# Monday, 07 November 2005

I love being "in the loop" on new software. It can get kind of surreal, though, when you are at a conference and find a technology kind of boring and everyday, and skip the sessions on it since it's "old stuff" ... and then realize it isn't even shipping yet! I spend time in betas, alphas, SDRs and the like, so it's often years between the time I start working with a technology and the time it releases. When I speak at Tech Ed I tend to hang at the cabanas where I can hear real stories of how people are using the tools and the problems they need help with.

The PDC though, that's a different story. Well, it should be anyway. The PDC is all about the future, right? The workflow announcement was a big deal, but the material wasn't new to me. I began to worry if I would ever attend a session that sparked my interest in something new and important. Then I found it. I've waited to blog it until the sessions were available online.

What Jan Gray had to say in FUN302 and Herb Sutter introduced in TLN309 totally grabbed me. Because clock speed cannot go up any more we all have to write concurrent code even if it's single user:

And since most people can't write good concurrent code (see Herb's Singleton example which still doesn't work even with the double checking and the volatile keyword, at least not all the time) then the OS or the framework or the language needs to do it for them. Herb and Jan are proposing different approaches, but these two guys are among the smartest I know (I've known Jan almost 30 years) and if they are both into this, then I need to learn more about it. You probably do too.

Kate

Monday, 07 November 2005 10:19:27 (Eastern Standard Time, UTC-05:00)  #    Comments [1]
Monday, 07 November 2005 14:36:18 (Eastern Standard Time, UTC-05:00)
Silicon has a limit to how fast it can "ramp", so no matter how fancy CPU engineering gets, there's a limit to how fast silicon transistors can switch between 0/1 (I think that 5Ghz is the theoretical ceiling).

We're already seeing the commercial response to this will AMD & Inter pushing their Dual-Core CPUs, and AMD is starting to implement their multi-CPU motherboards using NUMA.

Linear programming may soon be a thing of the past (at least for those of us who can afford to upgrade our CPUs ;) )
Comments are closed.