CPU's have API's?

[Carlos1010]

Hi all,

Just a question here for Linus's video at 3:50 he explains that open GL is for CPU's too? Does that mean that there are CPU's that have DX12 API? I really thought that only GPU's have an API. And just correct me if i'm wrong, is a API the part of the GPU that moves all the data and packets? The more efficient and faster the API the faster the GPU? Is that right? Thanks in advanced! 

 

[79 Potatoes]

CPUs and GPUs don't have APIs but you can utilize them with some APIs. IDK exactly what you mean, do you mean if there's an instruction limitation or something like in GPUs for different DX versions? Probably, but you'd have to get a really old CPU, or maybe not... other than GPU the only depending factor I can think for different DX versions is OS

[adm0n]

If you fire up CPU-z it'll tell you which instruction sets are supported by your CPU. Most common ones are SSE4.x and such. And there was actually a problem at the lunch of No Mans Sky which made it impossible to play the game if your CPU did not support SSE4.2 if I remember that right : D

 

I hope that is what you were asking for.

[minibois]

An API like DX12 is all about letting the software made on it talk with all hardware more efficiently. Not only GPU's.

You must have heard about AOTS' CPU benchmarks too, which performs quite a bit better on CPU's that have more than 4 cores.

[ARikozuM]

32 minutes ago, Carlos1010 said:

-snip-

Everything has an API. Your brain has an API that allows it to interact with its hardware.

 

Jokes aside. The API is in the programs, not the CPU/GPU. The API is what programmers use to interact with the system on a deeper level.

 

https://en.wikipedia.org/wiki/Application_programming_interface

[Mark77]

A more proper way of referring to the "API" on a CPU is its "ISA" or "Instruction Set Architecture".  Basically its a set of commands or instructions which are provided by the CPU/GPU for doing manipulations on data. 

 

In recent history, there have been two major philosophies for CPU's.  "RISC" and "CISC". 

 

"RISC" is when a CPU only provides very basic instructions, but performs them wickedly fast.  Higher level computations thus are created through a permutation of simple instructions.

 

"CISC" is when a CPU provides very complex instructions, but requires a lot of specialized circuitry to perform those instructions.  This approach tends to limit the clock speed that can be used by such a CPU.

 

x86 traditionally has been known as a "CISC" architecture, while platforms like the Digital Alpha (AXP) have been known as "RISC".  However, one of the major successes of Intel over the past number of years has been their ability to create a hybrid of a CISC and RISC core.  Essentially leveraging the advantages and speed of the RISC philosophy, with the highly complex x86 "CISC" instruction set.   In light of Intel's optimization and overall approach, the whole "CISC vs RISC" debate has largely been put to rest, with the 'answer' to optimal performance being somewhere inbetween the two philosophical extremes.