what are the numbers in the operation names when profiling an application
What are the numbers in "fusion_2", "fusion_4"? Where do they come from? Thank you!
Solution
These numbers exist to de-duplicate the names of generated HLO operations. The first fusion operation created by the compiler is called fusion, the next is fusion_2, then fusion_3, and so on.
Note that the order of creation does not necessarily match the order of execution.
- How to trigger XDebug profiler for a command line PHP script?
- Profile Generate vs PG
- Chrome timeline - how can I determine the cause of a "Recalculate Style" log entry?
- How to measure time taken by a function to execute
- gcc: undefined reference to _mcount (gprof instrumentation)
- How do I analyze a .hprof file?
- Android Methods calling tracking
- Does the Native VBA Compiler Optimize?
- VisualVM launcher error
- How to debug a java app that uses a JVMTI agent to set breakpoint (Unable to get necessary JVMTI capabilities)
- C++ Threadpool: Why doesn't it scale perfectly on Android & Arm Tablet
- Are there known false positives issues with Valgrind?
- How do I measure the execution time of JavaScript code with callbacks?
- Which eslint rules in my config are slow?
- dotMemory The resource loader failed to find MUI file. When try to profile local IIS
- Is is possible to use Profiling API right from C#?
- VisualVM socket.read
- Profiling jetty with visualvm is super slow
- Problems with using gperftools on Mac OS X
- How to determine a number of queries SQLAlchemy does during user request?
- How do I profile memory usage in Python?
- Javascript memory profiler for Firefox
- How do I profile a Python script?
- Profiling a running Java application in command line
- Is there a way to profile my Django application?
- What should I check: cpu time or wall time?
- Profiling c++ on mac os x
- Get variable stored at specific memory address in Java
- How to use linux `perf` tool to generate "Off-CPU" profile
- Unknown symbols in perf report output