dpctl.utils Functions

Other Functions

dpctl.utils.intel_device_info(sycl_device)[source]

For Intel(R) GPU devices returns a dictionary with device architectural details, and an empty dictionary otherwise. The dictionary contains the following keys:

device_id: 32-bits device PCI identifier gpu_eu_count: Total number of execution units gpu_hw_threads_per_eu: Number of thread contexts in EU gpu_eu_simd_width: Physical SIMD width of EU gpu_slices: Total number of slices gpu_subslices_per_slice: Number of sub-slices per slice gpu_eu_count_per_subslice: Number of EUs in subslice max_mem_bandwidth: Maximum memory bandwidth in bytes/second

Unsupported descriptors are omitted from the dictionary. Descriptors other than PCI identifier are supported only for SyclDevices with Leve-Zero backend.

dpctl.utils.intel_device_info_device_id(device: dpctl.SyclDevice) → int: Get ext_intel_device_id for the device, zero if not an intel device

dpctl.utils.intel_device_info_gpu_eu_count(device: dpctl.SyclDevice) → int: Returns the number of execution units (EUs) associated with the Intel GPU.

dpctl.utils.intel_device_info_gpu_eu_count_per_subslice(device: dpctl.SyclDevice) → int: Returns the number of EUs per subslice of GPU.

dpctl.utils.intel_device_info_gpu_eu_simd_width(device: dpctl.SyclDevice) → int: Returns the physical SIMD width of the execution unit (EU).

dpctl.utils.intel_device_info_gpu_hw_threads_per_eu(device: dpctl.SyclDevice) → int: Returns the number of hardware threads in EU.

dpctl.utils.intel_device_info_gpu_slices(device: dpctl.SyclDevice) → int: Returns the number of slices in the GPU device, or zero.

dpctl.utils.intel_device_info_gpu_subslices_per_slice(device: dpctl.SyclDevice) → int: Returns the number of subslices per slice.

dpctl.utils.intel_device_info_max_mem_bandwidth(device: dpctl.SyclDevice) → int: Returns the maximum memory bandwidth in units of bytes/second.

dpctl.utils.onetrace_enabled()[source]

Enable onetrace collection for kernels executed in this context.

N.B.: Proper working of this utility assumes that Python interpreter has been launched by onetrace tool from intel/pti-gpu project.

Example:

Launch the Python interpreter using onetrace tool:

$ onetrace --conditional-collection -v -t --demangle python app.py

Now using the context manager in the Python sessions enables data collection and its output for every offloaded kernel

import dpctl.tensor as dpt
from dpctl.utils import onetrace_enabled

# onetrace output reporting on execution of the kernel
# should be seen, starting with "Device Timeline"
with onetrace_enabled():
    dpt.arange(100, dtype='int16')