21 Feb
2012
21 Feb
'12
2:06 a.m.
On Tue, 21 Feb 2012, Alexey Toptygin wrote:
Date: Tue, 21 Feb 2012 01:32:17 +0000 (UTC)
From: Alexey Toptygin <alexeyt at freeshell.org>
Reply-To: "General Discussion: On-Topic and Off-Topic Posts"
<cctalk at classiccmp.org>
To: "General Discussion: On-Topic and Off-Topic Posts" <cctalk at
classiccmp.org>
Subject: Re: non-x86 multicore - Re: CPU monoculture
On Mon, 20 Feb 2012, Peter C. Wallace wrote:
It all has to do with real time and latency. X86s (and their assosciated
chipsets) have great throughput but dreadful latency (this is a common
tradeoff) especially for hardware replacement type tasks. The XMOS chips
have low latency (ns) and independent I/O available on multiple cores, the X86
cores share a (high latency/buffered) common data path
As I said before, the XMOS chips are designed for multiple coupled tasks with
(somewhat in FPGA space if the signal frequencies are not above perhaps
a few 10 of MHz). It is designed to simulate hardware in software. The x86
cannot do this very well, so as I said simply does not play in the same
space at all.
Peter Wallace
Much much faster and lower power for multiple low
latency coupled tasks
Right. So call me ignorant, and give me some examples of multiple low latency
coupled tasks that can't be easily implemented as a single low latency task.
Why do you need 2+ cores talking to each other? Why not 1 core that is 2+
times as fast?
Alexey