From: turin (turin@hell.com)
Date: Fri Mar 31 2006 - 16:19:01 MST
First I'd like to know how much is an exaflop and where that number comes from.
--- philgoetz@gmail.com wrote:
From: "Philip Goetz" <philgoetz@gmail.com>
<To: sl4@sl4.org
<Subject: Re: BLUE GENE and confusion about flops and tops and whistles
<Date: Fri, 31 Mar 2006 16:38:01 -0500
<
<Tops != Tflops
<Hz = cycles per second
<ops = operations per second
<flops = floating point operations per second
<
<The 6502 ran at 1.023MHz, doing about 300 Kops, but only 1.7 Kflops.
<It took 2-5 cycles per operation, and around 600 cycles to do a floating point
<16-bit multiply or divide.
<That's an extreme difference compared to today's pipelined architectures,
<but you generally can't interchange flops and ops. Or Hertz, for that matter.
This is precisely my problem. What is the most useful way of talking about the processing power of the central nervous system? What is the best way to talk about the brain in terms of how we talk about computers now? Does it even make sense to talk about flops in relationship to what the brain does? What should we consider an "operation" in terms of neurological firing? Can a single neuron firing accomplsih because of brain architecture more than 1 operation at one time because of some sort of superposition in neural architecture. are there shorcuts?
I find it very difficult to try to imagine what the brain does in terms of what computers do now, and I am curious if there is even a general consensus about this or if even these questions are up in the air so to speak
EOF
This archive was generated by hypermail 2.1.5 : Wed Jul 17 2013 - 04:00:56 MDT