Thinking Parallel

Some more news that have caught my attention recently. As always, presented in no particular order and with no guarantee that they are even remotely new – all I am guaranteeing is that I had not seen these articles before and that I found them interesting. Your experience may of course vary :).

• Randall Hyde writes about Knuth’s famous quote about premature optimization and puts it into perspective. Good read, especially if you have been taught to ignore performance issues because they will go away by itself with faster computers.
• Not exactly new, but there is an article on lock-free and wait-free algorithms and data-structures up here. It includes a call to action to create a portable lock-free lib under a liberal license, an endeavor that seems very useful to me. If only there was more time to work on all the interesting projects out there…
• Some of you may have noticed by now that I am very interested in project management and software development processes. Therefore I was delighted to have found this Google Tech Talk, where Ken Schwaber explains Scrum. The process itself can be explained in five minutes, yet the video takes an hour, but also includes very valuable insights into software development not related to Scrum.
• Joe Duffy explains why fair locks are not a good idea in general – very nice read if you are interested in locking stuff.
• CNET reports that Sun decided to open-source Fortress. Although obviously still in its alpha-stages, Fortress was designed to be a successor to Fortran. I have not looked at it more closely, but the birth of a new language is always exciting, especially when it is backed by a big name in the industry.
• The ACM Queue has a very detailed an interesting article describing Transactional Memory. I have pointed out an article describing transactional memory before but this one is much more detailed and better. Obviously, at least some people think this is going to be the future (Tim Sweeney of Epic Games:”transactions are the only plausible solution to concurrent mutable state.”). I admit that this way of programming looks very nice and easy. Unfortunately, I have not seen an actual system offering these capabilities and until then (and until I see some performance numbers on one of these systems) I will restrain myself from forming an opinion.