(originally posted here∞)

I think the vast majority of software development will not be affected at all by dual-core CPUs. Multi-CPU machines have been at least somewhat common for decades; meanwhile, coders are about the most stubborn group of people ever found.

It's not just recalcitrance, though. Good tools -- debuggers that don't magically alter race conditions, sensible thread-parallel language constructs, and much more -- simply don't exist yet. They may never. It's a proverbial "hard problem" in computer science.

The very few engineers who do specialize in flexible, performance-critical code will optimize for dual-core, of course. However, since they basically exist to write special cases, don't be surprised if they give dual XPs a different codepath from an Opteron 875. Frankly, considering XPs don't support SSE2, I find that scenario highly likely.

On the plus side, much of the above work is done on platforms: OS's, compilers, JIT runtimes, unified drivers, etc. Thus, their work will be felt a lot more loudly than their numbers...eventually (once everyone recompiles).

Finally, keep in mind that most multithreaded programming isn't done because the # of CPUs increases. Lots of applications that happen to scale well -- ranging from Apache to explorer.exe -- would be dramatically slower if they only had 1 thread, even on a single-CPU desktop.

---

Back to CollectedWritings