an embedded high-precision performance profiler for Clojure. clj-async-profiler
presents the profiling results as an interactive
flamegraph. You can navigate the
flamegraph, query it, change parameters and adapt the results for easier
clj-async-profiler is based upon
async-profiler which is
a low overhead sampling profiler for Java.
Why use clj-async-profiler?
Compared to other profiling tools, clj-async-profiler has several pronounced
strong points to make it your primary go-to profiler.
High precision. Historically, many Java profilers suffered from an issue
bias. It meant that
they produced skewed and incorrect results when profiling certain code, and
hence you would focus on optimizing already fast parts of your programs while
the actual slow parts remained slow to the profiler. clj-async-profiler
doesn't have this problem.
Low overhead. With its single-digit percentage overhead,
clj-async-profiler is suitable to be used on production servers to obtain the
performance profile in real production scenarios.
Embeddable. Adding the profiler is as easy as appending a single
dependency. It is convenient both during development and when profiling in
Programmatically controllable. Again, in production scenarios, being able
to drive the profiler from your application code enables multiple interesting
usage patterns, like starting the profiler on schedule, or as a reaction to
Convenient presentation. Flamegraphs as profile representation are very
descriptive and demonstrative. The underlying raw data is in plain text which
makes it malleable to extra processing and transformation.
Not just CPU profiling. clj-async-profiling also supports profiling of
allocations (shows which parts of you code allocate the most
objects/memory), locks, context switches, and other events supported by your