SPEC CPU2006 announced

[Mark McIntyre]

I think Usenet works and remains, because it didn't have the restrictive
Acceptable Use Policies from the Defense Communications Agency which
took over the ARPAnet, the academic IBM mainframe mentality of BITNET,
the staid academic mentality of CSnet, throw JANET and Alvey in there,
etc.

You're mistaken about this, but then your entire post is essentially
offtopic in all the groups bar one that you posted to. The group I'm
reading it in, like most others on usenet, DOES have an AUP, which you
breached.

That said, I simply ignored the initial posting. If the debate
continues I'll killfile both the thread and anyone supporting the
original posting, in order to save my energy.
--
Mark McIntyre

"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it."
--Brian Kernighan

 

[Andy Glew]

You're mistaken about this, but then your entire post is essentially
offtopic in all the groups bar one that you posted to. The group I'm
reading it in, like most others on usenet, DOES have an AUP, which you
breached.

Let me continue the thread, so that Mark McIntyre will killfile me, too.

Although the original topic, as to whether the new SPEC announcement
should be forbidden to be posted to these newsgroups, I think is relevant to all of
the newsgroups listed:
comp.arch,comp.benchmarks,comp.lang.c++,comp.lang.fortran,comp.lang.c

SPEC is clearly relevant to the comp.arch newsgroup.

Yes, it sucks that you have to pay a nominal fee. This fee irritated
me while I was a student at university, it irritated me while I was
between jobs at AMD and Intel, getting in the way of "freelance"
computer architecture research. Heck, it irritates me at Intel,
because I cannot freely copy the benchmark source code around.

But, while the fee irritates the heck out of me, SPEC is clearly one
of the most important benchmark suites. comp.arch must pay attention.

I argue the same thing goes for comp.benchmarks.

Although I am no longer a regular on comp.lang.c++, comp.lang.fortran,
comp.lang.c, I once was. If these newsgroups are still intended to
serve as a forum for compiler developers, then SPEC is relevant. SPEC
is one of the most widespread set of benchmarks for these programming
languages; compiler developers have no choice but to pay attention.

Now, if these newsgroups are no longer intended for professional
compiler developers, by all means ban SPEC announcements. I.e. if
comp.lang.c is now comp.lang.c.trivial-user-questions, ban SPEC.

Similarly, if none of these groups care about the performance of the
code generated.

But while we are talking about the language newsgroups
comp.lang.c++, comp.lang.fortran, comp.lang.c,
note one thing:

THE LANGUAGE SPECIFICATIONS COST $$$$.

An announcement of a new version of the C, C++, or FORTRAN language
standards can be construed as an offer for sale in exactly the same
way that the SPEC announcement is.

The language specs are not public domain. They are copyright by varius
bodies - ANSI, ISO. You are supposed to buy your copy. You are not
supposed to give a copy of a PDF to a friend. You are not supposed to
post them on the Internet.

I myself spend the, what, 80$ or so, last time I checked, to purchase
the C and C++ standards.

---

For logical consistency, if you ban the SPEC announcement, you should
also ban announcements of new language specs to these newsgroups.

That would clearly be stupid.

===

Finally, and once again, I agree that these nominal charges are
annoying. I am not so far from being a student who could not afford to
pay for a copy of the SPEC benchmarks, or buy a copy of the C++
standard, myself.

I suspect that the non-availability of a legal copy of the C++
language spec on the web, suitable for online reference, is in some
small way a contributing factor to the slow decline of this language.

I love languages that have manuals, etc., on the web: Java, Python, Perl.

But the cost is fairly nominal. The standards organizations - ANSI,
ISO, SPEC - think that they need to recoup their costs in some small
way by these charges.

Let's move on to discussing more interesting topics.

 

[Andy Glew]

comp.lang.c in particular views commercial postings of any kind
whatsoever as inappropriate, a fact prospective posters would do
well to remember.

So, an announcement of a new ANSI C standard is not supposed the be posted on
comp.lang.c? Remember, you have to pay $$$ to get a copy of the
standard.

 

[=?ISO-8859-15?Q?Juli=E1n?= Albo]

Although the original topic, as to whether the new SPEC announcement
should be forbidden to be posted to these newsgroups, I think is relevant
to all of the newsgroups listed:
comp.arch,comp.benchmarks,comp.lang.c++,comp.lang.fortran,comp.lang.c

I don't care. I will not visit his site nor recommend it to anyone, just
because the way used to advertise it. The same I do with all spam, even if
the product looks promising or if the sender is a non-profit, a charity
organization, o even Santa Klaus.

 

[Mark McIntyre]

Let me continue the thread, so that Mark McIntyre will killfile me, too.

Your wish is granted. Stupidity has its own reward.

Although the original topic, as to whether the new SPEC announcement
should be forbidden to be posted to these newsgroups, I think is relevant to all of
the newsgroups listed:
comp.arch,comp.benchmarks,comp.lang.c++,comp.lang.fortran,comp.lang.c

The topic is nothing at all to do with the C language, nor with C++,
nor with Fortran. Benchmarking is system specific ,and
system-specific features are nontopical in these groups.

--
Mark McIntyre

"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it."
--Brian Kernighan

 

[Richard Tobin]

Benchmarking is system specific ,and
system-specific features are nontopical in these groups.

In your opinion. Not all comp.lang.c regulars agree.

Incidentally, if you killfile Eugene Miya and Andy Glew, you are losing
two people who have both contributed far more to Usenet over the years
than you have.

-- Richard

 

[Default User]

So, an announcement of a new ANSI C standard is not supposed the be
posted on comp.lang.c? Remember, you have to pay $$$ to get a copy
of the standard.

When has ISO done that?




Brian

 

[Richard Maine]

When has ISO done that?

Pretty much forever. Except that it usually isn't ISO directly, but
rather the national organizations. Note that the money doesn't go to the
people who actually do the technical and editorial work on the standard.
In fact, those people have to pay for the privilege of donating their
work. The money from both the purchasers and the writers of the
standards all goes to run the bureaucracies involved - not to do the
technical work. Even the committee members don't get a free copy of what
they wrote. I was editor of the f95 and f2003 Fortran standards, so I
speak with pretty specific knowledge on that.

For Fortran 95, the charge was not exactly nominal, I might add. It was
something like $200 US for a long time for a paper copy, which was the
only form it was officially available in. I still think it might be
about that much for a paper copy, but electrinic ones are now available
at a much more nominal cost, but still not free.

Similar principles apply to the C standard, though it got an official
electronic copy available before Fortran did.


P.S. I've participated heavily in comp.lang.fortran for about 15 years
and I consider discussions of benchmarking in general, and SPEC in
specific, on topic for it at least. Such discussions are indeed, quite
common. But then clf isn't rabidly anti-commercial. We tend to get
annoyed at repetitive ads that have no new content (such as a vendor
that used to post an ad about every week). And marketspeak tends to get
ridiculed. But factual announcements of new products go over pretty well
as long as they aren't full of hype. From my 15 years of participation,
I can confidently say that this is representative of the majority of clf
(certainly not 100%, but a pretty clear majority).

 

[Default User]

Pretty much forever. Except that it usually isn't ISO directly, but
rather the national organizations.

You misunderstood me. I know ISO sells the Standards. I asked when they
or their associated groups posted ads to comp.lang.c.




Brian (doesn't recall any)

 

[Richard Maine]

You misunderstood me. I know ISO sells the Standards. I asked when they
or their associated groups posted ads to comp.lang.c.

Well, we post quite a lot of anouncements about such things as new
versions of the standards. We even post when review drafts of them come
out. The original post of this thread was announcement of a new version
of SPEC. The announcements of new versions of the language standards
are very similar in tone to the subjet SPEC announcement. In neither
case are such things as purchase costs specifically mentioned, but if
you want to consider the SPEC announcement as an advertisement, then I
don't see how you can consider the language standard annoucements any
differently. They sure look the same to me.

But I don't know why I bother to respond. If you don't like the SPEC
announcement being posted, you are of course free to dislike it. That
opinion appears to be in a small enough minority that it isn't likely to
have much effect, which in my opinion is good, because I like seeing
announcements like that.

 

[Default User]

Well, we post quite a lot of anouncements about such things as new
versions of the standards.

Possibly. I don't recall anything of the sort where a specific product
is being pushed.

But I don't know why I bother to respond. If you don't like the SPEC
announcement being posted, you are of course free to dislike it.

I'm terribly upset, I just think that certain people have discounted
the feelings of those that do object. I think the posts were marginal
at best. A more focused posting to groups more likely to be interested
would have been better. ANY thread started to that many groups is
highly suspect.




Brian (willing to call it a day if you are)

 

[John Mashey]

Yes, but this understates the issue. A serious proposal follows.

In your opinion. Not all comp.lang.c regulars agree.

Incidentally, if you killfile Eugene Miya and Andy Glew, you are losing
two people who have both contributed far more to Usenet over the years
than you have.

1) Computer architecture, language design, compiler design, OS design,
and benchmarking/performance analysis are rather strongly
inter-related, in the real world anyway. In particular, the SPEC CPU
benchmarks have long been intertwined with architecture and compilers.

2) People like Andy Glew, Eugene Miya not only have contributed to
Usenet, but actually have a raft of real-world *contributions* to the
discipline included within the posted newsgroups, as have several
posters. Interestingly, people that I either know directly, or by
real-world reputation/contributions (like Richard Maine: standards
efforts are *really* hard work and under-appreciated), and/or by long
history of sensible posts on Usenet, seem to think the SPEC
announcement was fine.

3) Now, for a SERIOUS OFFER to actually DO something, for those who are
unhappy. This discussion has seemed fairly wasteful, i.e., complaints
against a short SPEC announcement, given the vast amount of drivel
posted every day. Of course, I might be biased. As noted, while it's
been years since I've been officially involved with SPEC, I'm one of
its fathers, and it hate to see it acting improperly, if that's what it
was.

If I can get convinced (by Sept 1) that this was out of line, I will
talk to the SPEC folks about it. I make no guarantee that they would
do anything different, but at least the complaint will get heard.

4) So, what would convince me? The negatives have to outweigh the
existing positives, and I make no pretense at counting votes; I listen
harder to some people than others, so if DMR or BWK told me they really
hated seeing SPEC announced in comp.lang.c, I'd listen pretty hard

5) Unfortunately, I don't know any of the people who seem upset about
this:

{Flash (Mark) Gordon, Julian Albo, Mark McIntyre, "Default User",
CBFalconer, Al Balmer} seem unhappy about this, to various degrees, up
to:

"If you really are in those positions you should know better than to
bring your companies in to disrepute by spamming. "

"I will not visit his site nor recommend it to anyone, just
because the way used to advertise it."

"I'm terribly upset, I just think that certain people have discounted
the feelings of those that do object. "

I'm sad to see (even anonymous) users getting terribly upset by
SPEC.

6) I've done a quick rummage, so I've found:
a) Added up, according to Google profiles, they have written about
50-60,000 posts over the last few years. Presumably, given email
address changes, this is a lower limit, although I'm not exactly sure
how they count.

b) I'm trying to understand the weight I should assign to these
opinions, as compared to opinions held by people who I know have
actually contributed a lot to the relevant disciplines. I don't want
to discount the complaints just because I don't happen to know these
folks, and don't have the time to rummage through thousands of posts.
I spent a little time looking for real-world references, but as usual,
many names are nontrival. CB Falconer was at least easy to find.
"Default User" was of course hard

7) So, I would love it if anybody who feels strongly negative about the
original announcement would kindly post a little more background to
help me understand where their opinion comes from and how to weight it
(compared to Andy, Eugene, etc).

This is a serious question; I'm *not* trying to give anybody a hard
time, I'm just trying to understand whether this is a real issue or
not, and given the joint posting rate of ~10,000/year, or
1600/person/year, one more post apiece is just a drop in the ocean.

8) Here are things that would be plusses:

a) Work on either official or adhoc language/OS standards committees.

b) Work on relevant conference committees.

c) Relevant refereeing work.

d) Relevant refereed papers, invited talks, published books

e) Contribution of benchmarks to any of the benchmarking groups

f)* Especially insightful postings on relevant topics, i.e., ones that
are often back-referenced by others, copied onto others' websites,
edited into journals, etc.

g)* Creation of relevant software that gets widespread external use;
obviously, given the topics, compiler-writing and benchmark creation
would be relevant.

(* f) and g) are inherently fuzzy, sorry. The bar to posting is zero,
and these days, the bar to creating open-source software is not so high
either, so whether or not something actually has impact as a
contribution is not easy to tell.)

Obviously, just posting a lot in the relevant newsgroups is *not*
apriori interesting, since Sturgeon's Law ("90% of everything is crud")
rules, which of course makes the (rare) good stuff all the more
valuable.

h) Usenet longevity is relevant (and is hard to determine
mechanically, given email address changes). Also, SPEC familiarity is
somewhat relevant, i.e., it is a real plus to be familiar with the
relevant benchmarking in the 1980s or earlier, before SPEC started. (A
few of the complaints appeared, on the surface, to be from people who
didn't know what SPEC was.)

OF COURSE, very few people would have done all these, or even a small
subset, and there may well be other relevant items.

====
The worsening S/N ratio keeps me away most of the time, but if there's
a real complaint against SPEC on this , I'd like to do understand it.
I'm hoping NOT to generate another giant flurry of posts.

 

[Christopher Benson-Manica]

So, an announcement of a new ANSI C standard is not supposed the be posted on
comp.lang.c? Remember, you have to pay $$$ to get a copy of the
standard.

I did note that elsethread, and it's a valid point. It's unfortunate
that comp.lang.c does not have an official charter, the existence of
which would render these periodic topicality debates irrelevant.

 

[Anton Ertl]

This discussion has seemed fairly wasteful, i.e., complaints
against a short SPEC announcement, given the vast amount of drivel
posted every day.

Ouch, that was a classical defense used by spammers.

There is a fairly long tradition in Usenet against commercial
postings. There are some gray areas here, but apparently some people
felt that the announcement crossed the line (it did not raise my
hairs, but maybe I am desensitized). This tradition is
well-justified, because if it were not there, the usual drivel would
drown in a sea of commercials (actually, IIRC some time ago Usenet
traffic as seen by newsfeeds was 20% regular postings, 40% spam, and
40% spam-cancels; I think cleanfeed fixed that).

Also, it is much easier to get a consensus that some posting was
commerial than consensus that some posting was drivel, so a ban of
drivel would just produce a discussion whether some posting was drivel
or not.

One other thing: The complaints seem to come from the comp.lang.c
camp. My impression of this newsgroup is that the only on-topic
postings there are postings explaining why some other posting was
off-topic. So if somebody (cross)posts there, they should expect such
a reaction. Cross-posting without setting a Followup-To: one group
(or "poster") is bad form in any case, but particularly dumb if the
cross-posting goes to a group like comp.lang.c. In the present case
even I feel that the posting is off-topic there (and actually on-topic
only in comp.benchmarks, so at least the followups should have been
set there).

Followups set to comp.arch (This meta-discussion is just as off-topic
here as anywhere else (except maybe comp.lang.c), but at least I have
subscribed this newsgroup at the moment).

- anton

 

[Jerry Coffin]

Hi John,

[ ... ]

If I can get convinced (by Sept 1) that this was out of line, I will
talk to the SPEC folks about it. I make no guarantee that they would
do anything different, but at least the complaint will get heard.

I, for one, don't mind going on record as saying that I think it was and
is relevant to enough that posting it in the language-oriented
newsgroups was quite reasonable, even if it may be open to some argument
whether it's truly topical.

[ ... ]

f)* Especially insightful postings on relevant topics, i.e., ones that
are often back-referenced by others, copied onto others' websites,
edited into journals, etc.

While the email address given in your header is different from the one
on your card you gave me last October, I'm willing to believe that's an
attempt at spam control. It takes the kind of person who could help
start up SPEC to even consider taking on a task like figuring out which
posts are especially insightful and relevant!

Figuring out what copies onto web sites and such are really meaningful
is almost certain to be a serious challenge as well. Googling for my own
name turns up over two MILLION instances. My guess is that the vast
majority of those are simply blind, automated copies from Usenet onto
various web sites. I suspect the number of people who've posted as much
for as long as I have is fairly small, and I'd like to think I produce a
higher percentage of signal than many. Nonetheless, I'd bet at least 99%
of those are either duplicates of each other, or something that
pertained primrarliy to a conversation at the time, and barely worth
reading out of context. Many are probably both: of little value, and
still duplicated endlessly.

 

[Jerry Coffin]

[ ... ]

I did note that elsethread, and it's a valid point. It's unfortunate
that comp.lang.c does not have an official charter, the existence of
which would render these periodic topicality debates irrelevant.

Yes and no. If defined in sufficient detail, it might render all
discussion of what is or is not topical irrelevant (though it's hard to
imagine a charter with sufficient detail that at least some things
wouldn't still be open to question). That would still leave the
possibility of discussing whether something should or should not be
considered topical -- i.e. whether the charter is considered graven in
stone and never open to any change, or it attempts to change over time
to suit the needs of its audience.

Any real charter generally has to compromise between the two -- a
general statement of topic leaves a large grey area, but rarely requires
alteration. A more specific statement of topic leaves reduces the grey
area, but will also (usually) reuqire more frequent modification.

 

[=?ISO-8859-1?Q?Jan_Vorbr=FCggen?=]

Yes, it sucks that you have to pay a nominal fee. This fee irritated

me while I was a student at university, it irritated me while I was
between jobs at AMD and Intel, getting in the way of "freelance"
computer architecture research. Heck, it irritates me at Intel,
because I cannot freely copy the benchmark source code around.

As I noted elsethread, that irritation - in that SPEC can not and will not
offer it for anonymous download - will never go away for legal reasons:
There are copyright issues, and SPEC particularly wants to make sure that
the run rules and the publication rules are obeyed. The latter alone make
it imperative that you sign a contract with them.

Now, they could lower the fee, sure. They already have done so several
times in the past. I do think the $200 is about adequate to cover their
administrative costs for one licensee, and you sure wouldn't want SPEC
to go down because of a DDoS attack, would you? And frankly, anybody to
whom the $200 rate isn't applicable - the additional $600 is just in noise.
Heck, you could re-use runspec as your benchmark harness and save developing
one yourself - that's substantially more in savings.

Jan

 

[=?ISO-8859-1?Q?Jan_Vorbr=FCggen?=]

The topic is nothing at all to do with the C language, nor with C++,

nor with Fortran. Benchmarking is system specific ,and
system-specific features are nontopical in these groups.

Thanks, man, for making my morning! Thank God the coffee wasn't ready
yet, otherwise I would now need a new keyboard.

To keep with your apodictical style, you're a disgrace to whatever groups
you _do_ post in.

Jan

 

[glen herrmannsfeldt]

Andy Glew wrote:
(snip)

Although I am no longer a regular on comp.lang.c++, comp.lang.fortran,
comp.lang.c, I once was. If these newsgroups are still intended to
serve as a forum for compiler developers, then SPEC is relevant. SPEC
is one of the most widespread set of benchmarks for these programming
languages; compiler developers have no choice but to pay attention.

I definitely agree here!

Now, if these newsgroups are no longer intended for professional
compiler developers, by all means ban SPEC announcements. I.e. if
comp.lang.c is now comp.lang.c.trivial-user-questions, ban SPEC.

This is especially funny, knowing what comp.lang.c considered on topic.

Similarly, if none of these groups care about the performance of the
code generated.

Well, that does sound like comp.lang.c Some of the others do care,
though some call it a "quality of implementation" issue.

But while we are talking about the language newsgroups
comp.lang.c++, comp.lang.fortran, comp.lang.c,
note one thing:

THE LANGUAGE SPECIFICATIONS COST $$$$.

For a completely unrelated reason this came up last week.

(snip)

The language specs are not public domain. They are copyright by varius
bodies - ANSI, ISO. You are supposed to buy your copy. You are not
supposed to give a copy of a PDF to a friend. You are not supposed to
post them on the Internet.

Even freely available documents can still be copyright, such that
you are not supposed to deliver them. IBM has many of their
manuals available for free on the web, and also the past copies
of their journals. Still, you are not supposed to redistribute
them, though I don't know that they actually try to stop anyone.

(snip)

Finally, and once again, I agree that these nominal charges are
annoying. I am not so far from being a student who could not afford to
pay for a copy of the SPEC benchmarks, or buy a copy of the C++
standard, myself.

For IEEE 802.3 (or maybe others), they now have a policy to release
them free after six months. (Still under copyright, and careful not
to lose the rights to them.) That seems reasonable to me, as anyone
building real products won't want to be six months behind.

(snip regarding C++)

I love languages that have manuals, etc., on the web: Java, Python, Perl.

Well, Sun has released many things free, RPC, XDR, and NFS fairly
early. It might be that they require copyright acknowledgment,
but they realized that to be standards they had to be free.
Then Java and StarOffice, which probably wouldn't be near as popular
if they cost more.

But the cost is fairly nominal. The standards organizations - ANSI,
ISO, SPEC - think that they need to recoup their costs in some small
way by these charges.

I wouldn't mind if it was a little less. Still, I don't have enough
time now to actually run them myself. If I did have the time, it would
probably mean I was between jobs and wouldn't have the money.

They could have a discount for unemployed people, like other
groups have done.

Let's move on to discussing more interesting topics.

This one I agree with!!

-- glen

 

[CBFalconer]

.... snip ...

5) Unfortunately, I don't know any of the people who seem upset
about this:

{Flash (Mark) Gordon, Julian Albo, Mark McIntyre, "Default User",
CBFalconer, Al Balmer} seem unhappy about this, to various
degrees, up to:

Hardly seriously upset. I simply posed the question. I would have
been quite satisfied if the original announced "there is a charge,
to cover expenses".