Could "hypothreading" ever be a thing? Ie: two cores

seems your brain was hypothreading when it decided to post this

>>62072816
/thread

could 2 + 2 = 5?

>>62072816
>>62072823
>>62072838
>the absolute state of /g/

>>62072785
People thinking they have great ideas while having no idea what they're talking about.

Its already hard enough to keep track of your shit while hyperthreading. The definition of an explicit single core load is that the workload cannot be split into two parts. If you can split it, then you would just two separate cores - aka Hyperthreading or Multithreading.

You didnt put alot of thought behind this did you.

>>62073183
I didn't. It was just kind of a random thought I had.

>>62072838
>could 2 + 2 = 5?
Only for very large values of 2.

>>62072785
A single core already has multiple execution units inside, that's why we have massive data and instruction caches. It's to extract as much parallelism as possible from a single thread of instructions.

jesus christ all of you fuckwits that don't know anything about anything need to stop fucking posting
op, you just came up with instruction level parallelism, the closest to your idea is probably superscalar, in which the data dependencies are checked with an algorithm implemented in hardware so that instructions which don't need to be executed sequentially are executed in parallel on multiple of the several execution units (fpus, etc) if possible.
through this, they achieve per-core performance that is physically impossible with modern technology, sometimes upwards of (effectively) 30 giga-cycles per second.
it is a fairly complex idea though, and i hope that we can eventually come up with a simpler method of acquiring performance, but for now true (or "high level") paralleliam is too complicated, so this is a nice stopgap.

>>62072838
Only if you truly believe in the party

>>62073641
So you can rephrase OP's question in an interesting way.

Would it be possible to design a processor architecture with a unified set of execution units which can be assigned to as many or as few threads as the workload demands, instead of discreet cores?

>>62073386
This.

>>62074088
You'd need a pretty complex scheduler for a design like that I'd assume.

Also, most single threaded workloads are already only barely saturating all available execution ports on modern CPUs. There is a limit to instruction level parallelism due to data and control dependencies unfortunately.
That's also the reason why SMT works pretty well. The virtual threads use either unused execution ports or pipeline stalls, but most of SMT's performance comes from the unused ports.

>>62076835
Man, it has gotten to where it's weird to see someone know some shit here rather than just shitposters