Programmable Devices
CPLDs, FPGAs, SoC FPGAs, Configuration, and Transceivers
20641 Discussions

Next gen processor FS90, right direction?

sf90p
Beginner
1,038 Views
OK I'm proposing my new design, the FS90 Processor.

Current CPU, GPU's a lot of components.

OK here's the story, I would like a serial based fractional math based processor as fractional math is simpler and faster than decimal math and a serial based processor would be made of less components.

So heres the answer, if you rotate a memory data table 90 deg in memory you can have simultaneus single bit serial data access (64 channels using 64 bit memory) using standard memory, great for graphics. Add in an arithmetic matrix to the memory, and some command ROM and a complete computer may be possible.

My FS90 Processor, Fractional Serial (rotated memory by 90 deg)
fs90processor.com


So, any thoughts?
0 Kudos
12 Replies
BoonBengT_Intel
Moderator
971 Views

Hi @sf90p,


Thank you for posting in Intel community forum and hope all is well.

This does not seems to be Intel FPGA product, but more towards processors.


Hence allow me to route you to the correct channels and relevant support will get back to you.

Hope that clarify.


Best Wishes

BB


0 Kudos
DeividA_Intel
Moderator
946 Views

Hello sf90p,  

  


Thank you for posting on the Intel® communities. I do appreciate your feedback related to this new processor technology and architecture.


I will close this thread since support is not needed, however, we will keep your feedback in mind. Feel free to open a new thread if you need support or if you would like to share your ideas.



Please keep in mind that this thread will no longer be monitored by Intel.  


Regards,  

Deivid A.  

Intel Customer Support Technician  


0 Kudos
sf90p
Beginner
926 Views

OK, let's look at an engineering pont over view, cm, mm vs inches and fractions of.

Computers work in base 2, so does 1/2 inch, 1/4, 1/8 1/16, 1/32, 1/64, 1/128 & 1/256.

Whilst mm goes in dividing by 10 or base 10, also 0.5 or 1/2cm, or 500/1000, 1000/10000, 2000/10000, it all goes strange when using computer maths.

My point is fractional math is faster, less steps involved in calculating the answer hence is faster.

Serial based math computing is easy!

 

I think we need to move away from base 10 in as far as human & computer interaction is concerned and work with whichever is faster as computers are very fast nowdays and translating from and to human form is only necessary in print commands when a person needs to read or input its value.

 

Let's work in base 2 systems.

 

I'm wearing a "Base 2" t-shirt!

 

Let's store as w.f.o whole.fraction.of or fraction.of like 123/32

0 Kudos
AlHill
Super User
915 Views

Systems are already base-2.

And, you are abusing that edit button.

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

0 Kudos
sf90p
Beginner
903 Views

You are correct, a base 10 number system stored in base 2. I'm calling for a base 10 sizeing with base 2, stored in a base 2.

It's closer to the metal. It's about fractional math not decimal math on the metal.

0 Kudos
ak6dn
Valued Contributor III
843 Views

What percentage of most applications are spent doing fractional math?

I suspect it is in the single digits, i.e. less than 10%. Probably closer to 1%.

Optimizing 1% of your code to execute 10X (or even 100X) faster is not going to be noticeable.

What you have proposed might make an interesting technical paper or senior project, but is not a practical commercial application.

 

0 Kudos
sf90p
Beginner
832 Views

Yes though 10-100 times faster is a noticeable thing. It's a choice I guess though converting a lot would result in a rounding issue.

Popular may be popular.

Also have to look at practical advantages, half way along an edge is just simple, so is scaling so 100 pillars on a building is just x = n * (buildingwidth/pillarcount.) is something we already do.

 

Not sure what PI equals to as a fraction to 64 bit places.

0 Kudos
AlHill
Super User
826 Views

This is a really silly discussion.

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

0 Kudos
sf90p
Beginner
819 Views

Not really a silly discussion, 100 times faster and I would be seeing rainfall with puddle reflections in my games.

 

It's wrong to hardcode and decimal math is a form of hardcode format.

Is it a way cogs really turn? You could say a fraction of or a distance, eithway fraction math may work and be faster than decimal math?

 

There's probably a Paper out there...

0 Kudos
AlHill
Super User
818 Views

A matter of opinion.  However, this is not the place for such a discussion.  Take it somewhere else.

 

Doc (not an Intel employee or contractor)
[Maybe Windows 12 will be better]

0 Kudos
ak6dn
Valued Contributor III
820 Views

Absolutely silly. And completely unrelated to FPGAs, mostly. Not even talking about implementing such code in an FPGA. Just someone's pet project.

 

"Yes though 10-100 times faster is a noticeable thing." ... No it is not, if you only do it 1% of the time.

0 Kudos
sf90p
Beginner
810 Views

Then keep it to FPGA"s then and I would not have to answer your own questions lol

I was kind of looking for others...

 

Soon in a header file not far away. .. 

_Win_64, _OSX, _Linux, _Android... 

_Posix, _ANSI

_DECIMAL._64, _FRACTIONAL_64

 

I will give it a go, use one of the memory serial channels for one core and expand in to graphics once I have a command prompt and is confible.

0 Kudos
Reply