reverse multi-threading

Rubycon

Madame President
Aug 10, 2005
17,768
485
126
Isn't this a double edge sword though?

One of the things I found nice about SMP systems is if you have a really badly misbehaving app that grabs every cpu cycle the system is still responsive so it can be killed. If everything reverses back in to uniprocessor mode wouldn't the same pitfall exist?
 

Madwand1

Diamond Member
Jan 23, 2006
3,309
0
76
For the readers who haven't waded through the other thread (focusing in no small degree on hyperthreading not the reverse), let me link the Intel paper that probably sheds the greatest light on this subject:

http://www.intel.com/technology/magazine/research/speculative-threading-1205.htm

As to pitfalls, I'm sure there will be plenty, not least the continued encouragement of developer laziness and postponement of threaded design to future compiler + hardware solutions.

However, I don't think that this in itself will be any worse than say a badly written app + OS scheduling that would let such an application "take over" multiple CPU's on a non-reverse-hyperthreading machine. Moreover, not all apps will benefit from this tech and "take over" all CPU's -- just those ones that are built to incorporate this feature during build time.