Programming Linux on Cell

Nora writes "developerWorks has posted a slightly expanded version of the paper presented at LinuxTag today by IBM's Linux-for-Cell maintainer Arnd Bergmann (which was mentioned on Slashdot a few weeks ago). I searched the LinuxTag site to post a link to their copy of the paper also, but I couldn't find one (but my German is not so good, so it's probably my fault). In addition to the abovementioned paper on the SPU file system, we have also published an interview with Arnd Bergmann in which we learn more about programming for Linux on Cell, and programming for Cell in general -- and also that there is no such thing as the so-called 'Cell Workstation' that some of us have been looking forward to for a long time (apparently, it's just the Blade board prototype that many of us have already seen). And of course, much more. "

Comments

Too little, too late

[user9918277462]

With the recent conversion by Apple, unfortunately the world seems to be standardizing on the x86 architecture (or its 64-bit counterpart, AMD64) for general purpose computing, at least in the medium term.

There are many arguments why, but the main is that hardware development is expensive and essentially serial. Software development can be parallelized (as Open Source has done) but this does not apply to processors. Chip fabbing doesn't scale. It makes economic sense to have a single dominant computer architecture, manufactured globally and powering everything from a PC to a gaming console to a cellphone.

The Cell processor is a dead end, just like the overpriced PPC OpenWorkstation products and the overrated Alpha line.

Bottom line: At modern clockspeeds architecture is irrelevant.