Not that I am any kind of an AI expert, but I had recently learned of newly developed processors that include AI-specific hardware enhancements and decided to figure out what they had actually done and how it fit in. As you can see from previous posts below, I am in the habit of occasionally providing for the web site The Next Platform deep dives on some of the higher-level concepts which these folks address. What I learned was interesting enough that I thought I could explain via such an article. That article DOING THE MATH ON CPU NATIVE AI INFERENCE was just published today. Enjoy.
This, another article on The Next Platform, is pretty esoteric but it is also my idea of a fun investigation into computer architecture. Enjoy.
I’ve been thinking that I’ve been becoming technically well past my prime. Perhaps not. As an adjunct professor at WSU in computer science, I’d been worried about whether the Zoom infrastructure would be able to take on the load of all of us suddenly showing up to keep our education – and indeed much of our economy – infrastructure flowing. So, I did some research to find out how it is that they pulled it off and in the process got an article on the subject published. Here it is. Enjoy.
After quite some time (since 01/2017), I again have had the pleasure of having an article published in the web site The Next Platform. This one, as was true for a number of the preceding articles, deals with the use of a persistent memory attached directly to the processor chips in much the same way is RAM has been for years. That introduces a number of changes required in software, in this case within the kernel of operating systems.
You can find this article at https://www.nextplatform.com/2019/05/20/those-without-persistent-memory-are-fated-to-repeat-it/
As you will see in the page with link provided below, I happen to be part of a team at a local university which mentors math students in something similar to a study hall. What's interesting about it is that the concepts used there and outlined in the referenced page have what appears to be a means of improving productivity throughout our school systems. Read on. I'll let you see the opportunity for yourself.
A good friend of mine suggested after reading my previous post that that was interesting and all, but what we really needed was an outline on what should/would go into a Single-Payer bill. That is, what would Single-Payer really be. A sat on that for way too long, but it began to dawn on me that even those who are pushing Single-Payer haven't thought that through to much. Part of the problem is that, although Single-Payer can be feasible in a number of different forms, there really are a lot of variations on that theme. It happens that the devil is in the details. It also struck me that, even though there are a lot of folks backing Single-Payer, a lot more would provide their buy-in if their interests were supported. So, not only do we need some description of what Single-Payer could be, but we need to start a nation-wide discussion concerning what we want from it. Presumptuous, I know, but we need to start what should be a completely non-partisan process of creating Single-Payer somewhere.
I will admit to being politically what someone might call fiscally conservative. Still, when someone suggests an alternative solution like single-payer healthcare coverage to our current health care dilemma, I am willing to look into it. So, me, along with a few others, thought we'd attempt to outline the essentials of what was learned, and attempt to do it in a politically nonpartisan manner. We are sponsored by no one. I hope you will find it enlightening. I should add, as you read through this page - and the links found therein - we worked only to provide information in context; the decision is ultimately yours.
Enjoy.... A Layman's Study of Single-Payer
Normally when I write a new article it is because I found the underlying concepts of something new to be interesting. This time around I thought I'd make an attempt to more clearly explain something that has been around for quite a while, but at the same time seems to scare folks. That is simply how to program in a way that uses more of the many processors available in even the most inexpensive systems. Here is that set of articles:
- The Essentials Of Multiprocessor Programming ... (January 5, 2017)
- What Is SMP Without Shared Memory? ... (January 11, 2017)
- The Essence Of Multi-Threaded Applications ... (January 17, 2017)
- Multi-Threaded Programming By Hand Versus OpenMP ... (January 23, 2017)
For a lot of years now computer systems with a lot of processors, each often with their own cache, have used a common cache coherence protocol to allow all of them to access their shared memory and the contents of each others cache. Esoteric enough for you? What this has meant, though, is that all of the processors in such systems are essentially identical processors. High capacity systems can be built by having a lot of such homogenous processors, but it happens that there tends to be a limit in how large these system can get. More recently, specialized compute accelerators have been showing up on the market, with these producing outstanding performance in specialized situations, but being different, they have not been able to access the system's memory in the same way as the homogenous SMP's processors. Wanting the best of both words, a consortium of companies have formed the CCIX with the intent of creating still faster heterogeneous systems of both generic processors and accelerators. Outlining what they are up against is largely the purpose of two articles I've recently had published on TheNextPlatform ....
- Drilling Into The CCIX Coherence Standard ... (July 13, 2016)
- Weaving Accelerators Into The Memory Complex ... (July 14, 2016)
After having published on HPE's "The Machine" in a preceding set of articles, I was provided the opportunity to follow some of the work being done by Dhruva Chakrabarti and his team at Hewlett-Packard Labs in the development of a programming model in support of persistent memory. This resulted in a two-part series attempting to help describe such a persistent-memory programming model. These can be found here:
- Programming For Persistent Memory Takes Persistence (April 21, 2016)
- First Steps In The Program Model For Persistent Memory (April 25, 2016)