OSDev.org

But you still haven't told me what advantages it has. Why would I choose it rather than the, as you have told me, more efficient x86? And why doesn't anyone make them?

Hi,

Geri wrote:no. my system is good, and i am poroud of it.

You system is not good, and everyone with any common sense has told you why it will never be good, starting from your first posts on these forums 2 years ago, and including every other forum that you've mentioned SUBLEQ and/or Dawn on.

The only thing you can possibly be proud of is your ability to remain a deluded fool.


Cheers,

Brendan

iansjack wrote:But you still haven't told me what advantages it has. Why would I choose it rather than the, as you have told me, more efficient x86? And why doesn't anyone make them?

nobody were able to make an operating system for it, thats why it never manufactured. but now this is finally changed. the most advantages of subleq is the small transistor amout and power consumption, which makes the architecture very efficient, and can be designed from fragments of the cost of a generic cpu, and will dissipate very small power. its small, so it makes possible to have a lot of cpu cores. its easy to understand its architecture, so you dont have to deal with understanding the isa. subleq is possibly the future of the computer industry, it also removes the extremely oversized corporation mammoths and they peons from the game. subleq is the freedom that this industry is searching for.

You seem to be swapping ease of design for very inefficient software. As the design for very powerful processors that are power efficient already exist, I don't see that as an advantage. Perhaps if I were building my own processor, and didn't require good performance; but as things are I can buy processors with as much power as I need that are easy to program. So, other than as an intellectual exercise, I see no profit in using such an unsophisticated design.

You seem to strangely be really emphasizing fewer transistors being involved as a big plus, even though transistor counts haven't been a limitation in processor design for quite a while now. You have multiple cores but no way to synchronize between them, the instruction set is even less dense than typical RISC designs, making the same amount of cache space do less. Power consumption will be better but at the cost of a ton of performance. Similarly, a large part of the cost of making a CPU isn't in the design, but in the production. Additionally, any RISC architecture is far easier to understand than SUBLEQ, simply because each instruction maps to a logical operation instead of relying on one instruction to implement all operations. It's sort of like how after a point, it's easier to understand a digital circuit by various different logic gates than by individual transistors.

Hi,

Geri wrote:

iansjack wrote:But you still haven't told me what advantages it has. Why would I choose it rather than the, as you have told me, more efficient x86? And why doesn't anyone make them?

nobody were able to make an operating system for it,

Anyone could've made an OS for it, or ported an existing OS to it.

Geri wrote: thats why it never manufactured.

CPU designers don't work like that - they create CPUs despite knowing there's no software for the CPU, in the same way that OS developers create an OS despite knowing there's no applications or drivers for their OS.

The fact is that virtually every CPU designer knows about and has thought about OISC and SUBLEQ (in the same way that it'd be virtually impossible to find a programmer that hasn't heard about "Hello World"); and every single one of them knows that it has zero practical value; and this is why it has never been manufactured and will never be manufactured.

Geri wrote:but now this is finally changed.

No, it has not changed. In a hypothetical world that can never exist, if someone did create a SUBLEQ CPU they would look at Dawn and see that it's extremely poor and then port something like Linux to SUBLEQ.

Geri wrote:the most advantages of subleq is the small transistor amout and power consumption

No, this is not an advantage. Transistors cost less than dirt and "less transistors" is irrelevant. For any specific performance requirement, SUBLEQ has inferior power consumption because you need a significantly higher number of instructions to get anything done.

Geri wrote:which makes the architecture very efficient

No, it makes SUBLEQ a retarded joke.

Geri wrote:and can be designed from fragments of the cost of a generic cpu, and will dissipate very small power.

No, it's still an expensive waste of power that is inferior.

Geri wrote:its small, so it makes possible to have a lot of cpu cores.

Except, we already have more CPU cores than software developers are able to make effective use of (due to Amdahl's law, etc).

Geri wrote:its easy to understand its architecture, so you dont have to deal with understanding the isa.

People that actually buy computers don't care how complex it is, they care about features, performance, power consumption and price. Because your architecture is a deliberately crippled joke; it's extremely complicated for hardware and software developers to get acceptable features, performance, power consumption and price out of your architecture.

Geri wrote:subleq is possibly the future of the computer industry

No, that is not possible.

Geri wrote:it also removes the extremely oversized corporation mammoths and they peons from the game. subleq is the freedom that this industry is searching for.

No, the industry is not searching for this freedom (the industry is searching for ways to increase profit and prevent competition), and if the industry was searching for this freedom they would've switched to something like RISC-V already.


Cheers,

Brendan

Brendan wrote: Anyone could've made an OS for it, or ported an existing OS to it.

Anyone could've made a rocket and go to mars.

Brendan wrote: CPU designers don't work like that - they create CPUs despite knowing there's no software for the CPU, in the same way that OS developers create an OS despite knowing there's no applications (....)

chicken vs egg

Brendan wrote:No, it has not changed.

well i made the operating system, so its changed. and porting linux and g++ is not an option, they are designed to run on much complex processors.

Brendan wrote: No, this is not an advantage. Transistors cost less than dirt

a cpu sized a piece of dirt costs 0.2-1 usd, a cpu sized your palm costs 20k usd, more transistor = bigger cpu = big price.

Brendan wrote: you need a significantly higher number of instructions to get anything done.

feeding billions of transistors with electricity just becouse your complex instruction set is less energy efficient than executing more opcode for a certain type of task.

Brendan wrote:No, it makes SUBLEQ a retarded joke.

if you dont understand subleq that will not necessary make it a retarded joke.

Brendan wrote: No, it's still an expensive waste of power that is inferior.

great criticism on x86 and arm.

Brendan wrote: Except, we already have more CPU cores than software developers are able to make effective use of (due to Amdahl's law, etc).

this is a relative limit, that warns the programmers of designing proper algorythmics without long single-thread presetups to be able to access multi-cpu efficiently. it has nothing to do what and how developers are able to use the cpu cores in general.

Brendan wrote: People that actually buy computers don't care how complex it is, they care about features, performance, power consumption and price.

subleq is very efficient in these.

Brendan wrote: No, the industry is not searching for this freedom (the industry is searching for ways to increase profit and prevent competition)

the people are dictating, not the industry. x86 and arm canot be sustained in the future, these platforms will not be able to offer better alternatives than the current. modern i7-s are barely faster+efficient than the i7-s from 5 years before. the architecture hit a wall, and you will not be able to break it any more.

hgoel: we are still having 8 (virtual) cored cpus generally on x86 and both on arm. on this platforms we are also having integrated gpus as x86s and arm considered inefficient to do the software 3d renders - cpus are not concentrating enoguh computing performance on they area. as we can see, the performance is not enough dense (thats why we need igp/gpu). current risc cpus (like arm) are also falling to the same design trap. on the current areas we should be able to get 50-500x more efficiency out from the chips (even much bigger than we can get on gpus) and subleq is capable of this. there are still problems - as you mentioned - with the design side to make the synchronization beethwen efficient, the instruction set will not magically solve that issue - but thats a different issue, its not a question of the instruction set and not a question of operating system, so i cant really do anything much about it (excpet the fact i allow long sync times on the platform, as i alreday do).

Hi,

Geri, for the last few months I've been trying to figure out if you're a troll, or if you're actually so stupid that you genuinely believe what you say. I'm still unable to decide. However; I think we've gone well beyond the point where it matters.

I will give you 7 days to show any sign of improvement. If nothing changes; I will either ban you from these forums for being a troll, or I will ban you from these forums for being an ignorant fool that refuses to learn. The exact reason will be chosen by rolling a dice.


Cheers,

Brendan