My Photo

About Me

  • Hello you. I'm a 38-year old MSc student, studying Advanced Computer Science at Sussex University. I'm especially interested in Internet and mobile software, sensors and pervasive computing, user interfaces, and the process of developing great software.

    Before that I spent 11 years running Future Platforms, a software company I co-founded which makes lovely things for mobile phones, and which I sold in 2011.

    I read a lot, write here, and practice Aikido and airsoft. I live in Brighton, a seaside town on the south coast of the UK, with two cats and a clown.


Stalk Me

  • Email me:
    twhume at gmail dot com
Blog powered by Typepad

« Output from the Master's | Main | Needz launches »

September 05, 2012



Hey, I really enjoyed this talk! I had a thought recently: was the optimisation of the number of instructions to the JVM, or the actual processor instructions? I think it's probably the former. It'd be interesting to see how the JVM instructions map to actual instructions on the bare metal. I think you might have mentioned about getting a real world benchmark, but it was too difficult with a virtual machine.

All the best,

tom hume

Alex, glad you enjoyed it. I was talking about the former, generating JVM instructions. It's hard-to-impossible to measure machine instructions with a modern JVM, I suspect: byte code to machine code may not be a 1:1 mapping, and may be an inconsistent mapping if a JIT compiler is in the mix.

The comments to this entry are closed.