Deep Neural Network Library (DNNL)  1.1.0
Performance library for Deep Learning
Profiling with Intel(R) VTune Amplifier

DNNL uses just-in-time compilation (JIT) to generate optimal code for some functions based on input parameters and instruction set supported by the system. The library uses Intel SEAPI interface to annotate the generated code for Intel VTune Amplifier, so that it can correctly attribute processor monitoring unit (PMU) events to the code.

The behavior is controlled with build time option DNNL_ENABLE_JIT_PROFILING

Option Possible Values (defaults in bold) Description
DNNL_ENABLE_JIT_PROFILING ON, OFF Enables integration with Intel(R) VTune(TM) Amplifier