<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
</head>
<body text="#000000" bgcolor="#FFFFFF">
On 01/31/2018 09:59 AM, David Collier-Brown via talk wrote:<br>
<blockquote type="cite"
cite="mid:415acc9c-1d9f-4a74-8151-89458b1d3947@rogers.com">
<meta http-equiv="Content-Type" content="text/html; charset=utf-8">
<div class="moz-cite-prefix">
<pre wrap="">On Tue, Jan 30, 2018 at 03:49:54PM -0500, D. Hugh Redelmeier via talk wrote:
</pre>
<blockquote type="cite">
<pre wrap="">Another example: ARM is just now (more than 30 years on) targeting
datacentres. Interestingly, big iron has previously mostly been
replaced by co-ordinated hordes of x86 micros.</pre>
</blockquote>
There are two different cases to consider when doing data
centers: <br>
<ul>
<li>uniprocessors for individual tasks or trivially
parallelizable ones <br>
</li>
<li>multiprocessors for things that aren't parallelizable</li>
</ul>
<p>Anybody can provide the first. The second is harder.<br>
</p>
</div>
</blockquote>
Its worse than that.<br>
You have UMA/NUMA and Cache consistency issues and these are the
kinds of things that are facing the Intel/AMD designers now.<br>
<br>
<blockquote type="cite"
cite="mid:415acc9c-1d9f-4a74-8151-89458b1d3947@rogers.com">
<div class="moz-cite-prefix">
<p> </p>
<p>Mips had three MMUs, one of which was for each of the above
cases and one was a trivial one for embedded, so 32-CPU Mips
machines were available.</p>
<p>IBM and Sun spent lots of money designing backplanes that
could support >= 32 sockets: Sun when so far as to license
a Cray design when their in-house scheme failed to scale.</p>
</div>
</blockquote>
Hypertransport was an outgrowth of an AMD/DEC project to develop a
processor interconnect.<br>
Its the thing that allowed AMD to build systems with up to 8
processors in a NUMA architecture while Intel was stuck with 2
processor systems using a shared memory bus.<br>
Intel now has QPI so that they can have more processors in a system.<br>
I believe the theoretical limit for Hypertransport systems was 64
processors but I am not sure anybody ever built a system that big.<br>
<br>
The big problem is that once you get that many processors in a box
you have heat dissipation issues and all these interconnect buses
have distance limitations so there is no putting them on the
backplane.<br>
<br>
<blockquote type="cite"
cite="mid:415acc9c-1d9f-4a74-8151-89458b1d3947@rogers.com">
<div class="moz-cite-prefix">
<p>Until and unless chip vendors spend significant time and
money on MMUs and backplanes, they won't have an offering in
the second case, and will have chosen to limit themselves to a
large but limited role in the datacentre.</p>
<br>
</div>
</blockquote>
The biggest problem with multiprocessor systems is synchronization.<br>
<br>
For the most part software uses some kind of memory locking
instructions to manage concurrency but as more processors are added
it becomes difficult to insure that memory reads and writes can
remain atomic.<br>
<br>
<br>
Once you get into the realm of supercomputers your using some kind
of bus where your passing data between separate processors.<br>
<br>
<br>
<pre class="moz-signature" cols="72">--
Alvin Starr || land: (905)513-7688
Netvel Inc. || Cell: (416)806-0133
<a class="moz-txt-link-abbreviated" href="mailto:alvin@netvel.net">alvin@netvel.net</a> ||
</pre>
</body>
</html>