Re: Intel Itanium's very survival in doubt - inquirer article

From: Andrew Harrison (andrew_remove_.harrison_at_s_u_n.com)
Date: 07/13/04 

Date: Tue, 13 Jul 2004 17:30:11 +0100

John Santos wrote:

> On Fri, 9 Jul 2004, Andrew Harrison wrote:
>
>
>>John Santos wrote:
>>
>>
>>>On Thu, 8 Jul 2004, Andrew Harrison wrote:
>>
>>>All emulation performance data ever provided by anyone for
>>>anything in the history of the universe has this limitation.
>>>
>>
>>Sorry but that is entirely wrong. If for example HP had
>>run the HP-PA SPECint executables on Itanium then we would
>>be able to see how a workload which is 99.9% user runs.
>
>
> You take my quote out of context and then say it is entirely wrong.
>
> You snipped the statement I was responding to, and didn't even
> have the courtesy to indicate that you had snipped anything.
>
> My statement was and is perfectly true in the context in
> which I made it.
>

Sorry but it wasn't true.

I said that the benchmark data supplied by HP required
the consumer of this data to understand the ratio of
kernel/user that bind uses and then map it to their
application in order for HP's data to be usefull.

You reply that this is the status quo is incorrect.

Had HP chosen to publish the results of running SPECint
or SPECfp then customers would at least have access to
data based on a test that doesn't have an arbitrary and
unpublished mix of kernel and user. SPECint/SPECfp are
designed specifically to avoid using OS services.

This would have told customers exactly how fast the
emulation is for CPU bound applications and thats a
much easier starting point for them in the journey
they will need to undertake to get to a reasonable
approximation to how fast their application will run
emulated on Itanium.

> Having written both software and hardware emulators, and
> having benchmarked them hundreds of times, I think I know
> a hell of a lot more about the subject than you do.
>

In the context no comment is probably the nicest thing
I can say.
>
>>>>Regretably all the emulation performance data I have seen
>>>>provided by HP falls into the bind category, workloads that
>>>>are predominantly kernel based where the kernel is a native
>>>>IA64 port.
>>>
>>>
>>>Here is where you make an unsupported claim. Is BIND kernel-bound?
>>>You have posted no data to support this, you just assume it.
>>>
>>
>>Yes I have, just not in the particular posting you responded to.
>>
>>One of the Bind examples (running on HP-UX) includes top
>>output with shows a roughly 60/40 kernel/user split.
>
>
> The 60/40 split is useful information. However, it wasn't in
> the post I responded to, or in any post I've seen. Since you
> were assuming it, you should have cited it in the post I did see.
>

Why I am not a fan for repeating everything that has gone before
my posting in a thread.

> This is Usenet. Posts get lost. Posts get propagated out-of-order.
> People don't read it every day. Don't assume everyone has seen
> or has even had the opportunity to see every post. If you are
> basing a claim on something said in another thread or subthread,
> or even on something long-since snipped, then you should
> recapitulate.
>
> Please learn some basics of newsgroup etiquette. You've been
> posting long enough that no one should cut you any newbie slack.
>

Again you appear to be in the wrong here as well, newsgroup
etiquette in fact suggests that it bad practice to re-post
everything that has gone before your posting in the thread.

Regards
Andrew Harrison

Relevant Pages

  • re: [meta]: definition: co-moron
    ... If you are not responding in context, ... >> but only to identify the moron, ... activity on posting record, some unambiguous as I interpret it. ...
    (comp.theory)
  • re: [meta]: definition: co-moron
    ... If you are not responding in context, ... >> but only to identify the moron, ... activity on posting record, some unambiguous as I interpret it. ...
    (sci.math)
  • re: [meta]: definition: co-moron
    ... If you are not responding in context, ... >> but only to identify the moron, ... activity on posting record, some unambiguous as I interpret it. ...
    (sci.logic)
  • Re: XML::Simple Example perl help needed
    ... paragraph, as in this example. ... Top posting means to respond at the ... above the text you are responding to (as you ... of context your top post is and the harder it is for people to figure ...
    (comp.lang.perl.misc)
  • Re: dloyd lavies cyber terroists
    ... I am responding that your said you were tired of the BS, ... when you posted about my wife and certain book you knew or should have known that the filth mongers would run with it ... You don't restrain your posting. ... cause therefore I will attack you and contiue to seek to obscure you ...
    (rec.radio.amateur.policy)
Loading