Tuesday, October 18, 2016

Digest for comp.lang.c++@googlegroups.com - 25 updates in 5 topics

Ian Collins <ian-news@hotmail.com>: Oct 18 06:24PM +1300

On 10/18/16 10:39 AM, Jerry Stuckle wrote:
 
> No, I'm not going to waste my time looking for something you're too lazy
> to find yourself (hint: it's in this thread) and couldn't understand,
> anyway.
 
Well it's clear that one one did make such a claim. No one using an
agile methodology would be daft enough to do so. I can only assume that
your misinterpreted what someone said, give you won't give a quote.
 
 
> your computer and definitely not what actually can be accomplished. You
> won't get anywhere near 10GB throughput, even to another computer on
> your LAN.
 
What's beyond the computer is the link! I'm happy with the throughput I
get (about 6G disk to disk). If I need more, I go RAM to RAM.
 
> But then you've already backpedaled and admitted you don't have anywhere
> near a 10GB internet link.
 
Nowhere in this thread or anywhere else have I said anything other than
my Internet connection is rubbish, which it is.
 
> And you can't use anywhere near the 10GB link you claim to have.
 
And why do you thing that?
 
--
Ian
Ian Collins <ian-news@hotmail.com>: Oct 18 06:27PM +1300

On 10/18/16 10:32 AM, Jerry Stuckle wrote:
>> this in his office?
 
> Maybe you know of a system which can support 10G in other than very
> short bursts?
 
Any decent dual Xeon system.
 
> because an SSD can theoretically sustain a few GB/s doesn't mean
> anything. Real speed is much less than that, and by the time you throw
> processing time in there, you're talking much less than 1GB/s sustained.
 
These days we have this novel concept called "stripes" and hybrid storage.
 
--
Ian
David Brown <david.brown@hesbynett.no>: Oct 18 10:49AM +0200

On 17/10/16 23:34, Jerry Stuckle wrote:
 
>> While it may be uncommon, it is not unusual - particularly at my
>> CPOE, where we develop 10/25/40/100Gbit networking gear.
 
> There's a difference when you're developing such gear.
 
I believe Ian is also developing such devices - that is why he has such
high speed links in his office. What do /you/ think he is doing with
his 10Gb cards?
 
> But that still
> doesn't mean you have a 10G link - other than for testing.
 
Actually, that is /exactly/ what it means to have a 10 Gb link. No one
is claiming to have a 3600 TB per hour sustained transfer speed - they
are claiming to have a 10 Gb link. (Though I have no doubt that Scott's
and Ian's workstations can happily sustain that speed, at least for test
data.)
 
>> will drive four 40Gb ethernet links.
 
> How many of those are in people's offices? Or used to anywhere near
> their capacity?
 
I would not be surprised if Scott and Ian have such devices in their
offices - because they /develop/ systems with high speed networks.
 
Note the key difference here - they have not claimed to be /users/ of
10Gb networks, but /developers/. It is absolutely true that such
networks are mostly found in big server rooms, or specialised systems,
and are of little use to most desktop users. But Ian and Scott are not
"most desktop users".
Ian Collins <ian-news@hotmail.com>: Oct 18 09:54PM +1300

On 10/18/16 09:49 PM, David Brown wrote:
> networks are mostly found in big server rooms, or specialised systems,
> and are of little use to most desktop users. But Ian and Scott are not
> "most desktop users".
 
Indeed, but such systems are of use to developers using distributed
builds for large C++ code bases :)
 
--
Ian
scott@slp53.sl.home (Scott Lurndal): Oct 18 12:40PM


>Maybe you know of a system which can support 10G in other than very
>short bursts?
 
http://b2b.gigabyte.com/products/product-page.aspx?pid=5460#ov
 
This system has eight (count them, 8!) 40Gb LAN ports. It can
drive all of them simulataneously at line rate (which is half
the actual capacity of the installed SoC's).
 
Those 40Gb ports could instead be configured as 4x8x10Gb (32 10Gb ports).
scott@slp53.sl.home (Scott Lurndal): Oct 18 12:46PM

>because an SSD can theoretically sustain a few GB/s doesn't mean
>anything. Real speed is much less than that, and by the time you throw
>processing time in there, you're talking much less than 1GB/s sustained.
 
You may want to look up NVME and PCI-Express solid state storage.
 
You may also want to consider that most servers have more than a single
drive, use point-to-point intereconnects and in many cases have
PCI gen 3 root complexes on the processor itself providing 10s to 100s
of gigabytes/second of bisectional bandwidth.
 
This box <http://b2b.gigabyte.com/products/product-page.aspx?pid=5460#sp>
has 16x2x4 (128) SATA controllers; each of which can use a port multiplier
providing access to well over a thousand drive units (ssd or spinning rust).
Ian Collins <ian-news@hotmail.com>: Oct 19 09:02AM +1300

On 10/19/16 01:40 AM, Scott Lurndal wrote:
 
>> Maybe you know of a system which can support 10G in other than very
>> short bursts?
 
> http://b2b.gigabyte.com/products/product-page.aspx?pid=5460#ov
 
Scott, performance wise, how does native compiling on those beasts
compare to cross-compiling on Intel?
 
--
Ian
scott@slp53.sl.home (Scott Lurndal): Oct 18 08:16PM


>> http://b2b.gigabyte.com/products/product-page.aspx?pid=5460#ov
 
>Scott, performance wise, how does native compiling on those beasts
>compare to cross-compiling on Intel?
 
with 96 cores,
 
make -j 128
 
is sweet.
 
(The ARM64 GCC is quite good at this point in time, both in terms
of compile speed and object code & optimizer quality).
 
scott
Ian Collins <ian-news@hotmail.com>: Oct 19 09:28AM +1300

On 10/19/16 09:16 AM, Scott Lurndal wrote:
 
> is sweet.
 
> (The ARM64 GCC is quite good at this point in time, both in terms
> of compile speed and object code & optimizer quality).
 
Only 66 days until Christmas :)
 
--
Ian
Jerry Stuckle <jstucklex@attglobal.net>: Oct 18 04:35PM -0400

On 10/18/2016 1:27 AM, Ian Collins wrote:
 
>> Maybe you know of a system which can support 10G in other than very
>> short bursts?
 
> Any decent dual Xeon system.
 
No, not even a dual Xeon system can maintain 10G for anything more than
a short burst. Try transferring 100GB from your system to another one
across your 10GB line and see how long it takes.
 
>> anything. Real speed is much less than that, and by the time you throw
>> processing time in there, you're talking much less than 1GB/s sustained.
 
> These days we have this novel concept called "stripes" and hybrid storage.
 
Sure. But that still doesn't mean it can maintain that rate in other
than short bursts. Published rates are maximums, which can virtually
never be realized in actual operation. And never for any length of time.
 
--
==================
Remove the "x" from my email address
Jerry Stuckle
jstucklex@attglobal.net
==================
Jerry Stuckle <jstucklex@attglobal.net>: Oct 18 04:37PM -0400

On 10/18/2016 8:40 AM, Scott Lurndal wrote:
> drive all of them simulataneously at line rate (which is half
> the actual capacity of the installed SoC's).
 
> Those 40Gb ports could instead be configured as 4x8x10Gb (32 10Gb ports).
 
Yes, it has 8 arm processors and 8 40G ports. But nothing says it can
maintain anything near 40G on any of the ports for other than short bursts.
 
There are specs. Then there is the real world. Do you know the difference?
 
--
==================
Remove the "x" from my email address
Jerry Stuckle
jstucklex@attglobal.net
==================
Jerry Stuckle <jstucklex@attglobal.net>: Oct 18 04:39PM -0400

On 10/18/2016 8:46 AM, Scott Lurndal wrote:
 
> This box <http://b2b.gigabyte.com/products/product-page.aspx?pid=5460#sp>
> has 16x2x4 (128) SATA controllers; each of which can use a port multiplier
> providing access to well over a thousand drive units (ssd or spinning rust).
 
Sure, they may have more than a single drive. But that makes no
difference when transferring data - they don't take a byte (or word)
from the first drive, then a byte (or word) from the second drive, and
so on.
 
And port multipliers allow for access to more drives - but don't speed
up the bit rate.
 
--
==================
Remove the "x" from my email address
Jerry Stuckle
jstucklex@attglobal.net
==================
Jerry Stuckle <jstucklex@attglobal.net>: Oct 18 04:42PM -0400

On 10/18/2016 1:24 AM, Ian Collins wrote:
 
> Well it's clear that one one did make such a claim. No one using an
> agile methodology would be daft enough to do so. I can only assume that
> your misinterpreted what someone said, give you won't give a quote.
 
Too lazy to look it up yourself, I see. It's clear you don't care about
what others claim. You're just bent on proving me wrong, aren't you?
 
But you haven't. Not now, not ever.
 
> my Internet connection is rubbish, which it is.
 
>> And you can't use anywhere near the 10GB link you claim to have.
 
> And why do you thing that?
 
Because I understand how links work and what it takes to drive a 10GB
link. Unlike you, who claims to, but obviously have no real idea.
 
How long does it take you to transfer a 100GB file across your 10GB link?
 
--
==================
Remove the "x" from my email address
Jerry Stuckle
jstucklex@attglobal.net
==================
scott@slp53.sl.home (Scott Lurndal): Oct 18 08:49PM


>> Those 40Gb ports could instead be configured as 4x8x10Gb (32 10Gb ports).
 
>Yes, it has 8 arm processors and 8 40G ports. But nothing says it can
>maintain anything near 40G on any of the ports for other than short bursts.
 
8 arm SoCs, each of which has 48 processors (4 small form factor servers,
each with two SoCs connected via a coherent interconnect similar to HT or QPI).
 
That's 24 cores per 40Gb port.
 
 
>There are specs. Then there is the real world. Do you know the difference?
 
Probably better than you do, in this case.
 
These boxes are tested to ensure that they'll support line rate. The
bandwith of the internal chip communcations structure is sufficient to support
line rate. The 16 SATA controllers per SoC ensure that data can be made available
at line rate (although a few NVME cards would suffice).
 
The customers of the systems are quite pleased because they _do_ get
sustained line-rate from the SoC.
 
So yes, in the real world, the system can drive 4Gbytes/sec of traffic on
each 40Gb port until the cows come home without breaking a sweat.
"Rick C. Hodgin" <rick.c.hodgin@gmail.com>: Oct 18 11:51AM -0700

I'm using Visual Studio 2010 for a particular application, and its VC++
compiler has the requirement that this code declared with a static
global anonymous union:
 
// Fails:
union {
int a1;
HWND w1;
};
 
// Succeeds:
static union {
int a2;
HWND w2;
};
 
int APIENTRY WinMain(HINSTANCE hInst, HINSTANCE hPrevInst,
LPSTR cmdLine, int nShow)
{
// Code here
}
 
Is it a C++ requirement that global anonymous unions be declared static,
or just Microsoft's VC++ compiler requirement? If it's a C++ requirement,
why is it a requirement? What would be the difference between a global
anonymous union and a static global anonymous union? How would the
compiler see them as different?
 
Best regards,
Rick C. Hodgin
David Brown <david.brown@hesbynett.no>: Oct 18 09:17PM +0200

On 18/10/16 20:51, Rick C. Hodgin wrote:
> why is it a requirement? What would be the difference between a global
> anonymous union and a static global anonymous union? How would the
> compiler see them as different?
 
It is a C++ thing - gcc gives the error "namespace-scope anonymous
aggregates must be static" for the first version.
 
Both versions are valid C. Both versions are utterly useless (in C and
C++), because they don't define or declare a variable, a typedef, or a
union tag.
 
I am not sure why C++ rejects the first version - since the union is not
named, and does not declare anything, it has no effect that I can see.
Thus I don't see why the first version is "worse" than the second one.
Ian Collins <ian-news@hotmail.com>: Oct 19 08:17AM +1300

On 10/19/16 07:51 AM, Rick C. Hodgin wrote:
> };
 
> Is it a C++ requirement that global anonymous unions be declared static,
> or just Microsoft's VC++ compiler requirement?
 
C++. See section 9.5 pp5&6 of the C++11 standard.
 
> why is it a requirement? What would be the difference between a global
> anonymous union and a static global anonymous union? How would the
> compiler see them as different?
 
As a guess I'd say it has to be static because the members are in the
scope in which the union is declared.
 
--
Ian
Ian Collins <ian-news@hotmail.com>: Oct 19 08:41AM +1300

On 10/19/16 08:17 AM, David Brown wrote:
 
 
> I am not sure why C++ rejects the first version - since the union is not
> named, and does not declare anything, it has no effect that I can see.
> Thus I don't see why the first version is "worse" than the second one.
 
It introduces its member variables into the scope in which it is
declared, so we can write:
 
static union
{
int a1;
int w1;
};
 
int main()
{
a1 = 42;
}
 
whether we would want to is another matter!
 
--
Ian
woodbrian77@gmail.com: Oct 17 07:56PM -0700

On Monday, October 17, 2016 at 1:07:23 AM UTC-5, Juha Nieminen wrote:
> woodbrian77@gmail.com wrote:
> > Please don't swear here.
 
> Please don't be a retard here.
 
The dictionary says that word is "often offensive".
 
http://www.merriam-webster.com/dictionary/retard
 
 
"For the scientist who has lived by his faith in the power of
reason, the story ends like a bad dream. He has scaled the
mountains of ignorance, he is about to conquer the highest peak;
as he pulls himself over the final rock, he is greeted by a band
of theologians who have been sitting there for centuries."
― Robert Jastrow, God and the Astronomers
 
And there's no point in sulking when our understanding of a
subject proves to be inadequate.
 
 
Brian
Ebenezer Enterprises - "The rich and the poor meet together;
the L-RD is the Maker of them all." Proverbs 22:2
 
http://webEbenezer.net
Juha Nieminen <nospam@thanks.invalid>: Oct 18 07:16AM

>> > Please don't swear here.
 
>> Please don't be a retard here.
 
> The dictionary says that word is "often offensive".
 
Excellent. Mission accomplished.
 
--- news://freenews.netfront.net/ - complaints: news@netfront.net ---
woodbrian77@gmail.com: Oct 18 11:05AM -0700

On Tuesday, October 18, 2016 at 2:16:17 AM UTC-5, Juha Nieminen wrote:
 
> >> Please don't be a retard here.
 
> > The dictionary says that word is "often offensive".
 
> Excellent. Mission accomplished.
 
I wonder if you are like Thomas Nagel when he writes:
 
"I want atheism to be true and am made uneasy by the fact
that some of the most intelligent and well-informed people
that I know are religious believers. It isn't just that
I don't believe in God and, naturally hope that I'm right
in my belief. It's that I hope there is no God! I don't
want there to be a God."
 
I'm quoting there from "Undeniable" by Doug Axe -
https://www.amazon.com/Undeniable-Biology-Confirms-Intuition-Designed/dp/0062349589#reader_0062349589
 
Sorry if you don't like the way things are going. Sulking
about it won't help you.
 
 
Brian
Ebenezer Enterprises
http://webEbenezer.net
woodbrian77@gmail.com: Oct 18 11:38AM -0700


> I'm quoting there from "Undeniable" by Doug Axe -
> https://www.amazon.com/Undeniable-Biology-Confirms-Intuition-Designed/dp/0062349589#reader_0062349589
 
> Sorry if you don't like the way things are going.
 
In many ways I don't like the way things are going.
I agree with those who think things are going the wrong
way -- over 70% in the US. But for me things are finally
getting better. So while there's a lot going wrong, at
least my efforts are being rewarded. Perhaps you could
be happy with me about that. You've witnessed my
perseverance here in the face of many lies and smears
over the years.
 
 
Brian
Ebenezer Enterprises - "Blessed are you when people insult
you and persecute you, and falsely say all kinds of evil
against you because of Me." Matthew 5:11
 
http://webEbenezer.net
"Öö Tiib" <ootiib@hot.ee>: Oct 18 12:19AM -0700

On Monday, 17 October 2016 22:54:14 UTC+3, David Brown wrote:
> it would be useful or cost-effective for any given usage. And it seems
> to be very strongly biased towards MSVC usage (that's just a comment,
> not a criticism).
 
I think OP was just a spam and no discussion with tool manufacturers
will follow. IOW, the usual.
David Brown <david.brown@hesbynett.no>: Oct 18 11:07AM +0200

On 18/10/16 09:19, Öö Tiib wrote:
>> not a criticism).
 
> I think OP was just a spam and no discussion with tool manufacturers
> will follow. IOW, the usual.
 
I expect you are right (the OP is from the company that makes this
software) - in which case it's a kind of reverse advertising. "See how
rude and thoughtless we are - now you definitely won't want to buy
anything from us."
 
But /maybe/ this spammer will be different, and read this thread, and
follow up with an apology for spamming and then contribute to the group.
That would be /real/ advertising.
 
No harm in a bit of wildly unrealistic optimism on occasion :-)
Juha Nieminen <nospam@thanks.invalid>: Oct 18 07:18AM

> There are a lot of hypocrites in this world
 
So you decided to give us a demonstration, you self-righteous,
self-entitled dishonest hypocrite.
 
--- news://freenews.netfront.net/ - complaints: news@netfront.net ---
You received this digest because you're subscribed to updates for this group. You can change your settings on the group membership page.
To unsubscribe from this group and stop receiving emails from it send an email to comp.lang.c+++unsubscribe@googlegroups.com.
Posted by at

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)