Re: New CPU Architecture !
by nameta9 » Sun Oct 13, 2013 8:01 pm
CPU design hasn't changed at all in 50 years, I
was looking at old computers and CPUs, it is still exactly the same,
just a jump instruction, a move and a logical - mathematical operation,
etc. Very little has changed, funny how the most "innovative" field is
in reality the most conservative, where the basics have been put in
place decades ago with no real change (and even the higher level stuff,
Android is just Unix made 40 years ago and so forth).
Only the integration has change and improved, just Silicon Science and lithography and such, but Moore's law is running out, another 4 years and Moore's law will be no more...
There is now room to challenge it all with billions of transistors available! so for example get rid of all move and logical operations and set them in a memory all calculated ahead of time: if you have 10 registers and another 10 registers and 10 operations (add, sub, xor) just avoid all moves between registers and put them all cross connected but simply equal to one long register having the contents of the first and the second register and all ten operations, you would get just a thousand registers with everything all done ahead of time, no need to move between registers:
10 registers time 10 registers times 10 operations gives you 1,000 memory slots:
1to 10 registers , then 11 to 20 registers then 1 to 10 operations...
register 1 | register 11 | and
register 2 | register 11 | and
and so forth for all of the combinations.
Then the branches and sequence would just be signals to access any of the 1000 combinations and maybe just a few moves to record back the result into register 1 and 11 and so forth. Or a recursive set of all the results giving you 1000 registers times 1000 reg. times 10 operations giving you 10 million registers (now composed of register 1, 11, register 2, 11 and the and of the previous and results), and then some moves and sequence - branch signals and so forth.
And all kinds of permutations on these ideas, there could be so many new kinds of experimental architectures, I can't believe that in such a field we are still stuck with the same old instructions and CPU designs that we have been doing for 50 years, I can't believe it, we need new bold minds and experiments, we need creative ideas, we need new designs, new architectures, just do it man, do it, break away from the past, do it man !!!!
TARD
Only the integration has change and improved, just Silicon Science and lithography and such, but Moore's law is running out, another 4 years and Moore's law will be no more...
There is now room to challenge it all with billions of transistors available! so for example get rid of all move and logical operations and set them in a memory all calculated ahead of time: if you have 10 registers and another 10 registers and 10 operations (add, sub, xor) just avoid all moves between registers and put them all cross connected but simply equal to one long register having the contents of the first and the second register and all ten operations, you would get just a thousand registers with everything all done ahead of time, no need to move between registers:
10 registers time 10 registers times 10 operations gives you 1,000 memory slots:
1to 10 registers , then 11 to 20 registers then 1 to 10 operations...
register 1 | register 11 | and
register 2 | register 11 | and
and so forth for all of the combinations.
Then the branches and sequence would just be signals to access any of the 1000 combinations and maybe just a few moves to record back the result into register 1 and 11 and so forth. Or a recursive set of all the results giving you 1000 registers times 1000 reg. times 10 operations giving you 10 million registers (now composed of register 1, 11, register 2, 11 and the and of the previous and results), and then some moves and sequence - branch signals and so forth.
And all kinds of permutations on these ideas, there could be so many new kinds of experimental architectures, I can't believe that in such a field we are still stuck with the same old instructions and CPU designs that we have been doing for 50 years, I can't believe it, we need new bold minds and experiments, we need creative ideas, we need new designs, new architectures, just do it man, do it, break away from the past, do it man !!!!
TARD
No comments:
Post a Comment