pi_result | piEnqueueMemUnmap (pi_queue command_queue, pi_mem memobj, void *mapped_ptr, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
pi_result | piextKernelSetArgMemObj (pi_kernel kernel, pi_uint32 arg_index, const pi_mem_obj_property *arg_properties, const pi_mem *arg_value) |
pi_result | piextKernelSetArgSampler (pi_kernel kernel, pi_uint32 arg_index, const pi_sampler *arg_value) |
pi_result | piextUSMHostAlloc (void **result_ptr, pi_context context, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment) |
Allocates host memory accessible by the device. More... | |
pi_result | piextUSMDeviceAlloc (void **result_ptr, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment) |
Allocates device memory. More... | |
pi_result | piextUSMSharedAlloc (void **result_ptr, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t size, pi_uint32 alignment) |
Allocates memory accessible on both host and device. More... | |
pi_result | piextUSMPitchedAlloc (void **result_ptr, size_t *result_pitch, pi_context context, pi_device device, pi_usm_mem_properties *properties, size_t width_in_bytes, size_t height, unsigned int element_size_bytes) |
Allocates memory accessible on device. More... | |
pi_result | piextUSMFree (pi_context context, void *ptr) |
Indicates that the allocated USM memory is no longer needed on the runtime side. More... | |
pi_result | piextUSMEnqueueMemset (pi_queue queue, void *ptr, pi_int32 value, size_t count, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
USM Memset API. More... | |
pi_result | piextUSMEnqueueMemcpy (pi_queue queue, pi_bool blocking, void *dst_ptr, const void *src_ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
USM Memcpy API. More... | |
pi_result | piextUSMEnqueuePrefetch (pi_queue queue, const void *ptr, size_t size, pi_usm_migration_flags flags, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
Hint to migrate memory to the device. More... | |
pi_result | piextUSMEnqueueMemAdvise (pi_queue queue, const void *ptr, size_t length, pi_mem_advice advice, pi_event *event) |
USM Memadvise API. More... | |
pi_result | piextUSMGetMemAllocInfo (pi_context context, const void *ptr, pi_mem_alloc_info param_name, size_t param_value_size, void *param_value, size_t *param_value_size_ret) |
API to query information about USM allocated pointers Valid Queries: PI_MEM_ALLOC_TYPE returns host/device/shared pi_host_usm value PI_MEM_ALLOC_BASE_PTR returns the base ptr of an allocation if the queried pointer fell inside an allocation. More... | |
pi_result | piextUSMEnqueueFill2D (pi_queue queue, void *ptr, size_t pitch, size_t pattern_size, const void *pattern, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
USM 2D fill API. More... | |
pi_result | piextUSMEnqueueMemset2D (pi_queue queue, void *ptr, size_t pitch, int value, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
USM 2D Memset API. More... | |
pi_result | piextUSMEnqueueMemcpy2D (pi_queue queue, pi_bool blocking, void *dst_ptr, size_t dst_pitch, const void *src_ptr, size_t src_pitch, size_t width, size_t height, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
USM 2D Memcpy API. More... | |
pi_result | piextUSMImport (const void *ptr, size_t size, pi_context context) |
Import host system memory into USM. More... | |
pi_result | piextUSMRelease (const void *ptr, pi_context context) |
Release host system memory from USM. More... | |
pi_result | piextEnqueueDeviceGlobalVariableWrite (pi_queue queue, pi_program program, const char *name, pi_bool blocking_write, size_t count, size_t offset, const void *src, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
Device global variable. More... | |
pi_result | piextEnqueueDeviceGlobalVariableRead (pi_queue queue, pi_program program, const char *name, pi_bool blocking_read, size_t count, size_t offset, void *dst, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
API reading data from a device global variable to host. More... | |
pi_result | piextEnqueueReadHostPipe (pi_queue queue, pi_program program, const char *pipe_symbol, pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
Plugin. More... | |
pi_result | piextEnqueueWriteHostPipe (pi_queue queue, pi_program program, const char *pipe_symbol, pi_bool blocking, void *ptr, size_t size, pi_uint32 num_events_in_waitlist, const pi_event *events_waitlist, pi_event *event) |
Write to pipe of a given name. More... | |
pi_result | piextPluginGetOpaqueData (void *opaque_data_param, void **opaque_data_return) |
API to get Plugin internal data, opaque to SYCL RT. More... | |
pi_result | piTearDown (void *PluginParameter) |
API to notify that the plugin should clean up its resources. More... | |
pi_result | piPluginGetLastError (char **message) |
API to get Plugin specific warning and error messages. More... | |
pi_result | piPluginGetBackendOption (pi_platform platform, const char *frontend_option, const char **backend_option) |
API to get backend specific option. More... | |
pi_result | piGetDeviceAndHostTimer (pi_device Device, uint64_t *DeviceTime, uint64_t *HostTime) |
Queries device for it's global timestamp in nanoseconds, and updates HostTime with the value of the host timer at the closest possible point in time to that at which DeviceTime was returned. More... | |
pi_result | piextCommandBufferCreate (pi_context context, pi_device device, const pi_ext_command_buffer_desc *desc, pi_ext_command_buffer *ret_command_buffer) |
API to create a command-buffer. More... | |
pi_result | piextCommandBufferRetain (pi_ext_command_buffer command_buffer) |
API to increment the reference count of the command-buffer. More... | |
pi_result | piextCommandBufferRelease (pi_ext_command_buffer command_buffer) |
API to decrement the reference count of the command-buffer. More... | |
pi_result | piextCommandBufferFinalize (pi_ext_command_buffer command_buffer) |
API to stop command-buffer recording such that no more commands can be appended, and makes the command-buffer ready to enqueue on a command-queue. More... | |
pi_result | piextCommandBufferNDRangeKernel (pi_ext_command_buffer command_buffer, pi_kernel kernel, pi_uint32 work_dim, const size_t *global_work_offset, const size_t *global_work_size, const size_t *local_work_size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a kernel execution command to the command-buffer. More... | |
pi_result | piextCommandBufferMemcpyUSM (pi_ext_command_buffer command_buffer, void *dst_ptr, const void *src_ptr, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a USM memcpy command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferCopy (pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer, size_t src_offset, size_t dst_offset, size_t size, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a mem buffer copy command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferCopyRect (pi_ext_command_buffer command_buffer, pi_mem src_buffer, pi_mem dst_buffer, pi_buff_rect_offset src_origin, pi_buff_rect_offset dst_origin, pi_buff_rect_region region, size_t src_row_pitch, size_t src_slice_pitch, size_t dst_row_pitch, size_t dst_slice_pitch, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a rectangular mem buffer copy command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferRead (pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset, size_t size, void *dst, pi_uint32 num_sync_points_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a mem buffer read command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferReadRect (pi_ext_command_buffer command_buffer, pi_mem buffer, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, void *ptr, pi_uint32 num_events_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a rectangular mem buffer read command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferWrite (pi_ext_command_buffer command_buffer, pi_mem buffer, size_t offset, size_t size, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a mem buffer write command to the command-buffer. More... | |
pi_result | piextCommandBufferMemBufferWriteRect (pi_ext_command_buffer command_buffer, pi_mem buffer, pi_buff_rect_offset buffer_offset, pi_buff_rect_offset host_offset, pi_buff_rect_region region, size_t buffer_row_pitch, size_t buffer_slice_pitch, size_t host_row_pitch, size_t host_slice_pitch, const void *ptr, pi_uint32 num_events_in_wait_list, const pi_ext_sync_point *sync_point_wait_list, pi_ext_sync_point *sync_point) |
API to append a rectangular mem buffer write command to the command-buffer. More... | |
pi_result | piextEnqueueCommandBuffer (pi_ext_command_buffer command_buffer, pi_queue queue, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
API to submit the command-buffer to queue for execution, returns an error if the command-buffer is not finalized or another instance of the same command-buffer is currently executing. More... | |
pi_result | piextMemUnsampledImageHandleDestroy (pi_context context, pi_device device, pi_image_handle handle) |
API to destroy bindless unsampled image handles. More... | |
pi_result | piextMemSampledImageHandleDestroy (pi_context context, pi_device device, pi_image_handle handle) |
API to destroy bindless sampled image handles. More... | |
pi_result | piextMemImageAllocate (pi_context context, pi_device device, pi_image_format *image_format, pi_image_desc *image_desc, pi_image_mem_handle *ret_mem) |
API to allocate memory for bindless images. More... | |
pi_result | piextMemMipmapGetLevel (pi_context context, pi_device device, pi_image_mem_handle mip_mem, unsigned int level, pi_image_mem_handle *ret_mem) |
API to retrieve individual image from mipmap. More... | |
pi_result | piextMemImageFree (pi_context context, pi_device device, pi_image_mem_handle memory_handle) |
API to free memory for bindless images. More... | |
pi_result | piextMemMipmapFree (pi_context context, pi_device device, pi_image_mem_handle memory_handle) |
API to free mipmap memory for bindless images. More... | |
pi_result | piextMemUnsampledImageCreate (pi_context context, pi_device device, pi_image_mem_handle img_mem, pi_image_format *image_format, pi_image_desc *image_desc, pi_mem *ret_mem, pi_image_handle *ret_handle) |
API to create bindless image handles. More... | |
pi_result | piextMemSampledImageCreate (pi_context context, pi_device device, pi_image_mem_handle img_mem, pi_image_format *image_format, pi_image_desc *image_desc, pi_sampler sampler, pi_mem *ret_mem, pi_image_handle *ret_handle) |
API to create sampled bindless image handles. More... | |
pi_result | piextBindlessImageSamplerCreate (pi_context context, const pi_sampler_properties *sampler_properties, float min_mipmap_level_clamp, float max_mipmap_level_clamp, float max_anisotropy, pi_sampler *result_sampler) |
API to create samplers for bindless images. More... | |
pi_result | piextMemImageCopy (pi_queue command_queue, void *dst_ptr, void *src_ptr, const pi_image_format *image_format, const pi_image_desc *image_desc, const pi_image_copy_flags flags, pi_image_offset src_offset, pi_image_offset dst_offset, pi_image_region copy_extent, pi_image_region host_extent, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
API to copy image data Host to Device or Device to Host. More... | |
pi_result | piextMemImageGetInfo (const pi_image_mem_handle mem_handle, pi_image_info param_name, void *param_value, size_t *param_value_size_ret) |
API to query an image memory handle for specific properties. More... | |
pi_result | piextMemImportOpaqueFD (pi_context context, pi_device device, size_t size, int file_descriptor, pi_interop_mem_handle *ret_handle) |
API to import external memory in the form of a file descriptor. More... | |
pi_result | piextMemMapExternalArray (pi_context context, pi_device device, pi_image_format *image_format, pi_image_desc *image_desc, pi_interop_mem_handle mem_handle, pi_image_mem_handle *ret_mem) |
API to map an interop memory handle to an image memory handle. More... | |
pi_result | piextMemReleaseInterop (pi_context context, pi_device device, pi_interop_mem_handle memory_handle) |
API to destroy interop memory. More... | |
pi_result | piextImportExternalSemaphoreOpaqueFD (pi_context context, pi_device device, int file_descriptor, pi_interop_semaphore_handle *ret_handle) |
API to import an external semaphore in the form of a file descriptor. More... | |
pi_result | piextDestroyExternalSemaphore (pi_context context, pi_device device, pi_interop_semaphore_handle sem_handle) |
API to destroy the external semaphore handle. More... | |
pi_result | piextWaitExternalSemaphore (pi_queue command_queue, pi_interop_semaphore_handle sem_handle, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
API to instruct the queue with a non-blocking wait on an external semaphore. More... | |
pi_result | piextSignalExternalSemaphore (pi_queue command_queue, pi_interop_semaphore_handle sem_handle, pi_uint32 num_events_in_wait_list, const pi_event *event_wait_list, pi_event *event) |
API to instruct the queue to signal the external semaphore handle once all previous commands have completed execution. More... | |
#define __SYCL_PI_DEVICE_BINARY_TARGET_NVPTX64 "nvptx64" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV32 "spir" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64 "spir64" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_FPGA "spir64_fpga" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_GEN "spir64_gen" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_X86_64 "spir64_x86_64" |
#define __SYCL_PI_DEVICE_BINARY_TARGET_UNKNOWN "<unknown>" |
Target identification strings for pi_device_binary_struct.DeviceTargetSpec.
A device type represented by a particular target triple requires specific binary images. We need to map the image type onto the device target triple
#define __SYCL_PI_PROGRAM_METADATA_GLOBAL_ID_MAPPING "@global_id_mapping" |
#define __SYCL_PI_PROGRAM_METADATA_TAG_REQD_WORK_GROUP_SIZE "@reqd_work_group_size" |
#define __SYCL_PI_PROPERTY_SET_DEVICELIB_REQ_MASK "SYCL/devicelib req mask" |
#define __SYCL_PI_PROPERTY_SET_KERNEL_PARAM_OPT_INFO "SYCL/kernel param opt" |
#define __SYCL_PI_PROPERTY_SET_PROGRAM_METADATA "SYCL/program metadata" |
#define __SYCL_PI_PROPERTY_SET_SPEC_CONST_DEFAULT_VALUES_MAP "SYCL/specialization constants default values" |
#define __SYCL_PI_PROPERTY_SET_SPEC_CONST_MAP "SYCL/specialization constants" |
#define __SYCL_PI_PROPERTY_SET_SYCL_ASSERT_USED "SYCL/assert used" |
#define __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_GLOBALS "SYCL/device globals" |
#define __SYCL_PI_PROPERTY_SET_SYCL_DEVICE_REQUIREMENTS "SYCL/device requirements" |
#define __SYCL_PI_PROPERTY_SET_SYCL_EXPORTED_SYMBOLS "SYCL/exported symbols" |
#define __SYCL_PI_PROPERTY_SET_SYCL_HOST_PIPES "SYCL/host pipes" |
#define __SYCL_PI_PROPERTY_SET_SYCL_MISC_PROP "SYCL/misc properties" |
#define _PI_CONCAT | ( | a, | |
b | |||
) | _PI_STRING_HELPER(a.b) |
#define _PI_H_VERSION_STRING _PI_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR) |
#define _PI_PLUGIN_VERSION_CHECK | ( | PI_API_VERSION, | |
PI_PLUGIN_VERSION | |||
) |
#define _PI_PLUGIN_VERSION_STRING | ( | PluginVersion | ) | _PI_TRIPLE_CONCAT(_PI_H_VERSION_MAJOR, _PI_H_VERSION_MINOR, PluginVersion) |
#define _PI_STRING_SUBSTITUTE | ( | X | ) | _PI_STRING_HELPER(X) |
#define _PI_TRIPLE_CONCAT | ( | a, | |
b, | |||
c | |||
) | _PI_STRING_HELPER(a.b.c) |
#define PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT "pi_ext_intel_devicelib_assert" |
using _pi_offload_entry = _pi_offload_entry_struct * |
using pi_bitfield = pi_uint64 |
using pi_buffer_region = pi_buffer_region_struct * |
using pi_command_type = _pi_command_type |
using pi_context = _pi_context * |
using pi_context_info = _pi_context_info |
using pi_context_properties = intptr_t |
using pi_device = _pi_device * |
using pi_device_affinity_domain = pi_bitfield |
using pi_device_binaries = pi_device_binaries_struct * |
using pi_device_binary = pi_device_binary_struct * |
using pi_device_binary_type = uint8_t |
using pi_device_fp_config = pi_bitfield |
using pi_device_info = _pi_device_info |
using pi_device_partition_property = intptr_t |
using pi_device_type = _pi_device_type |
using pi_event_info = _pi_event_info |
using pi_event_status = _pi_event_status |
using pi_ext_command_buffer = _pi_ext_command_buffer * |
using pi_ext_sync_point = pi_uint32 |
using pi_image_desc = _pi_image_desc |
using pi_image_format = _pi_image_format |
using pi_image_handle = pi_uint64 |
using pi_image_info = _pi_image_info |
using pi_image_mem_handle = void * |
using pi_image_offset = pi_image_offset_struct * |
using pi_image_region = pi_image_region_struct * |
using pi_interop_mem_handle = pi_uint64 |
using pi_interop_semaphore_handle = pi_uint64 |
using pi_kernel = _pi_kernel * |
typedef _pi_kernel_info pi_kernel_info |
using pi_map_flags = pi_bitfield |
typedef uint32_t pi_mem_access_flag |
using pi_mem_advice = _pi_mem_advice |
using pi_mem_alloc_info = _pi_mem_alloc_info |
using pi_mem_flags = pi_bitfield |
using pi_mem_info = _pi_mem_info |
using pi_mem_obj_access = _pi_mem_obj_access |
using pi_mem_properties = pi_bitfield |
using pi_mem_type = _pi_mem_type |
using pi_native_handle = uintptr_t |
using pi_peer_attr = _pi_peer_attr |
using pi_platform = _pi_platform * |
using pi_platform_info = _pi_platform_info |
using pi_plugin = _pi_plugin |
using pi_profiling_info = _pi_profiling_info |
using pi_program = _pi_program * |
using pi_program_info = _pi_program_info |
using pi_queue_info = _pi_queue_info |
using pi_queue_properties = pi_bitfield |
using pi_result = _pi_result |
using pi_sampler = _pi_sampler * |
using pi_sampler_info = _pi_sampler_info |
using pi_sampler_properties = pi_bitfield |
using pi_usm_mem_properties = pi_bitfield |
using pi_usm_type = _pi_usm_type |
enum _pi_command_type |
enum _pi_context_info |
enum _pi_device_info |
enum _pi_device_type : pi_uint64 |
enum _pi_event_info |
enum _pi_event_status |
enum _pi_image_copy_flags |
enum _pi_image_info |
enum _pi_kernel_exec_info |
enum _pi_kernel_info |
enum _pi_mem_advice |
enum _pi_mem_alloc_info |
enum _pi_mem_info |
enum _pi_mem_obj_access |
enum _pi_mem_type |
enum _pi_peer_attr |
enum _pi_platform_backend |
enum _pi_platform_info |
enum _pi_profiling_info |
enum _pi_program_info |
enum _pi_queue_info |
enum _pi_result |
enum _pi_sampler_info |
enum _pi_usm_capabilities : pi_bitfield |
enum _pi_usm_type |
enum pi_property_type |
pi_result piContextCreate | ( | const pi_context_properties * | properties, |
pi_uint32 | num_devices, | ||
const pi_device * | devices, | ||
void(*)(const char *errinfo, const void *private_info, size_t cb, void *user_data) | pfn_notify, | ||
void * | user_data, | ||
pi_context * | ret_context | ||
) |
Referenced by sycl::_V1::detail::context_impl::context_impl().
|
inline |
Definition at line 119 of file pi_cuda.cpp.
References die(), fixupInfoValueTypes(), HANDLE_ERRORS, sycl::_V1::detail::memcpy(), PI_ASSERT, PI_CONTEXT_INFO_DEVICES, PI_CONTEXT_INFO_NUM_DEVICES, PI_CONTEXT_INFO_REFERENCE_COUNT, PI_EXT_CONTEXT_INFO_ATOMIC_FENCE_ORDER_CAPABILITIES, PI_EXT_CONTEXT_INFO_ATOMIC_FENCE_SCOPE_CAPABILITIES, PI_EXT_CONTEXT_INFO_ATOMIC_MEMORY_ORDER_CAPABILITIES, PI_EXT_CONTEXT_INFO_ATOMIC_MEMORY_SCOPE_CAPABILITIES, PI_EXT_ONEAPI_CONTEXT_INFO_USM_FILL2D_SUPPORT, PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMCPY2D_SUPPORT, PI_EXT_ONEAPI_CONTEXT_INFO_USM_MEMSET2D_SUPPORT, pi2ur::piContextGetInfo(), setErrorMessage(), and urContextGetInfo().
Referenced by sycl::_V1::checkContextSupports(), sycl::_V1::detail::context_impl::context_impl(), sycl::_V1::detail::MemoryManager::copy_2d_usm(), sycl::_V1::detail::createBinaryProgram(), sycl::_V1::detail::get_context_info(), piContextGetInfo(), and piPluginInit().
|
inline |
Definition at line 152 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piContextRelease(), and urContextRelease().
Referenced by piContextRelease(), piPluginInit(), and sycl::_V1::detail::context_impl::~context_impl().
|
inline |
Definition at line 147 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piContextRetain(), and urContextRetain().
Referenced by sycl::_V1::detail::context_impl::context_impl(), sycl::_V1::detail::context_impl::get(), sycl::_V1::detail::context_impl::getNative(), piContextRetain(), and piPluginInit().
|
inline |
Returns requested info for provided native device Return PI_DEVICE_INFO_EXTENSION_DEVICELIB_ASSERT for PI_DEVICE_INFO_EXTENSIONS query when the device supports native asserts.
Definition at line 73 of file pi_cuda.cpp.
References checkDeviceExtensions(), getDeviceVersion(), HANDLE_ERRORS, max(), sycl::_V1::detail::memcpy(), PI_ASSERT, PI_DEVICE_INFO_ADDRESS_BITS, PI_DEVICE_INFO_ATOMIC_64, PI_DEVICE_INFO_AVAILABLE, PI_DEVICE_INFO_BACKEND_VERSION, PI_DEVICE_INFO_BUILD_ON_SUBDEVICE, PI_DEVICE_INFO_BUILT_IN_KERNELS, PI_DEVICE_INFO_COMPILER_AVAILABLE, PI_DEVICE_INFO_DEVICE_ID, PI_DEVICE_INFO_DOUBLE_FP_CONFIG, PI_DEVICE_INFO_DRIVER_VERSION, PI_DEVICE_INFO_ENDIAN_LITTLE, PI_DEVICE_INFO_ERROR_CORRECTION_SUPPORT, PI_DEVICE_INFO_EXECUTION_CAPABILITIES, PI_DEVICE_INFO_EXTENSIONS, PI_DEVICE_INFO_GLOBAL_MEM_CACHE_SIZE, PI_DEVICE_INFO_GLOBAL_MEM_CACHE_TYPE, PI_DEVICE_INFO_GLOBAL_MEM_CACHELINE_SIZE, PI_DEVICE_INFO_GLOBAL_MEM_SIZE, PI_DEVICE_INFO_GPU_EU_COUNT, PI_DEVICE_INFO_GPU_EU_COUNT_PER_SUBSLICE, PI_DEVICE_INFO_GPU_EU_SIMD_WIDTH, PI_DEVICE_INFO_GPU_HW_THREADS_PER_EU, PI_DEVICE_INFO_GPU_SLICES, PI_DEVICE_INFO_GPU_SUBSLICES_PER_SLICE, PI_DEVICE_INFO_HALF_FP_CONFIG, PI_DEVICE_INFO_HOST_UNIFIED_MEMORY, PI_DEVICE_INFO_IL_VERSION, PI_DEVICE_INFO_IMAGE2D_MAX_HEIGHT, PI_DEVICE_INFO_IMAGE2D_MAX_WIDTH, PI_DEVICE_INFO_IMAGE3D_MAX_DEPTH, PI_DEVICE_INFO_IMAGE3D_MAX_HEIGHT, PI_DEVICE_INFO_IMAGE3D_MAX_WIDTH, PI_DEVICE_INFO_IMAGE_MAX_ARRAY_SIZE, PI_DEVICE_INFO_IMAGE_MAX_BUFFER_SIZE, PI_DEVICE_INFO_IMAGE_SRGB, PI_DEVICE_INFO_IMAGE_SUPPORT, PI_DEVICE_INFO_LINKER_AVAILABLE, PI_DEVICE_INFO_LOCAL_MEM_SIZE, PI_DEVICE_INFO_LOCAL_MEM_TYPE, PI_DEVICE_INFO_MAX_CLOCK_FREQUENCY, PI_DEVICE_INFO_MAX_COMPUTE_UNITS, PI_DEVICE_INFO_MAX_CONSTANT_ARGS, PI_DEVICE_INFO_MAX_CONSTANT_BUFFER_SIZE, PI_DEVICE_INFO_MAX_MEM_ALLOC_SIZE, PI_DEVICE_INFO_MAX_MEM_BANDWIDTH, PI_DEVICE_INFO_MAX_NUM_SUB_GROUPS, PI_DEVICE_INFO_MAX_PARAMETER_SIZE, PI_DEVICE_INFO_MAX_READ_IMAGE_ARGS, PI_DEVICE_INFO_MAX_SAMPLERS, PI_DEVICE_INFO_MAX_WORK_GROUP_SIZE, PI_DEVICE_INFO_MAX_WORK_ITEM_DIMENSIONS, PI_DEVICE_INFO_MAX_WORK_ITEM_SIZES, PI_DEVICE_INFO_MAX_WRITE_IMAGE_ARGS, PI_DEVICE_INFO_MEM_BASE_ADDR_ALIGN, PI_DEVICE_INFO_NAME, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_CHAR, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_DOUBLE, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_FLOAT, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_HALF, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_INT, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_LONG, PI_DEVICE_INFO_NATIVE_VECTOR_WIDTH_SHORT, PI_DEVICE_INFO_OPENCL_C_VERSION, PI_DEVICE_INFO_PARENT_DEVICE, PI_DEVICE_INFO_PARTITION_AFFINITY_DOMAIN, PI_DEVICE_INFO_PARTITION_MAX_SUB_DEVICES, PI_DEVICE_INFO_PARTITION_PROPERTIES, PI_DEVICE_INFO_PARTITION_TYPE, PI_DEVICE_INFO_PCI_ADDRESS, PI_DEVICE_INFO_PLATFORM, PI_DEVICE_INFO_PREFERRED_INTEROP_USER_SYNC, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_CHAR, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_DOUBLE, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_FLOAT, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_HALF, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_INT, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_LONG, PI_DEVICE_INFO_PREFERRED_VECTOR_WIDTH_SHORT, PI_DEVICE_INFO_PRINTF_BUFFER_SIZE, PI_DEVICE_INFO_PROFILE, PI_DEVICE_INFO_PROFILING_TIMER_RESOLUTION, PI_DEVICE_INFO_QUEUE_PROPERTIES, PI_DEVICE_INFO_REFERENCE_COUNT, PI_DEVICE_INFO_SINGLE_FP_CONFIG, PI_DEVICE_INFO_SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS, PI_DEVICE_INFO_SUB_GROUP_SIZES_INTEL, PI_DEVICE_INFO_TYPE, PI_DEVICE_INFO_USM_CROSS_SHARED_SUPPORT, PI_DEVICE_INFO_USM_DEVICE_SUPPORT, PI_DEVICE_INFO_USM_HOST_SUPPORT, PI_DEVICE_INFO_USM_SINGLE_SHARED_SUPPORT, PI_DEVICE_INFO_USM_SYSTEM_SHARED_SUPPORT, PI_DEVICE_INFO_UUID, PI_DEVICE_INFO_VENDOR, PI_DEVICE_INFO_VENDOR_ID, PI_DEVICE_INFO_VERSION, PI_EXT_CODEPLAY_DEVICE_INFO_MAX_REGISTERS_PER_WORK_GROUP, PI_EXT_DEVICE_INFO_ATOMIC_FENCE_ORDER_CAPABILITIES, PI_EXT_DEVICE_INFO_ATOMIC_FENCE_SCOPE_CAPABILITIES, PI_EXT_DEVICE_INFO_ATOMIC_MEMORY_ORDER_CAPABILITIES, PI_EXT_DEVICE_INFO_ATOMIC_MEMORY_SCOPE_CAPABILITIES, PI_EXT_INTEL_DEVICE_INFO_ESIMD_SUPPORT, PI_EXT_INTEL_DEVICE_INFO_FREE_MEMORY, PI_EXT_INTEL_DEVICE_INFO_MAX_COMPUTE_QUEUE_INDICES, PI_EXT_INTEL_DEVICE_INFO_MEM_CHANNEL_SUPPORT, PI_EXT_INTEL_DEVICE_INFO_MEMORY_BUS_WIDTH, PI_EXT_INTEL_DEVICE_INFO_MEMORY_CLOCK_RATE, PI_EXT_ONEAPI_DEVICE_INFO_BFLOAT16_MATH_FUNCTIONS, PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_1D_USM_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_2D_USM_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SHARED_USM_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_BINDLESS_IMAGES_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_IMAGE_PITCH_ALIGN, PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_EXPORT_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_MEMORY_IMPORT_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_EXPORT_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_INTEROP_SEMAPHORE_IMPORT_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_IP_VERSION, PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_HEIGHT, PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_PITCH, PI_EXT_ONEAPI_DEVICE_INFO_MAX_IMAGE_LINEAR_WIDTH, PI_EXT_ONEAPI_DEVICE_INFO_MAX_WORK_GROUPS_3D, PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_ANISOTROPY_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_LEVEL_REFERENCE_SUPPORT, PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_MAX_ANISOTROPY, PI_EXT_ONEAPI_DEVICE_INFO_MIPMAP_SUPPORT, PI_MEMORY_ORDER_ACQ_REL, PI_MEMORY_ORDER_ACQUIRE, PI_MEMORY_ORDER_RELAXED, PI_MEMORY_ORDER_RELEASE, PI_MEMORY_ORDER_SEQ_CST, PI_MEMORY_SCOPE_DEVICE, PI_MEMORY_SCOPE_SUB_GROUP, PI_MEMORY_SCOPE_SYSTEM, PI_MEMORY_SCOPE_WORK_GROUP, PI_MEMORY_SCOPE_WORK_ITEM, PI_TO_UR_MAP_DEVICE_INFO, pi2ur::piDeviceGetInfo(), ur2piDeviceInfoValue(), UR_EXT_DEVICE_INFO_OPENCL_C_VERSION, urDeviceGetInfo(), OCLV::V2_0, OCLV::V2_1, and OCLV::V3_0.
Referenced by sycl::_V1::detail::applyAllowList(), sycl::_V1::detail::get_device_info_impl< ReturnT, Param >::get(), sycl::_V1::detail::get_device_info_impl< platform, Param >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< info::fp_config >, Param >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< info::fp_config >, info::device::single_fp_config >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::queue_profiling >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< memory_order >, info::device::atomic_memory_order_capabilities >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< memory_order >, info::device::atomic_fence_order_capabilities >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< memory_scope >, info::device::atomic_memory_scope_capabilities >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< memory_scope >, info::device::atomic_fence_scope_capabilities >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::ext_oneapi_bfloat16_math_functions >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< info::execution_capability >, info::device::execution_capabilities >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< info::partition_property >, info::device::partition_properties >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< info::partition_affinity_domain >, info::device::partition_affinity_domains >::get(), sycl::_V1::detail::get_device_info_impl< info::partition_affinity_domain, info::device::partition_type_affinity_domain >::get(), sycl::_V1::detail::get_device_info_impl< info::partition_property, info::device::partition_type_property >::get(), sycl::_V1::detail::get_device_info_impl< std::vector< size_t >, info::device::sub_group_sizes >::get(), sycl::_V1::detail::get_device_info_impl< range< Dimensions >, info::device::max_work_item_sizes< Dimensions > >::get(), sycl::_V1::detail::get_device_info_impl< ext::oneapi::experimental::architecture, ext::oneapi::experimental::info::device::architecture >::get(), sycl::_V1::detail::get_device_info_impl< id< 1 >, ext::oneapi::experimental::info::device::max_work_groups< 1 > >::get(), sycl::_V1::detail::get_device_info_impl< id< 2 >, ext::oneapi::experimental::info::device::max_work_groups< 2 > >::get(), sycl::_V1::detail::get_device_info_impl< id< 3 >, ext::oneapi::experimental::info::device::max_work_groups< 3 > >::get(), sycl::_V1::detail::get_device_info_impl< device, info::device::parent_device >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::usm_device_allocations >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::usm_host_allocations >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::usm_shared_allocations >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::usm_restricted_shared_allocations >::get(), sycl::_V1::detail::get_device_info_impl< bool, info::device::usm_system_allocations >::get(), sycl::_V1::detail::get_device_info_impl< uint32_t, ext::codeplay::experimental::info::device::max_registers_per_work_group >::get(), sycl::_V1::detail::get_device_info_impl< ext::oneapi::experimental::graph_support_level, ext::oneapi::experimental::info::device::graph_support >::get(), sycl::_V1::detail::device_impl::get_device_info_string(), sycl::_V1::detail::ProgramManager::getBuiltPIProgram(), sycl::_V1::detail::platform_impl::getPlatformFromPiDevice(), sycl::_V1::detail::ProgramManager::getProgramBuildLog(), sycl::_V1::detail::enqueue_kernel_launch::handleInvalidValue(), sycl::_V1::detail::enqueue_kernel_launch::handleInvalidWorkGroupSize(), sycl::_V1::detail::enqueue_kernel_launch::handleInvalidWorkItemSize(), sycl::_V1::detail::device_impl::has(), sycl::_V1::opencl::has_extension(), piDeviceGetInfo(), piKernelGetSubGroupInfo(), and piPluginInit().
|
inline |
Definition at line 80 of file pi_cuda.cpp.
References die(), HANDLE_ERRORS, PI_ASSERT, PI_DEVICE_PARTITION_BY_AFFINITY_DOMAIN, PI_DEVICE_PARTITION_BY_COUNTS, PI_DEVICE_PARTITION_EQUALLY, PI_EXT_INTEL_DEVICE_PARTITION_BY_CSLICE, pi2ur::piDevicePartition(), and urDevicePartition().
Referenced by sycl::_V1::detail::device_impl::create_sub_devices(), piDevicePartition(), and piPluginInit().
Definition at line 69 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piDeviceRelease(), and urDeviceRelease().
Referenced by sycl::_V1::detail::platform_impl::get_devices(), piDeviceRelease(), piPluginInit(), and sycl::_V1::detail::device_impl::~device_impl().
Definition at line 65 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piDeviceRetain(), and urDeviceRetain().
Referenced by sycl::_V1::detail::device_impl::get(), sycl::_V1::detail::device_impl::getNative(), piDeviceRetain(), and piPluginInit().
|
inline |
Definition at line 58 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_DEVICE_TYPE_ACC, PI_DEVICE_TYPE_ALL, PI_DEVICE_TYPE_CPU, PI_DEVICE_TYPE_GPU, pi2ur::piDevicesGet(), and urDeviceGet().
Referenced by sycl::_V1::detail::platform_impl::get_devices(), piDevicesGet(), and piPluginInit().
|
inline |
Definition at line 634 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueEventsWait(), and urEnqueueEventsWait().
Referenced by sycl::_V1::detail::MemoryManager::copy_2d_usm(), sycl::_V1::detail::MemoryManager::copy_usm(), sycl::_V1::detail::MemoryManager::fill_2d_usm(), sycl::_V1::detail::MemoryManager::fill_usm(), sycl::_V1::detail::MemoryManager::memset_2d_usm(), piEnqueueEventsWait(), piPluginInit(), and sycl::_V1::detail::Command::waitForEvents().
|
inline |
Definition at line 642 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueEventsWaitWithBarrier(), and urEnqueueEventsWaitWithBarrier().
Referenced by piEnqueueEventsWaitWithBarrier(), and piPluginInit().
|
inline |
Definition at line 531 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueKernelLaunch(), and urEnqueueKernelLaunch().
Referenced by piEnqueueKernelLaunch(), piPluginInit(), and sycl::_V1::detail::SetKernelParamsAndLaunch().
|
inline |
Definition at line 703 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueMemBufferCopy(), and urEnqueueMemBufferCopy().
Referenced by sycl::_V1::detail::copyD2D(), piEnqueueMemBufferCopy(), and piPluginInit().
|
inline |
Definition at line 714 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, PI_ASSERT, pi2ur::piEnqueueMemBufferCopyRect(), urEnqueueMemBufferCopyRect(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::copyD2D(), piEnqueueMemBufferCopyRect(), and piPluginInit().
|
inline |
Definition at line 727 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueMemBufferFill(), and urEnqueueMemBufferFill().
Referenced by sycl::_V1::detail::MemoryManager::fill(), piEnqueueMemBufferFill(), and piPluginInit().
|
inline |
Definition at line 739 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_MAP_READ, PI_MAP_WRITE, PI_MAP_WRITE_INVALIDATE_REGION, pi2ur::piEnqueueMemBufferMap(), and urEnqueueMemBufferMap().
Referenced by sycl::_V1::detail::memBufferMapHelper(), piEnqueueMemBufferMap(), and piPluginInit().
|
inline |
Definition at line 651 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueMemBufferRead(), and urEnqueueMemBufferRead().
Referenced by sycl::_V1::detail::copyD2H(), piEnqueueMemBufferRead(), and piPluginInit().
|
inline |
Definition at line 663 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, PI_ASSERT, pi2ur::piEnqueueMemBufferReadRect(), urEnqueueMemBufferReadRect(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::copyD2H(), piEnqueueMemBufferReadRect(), and piPluginInit().
|
inline |
Definition at line 677 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueMemBufferWrite(), and urEnqueueMemBufferWrite().
Referenced by sycl::_V1::detail::copyH2D(), piEnqueueMemBufferWrite(), and piPluginInit().
|
inline |
Definition at line 689 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, PI_ASSERT, pi2ur::piEnqueueMemBufferWriteRect(), urEnqueueMemBufferWriteRect(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::copyH2D(), piEnqueueMemBufferWriteRect(), and piPluginInit().
|
inline |
Definition at line 792 of file pi_cuda.cpp.
References pi_image_region_struct::depth, HANDLE_ERRORS, pi_image_region_struct::height, PI_ASSERT, pi2ur::piEnqueueMemImageCopy(), urEnqueueMemImageCopy(), pi_image_region_struct::width, pi_image_offset_struct::x, pi_image_offset_struct::y, and pi_image_offset_struct::z.
Referenced by sycl::_V1::detail::copyD2D(), piEnqueueMemImageCopy(), and piPluginInit().
|
inline |
Definition at line 801 of file pi_cuda.cpp.
References die(), PI_ASSERT, and pi2ur::piEnqueueMemImageFill().
Referenced by sycl::_V1::detail::MemoryManager::fill(), piEnqueueMemImageFill(), and piPluginInit().
|
inline |
Definition at line 766 of file pi_cuda.cpp.
References pi_image_region_struct::depth, HANDLE_ERRORS, pi_image_region_struct::height, PI_ASSERT, pi2ur::piEnqueueMemImageRead(), urEnqueueMemImageRead(), pi_image_region_struct::width, pi_image_offset_struct::x, pi_image_offset_struct::y, and pi_image_offset_struct::z.
Referenced by piEnqueueMemImageRead(), and piPluginInit().
|
inline |
Definition at line 778 of file pi_cuda.cpp.
References pi_image_region_struct::depth, HANDLE_ERRORS, pi_image_region_struct::height, PI_ASSERT, pi2ur::piEnqueueMemImageWrite(), urEnqueueMemImageWrite(), pi_image_region_struct::width, pi_image_offset_struct::x, pi_image_offset_struct::y, and pi_image_offset_struct::z.
Referenced by piEnqueueMemImageWrite(), and piPluginInit().
|
inline |
Definition at line 750 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEnqueueMemUnmap(), and urEnqueueMemUnmap().
Referenced by sycl::_V1::detail::memUnmapHelper(), piEnqueueMemUnmap(), and piPluginInit().
|
inline |
Create PI event object in a signalled/completed state.
context | is the PI context of the event. |
ret_event | is the PI even created. |
Definition at line 556 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piEventCreate(), and urEventCreateWithNativeHandle().
Referenced by piEventCreate(), and piPluginInit().
|
inline |
Definition at line 560 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_EVENT_INFO_COMMAND_EXECUTION_STATUS, PI_EVENT_INFO_COMMAND_QUEUE, PI_EVENT_INFO_COMMAND_TYPE, PI_EVENT_INFO_CONTEXT, PI_EVENT_INFO_REFERENCE_COUNT, pi2ur::piEventGetInfo(), and urEventGetInfo().
Referenced by sycl::_V1::detail::event_impl::event_impl(), sycl::_V1::detail::event_impl::flushIfNeeded(), sycl::_V1::detail::get_event_info(), piEventGetInfo(), and piPluginInit().
|
inline |
Definition at line 567 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_PROFILING_INFO_COMMAND_END, PI_PROFILING_INFO_COMMAND_QUEUED, PI_PROFILING_INFO_COMMAND_START, PI_PROFILING_INFO_COMMAND_SUBMIT, pi2ur::piEventGetProfilingInfo(), and urEventGetProfilingInfo().
Referenced by sycl::_V1::detail::get_event_profiling_info(), piEventGetProfilingInfo(), and piPluginInit().
Definition at line 594 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEventRelease(), and urEventRelease().
Referenced by sycl::_V1::detail::DeviceGlobalUSMMem::getZeroInitEvent(), sycl::_V1::detail::context_impl::initializeDeviceGlobals(), piEventRelease(), piPluginInit(), sycl::_V1::detail::DeviceGlobalMapEntry::removeAssociatedResources(), sycl::_V1::detail::event_impl::~event_impl(), and sycl::_V1::detail::OwnedPiEvent::~OwnedPiEvent().
Definition at line 592 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piEventRetain(), and urEventRetain().
Referenced by sycl::_V1::detail::MemoryManager::allocateInteropMemObject(), sycl::_V1::detail::event_impl::getNative(), sycl::_V1::detail::make_event(), sycl::_V1::detail::OwnedPiEvent::OwnedPiEvent(), piEventRetain(), and piPluginInit().
pi_result piEventSetCallback | ( | pi_event | event, |
pi_int32 | command_exec_callback_type, | ||
void(*)(pi_event event, pi_int32 event_command_status, void *user_data) | pfn_notify, | ||
void * | user_data | ||
) |
Referenced by piPluginInit().
Definition at line 588 of file pi_cuda.cpp.
References die(), and pi2ur::piEventSetStatus().
Referenced by piEventSetStatus(), and piPluginInit().
Definition at line 575 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piEventsWait(), and urEventWait().
Referenced by sycl::_V1::detail::memUnmapHelper(), piEventsWait(), piPluginInit(), sycl::_V1::detail::waitForEvents(), sycl::_V1::detail::Command::waitForEvents(), and sycl::_V1::detail::event_impl::waitInternal().
|
inline |
API to create samplers for bindless images.
context | is the pi_context |
device | is the pi_device |
sampler_properties | is the pointer to the sampler properties bitfield |
min_mipmap_level_clamp | is the minimum mipmap level to sample from |
max_mipmap_level_clamp | is the maximum mipmap level to sample from |
max_anisotropy | is the maximum anisotropic ratio |
result_sampler | is the returned sampler |
Definition at line 394 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_SAMPLER_ADDRESSING_MODE_CLAMP, PI_SAMPLER_ADDRESSING_MODE_CLAMP_TO_EDGE, PI_SAMPLER_ADDRESSING_MODE_MIRRORED_REPEAT, PI_SAMPLER_ADDRESSING_MODE_NONE, PI_SAMPLER_ADDRESSING_MODE_REPEAT, PI_SAMPLER_FILTER_MODE_LINEAR, PI_SAMPLER_FILTER_MODE_NEAREST, PI_SAMPLER_PROPERTIES_ADDRESSING_MODE, PI_SAMPLER_PROPERTIES_FILTER_MODE, PI_SAMPLER_PROPERTIES_MIP_FILTER_MODE, PI_SAMPLER_PROPERTIES_NORMALIZED_COORDS, pi2ur::piextBindlessImageSamplerCreate(), and urSamplerCreate().
Referenced by sycl::_V1::ext::oneapi::experimental::create_image(), and piextBindlessImageSamplerCreate().
|
inline |
API to create a command-buffer.
context | The context to associate the command-buffer with. |
device | The device to associate the command-buffer with. |
desc | Descriptor for the new command-buffer. |
ret_command_buffer | Pointer to fill with the address of the new command-buffer. |
Definition at line 1036 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferCreate(), and urCommandBufferCreateExp().
Referenced by sycl::_V1::ext::oneapi::experimental::detail::exec_graph_impl::createCommandBuffers(), piextCommandBufferCreate(), and piPluginInit().
|
inline |
API to stop command-buffer recording such that no more commands can be appended, and makes the command-buffer ready to enqueue on a command-queue.
command_buffer | The command_buffer to finalize. |
Definition at line 1051 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferFinalize(), and urCommandBufferFinalizeExp().
Referenced by sycl::_V1::ext::oneapi::experimental::detail::exec_graph_impl::createCommandBuffers(), and piextCommandBufferFinalize().
|
inline |
API to append a mem buffer copy command to the command-buffer.
command_buffer | The command-buffer to append onto. |
src_buffer | is the data to be copied |
dst_buffer | is the location the data will be copied |
src_offset | offset into src_buffer |
dst_offset | offset into dst_buffer |
size | is number of bytes to copy |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1074 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferMemBufferCopy(), and urCommandBufferAppendMembufferCopyExp().
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyD2D_cmd_buffer(), piextCommandBufferMemBufferCopy(), and piPluginInit().
|
inline |
API to append a rectangular mem buffer copy command to the command-buffer.
command_buffer | The command-buffer to append onto. |
src_buffer | is the data to be copied |
dst_buffer | is the location the data will be copied |
src_origin | offset for the start of the region to copy in src_buffer |
dst_origin | offset for the start of the region to copy in dst_buffer |
region | The size of the region to be copied |
src_row_pitch | Row pitch for the src data |
src_slice_pitch | Slice pitch for the src data |
dst_row_pitch | Row pitch for the dst data |
dst_slice_pitch | Slice pitch for the dst data |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1084 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, pi2ur::piextCommandBufferMemBufferCopyRect(), urCommandBufferAppendMembufferCopyRectExp(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyD2D_cmd_buffer(), piextCommandBufferMemBufferCopyRect(), and piPluginInit().
|
inline |
API to append a mem buffer read command to the command-buffer.
command_buffer | The command-buffer to append onto. |
buffer | is the data to be read |
offset | offset into buffer |
size | is number of bytes to read |
dst | is the pointer to the destination |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1096 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextCommandBufferMemBufferRead(), and urCommandBufferAppendMembufferReadExp().
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyD2H_cmd_buffer(), piextCommandBufferMemBufferRead(), and piPluginInit().
|
inline |
API to append a rectangular mem buffer read command to the command-buffer.
command_buffer | The command-buffer to append onto. |
buffer | is the data to be read |
buffer_offset | offset for the start of the region to read in buffer |
host_offset | offset for the start of the region to be written from ptr |
region | The size of the region to read |
buffer_row_pitch | Row pitch for the source buffer data |
buffer_slice_pitch | Slice pitch for the source buffer data |
host_row_pitch | Row pitch for the destination data ptr |
host_slice_pitch | Slice pitch for the destination data ptr |
ptr | is the location the data will be written |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1105 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, PI_ASSERT, pi2ur::piextCommandBufferMemBufferReadRect(), urCommandBufferAppendMembufferReadRectExp(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyD2H_cmd_buffer(), piextCommandBufferMemBufferReadRect(), and piPluginInit().
|
inline |
API to append a mem buffer write command to the command-buffer.
command_buffer | The command-buffer to append onto. |
buffer | is the location to write the data |
offset | offset into buffer |
size | is number of bytes to write |
ptr | is the pointer to the source |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1118 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextCommandBufferMemBufferWrite(), and urCommandBufferAppendMembufferWriteExp().
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyH2D_cmd_buffer(), piextCommandBufferMemBufferWrite(), and piPluginInit().
|
inline |
API to append a rectangular mem buffer write command to the command-buffer.
command_buffer | The command-buffer to append onto. |
buffer | is the location to write the data |
buffer_offset | offset for the start of the region to write in buffer |
host_offset | offset for the start of the region to be read from ptr |
region | The size of the region to write |
buffer_row_pitch | Row pitch for the buffer data |
buffer_slice_pitch | Slice pitch for the buffer data |
host_row_pitch | Row pitch for the source data ptr |
host_slice_pitch | Slice pitch for the source data ptr |
ptr | is the pointer to the source |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1127 of file pi_cuda.cpp.
References pi_buff_rect_region_struct::depth_scalar, HANDLE_ERRORS, pi_buff_rect_region_struct::height_scalar, PI_ASSERT, pi2ur::piextCommandBufferMemBufferWriteRect(), urCommandBufferAppendMembufferWriteRectExp(), pi_buff_rect_region_struct::width_bytes, pi_buff_rect_offset_struct::x_bytes, pi_buff_rect_offset_struct::y_scalar, and pi_buff_rect_offset_struct::z_scalar.
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copyH2D_cmd_buffer(), piextCommandBufferMemBufferWriteRect(), and piPluginInit().
|
inline |
API to append a USM memcpy command to the command-buffer.
command_buffer | The command-buffer to append onto. |
dst_ptr | is the location the data will be copied |
src_ptr | is the data to be copied |
size | is number of bytes to copy |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this memory operation. |
Definition at line 1065 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferMemcpyUSM(), and urCommandBufferAppendMemcpyUSMExp().
Referenced by sycl::_V1::detail::MemoryManager::ext_oneapi_copy_usm_cmd_buffer(), piextCommandBufferMemcpyUSM(), and piPluginInit().
|
inline |
API to append a kernel execution command to the command-buffer.
command_buffer | The command-buffer to append onto. |
kernel | The kernel to append. |
work_dim | Dimension of the kernel execution. |
global_work_offset | Offset to use when executing kernel. |
global_work_size | Global work size to use when executing kernel. |
local_work_size | Local work size to use when executing kernel. |
num_sync_points_in_wait_list | The number of sync points in the provided wait list. |
sync_point_wait_list | A list of sync points that this command must wait on. |
sync_point | The sync_point associated with this kernel execution. |
Definition at line 1055 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferNDRangeKernel(), and urCommandBufferAppendKernelLaunchExp().
Referenced by sycl::_V1::detail::enqueueImpCommandBufferKernel(), piextCommandBufferNDRangeKernel(), and piPluginInit().
|
inline |
API to decrement the reference count of the command-buffer.
After the command_buffer reference count becomes zero and has finished execution, the command-buffer is deleted.
command_buffer | The command_buffer to release. |
Definition at line 1047 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferRelease(), and urCommandBufferReleaseExp().
Referenced by piextCommandBufferRelease(), piPluginInit(), and sycl::_V1::ext::oneapi::experimental::detail::exec_graph_impl::~exec_graph_impl().
|
inline |
API to increment the reference count of the command-buffer.
command_buffer | The command_buffer to retain. |
Definition at line 1043 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextCommandBufferRetain(), and urCommandBufferRetainExp().
Referenced by piextCommandBufferRetain(), and piPluginInit().
|
inline |
Creates PI context object from a native handle.
NOTE: The created PI object takes ownership of the native handle. NOTE: The number of devices and the list of devices is needed for Level Zero backend because there is no possilibity to query this information from context handle for Level Zero. If backend has API to query a list of devices from the context native handle then these parameters are ignored.
nativeHandle | is the native handle to create PI context from. |
numDevices | is the number of devices in the context. Parameter is ignored if number of devices can be queried from the context native handle for a backend. |
devices | is the list of devices in the context. Parameter is ignored if devices can be queried from the context native handle for a backend. |
pluginOwnsNativeHandle | Indicates whether the created PI object should take ownership of the native handle. |
context | is the PI context created from the native handle. |
Definition at line 138 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextContextCreateWithNativeHandle(), and urContextCreateWithNativeHandle().
Referenced by sycl::_V1::ext::oneapi::level_zero::make_context(), sycl::_V1::detail::make_context(), piextContextCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI context object.
context | is the PI context to get the native handle of. |
nativeHandle | is the native handle of context. |
Definition at line 133 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextContextGetNativeHandle(), piextGetNativeHandle(), and urContextGetNativeHandle().
Referenced by sycl::_V1::detail::context_impl::getNative(), piextContextGetNativeHandle(), and piPluginInit().
|
inline |
Definition at line 127 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextContextSetExtendedDeleter(), and urContextSetExtendedDeleter().
Referenced by sycl::_V1::detail::pi::contextSetExtendedDeleter(), piextContextSetExtendedDeleter(), and piPluginInit().
|
inline |
API to destroy the external semaphore handle.
context | is the pi_context |
device | is the pi_device |
sem_handle | is the interop semaphore handle to the external semaphore to be destroyed |
Definition at line 481 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextDestroyExternalSemaphore(), and urBindlessImagesDestroyExternalSemaphoreExp().
Referenced by sycl::_V1::ext::oneapi::experimental::destroy_external_semaphore(), and piextDestroyExternalSemaphore().
|
inline |
Creates PI device object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI device from. |
platform | is the platform of the device (optional). |
device | is the PI device created from the native handle. |
Definition at line 101 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextDeviceCreateWithNativeHandle(), and urDeviceCreateWithNativeHandle().
Referenced by sycl::_V1::ext::oneapi::level_zero::make_device(), sycl::_V1::detail::make_device(), piextDeviceCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI device object.
device | is the PI device to get the native handle of. |
nativeHandle | is the native handle of device. |
Definition at line 95 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextDeviceGetNativeHandle(), piextGetNativeHandle(), and urDeviceGetNativeHandle().
Referenced by sycl::_V1::detail::device_impl::getNative(), piextDeviceGetNativeHandle(), and piPluginInit().
|
inline |
Selects the most appropriate device binary based on runtime information and the IR characteristics.
Definition at line 88 of file pi_cuda.cpp.
References __SYCL_PI_DEVICE_BINARY_TARGET_AMDGCN, __SYCL_PI_DEVICE_BINARY_TARGET_NATIVE_CPU, __SYCL_PI_DEVICE_BINARY_TARGET_NVPTX64, __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV32, __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64, __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_FPGA, __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_GEN, __SYCL_PI_DEVICE_BINARY_TARGET_SPIRV64_X86_64, __SYCL_PI_DEVICE_BINARY_TARGET_UNKNOWN, HANDLE_ERRORS, max(), pi2ur::piextDeviceSelectBinary(), and urDeviceSelectBinary().
Referenced by sycl::_V1::detail::compatibleWithDevice(), sycl::_V1::detail::getBinImageFromMultiMap(), sycl::_V1::detail::ProgramManager::getDeviceImage(), piextDeviceSelectBinary(), and piPluginInit().
Definition at line 1169 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextDisablePeerAccess(), and urUsmP2PDisablePeerAccessExp().
Referenced by piextDisablePeerAccess(), and piPluginInit().
Definition at line 1163 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextEnablePeerAccess(), and urUsmP2PEnablePeerAccessExp().
Referenced by piextEnablePeerAccess(), and piPluginInit().
|
inline |
API to submit the command-buffer to queue for execution, returns an error if the command-buffer is not finalized or another instance of the same command-buffer is currently executing.
command_buffer | The command-buffer to be submitted. |
queue | The PI queue to submit on. |
num_events_in_wait_list | The number of events that this execution depends on. |
event_wait_list | List of pi_events to wait on. |
event | The pi_event associated with this enqueue. |
Definition at line 1140 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextEnqueueCommandBuffer(), and urCommandBufferEnqueueExp().
Referenced by sycl::_V1::ext::oneapi::experimental::detail::exec_graph_impl::enqueue(), piextEnqueueCommandBuffer(), and piPluginInit().
|
inline |
API reading data from a device global variable to host.
queue | is the queue |
program | is the program containing the device global variable |
blocking_read | is true if the read should block |
name | is the unique identifier for the device global variable |
count | is the number of bytes to copy |
offset | is the byte offset into the device global variable to start copying |
dst | is a pointer to where the data must be copied to |
num_events_in_wait_list | is a number of events in the wait list |
event_wait_list | is the wait list |
event | is the resulting event |
Queue | is the queue |
Program | is the program containing the device global variable |
Name | is the unique identifier for the device global variable |
BlockingRead | is true if the read should block |
Count | is the number of bytes to copy |
Offset | is the byte offset into the device global variable to start copying |
Dst | is a pointer to where the data must be copied to |
NumEventsInWaitList | is a number of events in the wait list |
EventWaitList | is the wait list |
Event | is the resulting event |
queue | is the queue |
program | is the program containing the device global variable |
name | is the unique identifier for the device global variable |
blocking_read | is true if the read should block |
count | is the number of bytes to copy |
offset | is the byte offset into the device global variable to start copying |
dst | is a pointer to where the data must be copied to |
num_events_in_wait_list | is a number of events in the wait list |
event_wait_list | is the wait list |
event | is the resulting event |
Definition at line 970 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueReadGlobalVariableCache, clEnqueueReadGlobalVariableName, ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, pi2ur::piextEnqueueDeviceGlobalVariableRead(), and urEnqueueDeviceGlobalVariableRead().
Referenced by sycl::_V1::detail::memcpyFromDeviceGlobalDirect(), piextEnqueueDeviceGlobalVariableRead(), and piPluginInit().
|
inline |
Device global variable.
API for writing data from host to a device global variable.
queue | is the queue |
program | is the program containing the device global variable |
blocking_write | is true if the write should block |
name | is the unique identifier for the device global variable |
count | is the number of bytes to copy |
offset | is the byte offset into the device global variable to start copying |
src | is a pointer to where the data must be copied from |
num_events_in_wait_list | is a number of events in the wait list |
event_wait_list | is the wait list |
event | is the resulting event |
Device global variable.
Queue | is the queue |
Program | is the program containing the device global variable |
Name | is the unique identifier for the device global variable |
BlockingWrite | is true if the write should block |
Count | is the number of bytes to copy |
Offset | is the byte offset into the device global variable to start copying |
Src | is a pointer to where the data must be copied from |
NumEventsInWaitList | is a number of events in the wait list |
EventWaitList | is the wait list |
Event | is the resulting event |
Device global variable.
queue | is the queue |
program | is the program containing the device global variable |
name | is the unique identifier for the device global variable |
blocking_write | is true if the write should block |
count | is the number of bytes to copy |
offset | is the byte offset into the device global variable to start copying |
src | is a pointer to where the data must be copied from |
num_events_in_wait_list | is a number of events in the wait list |
event_wait_list | is the wait list |
event | is the resulting event |
Definition at line 961 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueWriteGlobalVariableCache, clEnqueueWriteGlobalVariableName, ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, pi2ur::piextEnqueueDeviceGlobalVariableWrite(), and urEnqueueDeviceGlobalVariableWrite().
Referenced by sycl::_V1::detail::context_impl::initializeDeviceGlobals(), sycl::_V1::detail::memcpyToDeviceGlobalDirect(), piextEnqueueDeviceGlobalVariableWrite(), and piPluginInit().
|
inline |
Plugin.
Read from pipe of a given name
queue | a valid host command-queue in which the read / write command will be queued. command_queue and program must be created with the same OpenCL context. |
program | a program object with a successfully built executable. |
pipe_symbol | the name of the program scope pipe global variable. |
blocking | indicate if the read and write operations are blocking or non-blocking |
ptr | a pointer to buffer in host memory that will hold resulting data from pipe |
size | size of the memory region to read or write, in bytes. |
num_events_in_waitlist | number of events in the wait list. |
events_waitlist | specify events that need to complete before this particular command can be executed. |
event | returns an event object that identifies this read / write command and can be used to query or queue a wait for this command to complete. |
Plugin.
Queue | is the queue |
Program | is the program containing the device variable |
PipeSymbol | is the unique identifier for the device variable |
Blocking | is true if the write should block |
Ptr | is a pointer to where the data will be copied to |
Size | is size of the data that is read/written from/to pipe |
NumEventsInWaitList | is a number of events in the wait list |
EventWaitList | is the wait list |
Event | is the resulting event |
Definition at line 981 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueReadHostPipeINTELCache, clEnqueueReadHostPipeName, die(), ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, and urEnqueueReadHostPipe().
Referenced by sycl::_V1::detail::enqueueReadWriteHostPipe(), and piPluginInit().
|
inline |
Write to pipe of a given name.
queue | a valid host command-queue in which the read / write command will be queued. command_queue and program must be created with the same OpenCL context. |
program | a program object with a successfully built executable. |
pipe_symbol | the name of the program scope pipe global variable. |
blocking | indicate if the read and write operations are blocking or non-blocking |
ptr | a pointer to buffer in host memory that holds data to be written to host pipe. |
size | size of the memory region to read or write, in bytes. |
num_events_in_waitlist | number of events in the wait list. |
events_waitlist | specify events that need to complete before this particular command can be executed. |
event | returns an event object that identifies this read / write command and can be used to query or queue a wait for this command to complete. |
Write to pipe of a given name.
Queue | is the queue |
Program | is the program containing the device variable |
PipeSymbol | is the unique identifier for the device variable |
Blocking | is true if the write should block |
Ptr | is a pointer to where the data must be copied from |
Size | is size of the data that is read/written from/to pipe |
NumEventsInWaitList | is a number of events in the wait list |
EventWaitList | is the wait list |
Event | is the resulting event |
Definition at line 1001 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueWriteHostPipeINTELCache, clEnqueueWriteHostPipeName, die(), ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, and urEnqueueWriteHostPipe().
Referenced by sycl::_V1::detail::enqueueReadWriteHostPipe(), and piPluginInit().
|
inline |
Creates PI event object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI event from. |
context | is the corresponding PI context |
pluginOwnsNativeHandle | Indicates whether the created PI object should take ownership of the native handle. |
event | is the PI event created from the native handle. |
Definition at line 604 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextEventCreateWithNativeHandle(), and urEventCreateWithNativeHandle().
Referenced by sycl::_V1::detail::make_event(), piextEventCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI event object.
event | is the PI event to get the native handle of. |
nativeHandle | is the native handle of event. |
Definition at line 598 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextEventGetNativeHandle(), and urEventGetNativeHandle().
Referenced by sycl::_V1::detail::event_impl::getNative(), piextEventGetNativeHandle(), and piPluginInit().
|
inline |
Retrieves a device function pointer to a user-defined function.
function_name
. function_pointer_ret
is set to 0 if query failed.program
must be built before calling this API. device
must present in the list of devices returned by get_device
method for program
.If a fallback method determines the function exists but the address is not available PI_ERROR_FUNCTION_ADDRESS_IS_NOT_AVAILABLE is returned. If the address does not exist PI_ERROR_INVALID_KERNEL_NAME is returned.
Definition at line 821 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clGetDeviceFunctionPointerCache, clGetDeviceFunctionPointerName, ExtFuncPtrCache, HANDLE_ERRORS, is_in_separated_string(), PI_ASSERT, PI_PROGRAM_INFO_KERNEL_NAMES, pi2ur::piextGetDeviceFunctionPointer(), and urProgramGetFunctionPointer().
Referenced by sycl::_V1::detail::program_impl::has_kernel(), piextGetDeviceFunctionPointer(), and piPluginInit().
|
inline |
API to import an external semaphore in the form of a file descriptor.
context | is the pi_context |
device | is the pi_device |
file_descriptor | is the file descriptor |
ret_handle | is the returned interop semaphore handle to the external semaphore |
Definition at line 473 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextImportExternalSemaphoreOpaqueFD(), and urBindlessImagesImportExternalSemaphoreOpaqueFDExp().
Referenced by sycl::_V1::ext::oneapi::experimental::import_external_semaphore(), and piextImportExternalSemaphoreOpaqueFD().
|
inline |
Creates PI kernel object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI kernel from. |
context | is the PI context of the kernel. |
program | is the PI program of the kernel. |
pluginOwnsNativeHandle | Indicates whether the created PI object should take ownership of the native handle. |
kernel | is the PI kernel created from the native handle. |
Definition at line 541 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextKernelCreateWithNativeHandle(), and urKernelCreateWithNativeHandle().
Referenced by sycl::_V1::detail::make_kernel(), piextKernelCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI kernel object.
kernel | is the PI kernel to get the native handle of. |
nativeHandle | is the native handle of kernel. |
Definition at line 551 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, piextGetNativeHandle(), pi2ur::piextKernelGetNativeHandle(), and urKernelGetNativeHandle().
Referenced by sycl::_V1::detail::kernel_impl::getNative(), piextKernelGetNativeHandle(), and piPluginInit().
|
inline |
Definition at line 347 of file pi_cuda.cpp.
References HANDLE_ERRORS, _pi_mem_obj_property::mem_access, PI_ACCESS_READ_ONLY, PI_ACCESS_READ_WRITE, PI_ACCESS_WRITE_ONLY, PI_ASSERT, PI_KERNEL_ARG_MEM_OBJ_ACCESS, pi2ur::piextKernelSetArgMemObj(), _pi_mem_obj_property::pNext, _pi_mem_obj_property::type, and urKernelSetArgMemObj().
Referenced by piextKernelSetArgMemObj(), piPluginInit(), and sycl::_V1::detail::SetArgBasedOnType().
|
inline |
Sets up pointer arguments for CL kernels.
An extra indirection is required due to CL argument conventions.
kernel | is the kernel to be launched |
arg_index | is the index of the kernel argument |
arg_size | is the size in bytes of the argument (ignored in CL) |
arg_value | is the pointer argument |
Definition at line 868 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clSetKernelArgMemPointerINTELCache, clSetKernelArgMemPointerName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextKernelSetArgPointer(), and urKernelSetArgPointer().
Referenced by piextKernelSetArgPointer(), piPluginInit(), and sycl::_V1::detail::SetArgBasedOnType().
|
inline |
Definition at line 354 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextKernelSetArgSampler(), and urKernelSetArgSampler().
Referenced by piextKernelSetArgSampler(), piPluginInit(), and sycl::_V1::detail::SetArgBasedOnType().
|
inline |
Creates PI mem object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI mem from. |
context | The PI context of the memory allocation. |
ownNativeHandle | Indicates if we own the native memory handle or it came from interop that asked to not transfer the ownership to SYCL RT. |
mem | is the PI mem created from the native handle. |
Definition at line 235 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemCreateWithNativeHandle(), and urMemBufferCreateWithNativeHandle().
Referenced by piextMemCreateWithNativeHandle(), piPluginInit(), and sycl::_V1::detail::SYCLMemObjT::SYCLMemObjT().
|
inline |
Gets the native handle of a PI mem object.
mem | is the PI mem to get the native handle of. |
nativeHandle | is the native handle of mem. |
Definition at line 231 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, piextGetNativeHandle(), pi2ur::piextMemGetNativeHandle(), and urMemGetNativeHandle().
Referenced by sycl::_V1::detail::buffer_impl::getNativeVector(), sycl::_V1::detail::memBufferCreateHelper(), sycl::_V1::detail::memReleaseHelper(), piextMemGetNativeHandle(), and piPluginInit().
|
inline |
API to allocate memory for bindless images.
context | is the pi_context |
device | is the pi_device |
flags | are extra flags to pass (currently unused) |
image_format | format of the image (channel order and data type) |
image_desc | image descriptor |
ret_mem | is the returning memory handle to newly allocated memory |
Definition at line 368 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemImageAllocate(), and urBindlessImagesImageAllocateExp().
Referenced by sycl::_V1::ext::oneapi::experimental::alloc_image_mem(), sycl::_V1::ext::oneapi::experimental::alloc_mipmap_mem(), and piextMemImageAllocate().
|
inline |
API to copy image data Host to Device or Device to Host.
queue | is the queue to submit to |
dst_ptr | is the location the data will be copied to |
src_ptr | is the data to be copied |
image_format | format of the image (channel order and data type) |
image_desc | image descriptor |
flags | flags describing copy direction (H2D or D2H) |
src_offset | is the offset into the source image/memory |
dst_offset | is the offset into the destination image/memory |
copy_extent | is the extent (region) of the image/memory to copy |
host_extent | is the extent (region) of the memory on the host |
num_events_in_wait_list | is the number of events in the wait list |
event_wait_list | is the list of events to wait on before copying |
event | is the returned event representing this operation |
Definition at line 421 of file pi_cuda.cpp.
References pi_image_region_struct::depth, HANDLE_ERRORS, pi_image_region_struct::height, pi2ur::pi2urImageCopyFlags(), pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemImageCopy(), urBindlessImagesImageCopyExp(), pi_image_region_struct::width, pi_image_offset_struct::x, pi_image_offset_struct::y, and pi_image_offset_struct::z.
Referenced by sycl::_V1::detail::MemoryManager::copy_image_bindless(), and piextMemImageCopy().
|
inline |
Creates PI image object from a native handle.
nativeHandle | is the native handle to create PI image from. |
context | The PI context of the memory allocation. |
ownNativeHandle | Indicates if we own the native memory handle or it came from interop that asked to not transfer the ownership to SYCL RT. |
ImageFormat | is the pi_image_format struct that specifies the image channnel order and channel data type that match what the nativeHandle uses |
ImageDesc | is the pi_image_desc struct that specifies the image dimension, pitch, slice and other information about the nativeHandle |
img | is the PI img created from the native handle. |
Definition at line 258 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemImageCreateWithNativeHandle(), and urMemImageCreateWithNativeHandle().
Referenced by piextMemImageCreateWithNativeHandle(), and sycl::_V1::detail::SYCLMemObjT::SYCLMemObjT().
|
inline |
API to free memory for bindless images.
context | is the pi_context |
device | is the pi_device |
memory_handle | is the handle to image memory to be freed |
Definition at line 411 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemImageFree(), and urBindlessImagesImageFreeExp().
Referenced by sycl::_V1::ext::oneapi::experimental::free_image_mem(), and piextMemImageFree().
|
inline |
API to query an image memory handle for specific properties.
mem_handle | is the handle to the image memory |
param_name | is the queried info name |
param_value | is the returned query value |
param_value_size_ret | is the returned query value size |
Definition at line 444 of file pi_cuda.cpp.
References HANDLE_ERRORS, _pi_image_format::image_channel_data_type, _pi_image_format::image_channel_order, pi2ur::pi2urImageInfoFlags(), PI_IMAGE_INFO_FORMAT, pi2ur::piextMemImageGetInfo(), pi2ur::ur2piImageFormat(), and urBindlessImagesImageGetInfoExp().
Referenced by sycl::_V1::ext::oneapi::experimental::get_image_channel_type(), sycl::_V1::ext::oneapi::experimental::get_image_num_channels(), sycl::_V1::ext::oneapi::experimental::get_image_range(), and piextMemImageGetInfo().
|
inline |
API to import external memory in the form of a file descriptor.
context | is the pi_context |
device | is the pi_device |
size | is the size of the external memory |
file_descriptor | is the file descriptor |
ret_handle | is the returned interop memory handle to the external memory |
Definition at line 453 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemImportOpaqueFD(), and urBindlessImagesImportOpaqueFDExp().
Referenced by sycl::_V1::ext::oneapi::experimental::import_external_memory< external_mem_fd >(), and piextMemImportOpaqueFD().
|
inline |
API to map an interop memory handle to an image memory handle.
context | is the pi_context |
device | is the pi_device |
image_format | format of the image (channel order and data type) |
image_desc | image descriptor |
mem_handle | is the interop memory handle to the external memory |
ret_mem | is the returned image memory handle to the externally allocated memory |
Definition at line 459 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemMapExternalArray(), and urBindlessImagesMapExternalArrayExp().
Referenced by sycl::_V1::ext::oneapi::experimental::map_external_memory_array(), and piextMemMapExternalArray().
|
inline |
API to free mipmap memory for bindless images.
context | is the pi_context |
device | is the pi_device |
memory_handle | is the handle to image memory to be freed |
Definition at line 416 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemMipmapFree(), and urBindlessImagesMipmapFreeExp().
Referenced by sycl::_V1::ext::oneapi::experimental::free_image_mem(), sycl::_V1::ext::oneapi::experimental::free_mipmap_mem(), and piextMemMipmapFree().
|
inline |
API to retrieve individual image from mipmap.
context | is the pi_context |
device | is the pi_device |
mip_mem | is the memory handle to the mipmap |
level | is the requested level of the mipmap |
ret_mem | is the returning memory handle to the individual image |
Definition at line 403 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemMipmapGetLevel(), and urBindlessImagesMipmapGetLevelExp().
Referenced by sycl::_V1::ext::oneapi::experimental::get_mip_level_mem_handle(), and piextMemMipmapGetLevel().
|
inline |
API to destroy interop memory.
context | is the pi_context |
device | is the pi_device |
memory_handle | is the handle to interop memory to be freed |
Definition at line 467 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemReleaseInterop(), and urBindlessImagesReleaseInteropExp().
Referenced by piextMemReleaseInterop(), and sycl::_V1::ext::oneapi::experimental::release_external_memory().
|
inline |
API to create sampled bindless image handles.
context | is the pi_context |
device | is the pi_device |
img_mem | is the handle to memory from which to create the image |
image_format | format of the image (channel order and data type) |
image_desc | image descriptor |
sampler | is the pi_sampler |
ret_mem | is the returning pi_mem image object |
ret_handle | is the returning memory handle to newly allocated memory |
Definition at line 385 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemSampledImageCreate(), and urBindlessImagesSampledImageCreateExp().
Referenced by sycl::_V1::ext::oneapi::experimental::create_image(), and piextMemSampledImageCreate().
|
inline |
API to destroy bindless sampled image handles.
context | is the pi_context |
handle | is the image handle |
Definition at line 439 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemSampledImageHandleDestroy(), and urBindlessImagesSampledImageHandleDestroyExp().
Referenced by sycl::_V1::ext::oneapi::experimental::destroy_image_handle(), and piextMemSampledImageHandleDestroy().
|
inline |
API to create bindless image handles.
context | is the pi_context |
device | is the pi_device |
img_mem | is the handle to memory from which to create the image |
image_format | format of the image (channel order and data type) |
image_desc | image descriptor |
ret_mem | is the returning pi_mem image object |
ret_handle | is the returning memory handle to newly allocated memory |
Definition at line 377 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, pi2ur::piextMemUnsampledImageCreate(), and urBindlessImagesUnsampledImageCreateExp().
Referenced by sycl::_V1::ext::oneapi::experimental::create_image(), and piextMemUnsampledImageCreate().
|
inline |
API to destroy bindless unsampled image handles.
context | is the pi_context |
device | is the pi_device |
handle | is the image handle |
Definition at line 434 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextMemUnsampledImageHandleDestroy(), and urBindlessImagesUnsampledImageHandleDestroyExp().
Referenced by sycl::_V1::ext::oneapi::experimental::destroy_image_handle(), and piextMemUnsampledImageHandleDestroy().
|
inline |
Definition at line 1175 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_PEER_ACCESS_SUPPORTED, PI_PEER_ATOMICS_SUPPORTED, pi2ur::piextPeerAccessGetInfo(), and urUsmP2PPeerAccessGetInfoExp().
Referenced by piextPeerAccessGetInfo(), and piPluginInit().
|
inline |
Creates PI platform object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI device from. |
platform | is the PI platform created from the native handle. |
Definition at line 42 of file pi_cuda.cpp.
References PI_ASSERT, pi2ur::piextPlatformCreateWithNativeHandle(), and urPlatformCreateWithNativeHandle().
Referenced by sycl::_V1::detail::make_platform(), piextPlatformCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI platform object.
platform | is the PI platform to get the native handle of. |
nativeHandle | is the native handle of platform. |
Definition at line 37 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, piextGetNativeHandle(), pi2ur::piextPlatformGetNativeHandle(), and urPlatformGetNativeHandle().
Referenced by sycl::_V1::detail::platform_impl::getNative(), piextPlatformGetNativeHandle(), and piPluginInit().
|
inline |
API to get Plugin internal data, opaque to SYCL RT.
Some devices whose device code is compiled by the host compiler (e.g. CPU emulators) may use it to access some device code functionality implemented in/behind the plugin.
opaque_data_param | - unspecified argument, interpretation is specific to a plugin |
opaque_data_return | - placeholder for the returned opaque data. |
Definition at line 1149 of file pi_cuda.cpp.
References pi2ur::piextPluginGetOpaqueData().
Referenced by sycl::_V1::detail::getPluginOpaqueData(), piextPluginGetOpaqueData(), and piPluginInit().
|
inline |
Creates PI program object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI program from. |
context | is the PI context of the program. |
pluginOwnsNativeHandle | Indicates whether the created PI object should take ownership of the native handle. |
program | is the PI program created from the native handle. |
Definition at line 327 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextProgramCreateWithNativeHandle(), and urProgramCreateWithNativeHandle().
Referenced by sycl::_V1::detail::make_kernel_bundle(), piextProgramCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI program object.
program | is the PI program to get the native handle of. |
nativeHandle | is the native handle of program. |
Definition at line 322 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, piextGetNativeHandle(), pi2ur::piextProgramGetNativeHandle(), and urProgramGetNativeHandle().
Referenced by sycl::_V1::detail::device_image_impl::getNative(), sycl::_V1::detail::program_impl::getNative(), piextProgramGetNativeHandle(), and piPluginInit().
|
inline |
Sets a specialization constant to a specific value.
Note: Only used when specialization constants are natively supported (SPIR-V binaries), and not when they are emulated (AOT binaries).
prog | the program object which will use the value |
spec_id | integer ID of the constant |
spec_size | size of the value |
spec_value | bytes of the value |
Definition at line 1028 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clSetProgramSpecializationConstantCache, clSetProgramSpecializationConstantName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextProgramSetSpecializationConstant(), and urProgramSetSpecializationConstants().
Referenced by sycl::_V1::detail::enableITTAnnotationsIfNeeded(), sycl::_V1::detail::program_impl::flush_spec_constants(), piextProgramSetSpecializationConstant(), piPluginInit(), and sycl::_V1::detail::setSpecializationConstants().
|
inline |
properties | points to a zero-terminated array of extra data describing desired queue properties. Format is {[PROPERTY[, property-specific elements of data]*,]* 0} |
Definition at line 162 of file pi_cuda.cpp.
References __SYCL_PI_CUDA_SYNC_WITH_DEFAULT, __SYCL_PI_CUDA_USE_DEFAULT_STREAM, HANDLE_ERRORS, PI_ASSERT, PI_EXT_ONEAPI_QUEUE_FLAG_DISCARD_EVENTS, PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_HIGH, PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_LOW, PI_EXT_QUEUE_FLAG_SUBMISSION_IMMEDIATE, PI_EXT_QUEUE_FLAG_SUBMISSION_NO_IMMEDIATE, PI_QUEUE_COMPUTE_INDEX, PI_QUEUE_FLAG_ON_DEVICE, PI_QUEUE_FLAG_ON_DEVICE_DEFAULT, PI_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE, PI_QUEUE_FLAG_PROFILING_ENABLE, PI_QUEUE_FLAGS, pi2ur::piextQueueCreate(), piQueueCreate(), and urQueueCreate().
Referenced by sycl::_V1::detail::queue_impl::createQueue(), piextQueueCreate(), piPluginInit(), piQueueCreate(), and pi2ur::piQueueCreate().
|
inline |
Creates PI queue object from a native handle.
NOTE: The created PI object takes ownership of the native handle.
nativeHandle | is the native handle to create PI queue from. |
nativeHandleDesc | provides additional properties of the native handle. |
context | is the PI context of the queue. |
device | is the PI device associated with the native device used when creating the native queue. This parameter is optional but some backends may fail to create the right PI queue if omitted. |
pluginOwnsNativeHandle | Indicates whether the created PI object should take ownership of the native handle. |
Properties | holds queue properties. |
queue | is the PI queue created from the native handle. |
Definition at line 193 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_EXT_ONEAPI_QUEUE_FLAG_DISCARD_EVENTS, PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_HIGH, PI_EXT_ONEAPI_QUEUE_FLAG_PRIORITY_LOW, PI_QUEUE_FLAG_ON_DEVICE, PI_QUEUE_FLAG_ON_DEVICE_DEFAULT, PI_QUEUE_FLAG_OUT_OF_ORDER_EXEC_MODE_ENABLE, PI_QUEUE_FLAG_PROFILING_ENABLE, pi2ur::piextQueueCreateWithNativeHandle(), and urQueueCreateWithNativeHandle().
Referenced by sycl::_V1::detail::make_queue(), piextQueueCreateWithNativeHandle(), and piPluginInit().
|
inline |
Gets the native handle of a PI queue object.
queue | is the PI queue to get the native handle of. |
nativeHandle | is the native handle of queue or commandlist. |
nativeHandleDesc | provides additional properties of the native handle. |
Definition at line 185 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, piextGetNativeHandle(), pi2ur::piextQueueGetNativeHandle(), and urQueueGetNativeHandle().
Referenced by sycl::_V1::detail::queue_impl::getNative(), piextQueueGetNativeHandle(), and piPluginInit().
|
inline |
API to instruct the queue to signal the external semaphore handle once all previous commands have completed execution.
command_queue | is the queue instructed to signal |
sem_handle | is the interop semaphore handle to signal |
num_events_in_wait_list | is the number of events in the wait list |
event_wait_list | is the list of events to wait on before this operation |
event | is the returned event representing this operation |
Definition at line 494 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextSignalExternalSemaphore(), and urBindlessImagesSignalExternalSemaphoreExp().
Referenced by piextSignalExternalSemaphore().
|
inline |
Allocates device memory.
result_ptr | contains the allocated memory |
context | is the pi_context |
device | is the device the memory will be allocated on |
properties | are optional allocation properties |
size | is the size of the allocation |
alignment | is the desired alignment of the allocation |
result_ptr | contains the allocated memory |
context | is the pi_context |
device | is the device the memory will be allocated on |
pi_usm_mem_properties | are optional allocation properties |
size_t | is the size of the allocation |
alignment | is the desired alignment of the allocation |
Definition at line 828 of file pi_cuda.cpp.
References sycl::_V1::ext::oneapi::experimental::alignment, sycl::_V1::ext::oneapi::experimental::detail::Alignment, ExtFuncPtrCacheT::clDeviceMemAllocINTELCache, clDeviceMemAllocName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextUSMDeviceAlloc(), and urUSMDeviceAlloc().
Referenced by sycl::_V1::detail::usm::alignedAllocInternal(), piextUSMDeviceAlloc(), and piPluginInit().
|
inline |
USM 2D fill API.
queue | is the queue to submit to |
ptr | is the ptr to fill |
pitch | is the total width of the destination memory including padding |
pattern | is a pointer with the bytes of the pattern to set |
pattern_size | is the size in bytes of the pattern |
width | is width in bytes of each row to fill |
height | is height the columns to fill |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
USM 2D fill API.
queue | is the queue to submit to |
ptr | is the ptr to fill |
pitch | is the total width of the destination memory including padding |
pattern | is a pointer with the bytes of the pattern to set |
pattern_size | is the size in bytes of the pattern |
width | is width in bytes of each row to fill |
height | is height the columns to fill |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
USM 2D fill API.
queue | is the queue to submit to |
ptr | is the ptr to fill |
pattern | is a pointer with the bytes of the pattern to set |
pattern_size | is the size in bytes of the pattern |
pitch | is the total width of the destination memory including padding |
width | is width in bytes of each row to fill |
height | is height the columns to fill |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Definition at line 909 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piextUSMEnqueueFill2D(), and urEnqueueUSMFill2D().
Referenced by sycl::_V1::detail::MemoryManager::fill_2d_usm(), piextUSMEnqueueFill2D(), and piPluginInit().
|
inline |
USM Memadvise API.
queue | is the queue to submit to |
ptr | is the data to be advised |
length | is the size in bytes of the memory to advise |
advice | is device specific advice |
event | is the event that represents this operation |
USM Memadvise API.
Queue | is the queue to submit to |
Ptr | is the data to be advised |
Length | is the size in bytes of the meory to advise |
Advice | is device specific advice |
Event | is the event that represents this operation |
queue | is the queue to submit to |
ptr | is the data to be advised |
length | is the size in bytes of the meory to advise |
advice | is device specific advice |
event | is the event that represents this operation |
Definition at line 902 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_MEM_ADVICE_CUDA_SET_PREFERRED_LOCATION, PI_MEM_ADVICE_CUDA_SET_READ_MOSTLY, PI_MEM_ADVICE_CUDA_UNSET_PREFERRED_LOCATION, PI_MEM_ADVICE_CUDA_UNSET_READ_MOSTLY, PI_MEM_ADVICE_RESET, pi2ur::piextUSMEnqueueMemAdvise(), and urEnqueueUSMAdvise().
Referenced by sycl::_V1::detail::MemoryManager::advise_usm(), piextUSMEnqueueMemAdvise(), and piPluginInit().
|
inline |
USM Memcpy API.
queue | is the queue to submit to |
blocking | is whether this operation should block the host |
src_ptr | is the data to be copied |
dst_ptr | is the location the data will be copied |
size | is number of bytes to copy |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Definition at line 881 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueMemcpyINTELCache, clEnqueueMemcpyName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextUSMEnqueueMemcpy(), and urEnqueueUSMMemcpy().
Referenced by sycl::_V1::detail::MemoryManager::copy_2d_usm(), sycl::_V1::detail::MemoryManager::copy_usm(), piextUSMEnqueueMemcpy(), and piPluginInit().
|
inline |
USM 2D Memcpy API.
queue | is the queue to submit to |
blocking | is whether this operation should block the host |
dst_ptr | is the location the data will be copied |
dst_pitch | is the total width of the destination memory including padding |
src_ptr | is the data to be copied |
src_pitch | is the total width of the source memory including padding |
width | is width in bytes of each row to be copied |
height | is height the columns to be copied |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
queue | is the queue to submit to |
blocking | is whether this operation should block the host |
dst_ptr | is the location the data will be copied |
dst_pitch | is the total width of the destination memory including padding |
src_ptr | is the data to be copied |
dst_pitch | is the total width of the source memory including padding |
width | is width in bytes of each row to be copied |
height | is height the columns to be copied |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Definition at line 934 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMEnqueueMemcpy2D(), and urEnqueueUSMMemcpy2D().
Referenced by sycl::_V1::detail::MemoryManager::copy_2d_usm(), piextUSMEnqueueMemcpy2D(), and piPluginInit().
|
inline |
USM Memset API.
queue | is the queue to submit to |
ptr | is the ptr to memset |
value | is value to set. It is interpreted as an 8-bit value and the upper 24 bits are ignored |
count | is the size in bytes to memset |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Queue | is the queue to submit to |
Ptr | is the ptr to memset |
Value | is value to set. It is interpreted as an 8-bit value and the upper 24 bits are ignored |
Count | is the size in bytes to memset |
NumEventsInWaitlist | is the number of events to wait on |
EventsWaitlist | is an array of events to wait on |
Event | is the event that represents this operation |
Definition at line 873 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clEnqueueMemFillINTELCache, clEnqueueMemFillName, ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMEnqueueMemset(), and urEnqueueUSMFill().
Referenced by sycl::_V1::detail::MemoryManager::fill_usm(), piextUSMEnqueueMemset(), and piPluginInit().
|
inline |
USM 2D Memset API.
queue | is the queue to submit to |
ptr | is the ptr to fill |
pitch | is the total width of the destination memory including padding |
value | the value to fill into the region in |
ptr | |
width | is width in bytes of each row to fill |
height | is height the columns to fill |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
queue | is the queue to submit to |
ptr | is the ptr to fill |
pitch | is the total width of the destination memory including padding |
pattern | is a pointer with the bytes of the pattern to set |
pattern_size | is the size in bytes of the pattern |
width | is width in bytes of each row to fill |
height | is height the columns to fill |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
queue | is the queue to submit to |
ptr | is the ptr to memset |
value | contains the byte to set with |
pitch | is the total width of the destination memory including padding |
width | is width in bytes of each row to memset |
height | is height the columns to memset |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Definition at line 922 of file pi_cuda.cpp.
References die(), and pi2ur::piextUSMEnqueueMemset2D().
Referenced by sycl::_V1::detail::MemoryManager::memset_2d_usm(), piextUSMEnqueueMemset2D(), and piPluginInit().
|
inline |
Hint to migrate memory to the device.
queue | is the queue to submit to |
ptr | points to the memory to migrate |
size | is the number of bytes to migrate |
flags | is a bitfield used to specify memory migration options |
num_events_in_waitlist | is the number of events to wait on |
events_waitlist | is an array of events to wait on |
event | is the event that represents this operation |
Queue | is the queue to submit to |
Ptr | points to the memory to migrate |
Size | is the number of bytes to migrate |
Flags | is a bitfield used to specify memory migration options |
NumEventsInWaitlist | is the number of events to wait on |
EventsWaitlist | is an array of events to wait on |
Event | is the event that represents this operation |
Queue | is the queue to submit to |
Ptr | points to the memory to migrate |
Size | is the number of bytes to migrate |
Flags | is a bitfield used to specify memory migration options |
NumEventsInWaitList | is the number of events to wait on |
EventsWaitList | is an array of events to wait on |
Event | is the event that represents this operation |
Definition at line 892 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMEnqueuePrefetch(), and urEnqueueUSMPrefetch().
Referenced by piextUSMEnqueuePrefetch(), piPluginInit(), and sycl::_V1::detail::MemoryManager::prefetch_usm().
|
inline |
Indicates that the allocated USM memory is no longer needed on the runtime side.
The actual freeing of the memory may be done in a blocking or deferred manner, e.g. to avoid issues with indirect memory access from kernels.
context | is the pi_context of the allocation |
ptr | is the memory to be freed |
Indicates that the allocated USM memory is no longer needed on the runtime side.
context | is the pi_context of the allocation |
ptr | is the memory to be freed |
Definition at line 863 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clMemBlockingFreeINTELCache, clMemBlockingFreeName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextUSMFree(), and urUSMFree().
Referenced by sycl::_V1::detail::usm::freeInternal(), piextUSMFree(), and piPluginInit().
|
inline |
API to query information about USM allocated pointers Valid Queries: PI_MEM_ALLOC_TYPE returns host/device/shared pi_host_usm value PI_MEM_ALLOC_BASE_PTR returns the base ptr of an allocation if the queried pointer fell inside an allocation.
Result must fit in void * PI_MEM_ALLOC_SIZE returns how big the queried pointer's allocation is in bytes. Result is a size_t. PI_MEM_ALLOC_DEVICE returns the pi_device this was allocated against
context | is the pi_context |
ptr | is the pointer to query |
param_name | is the type of query to perform |
param_value_size | is the size of the result in bytes |
param_value | is the result |
param_value_size_ret | is how many bytes were written |
API to query information about USM allocated pointers Valid Queries: PI_MEM_ALLOC_TYPE returns host/device/shared pi_host_usm value PI_MEM_ALLOC_BASE_PTR returns the base ptr of an allocation if the queried pointer fell inside an allocation.
Valid Queries: PI_MEM_ALLOC_TYPE returns host/device/shared pi_usm_type value PI_MEM_ALLOC_BASE_PTR returns the base ptr of an allocation if the queried pointer fell inside an allocation. Result must fit in void * PI_MEM_ALLOC_SIZE returns how big the queried pointer's allocation is in bytes. Result is a size_t. PI_MEM_ALLOC_DEVICE returns the pi_device this was allocated against
Context | is the pi_context |
Ptr | is the pointer to query |
ParamName | is the type of query to perform |
ParamValueSize | is the size of the result in bytes |
ParamValue | is the result |
ParamValueRet | is how many bytes were written |
Result must fit in void * PI_MEM_ALLOC_SIZE returns how big the queried pointer's allocation is in bytes. Result is a size_t. PI_MEM_ALLOC_DEVICE returns the pi_device this was allocated against
context | is the pi_context |
ptr | is the pointer to query |
param_name | is the type of query to perform |
param_value_size | is the size of the result in bytes |
param_value | is the result |
param_value_ret | is how many bytes were written |
Definition at line 945 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clGetMemAllocInfoINTELCache, clGetMemAllocInfoName, die(), ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, PI_MEM_ALLOC_BASE_PTR, PI_MEM_ALLOC_DEVICE, PI_MEM_ALLOC_SIZE, PI_MEM_ALLOC_TYPE, pi2ur::piextUSMGetMemAllocInfo(), ur2piUSMAllocInfoValue(), and urUSMGetMemAllocInfo().
Referenced by sycl::_V1::get_pointer_device(), sycl::_V1::get_pointer_type(), piextUSMGetMemAllocInfo(), and piPluginInit().
|
inline |
Allocates host memory accessible by the device.
result_ptr | contains the allocated memory |
context | is the pi_context |
properties | are optional allocation properties |
size | is the size of the allocation |
alignment | is the desired alignment of the allocation |
result_ptr | contains the allocated memory |
context | is the pi_context |
pi_usm_mem_properties | are optional allocation properties |
size_t | is the size of the allocation |
alignment | is the desired alignment of the allocation |
Definition at line 856 of file pi_cuda.cpp.
References sycl::_V1::ext::oneapi::experimental::alignment, sycl::_V1::ext::oneapi::experimental::detail::Alignment, ExtFuncPtrCacheT::clHostMemAllocINTELCache, clHostMemAllocName, ExtFuncPtrCache, HANDLE_ERRORS, pi2ur::piextUSMHostAlloc(), and urUSMHostAlloc().
Referenced by sycl::_V1::detail::usm::alignedAllocHost(), piextUSMHostAlloc(), and piPluginInit().
|
inline |
Import host system memory into USM.
ptr | start address of memory range to import |
size | is the number of bytes to import |
context | is the pi_context |
Definition at line 953 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMImport(), and urUSMImportExp().
Referenced by piextUSMImport(), piPluginInit(), and sycl::_V1::prepare_for_usm_device_copy().
|
inline |
Allocates memory accessible on device.
result_ptr | contains the allocated memory |
result_pitch | contains the returned memory pitch |
context | is the pi_context |
device | is the device the memory will be allocated on |
properties | are optional allocation properties |
width_in_bytes | is the width of the allocation in bytes |
height | is the height of the allocation in rows |
element_size_bytes | is the size in bytes of an element in the allocation |
Definition at line 846 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMPitchedAlloc(), and urUSMPitchedAllocExp().
Referenced by piextUSMPitchedAlloc(), and sycl::_V1::ext::oneapi::experimental::pitched_alloc_device().
|
inline |
Release host system memory from USM.
ptr | start address of imported memory range |
context | is the pi_context |
Definition at line 957 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextUSMRelease(), and urUSMReleaseExp().
Referenced by piextUSMRelease(), piPluginInit(), and sycl::_V1::release_from_usm_device_copy().
|
inline |
Allocates memory accessible on both host and device.
result_ptr | contains the allocated memory |
context | is the pi_context |
device | is the device the memory will be allocated on |
properties | are optional allocation properties |
size | is the size of the allocation |
alignment | is the desired alignment of the allocation |
result_ptr | contains the allocated memory |
context | is the pi_context |
device | is the device the memory will be allocated on |
pi_usm_mem_properties | are optional allocation properties |
size_t | is the size of the allocation |
alignment | is the desired alignment of the allocation |
Definition at line 837 of file pi_cuda.cpp.
References sycl::_V1::ext::oneapi::experimental::alignment, sycl::_V1::ext::oneapi::experimental::detail::Alignment, ExtFuncPtrCacheT::clSharedMemAllocINTELCache, clSharedMemAllocName, ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, PI_MEM_ALLOC_DEVICE_READ_ONLY, PI_MEM_ALLOC_FLAGS, PI_MEM_ALLOC_INITIAL_PLACEMENT_DEVICE, PI_MEM_ALLOC_INITIAL_PLACEMENT_HOST, PI_MEM_ALLOC_WRTITE_COMBINED, pi2ur::piextUSMSharedAlloc(), and urUSMSharedAlloc().
Referenced by sycl::_V1::detail::usm::alignedAllocInternal(), piextUSMSharedAlloc(), and piPluginInit().
|
inline |
API to instruct the queue with a non-blocking wait on an external semaphore.
command_queue | is the queue instructed to wait |
sem_handle | is the interop semaphore handle |
num_events_in_wait_list | is the number of events in the wait list |
event_wait_list | is the list of events to wait on before this operation |
event | is the returned event representing this operation |
Definition at line 486 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piextWaitExternalSemaphore(), and urBindlessImagesWaitExternalSemaphoreExp().
Referenced by piextWaitExternalSemaphore().
|
inline |
Queries device for it's global timestamp in nanoseconds, and updates HostTime with the value of the host timer at the closest possible point in time to that at which DeviceTime was returned.
Device | device to query for timestamp |
DeviceTime | pointer to store device timestamp in nanoseconds. Optional argument, can be nullptr |
HostTime | pointer to store host timestamp in nanoseconds. Optional argurment, can be nullptr in which case timestamp will not be written |
Definition at line 1158 of file pi_cuda.cpp.
References getDeviceVersion(), getPlatformVersion(), HANDLE_ERRORS, pi2ur::piGetDeviceAndHostTimer(), setErrorMessage(), urDeviceGetGlobalTimestamps(), and OCLV::V2_1.
Referenced by sycl::_V1::detail::device_impl::getCurrentDeviceTime(), sycl::_V1::detail::device_impl::isGetDeviceAndHostTimerSupported(), piGetDeviceAndHostTimer(), and piPluginInit().
|
inline |
Definition at line 335 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piKernelCreate(), and urKernelCreate().
Referenced by sycl::_V1::detail::ProgramManager::getOrCreateKernel(), piKernelCreate(), and piPluginInit().
|
inline |
Definition at line 502 of file pi_cuda.cpp.
References die(), HANDLE_ERRORS, PI_ASSERT, PI_KERNEL_GROUP_INFO_COMPILE_WORK_GROUP_SIZE, PI_KERNEL_GROUP_INFO_GLOBAL_WORK_SIZE, PI_KERNEL_GROUP_INFO_LOCAL_MEM_SIZE, PI_KERNEL_GROUP_INFO_NUM_REGS, PI_KERNEL_GROUP_INFO_PREFERRED_WORK_GROUP_SIZE_MULTIPLE, PI_KERNEL_GROUP_INFO_PRIVATE_MEM_SIZE, PI_KERNEL_GROUP_INFO_WORK_GROUP_SIZE, pi2ur::piKernelGetGroupInfo(), urKernelGetGroupInfo(), and urKernelGetInfo().
Referenced by sycl::_V1::detail::enqueueImpCommandBufferKernel(), sycl::_V1::detail::get_kernel_device_specific_info_helper(), sycl::_V1::detail::enqueue_kernel_launch::handleInvalidWorkGroupSize(), piKernelGetGroupInfo(), piPluginInit(), and sycl::_V1::detail::SetKernelParamsAndLaunch().
|
inline |
Definition at line 360 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_KERNEL_INFO_ATTRIBUTES, PI_KERNEL_INFO_CONTEXT, PI_KERNEL_INFO_FUNCTION_NAME, PI_KERNEL_INFO_NUM_ARGS, PI_KERNEL_INFO_PROGRAM, PI_KERNEL_INFO_REFERENCE_COUNT, pi2ur::piKernelGetInfo(), and urKernelGetInfo().
Referenced by sycl::_V1::detail::get_kernel_info(), sycl::_V1::detail::ProgramManager::getPiProgramFromPiKernel(), sycl::_V1::detail::enqueue_kernel_launch::handleInvalidWorkGroupSize(), piKernelGetInfo(), and piPluginInit().
|
inline |
API to query information from the sub-group from a kernel.
kernel | is the pi_kernel to query |
device | is the device the kernel is executed on |
param_name | is a pi_kernel_sub_group_info enum value that specifies the informtation queried for. |
input_value_size | is the size of input value passed in ptr input_value param |
input_value | is the ptr to the input value passed. |
param_value_size | is the size of the value in bytes. |
param_value | is a pointer to the value to set. |
param_value_size_ret | is a pointer to return the size of data in param_value ptr. |
All queries expect a return of 4 bytes in param_value_size, param_value_size_ret, and a uint32_t value should to be written in param_value ptr. Note: This behaviour differs from OpenCL. OpenCL returns size_t.
Definition at line 510 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_DEVICE_INFO_MAX_WORK_ITEM_DIMENSIONS, PI_DEVICE_INFO_MAX_WORK_ITEM_SIZES, PI_DEVICE_INFO_SUB_GROUP_SIZES_INTEL, PI_KERNEL_COMPILE_NUM_SUB_GROUPS, PI_KERNEL_COMPILE_SUB_GROUP_SIZE_INTEL, PI_KERNEL_MAX_NUM_SUB_GROUPS, PI_KERNEL_MAX_SUB_GROUP_SIZE, piDeviceGetInfo(), pi2ur::piKernelGetSubGroupInfo(), and urKernelGetSubGroupInfo().
Referenced by sycl::_V1::detail::get_kernel_device_specific_info_helper(), sycl::_V1::detail::get_kernel_device_specific_info_with_input(), piKernelGetSubGroupInfo(), and piPluginInit().
Definition at line 525 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piKernelRelease(), and urKernelRelease().
Referenced by piKernelRelease(), piPluginInit(), sycl::_V1::detail::kernel_impl::~kernel_impl(), and sycl::_V1::detail::KernelProgramCache::~KernelProgramCache().
Definition at line 520 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piKernelRetain(), and urKernelRetain().
Referenced by sycl::_V1::detail::kernel_impl::get(), sycl::_V1::detail::kernel_impl::getNative(), sycl::_V1::detail::make_kernel(), piKernelRetain(), and piPluginInit().
|
inline |
Definition at line 341 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piKernelSetArg(), urKernelSetArgLocal(), and urKernelSetArgValue().
Referenced by piKernelSetArg(), piPluginInit(), and sycl::_V1::detail::SetArgBasedOnType().
|
inline |
API to set attributes controlling kernel execution.
kernel | is the pi kernel to execute |
param_name | is a pi_kernel_exec_info value that specifies the info passed to the kernel |
param_value_size | is the size of the value in bytes |
param_value | is a pointer to the value to set for the kernel |
If param_name is PI_USM_INDIRECT_ACCESS, the value will be a ptr to the pi_bool value PI_TRUE If param_name is PI_USM_PTRS, the value will be an array of ptrs
Definition at line 1021 of file pi_cuda.cpp.
References die(), HANDLE_ERRORS, PI_ASSERT, PI_EXT_KERNEL_EXEC_INFO_CACHE_CONFIG, PI_EXT_KERNEL_EXEC_INFO_CACHE_DEFAULT, PI_EXT_KERNEL_EXEC_INFO_CACHE_LARGE_DATA, PI_EXT_KERNEL_EXEC_INFO_CACHE_LARGE_SLM, PI_TRUE, PI_USM_INDIRECT_ACCESS, PI_USM_PTRS, pi2ur::piKernelSetExecInfo(), urKernelSetExecInfo(), and USMSetIndirectAccess().
Referenced by sycl::_V1::detail::enqueueImpKernel(), sycl::_V1::detail::ProgramManager::getOrCreateKernel(), sycl::_V1::detail::kernel_impl::kernel_impl(), piKernelSetExecInfo(), and piPluginInit().
|
inline |
Definition at line 205 of file pi_cuda.cpp.
References ExtFuncPtrCacheT::clCreateBufferWithPropertiesINTELCache, clCreateBufferWithPropertiesName, die(), ExtFuncPtrCache, HANDLE_ERRORS, PI_ASSERT, PI_MEM_ACCESS_READ_ONLY, PI_MEM_FLAGS_ACCESS_RW, PI_MEM_FLAGS_HOST_PTR_ALLOC, PI_MEM_FLAGS_HOST_PTR_COPY, PI_MEM_FLAGS_HOST_PTR_USE, pi2ur::piMemBufferCreate(), and urMemBufferCreate().
Referenced by sycl::_V1::detail::memBufferCreateHelper(), piMemBufferCreate(), and piPluginInit().
|
inline |
Definition at line 813 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, PI_BUFFER_CREATE_TYPE_REGION, PI_MEM_ACCESS_READ_ONLY, PI_MEM_FLAGS_ACCESS_RW, PI_MEM_FLAGS_HOST_PTR_ALLOC, PI_MEM_FLAGS_HOST_PTR_COPY, PI_MEM_FLAGS_HOST_PTR_USE, pi2ur::piMemBufferPartition(), and urMemBufferPartition().
Referenced by sycl::_V1::detail::MemoryManager::allocateMemSubBuffer(), piMemBufferPartition(), and piPluginInit().
|
inline |
Definition at line 212 of file pi_cuda.cpp.
References die(), HANDLE_ERRORS, PI_ASSERT, PI_MEM_CONTEXT, PI_MEM_SIZE, pi2ur::piMemGetInfo(), and urMemGetInfo().
Referenced by sycl::_V1::detail::SYCLMemObjT::getBufSizeForContext(), piMemGetInfo(), piPluginInit(), and sycl::_V1::detail::SYCLMemObjT::SYCLMemObjT().
|
inline |
Definition at line 222 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::pi2urImageDesc(), PI_ASSERT, PI_MEM_ACCESS_READ_ONLY, PI_MEM_FLAGS_ACCESS_RW, PI_MEM_FLAGS_HOST_PTR_ALLOC, PI_MEM_FLAGS_HOST_PTR_COPY, PI_MEM_FLAGS_HOST_PTR_USE, pi2ur::piMemImageCreate(), and urMemImageCreate().
Referenced by sycl::_V1::detail::MemoryManager::allocateImageObject(), piMemImageCreate(), and piPluginInit().
|
inline |
Definition at line 758 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_IMAGE_INFO_DEPTH, PI_IMAGE_INFO_ELEMENT_SIZE, PI_IMAGE_INFO_FORMAT, PI_IMAGE_INFO_HEIGHT, PI_IMAGE_INFO_ROW_PITCH, PI_IMAGE_INFO_SLICE_PITCH, PI_IMAGE_INFO_WIDTH, pi2ur::piMemImageGetInfo(), and urMemImageGetInfo().
Referenced by sycl::_V1::detail::getImageInfo(), piMemImageGetInfo(), and piPluginInit().
Definition at line 220 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piMemRelease(), and urMemRelease().
Referenced by sycl::_V1::detail::memReleaseHelper(), piMemRelease(), piPluginInit(), and sycl::_V1::detail::SYCLMemObjT::updateHostMemory().
Definition at line 218 of file pi_cuda.cpp.
References HANDLE_ERRORS, PI_ASSERT, pi2ur::piMemRetain(), and urMemRetain().
Referenced by sycl::_V1::detail::buffer_impl::addInteropObject(), sycl::_V1::detail::buffer_impl::getNativeVector(), piMemRetain(), piPluginInit(), and sycl::_V1::detail::SYCLMemObjT::SYCLMemObjT().
|
inline |
Definition at line 30 of file pi_cuda.cpp.
References die(), fixupInfoValueTypes(), HANDLE_ERRORS, sycl::_V1::detail::memcpy(), PI_ASSERT, PI_EXT_PLATFORM_BACKEND_OPENCL, PI_EXT_PLATFORM_INFO_BACKEND, PI_PLATFORM_INFO_EXTENSIONS, PI_PLATFORM_INFO_NAME, PI_PLATFORM_INFO_PROFILE, PI_PLATFORM_INFO_VENDOR, PI_PLATFORM_INFO_VERSION, pi2ur::piPlatformGetInfo(), ur2piPlatformInfoValue(), and urPlatformGetInfo().
Referenced by sycl::_V1::detail::get_platform_info_string_impl(), sycl::_V1::opencl::has_extension(), piPlatformGetInfo(), piPluginInit(), and sycl::_V1::detail::platform_impl::platform_impl().
|
inline |
Definition at line 25 of file pi_cuda.cpp.
References Adapter, Adapters, HANDLE_ERRORS, pi2ur::piPlatformsGet(), urAdapterGet(), urInit(), urPlatformGet(), and AdapterHolder::Vec.
Referenced by sycl::_V1::detail::platform_impl::get_platforms(), piPlatformsGet(), and piPluginInit().
|
inline |
API to get backend specific option.
frontend_option | is a string that contains frontend option. |
backend_option | is used to return the backend option corresponding to frontend option. |
Definition at line 51 of file pi_cuda.cpp.
References HANDLE_ERRORS, pi2ur::piPluginGetBackendOption(), and urPlatformGetBackendOption().
Referenced by sycl::_V1::detail::platform_impl::getBackendOption(), piPluginGetBackendOption(), and piPluginInit().
|
inline |
API to get Plugin specific warning and error messages.
message | is a returned address to the first element in the message the plugin owns the error message string. The string is thread-local. As a result, different threads may return different errors. A message is overwritten by the following error or warning that is produced within the given thread. The memory is cleaned up at the end of the thread's lifetime. |
device,and | syncronized host timestamp |
Definition at line 47 of file pi_cuda.cpp.
References Adapter, ErrorMessage, ErrorMessageCode, pi2ur::piPluginGetLastError(), ur2piResult(), urAdapterGet(), urAdapterGetLastError(), and urAdapterRelease().
Referenced by sycl::_V1::detail::device_impl::getCurrentDeviceTime(), piPluginGetLastError(), and piPluginInit().
Definition at line 1186 of file pi_cuda.cpp.
References _PI_API, _PI_CL, _PI_LEVEL_ZERO_PLUGIN_VERSION_STRING, _PI_PLUGIN_VERSION_CHECK, _PI_UNIFIED_RUNTIME_PLUGIN_VERSION_STRING, Adapters, enableCUDATracing(), enableZeTracing(), HANDLE_ERRORS, syclcompat::detail::memset(), PI_ASSERT, piContextCreate(), piContextGetInfo(), piContextRelease(), piContextRetain(), piDeviceGetInfo(), piDevicePartition(), piDeviceRelease(), piDeviceRetain(), piDevicesGet(), piEnqueueEventsWait(), piEnqueueEventsWaitWithBarrier(), piEnqueueKernelLaunch(), piEnqueueMemBufferCopy(), piEnqueueMemBufferCopyRect(), piEnqueueMemBufferFill(), piEnqueueMemBufferMap(), piEnqueueMemBufferRead(), piEnqueueMemBufferReadRect(), piEnqueueMemBufferWrite(), piEnqueueMemBufferWriteRect(), piEnqueueMemImageCopy(), piEnqueueMemImageFill(), piEnqueueMemImageRead(), piEnqueueMemImageWrite(), piEnqueueMemUnmap(), piEventCreate(), piEventGetInfo(), piEventGetProfilingInfo(), piEventRelease(), piEventRetain(), piEventSetCallback(), piEventSetStatus(), piEventsWait(), piextCommandBufferCreate(), piextCommandBufferMemBufferCopy(), piextCommandBufferMemBufferCopyRect(), piextCommandBufferMemBufferRead(), piextCommandBufferMemBufferReadRect(), piextCommandBufferMemBufferWrite(), piextCommandBufferMemBufferWriteRect(), piextCommandBufferMemcpyUSM(), piextCommandBufferNDRangeKernel(), piextCommandBufferRelease(), piextCommandBufferRetain(), piextContextCreateWithNativeHandle(), piextContextGetNativeHandle(), piextContextSetExtendedDeleter(), piextDeviceCreateWithNativeHandle(), piextDeviceGetNativeHandle(), piextDeviceSelectBinary(), piextDisablePeerAccess(), piextEnablePeerAccess(), piextEnqueueCommandBuffer(), piextEnqueueDeviceGlobalVariableRead(), piextEnqueueDeviceGlobalVariableWrite(), piextEnqueueReadHostPipe(), piextEnqueueWriteHostPipe(), piextEventCreateWithNativeHandle(), piextEventGetNativeHandle(), piextGetDeviceFunctionPointer(), piextGetNativeHandle(), piextKernelCreateWithNativeHandle(), piextKernelGetNativeHandle(), piextKernelSetArgMemObj(), piextKernelSetArgPointer(), piextKernelSetArgSampler(), piextMemCreateWithNativeHandle(), piextMemGetNativeHandle(), piextPeerAccessGetInfo(), piextPlatformCreateWithNativeHandle(), piextPlatformGetNativeHandle(), piextPluginGetOpaqueData(), piextProgramCreateWithNativeHandle(), piextProgramGetNativeHandle(), piextProgramSetSpecializationConstant(), piextQueueCreate(), piextQueueCreateWithNativeHandle(), piextQueueGetNativeHandle(), piextUSMDeviceAlloc(), piextUSMEnqueueFill2D(), piextUSMEnqueueMemAdvise(), piextUSMEnqueueMemcpy(), piextUSMEnqueueMemcpy2D(), piextUSMEnqueueMemset(), piextUSMEnqueueMemset2D(), piextUSMEnqueuePrefetch(), piextUSMFree(), piextUSMGetMemAllocInfo(), piextUSMHostAlloc(), piextUSMImport(), piextUSMRelease(), piextUSMSharedAlloc(), _pi_plugin::PiFunctionTable, piGetDeviceAndHostTimer(), piKernelCreate(), piKernelGetGroupInfo(), piKernelGetInfo(), piKernelGetSubGroupInfo(), piKernelRelease(), piKernelRetain(), piKernelSetArg(), piKernelSetExecInfo(), piMemBufferCreate(), piMemBufferPartition(), piMemGetInfo(), piMemImageCreate(), piMemImageGetInfo(), piMemRelease(), piMemRetain(), piPlatformGetInfo(), piPlatformsGet(), piPluginGetBackendOption(), piPluginGetLastError(), piProgramBuild(), piProgramCompile(), piProgramCreate(), piProgramCreateWithBinary(), piProgramGetBuildInfo(), piProgramGetInfo(), piProgramLink(), piProgramRelease(), piProgramRetain(), piQueueCreate(), piQueueFinish(), piQueueFlush(), piQueueGetInfo(), piQueueRelease(), piQueueRetain(), piSamplerCreate(), piSamplerGetInfo(), piSamplerRelease(), piSamplerRetain(), piTearDown(), _pi_plugin::PiVersion, _pi_plugin::PluginVersion, SupportedVersion, urAdapterGet(), urInit(), and AdapterHolder::Vec.
Referenced by sycl::_V1::detail::pi::bindPlugin().
pi_result piProgramBuild | ( | pi_program | program, |
pi_uint32 | num_devices, | ||
const pi_device * | device_list, | ||
const char * | options, | ||
void(*)(pi_program program, void *user_data) | pfn_notify, | ||
void * | user_data | ||
) |
Referenced by sycl::_V1::detail::make_kernel_bundle(), and piPluginInit().
pi_result piProgramCompile | ( | pi_program | program, |
pi_uint32 | num_devices, | ||
const pi_device * | device_list, | ||
const char * | options, | ||
pi_uint32 | num_input_headers, | ||
const pi_program * | input_headers, | ||
const char ** | header_include_names, | ||
void(*)(pi_program program, void *user_data) | pfn_notify, | ||
void * | user_data | ||
) |
|
inline |
Definition at line 242 of file pi_cuda.cpp.
References CHECK_ERR_SET_NULL_RET, checkDeviceExtensions(), getDeviceVersion(), getPlatformVersion(), HANDLE_ERRORS, PI_ASSERT, pi2ur::piProgramCreate(), urProgramCreateWithIL(), and OCLV::V2_1.
Referenced by sycl::_V1::detail::createSpirvProgram(), piPluginInit(), and piProgramCreate().
|
inline |
Crea