Floating-point arithmetic: any optimization tips ?
posted 10 years ago
I'm writing an application that requires billions of floating-point arithmetic operations. I was wondering whether anyone could suggest some basic tips for optimizing Java/JVMs for such operations ? Or refer to some website/book on the subject ?
Up till now, I've come up with the following (very basic) facts: - Use primitive variables rather than wrapper classes (such as Double) - Run the jvm with '-server' flag (this gave me over 100% performance gain!)
Any other JVM-related tips would be greatly appreciated. Thanks.