Accurate Bytecode-level Profiling of Dynamically Optimized Code with Graal
Many profilers based on bytecode instrumentation yield wrong results in the presence of an optimizing dynamic compiler, either due to not being aware of optimizations such as stack allocation and method inlining, or due to the inserted code disrupting such optimizations. To solve these problems, we make the dynamic compiler Graal aware of the inserted profiling code and extend Graal with a set of query intrinsics for retrieving its optimization decisions within inserted profiling code. We demonstrate the technique with several concrete profilers, illustrating that it may enhance existing profiling techniques, enable new kinds of profilers, or even go beyond profiling.
Ph.D. student at University of Lugano
Tue 7 Jul
|13:50 - 14:15|
|14:15 - 14:40|
|14:40 - 15:05|
|15:05 - 15:30|