Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!


Forgot your password?

Submission + - Tiny cores are here, and they change programming (futurechips.org) 3

An anonymous reader writes: Intel is returning to in-order cores after two decades with Atom and Knights. ARM is already building in-order cores for iPhones, iPads, and Androids. IBM has switched to in-order cores after building generations of out-of-order cores. This indicates a clear trend that in-order cores are back in the mainstream. Highlighting the performance characteristics of in-order and out-of-order cores, Dr. Aater Suleman's article explains why programming for in-order cores is very different from programming for the now-traditional out-of-order cores. Thus, this new trend requires a change in compilers, tools, and programming techniques. Compilers need to get better at removing useless code and instruction scheduling. Programmers need to weigh new trade-offs and perform classic optimizations that have been forgotten. I liked this article particularly for the very simple code examples and a simple explanation of in-order and out-of-order differences. The message is clear: programmers and compilers need to understand in-order cores and target their code better.
This discussion was created for logged-in users only, but now has been archived. No new comments can be posted.

Tiny cores are here, and they change programming

Comments Filter:
  • I see his point that in-order cores are becoming common. I mean iPhone has an in-order core and they are common, hence in-order cores are common ... no debate. What I am thinking about is the importance of the optimizations Shouldn't all iPhone app developers know about these optimizations? This seems like it can be a big win for all cell phone developers to follow some of these techniques. I do notice some very slow apps on my iphone and wonder his arguments play a role.
  • The article makes a strong case for programming optimizations and Ilike the tutorials too. reminds me of programming back in the day. but surprisingly the case has been made for serial programs only. Isn'tonly multi-threading relevant to this argument... Don't get me wrong. I do see that improving single threading alsomakes stuff faster but how significant is that. Just Ahmdahl's or theparallel threads as well.
    • Nope dude. Wrong argument! Speeding up a single thread is like speeding up each individual thread, in addition to your Ahmdahl's (dunno what you really meant there). Make single-thread faster then all threads become faster which makes the program faster. Multi-threading is not a magic, just means there are multiple single threads running with each other so single thread is always relevant. LOL@nostalgic btw.

I THINK MAN INVENTED THE CAR by instinct. -- Jack Handley, The New Mexican, 1988.