The open source community is a valuable asset to software development. A place where people can share, learn, explore, and grow – an environment that strives on working together. In support of this community, Rogue Wave Software has released our ThreadSpotter product as an open source product with help from our partner, ParaTools.
The release of ThreadSpotter extends our participation in the open source community, adding to our commitment to build confidence in open source with products such as the OpenLogic scanning and governance platform. We appreciate that open source is vital to the success of application development and we want to continue our involvement in such a strong community.
The ThreadSpotter source code has been released under a LGPL license. You can get a pre-built binary in a LiveDVD (ISO) and a Virtual Machine (OVA) image at http://www.hpclinux.com. Or, if you’re attending this year’s Supercomputing 2014 conference, you can stop by the ParaTools booth to pick up a DVD. Once you like what you see, you can get fee-based support with a subscription.
ThreadSpotter will help you understand and improve the runtime performance of your applications by identifying areas where the program is using processor cache memory, and in some cases suggests ways to restructure the code to make more effective use of the cache memory.
Sometimes you may not think about cache behavior when you write algorithms, and several popular programming paradigms encourage programming styles that can lead to very inefficient use of cache memory. Good use of cache memory can easily make an order of magnitude (or a factor of 10) difference on program performance. It can also act as a limiting factor on scalability and the ability of the program to run on multicore processors.
ThreadSpotter excels where other tools and techniques for cache performance analysis fail. These other tools tend to give very low level information that comes directly from hardware counters, such as the raw number of cache misses. Interpreting low level hardware counter data takes significant expertise. ThreadSpotter points you towards the lines of code that are problematic and provides an easy explanation of what about that section of code causes it to make poor use of the cache.
Get to know ThreadSpotter, code with ThreadSpotter, and enjoy the gift of open source.