Intel patent exposure hyperthreading technology successor, more flexible scheduling, efficient processing of tasks

CTOnews.com August 15 news, YouTube channel Moore's Law Is Dead in the latest video, that Intel's Hyperthreading technology is about to usher in a successor.

MLID says the successor to hyper-threading technology is called "Rentable Unit", which is more flexible and feasible than traditional synchronous multithreading (SMT) technology.

Intel recently applied for a new technology patent to introduce a parallel scheduling scheme across hybrid architectures.

The new patent is called "METHODS AND APPARATUS TO SCHEDULE PARALLEL INSTRUCTIONS USING HYBRID CORES", and CTOnews.com is attached with a link so that interested users can read it in depth.

Intel adopts a mixed architecture design of large and small cores (performance + efficiency) on the Alder Lake (12th generation) processor. Through dynamic algorithms, complex tasks are handed over to the performance core (P core) and background tasks to efficiency (E core).

SMT is enabled in the P core, and even if the physical kernel has two logical threads, it can only execute one instruction at a time and cannot work in parallel.

In order to overcome the problem of parallelism, Intel introduced E-core, and this patent focuses on a new thread scheduling, which can greatly improve the performance of CPU.

CTOnews.com briefly combs how thread scheduling works as follows:

Rentable Unit receives instructions and divides them according to their complexity, available resources, and CPU structure.

Rentable Unit schedules P and E cores according to tasks to ensure that all cores are active.

The principle is to ensure that no core is idle, and if the P core is idle, it can also handle the task of the E core.