Improve rust version performance

[qarmin]

The main issue was that instead of using OnceLock (which was stabilized about half a year after the initial version was merged), a custom implementation was used.
That implementation relied on unwrap in a critical section, which significantly reduced performance.

Additionally, it caused the following warnings in Rust 1.91.0:

warning: creating a shared reference to mutable static
--> rust/smooth.rs:831:9
|
831 | / STATIC_DATA.0.call_once(|| {
832 | | STATIC_DATA.1 = Some(initStaticData());
833 | | });
| |__________^ shared reference to mutable static
|
= note: for more information, see <https://doc.rust-lang.org/edition-guide/rust-2024/static-mut-references.html>
= note: shared references to mutable statics are dangerous; it's undefined behavior if the static is mutated or if a mutable reference is created for it while the shared reference lives
warning: creating a shared reference to mutable static
--> rust/smooth.rs:834:9
|
834 | STATIC_DATA.1.as_ref().unwrap()
| ^^^^^^^^^^^^^^^^^^^^^^ shared reference to mutable static
|
= note: for more information, see <https://doc.rust-lang.org/edition-guide/rust-2024/static-mut-references.html>
= note: shared references to mutable statics are dangerous; it's undefined behavior if the static is mutated or if a mutable reference is created for it while the shared reference lives
warning: `opensimplex2` (lib) generated 4 warnings

I’ve also added simple stability tests to prevent accidental changes to the function output in the future.
An additional example was included to make it easier to verify performance.

Old implementation - ~640ms

New implementation - ~490ms

[EndrII]

Nice work, now is really faster !
But still slower than C., i think The Rust is good balance choose, but performance choose it should be C.

MarcoCiaramella C Impl 2D: 623 msec
Deprecated C Impl 2D: 617 msec
Rust Impl 2D: 686 msec

[EndrII]

good job

[CryZe]

Based on this PR I made some further improvements by precomputing the lookup tables entirely at compile time: CryZe@cebabc3

Feel free to pull it into this PR.

[gabrielcfvg]

Based on this PR I made some further improvements by precomputing the lookup tables entirely at compile time: CryZe@cebabc3

Feel free to pull it into this PR.

I was working in a similar path, making the gradients to be initialized at compile time so that don't need to initialize the context at runtime, and more important, remove the synchronization/verification that happens in every access to the gradients.

@CryZe did just that and better than I would be able to. With his improvements, I got a 7% speedup over the original pull request implementation and 6.5% over the deprecated C implementation that was benchmarked by @EndrII.

[EndrII]

and 6.5% over the deprecated C implementation that was benchmarked by @EndrII.

hm looks good, but i want to test it before merging

[EndrII]

 Compiling opensimplex2 v1.1.0 (/###/OpenSimplex2)
error: `#[panic_handler]` function required, but not found
error: could not compile `opensimplex2` (lib) due to 1 previous error

looks as optimistaions broken this library

[EndrII]

build is failed

[qarmin]

Library is now no_std by default, so --features std needs to be added to run/build command, to be able to use std functions like println

[gabrielcfvg]

and 6.5% over the deprecated C implementation that was benchmarked by @EndrII.

hm looks good, but i want to test it before merging

Be careful when profiling it, since the gradients are now constant, the used functions are pure, meaning that if the returned value is not used in some way like it wasn't in the original benchmarks, the compiler probably will wipe the function call away, making the timings be zero.

[EndrII]

wow, amassing work !

MarcoCiaramella C Impl 2D: 626 msec
Deprecated C Impl 2D: 617 msec
Rust Impl 2D: 602 msec

@CryZe@qarmin@gabrielcfvg Thank you very much!!!
today i will integrate this library to my project )