Larrabee delayed: anyone know what's happening? [Computer Architecture]

Prev: PEEEEEEP
Next: Texture units as a general function

From: Bill Todd on 25 Dec 2009 04:07

Robert Myers wrote:

....

> I could (and did) hang a complete 32-bit processor off a PC/XT bus.
> The 32-bit processor (which was a single board computer), depended on
> the host OS for I/O. If I used concurrent DOS, I could do whatever I
> wanted on the host 8-bit computer while the 32-bit mini ground away on
> problems formerly done at great expense on a Cray. Which was the
> channel?

Sounds like neither one was. The host had direct control of the I/O
devices (like a channel), but was not programmed by the ancillary mini
(as an actual channel would have been). It cooperated with the mini to
provide I/O facilities for it, rather than acted as an intelligent,
programmable servant to the mini.

Why should I care? The most natural way of looking at it
> (as you describe things) is to think of the "host" PC, which was
> utterly programmable, as a mainframe channel, except that it cost
> hundreds of thousands of dollars less and didn't come with that
> priceless baby-blue logo.

Why try to use channel terminology to describe a configuration that
doesn't really fit it? IIRC the XT (ISA) bus didn't support masters
other than the host system: if so, then (at least from the hardware
viewpoint) the host called the shots (as it does in the channel
approach) but also handled the I/O devices (as it does not in the
channel approach).

>
> Where's the magic?

I haven't been following the earlier discussion very closely. If you're
talking about the 'magic' of moving code to where the data resides,
that's sort of what channels allowed the host to do, though my
impression is that the main purpose for which channels were developed
was to off-load work from the host processor rather than to move code
closer to the data per se.

To put it another way, I suspect that the original mention of channels
was merely an observation that they exhibit some similarities to the
more general idea of 'moving code rather than data around', rather than
a suggestion that channels had already solved this general issue long
ago. Someone else observed that distributed databases often move
operations to the data rather than the reverse, though again only in a
specialized manner. I don't think that either does so at anything like
as fine a grain (I think 4 KB was used as an example) as Andy has been
talking about (though I'm not sure that working at this fine a grain
would be useful, at least for any applications that I'm familiar with).

Or am I (as usual) dealing with people who have
> drunk someone's Kool-Aid? How did DEC do similar things?

By and large it didn't: it handled peripheral devices much like PCs do,
with host-resident device drivers specific to the foibles of each
specific device. As devices became more intelligent it developed a
"Mass Storage Control Protocol" (MSCP) similar to (and I think somewhat
predating) SCSI to standardize such control.

When clusters came along (1983) it developed external "Hierarchical
Storage Controllers" (HSCs, using - I think - MSCP extensions) which
could be considered somewhat akin to channel controllers, save that they
were developed not to off-load the host but to facilitate concurrently
sharing the storage they controlled among multiple hosts (and AFAIK did
not allow arbitrary code to be downloaded to execute in them, though
they were intelligent and could independently handle things like RAID -
including NV mirrored write-back caches and cooperation with other HSCs
- and backup).

While in a stand-alone host it usually makes the most sense (at least
for performance) to cache data as close to where it's processed as
possible, in a shared environment being able to cache shared dirty data
where it's stored avoids the need for complex distributed caches among
the hosts (and centralizes the NVRAM protection rather than requiring it
in any host wishing to cache such data). For that matter, centralized
read-caching, while suboptimal in efficiency, is considerably simpler
than using a distributed cache among the hosts - but VAX clusters did
implement distributed read and write-back caching for situations in
which it was desirable.

Channels may have evolved to facilitate this kind of sharing among
multiple hosts as well, from what some people have said here. If so,
that would seem to constitute an example of evolutionary convergence
from disparate roots.

- bill

From: Bill Todd on 25 Dec 2009 04:24

Andy "Krazy" Glew wrote:
> Bill Todd wrote:
>> Andy "Krazy" Glew wrote:
>>
>> ....
>>
>>> At the moment my newsreader (I switched back to Thunderbird from
>>> Seamonkey) is my only non-cloud based app. The only app that ties me
>>> to a particular machine, which I cannot use when that machine is not
>>> around.
>>
>> You do know that you can place TBird's data (it's 'profile') anywhere
>> (e.g., somewhere in the cloud) such that you can use it from anywhere
>> you can run a (compatible) copy of TBird, right?
>>
>> - bill
>
> 1) I did not know that. Thanks, I will look into it.
>
> However
>
> 2) It doesn't help me, since I really want to read news from a machine
> that has nothing like Tbird installed on it.

You don't have to 'install' Tbird to run it (i.e., it has a 'portable'
variety) - though of course you do need a machine running Windows,
Linux, or OS/X to run the appropriate version of Tbird on. I suspect
you can run such a portable version and still redirect its profile
elsewhere (e.g., into the cloud where you can use it from anywhere),
though most people just keep the profile with the portable application
on a USB drive that they carry around with them and plug into whatever
machine they'd like to use at the time. I've never tried using, say, a
profile that originated on Windows with a Tbird executing on Linux (or
vice versa), so can't comment on whether that might raise issues. And
my guess would be that even if you could do that, you'd still need
separate executables for different OSs.

There are Javascript
> systems that can run tty apps in a browser window. I'm not aware of any
> way to run Tbird on machine 1 from a browser on machine 2.
>
> Am I missing that? Is there a non-X way of running Tbird?

Sorry - I have no experience in that area.

- bill

From: Anton Ertl on 25 Dec 2009 08:28

"Del Cecchi" <delcecchiofthenorth(a)gmail.com> writes:
>"Robert Myers" <rbmyersusa(a)gmail.com> wrote in message
>news:88f15511-25c5-4e07-801b-c15ab63b4d89(a)s31g2000yqs.googlegroups.com...
>On Dec 23, 9:12 pm, "Del Cecchi" <delcecchioftheno...(a)gmail.com>
>wrote:

There is something wrong with your quoting:

1) You don't mark the quoted material as such (by prefixing ">" to
every line).

2) You quote the full message you are replying to instead of just the
relevant parts.

Please fix these problems. I am pretty sure you know how to do it
right, but just in case you don't, you might want to read

http://www.complang.tuwien.ac.at/anton/mail-news-errors.html#quoting

Moreover, please use a working email address (as is required by your
Usenet provider <http://www.individual.net/rules.php>):

|A message that you sent could not be delivered to one or more of its
|recipients. This is a permanent error. The following address(es) failed:
|
| delcecchiofthenorth(a)gmail.com
| SMTP error from remote mail server after RCPT TO:<delcecchiofthenorth(a)gmail.com>:
| host gmail-smtp-in.l.google.com [209.85.220.17]:
| 550-5.1.1 The email account that you tried to reach does not exist. Please try
| 550-5.1.1 double-checking the recipient's email address for typos or
| 550-5.1.1 unnecessary spaces. Learn more at
| 550 5.1.1 http://mail.google.com/support/bin/answer.py?answer=6596 25si22237623fxm.51

I have killfiled this address now.

- anton
--
M. Anton Ertl Some things have to be seen to be believed
anton(a)mips.complang.tuwien.ac.at Most things have to be believed to be seen
http://www.complang.tuwien.ac.at/anton/home.html

From: Del Cecchi on 25 Dec 2009 11:10

"Anton Ertl" <anton(a)mips.complang.tuwien.ac.at> wrote in message
news:2009Dec25.142847(a)mips.complang.tuwien.ac.at...
> "Del Cecchi" <delcecchiofthenorth(a)gmail.com> writes:
>>"Robert Myers" <rbmyersusa(a)gmail.com> wrote in message
>>news:88f15511-25c5-4e07-801b-c15ab63b4d89(a)s31g2000yqs.googlegroups.com...
>>On Dec 23, 9:12 pm, "Del Cecchi" <delcecchioftheno...(a)gmail.com>
>>wrote:
>
> There is something wrong with your quoting:
>
> 1) You don't mark the quoted material as such (by prefixing ">" to
> every line).
>
> 2) You quote the full message you are replying to instead of just
> the
> relevant parts.
>
> Please fix these problems. I am pretty sure you know how to do it
> right, but just in case you don't, you might want to read
>
> http://www.complang.tuwien.ac.at/anton/mail-news-errors.html#quoting
>
> Moreover, please use a working email address (as is required by your
> Usenet provider <http://www.individual.net/rules.php>):
>
> |A message that you sent could not be delivered to one or more of
> its
> |recipients. This is a permanent error. The following address(es)
> failed:
> |
> | delcecchiofthenorth(a)gmail.com
> | SMTP error from remote mail server after RCPT
> TO:<delcecchiofthenorth(a)gmail.com>:
> | host gmail-smtp-in.l.google.com [209.85.220.17]:
> | 550-5.1.1 The email account that you tried to reach does not
> exist. Please try
> | 550-5.1.1 double-checking the recipient's email address for
> typos or
> | 550-5.1.1 unnecessary spaces. Learn more at
> | 550 5.1.1
> http://mail.google.com/support/bin/answer.py?answer=6596
> 25si22237623fxm.51
>
> I have killfiled this address now.
>
> - anton
> --
> M. Anton Ertl Some things have to be seen to be
> believed
> anton(a)mips.complang.tuwien.ac.at Most things have to be believed to
> be seen
> http://www.complang.tuwien.ac.at/anton/home.html

Sorry for that. The quoting pecularities seem to only by on Robert
Meyers for some odd reason. This quoting appears to me to be ok on
this reply, for example.

I will change the email. It is a hangover from the era before decent
spam filtering.

del

From: Del Cecchi on 25 Dec 2009 11:12

First | Prev | Next | Last
Pages: 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
Prev: PEEEEEEP
Next: Texture units as a general function