After 10 years as a Java developer, Doug transitioned to working on Azul's Java Virtual Machine.
Today, Doug continues his interest in building performance tools for developers as the
Lead Developer of Datadog's Java Application Performance Monitoring.
While Doug's passion for developing software remains, his true passion is in sharing his
interest in low-level details and JVM performance with others.
Over the years, Java developers have learned through trial-and-error the best ways to use Java's type system. But certain parts of the type system like wildcard generics and covariant return types are
under used; in-part, because they are not well understood.
Fortunately by going back to the mathematical roots of type systems, we can understand not only wildcard types and covariance, but also why we should prefer composition over inheritance and even how compilers perform some of their optimization magic.
Just like you need to understand execution plans to write efficient SQL, it helps to understand the JVM's execution engine to write efficient Java code.
In this talk, you'll learn how Java's Just-in-Time (JIT) compiler dissects your code, and by doing so you'll learn how to write efficient code more efficiently, but also gain a better understanding of concurrency challenges as well.
Early releases of Java performed poorly, but those issues largely disappeared long ago with the introduction of HotSpot. However, much of the performance advice for Java persists through hearsay from those early days.
In this talk, we'll forget the hearsay and take an objective look using benchmarking and profiling tools to find out which optimizations matter today and just as importantly those that don't.
Most of us don't want to go back to the days of malloc and free, but the garbage collector isn't always our friend.
In this presentation, you'll learn about the different garbage collection strategies used in JVMs, how to monitor garbage collection, analyze memory dumps, and why you might want to use one collection strategy instead of another.
HotSpot promises to do wonders for us by Just-in-Time (JIT) compiling the “right” code for us, but how does it makes those decisions? And, perhaps more importantly, what happens when it's wrong?
In this talk, you'll learn through real code examples just how the JVM decides to compile your code, deoptimize your code, and stop-the-world for a GC.
Unlike earlier languages, Java had a well-defined threading and memory model from its inception, and over the years additional packages have been added to the standard library to help solve concurrency problems.
Despite all this, Java's memory model still remains poorly understood and
its concurrency libraries under utilized.
In this talk, we'll discuss concurrency in detail starting from the low-level
of the hardware up to the high-level of concurrent collections and Java 7's
new parallel fork join library.
Early releases of Java performed poorly, but those issues largely disappeared long ago with the introduction of HotSpot. However, much of the performance advice for Java persists through hearsay from those early days.
In this talk, we'll forget the hearsay and take an objective look using benchmarking and profiling tools to find out which optimizations matter today and just as importantly those that don't.
HotSpot promises to do wonders for us by Just-in-Time (JIT) compiling the “right” code for us, but how does it makes those decisions? And, perhaps more importantly, what happens when it's wrong?
In this talk, you'll learn through real code examples just how the JVM decides to compile your code, deoptimize your code, and stop-the-world for a GC.