Products & Services
|
Intel® Trace Collector
Maximize Efficient Application Usage
Intel Trace Collector is a low-overhead tracing library that performs
event-based tracing in applications. You can analyze the collected trace data
for performance hotspots and bottlenecks. The product is completely thread safe
and integrates with C++, Java and multi-threaded processes with and without MPI.
Binary instrumentation and fail-safe mode are important other new features.
Features and Benefits
-
Automatically instrument MPI and routine entry/exit points. The traditional
application programming interface (API) is also available to instrument
application source code.
-
Trace concurrent behavior of parallel applications and calculate statistics
on demand for specific time intervals and processes.
-
Reduce runtime overhead through scalability and memory handling. Trace data
is cached in memory.
-
Monitor the application runtime with the Intel Trace Collector or use it in a
purely statistical mode with counters. It calculates statistics regarding
function calls, messages, and collective operations at runtime and stores them
without event data.
-
Capture application statistics displays with trace data.
-
Be assured of completely thread-safe tracing of multi-threaded MPI
applications, event-based tracing to non-MPI applications supporting Java
processes, and multi-threaded processes without MPI.
-
Automatically monitor function entry/exit points through binary
instrumentation of the executables. This leads to a detailed program analysis of
your code and application runtime. In addition, with instrumentation
capabilities you can trace C++ and Java class hierarchies.
-
Use the extended APIs to trace other programming models and libraries.
-
Record event trace data in highly-scalable Structured Trace Format (STF).
|
|
