Augmenting profile data with information gathered from a JIT compiler

摘要

A method, executed by a computer, for augmenting a first performance profile with data extracted from a Just-in-Time compiler, the Just-in-Time compiler compiling bytecodes into machine instructions and generating the first performance profile, the bytecodes having an associated original call structure includes: tracking “in-lining” optimizations performed by a Just-in-Time compiler compiling bytecodes into machine instructions; extracting data associated with the tracked “in-lining” optimizations; storing the extracted data in a second profile; and augmenting the first performance profile with the extracted data associated with the tracked “in-lining” optimizations, the extracted data comprising call paths corresponding to the original call structure associated with the bytecodes. A corresponding computer program product and computer system are also disclosed herein.

主权项

1. A method, executed by a computer, for augmenting a first performance profile with data extracted from a Just-in-Time compiler, the Just-in-Time compiler compiling bytecodes into machine instructions and generating the first performance profile, the bytecodes having an associated original call structure, the method comprising:
tracking “in-lining” optimizations that collapse the associated original call structure as performed by a Just-in-Time compiler compiling bytecodes into machine instructions; extracting data associated with the tracked “in-lining” optimizations that collapse the associated original call structure to provide extracted data that specifies functions within the associated original call structure that are collapsed by the tracked “in-lining” optimizations; storing the extracted data that specifies functions within the associated original call structure that are collapsed by the tracked “in-lining” optimizations in a second performance profile; and augmenting the first performance profile with the extracted data that specifies functions within the associated original call structure that are collapsed by the tracked “in-lining” optimizations.