DSPRelated.com
Forums
Online · Live + on demand
February 9-11, 2027
Price goes up Jan 1: register now to lock in today's rate
Register Now →

the performance of x86 processor and DSP

Started by jogging ●August 6, 2009
Jerry Avins  <jya@ieee.org> wrote:

>Rune Allnor wrote:
>> I don't think it would qualifty - they use the >> decimal number system. You cold of course go weird >> and use the babylonian dodecal system, but then the
>I assume that's a European version of duodecimal.
This is tangential, but ages ago I encountered a computer that used what they called bi-quinary arithmetic (which is subtly different from binary-coded decimal, but close). Steve
On 7 Aug, 18:01, Jerry Avins <j...@ieee.org> wrote:
> Rune Allnor wrote: > > &#4294967295; &#4294967295;... > > > I don't think it would qualifty - they use the > > decimal number system. You cold of course go weird > > and use the babylonian dodecal system, but then the > > I assume that's a European version of duodecimal.
It's Greek'ish for 'base 12': http://en.wikipedia.org/wiki/Dodecahedron
> > number representations would be piecewise continuous, > > not discrete. > > Please amplify on or provide a link to the notion of a continuous rather > than discrete number system.
It was a bad attempt to hint on the 12-base method of counting on the fingers. In the decimal system people count each finger as 1, so since we have 10 fingers the decimal system naturally emerges. The babylonians apparently used the thumb on one hand to count each bone in the other fingers. The four fingers have three bones each (at least at the outset), so a base-12 system naturally emerges. Of course, the babylonians ended up with a base-60 number system, which is natural if one uses the same thumb to count the fingers on the other hand. Don't know how the abacus works, but the hands and fingers ought to be quite a useful tools, if used as an abacus. My awful attempt at a joke was based on that the bones in question are connected in triplets. But then, it's been close to 30C inside, for a few days now. WAY too hot for me. Rune
Steve Pope wrote:
> Jerry Avins <jya@ieee.org> wrote: > >> Rune Allnor wrote: > >>> I don't think it would qualifty - they use the >>> decimal number system. You cold of course go weird >>> and use the babylonian dodecal system, but then the > >> I assume that's a European version of duodecimal. > > This is tangential, but ages ago I encountered a computer > that used what they called bi-quinary arithmetic > (which is subtly different from binary-coded decimal, > but close).
TI made a counter chip that could be configures either BCD or biquinary. Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Rune Allnor wrote:
> On 7 Aug, 18:01, Jerry Avins <j...@ieee.org> wrote: >> Rune Allnor wrote: >> >> ... >> >>> I don't think it would qualifty - they use the >>> decimal number system. You cold of course go weird >>> and use the babylonian dodecal system, but then the >> I assume that's a European version of duodecimal. > > It's Greek'ish for 'base 12': > > http://en.wikipedia.org/wiki/Dodecahedron > >>> number representations would be piecewise continuous, >>> not discrete. >> Please amplify on or provide a link to the notion of a continuous rather >> than discrete number system. > > It was a bad attempt to hint on the 12-base method > of counting on the fingers. In the decimal system > people count each finger as 1, so since we have 10 > fingers the decimal system naturally emerges. > > The babylonians apparently used the thumb on one hand > to count each bone in the other fingers. The four fingers > have three bones each (at least at the outset), so > a base-12 system naturally emerges. > > Of course, the babylonians ended up with a base-60 > number system, which is natural if one uses the same > thumb to count the fingers on the other hand. Don't > know how the abacus works, but the hands and fingers > ought to be quite a useful tools, if used as an abacus. > > My awful attempt at a joke was based on that the bones > in question are connected in triplets. But then, it's > been close to 30C inside, for a few days now. WAY too > hot for me.
Whew! That's a relief! (The missing smiley, not the heat. You have mt sympathy for that.) I understand that there's a town in France where six-fingered people are common. Maybe in Babylon too? Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
On Aug 7, 10:15&#4294967295;pm, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
> jogging wrote: > >>> can x86 processors replace DSPs? > > >>This stupid question comes to anybody's mind at one point of time. The > >>answer is "it depends". > > > Hi, Vladimir > > Can you share your experience when you encountered this stupid > > question? > > I think it is not possible for an engineer to learn every computing > > platform. > > But you are not familiar with every computing platform, how can you > > make a better choice? > > The strong points of x86 is that they are supported by the cheap off the > shelf software, cheap off the shelf hardware and cheap off the street > programmers. But those advantages naturally limit your choices to > PC-compatible architecture and the desktop OS like Windows or Linux. >
Besides these, there are a lot of open source projects which some people use them directly in their project.
> The top of the line x86s are very inconvenient and demanding from the > hardware standpoint, so rolling your own board is hardly an option. > The industrial PCs aren't cheap. The I/O cards aren't cheap either. > > The idea of using a common x86 PC for the real time signal processing > seems very attractive initially, and many people fall into this trap. > Even if you managed to master the particular OS + drivers + software + > hardware bundle, everything will change completely within a year or so. > &#4294967295; You will have to keep up with the continuous software and hardware > changes. Also, the reliability of the common PC sucks. > > However there are the application areas (like IP multimedia servers, for > example), which are traditionally covered by PC x86 solutions.
For IP multimedia servers, I learn that some companies such as radvision use DSP boards. In my opinion solutions using DSP is more challenging.
> > Now return back to your cubicle. The x86 won't replace DSPs as they > didn't replace the common microcontrollers before, although there is a > significant penetration in some application areas. > > Vladimir Vassilevsky > DSP and Mixed Signal Design Consultanthttp://www.abvolt.com
Best Regards Jogging
On Aug 7, 11:17&#4294967295;pm, Jerry Avins <j...@ieee.org> wrote:
> Vladimir Vassilevsky wrote: > > > jogging wrote: > > >>>> can x86 processors replace DSPs? > > >>> This stupid question comes to anybody's mind at one point of time. The > >>> answer is "it depends". > > >> Hi, Vladimir > >> Can you share your experience when you encountered this stupid > >> question? > >> I think it is not possible for an engineer to learn every computing > >> platform. > >> But you are not familiar with every computing platform, how can you > >> make a better choice? > > > The strong points of x86 is that they are supported by the cheap off the > > shelf software, cheap off the shelf hardware and cheap off the street > > programmers. But those advantages naturally limit your choices to > > PC-compatible architecture and the desktop OS like Windows or Linux. > > > The top of the line x86s are very inconvenient and demanding from the > > hardware standpoint, so rolling your own board is hardly an option. > > I know one case where a commercial Pentium board was used with > proprietary software; no OS between it and the hardware. It worked well > but was a mistake anyway. The design was decided on because it was > believed that it would be replicated several times. Instead, it was a > technically successful one-off.
The design was a technically successful one off. Why do you call it a mistake?
> > > The industrial PCs aren't cheap. The I/O cards aren't cheap either. > > > The idea of using a common x86 PC for the real time signal processing > > seems very attractive initially, and many people fall into this trap. > > Even if you managed to master the particular OS + drivers + software + > > hardware bundle, everything will change completely within a year or so. > > &#4294967295;You will have to keep up with the continuous software and hardware > > changes. Also, the reliability of the common PC sucks. > > > However there are the application areas (like IP multimedia servers, for > > example), which are traditionally covered by PC x86 solutions. > > > Now return back to your cubicle. The x86 won't replace DSPs as they > > didn't replace the common microcontrollers before, although there is a > > significant penetration in some application areas. > > I'm sure you realize that you were addressing a seasoned professional > who makes his living doing DSP on x86 hardware. What gives with you? > > Jerry > -- > Engineering is the art of making what you want from things you can get. > &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;
Best Regards Jogging
jogging wrote:
> On Aug 7, 11:17 pm, Jerry Avins <j...@ieee.org> wrote:
...
>> I know one case where a commercial Pentium board was used with >> proprietary software; no OS between it and the hardware. It worked well >> but was a mistake anyway. The design was decided on because it was >> believed that it would be replicated several times. Instead, it was a >> technically successful one-off. > > The design was a technically successful one off. Why do you call it a > mistake?
Using a commercial motherboard was intended to save costs in a moderate-volume product run. For a one-off, the development cost was excessive. Doesn't what I wrote imply that? ... Jerry -- Engineering is the art of making what you want from things you can get. &#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;&#4294967295;

Jerry Avins wrote:

>> On Aug 7, 11:17 pm, Jerry Avins <j...@ieee.org> wrote:
>>> I know one case where a commercial Pentium board was used with >>> proprietary software; no OS between it and the hardware. It worked well >>> but was a mistake anyway.
Exactly.
> The design was decided on because it was >>> believed that it would be replicated several times. Instead, it was a >>> technically successful one-off. > Using a commercial motherboard was intended to save costs in a > moderate-volume product run.
It could be fine for the one-time use such as R&D work in the lab, but it is very impractical for production.
> For a one-off, the development cost was > excessive. Doesn't what I wrote imply that?
Some people think that if they stuff an evaluation board with the opensource junk from the Internet, and package it into the radioshack box, that makes a successfull product :))) VLV
On Aug 9, 1:15&#4294967295;am, Vladimir Vassilevsky <nos...@nowhere.com> wrote:
> Jerry Avins wrote: > >> On Aug 7, 11:17 pm, Jerry Avins <j...@ieee.org> wrote: > >>> I know one case where a commercial Pentium board was used with > >>> proprietary software; no OS between it and the hardware. It worked well > >>> but was a mistake anyway. > > Exactly. > > > The design was decided on because it was > >>> believed that it would be replicated several times. Instead, it was a > >>> technically successful one-off. > > Using a commercial motherboard was intended to save costs in a > > moderate-volume product run. > > It could be fine for the one-time use such as R&D work in the lab, but > it is very impractical for production. > > > For a one-off, the development cost was > > excessive. Doesn't what I wrote imply that? > > Some people think that if they stuff an evaluation board with the > opensource junk from the Internet, and package it into the radioshack > box, that makes a successfull product :))) > > VLV
If a product is table, the performance is great and the cost is reasonable, a product is successful. Customers don't care about the technology used. They care abort performance, cost and maybe UI. Best Regards Jogging
>> Jerry Avins wrote: >> >> On Aug 7, 11:17 pm, Jerry Avins <j...@ieee.org> wrote: >> > For a one-off, the development cost was >> > excessive. Doesn't what I wrote imply that? > >If a product is table, the performance is great and the cost is >reasonable, a product is successful. Customers don't care about the >technology used. They care abort performance, cost and maybe UI. > >Best Regards >Jogging
If it doesn't yield a profit, it generally doesn't matter how well the technology worked. Sometimes there are other factors, like making the company name better known to the general public, in spite of a net loss (e.g., the DLP). Even then, it is hoped that those factors make money in the long run. Development cost is one-time, and can be expected to be amortized over many sales. Very simple economics... Michael