By: David Kanter (dkanter.delete@this.realworldtech.com), July 25, 2012 9:45 pm
Room: Moderated Discussions
> I think
> what Paul is getting at is that you're presenting a limited and artificial
> characterization of efficiency of modern chips intended for high FP
> loads.
Yes, that's true.
> For example you repeatedly castigate Wilco for bringing up Dhrystone
> and CoreMark in defense of ARM processors, as being unrepresentative. Yet here
> you're essentially presenting efficiency data for computational kernels, not
> real world parallel FP applications.
> An uncharitable person would call you a
> hypocrite. You are a hypocrite.
There's a huge distinction. Wilco uses this to discuss performance and routinely extrapolates from Dhrystone to real workloads that matter.
I would never extrapolate from Dhrystone to real workloads without some sophisticated assessments of performance (e.g. a regression framework to break the workload down into compute and memory bound aspects).
> If you don't like SPECfp_rate then at least
> man up, get some Nvidia, Intel, and AMD/ATI chips and benchmark them with either
> SPLASH-2 (old), SPEChpc2002, or (preferably) PARSEC.
>
> Be a man.
I like SPECfp_rate...but it doesn't run on GPUs, and AFAIK, SPEChpc2002 doesn't either. Which makes it hard to compare.
DK
> what Paul is getting at is that you're presenting a limited and artificial
> characterization of efficiency of modern chips intended for high FP
> loads.
Yes, that's true.
> For example you repeatedly castigate Wilco for bringing up Dhrystone
> and CoreMark in defense of ARM processors, as being unrepresentative. Yet here
> you're essentially presenting efficiency data for computational kernels, not
> real world parallel FP applications.
> An uncharitable person would call you a
> hypocrite. You are a hypocrite.
There's a huge distinction. Wilco uses this to discuss performance and routinely extrapolates from Dhrystone to real workloads that matter.
I would never extrapolate from Dhrystone to real workloads without some sophisticated assessments of performance (e.g. a regression framework to break the workload down into compute and memory bound aspects).
> If you don't like SPECfp_rate then at least
> man up, get some Nvidia, Intel, and AMD/ATI chips and benchmark them with either
> SPLASH-2 (old), SPEChpc2002, or (preferably) PARSEC.
>
> Be a man.
I like SPECfp_rate...but it doesn't run on GPUs, and AFAIK, SPEChpc2002 doesn't either. Which makes it hard to compare.
DK
