Algorithm complexity and modern CPU’s

Jan Gray wrote an excellent essay … My favorite quote from this piece is “This scenario is so bad and so common that the microprocessor vendors use 80% of their transistor budgets for on-chip caches — Intel as glorified SRAM vendor.”. via [iunknown.com]

This echos a favorite theme of mine: most discussions of performance ignore the differing costs of memory access. To the extent that people think about it, they split the world into RAM and disk, when it’s much more productive to think of a series of ziggurat-like steps that extend all the way from the CPU registers to the Internet and offline storage. (BTW, the essay is a year old, which makes some of the references, especially to C#’s closure support, a little odd).

Most important question of the week

http://blogs.gotdotnet.com/BradA/commentview.aspx/ddedfad0-c94f-4873-9c8c-4005cd8faaa5 Should the BCL contain both generics-based collection classes (List<T> ) AND object-containing classes? (my initial gut is that generics-only is the way to go. Better to accept some pain now than 5 years from now. )

Blogged on a Tablet PC

Conference Economics

Martin Spedding clarifies that he’s criticizing PNC on a strictly economic basis. This is perfectly valid: consider that the benefits of a conference accrue primarily to the individual while the cost is borne primarily by companies. And the value of a conference is delivered primarily by speakers and attendees, but the profit is enjoyed primarily by the conference organizer. Something is awry in the was that conferences work.

Blogged on a Tablet PC