Introduce fast and deterministic RNG

[Cuda-Chen]

Use SplitMix64 for fast and deterministic RNG for cross compiler/platform support.

[howjmay]

I am curious. Doesn't it still generate the same random number sequence?
I concerned about we may not cover all the possible values with the previous RNG function.
But maybe having a consistent random number sequence is better?

[Cuda-Chen]

I am curious. Doesn't it still generate the same random number sequence? I concerned about we may not cover all the possible values with the previous RNG function. But maybe having a consistent random number sequence is better?

After searching, I realize that "consistent random number sequence" means the RNG always produces the same random number sequence every time the test program is run. Is that what we are going to discuss about?

[howjmay]

I am curious. Doesn't it still generate the same random number sequence? I concerned about we may not cover all the possible values with the previous RNG function. But maybe having a consistent random number sequence is better?

After searching, I realize that "consistent random number sequence" means the RNG always produces the same random number sequence every time the test program is run. Is that what we are going to discuss about?

Yes that is exactly what I mentioned. So do we need to have a consistent random number sequence or maybe not?

[jserv]

After searching, I realize that "consistent random number sequence" means the RNG always produces the same random number sequence every time the test program is run. Is that what we are going to discuss about?

Yes that is exactly what I mentioned. So do we need to have a consistent random number sequence or maybe not?

For the purpose of CI pipeline, we shall stick to deterministic random number generation.

[howjmay]

After searching, I realize that "consistent random number sequence" means the RNG always produces the same random number sequence every time the test program is run. Is that what we are going to discuss about?

Yes that is exactly what I mentioned. So do we need to have a consistent random number sequence or maybe not?

For the purpose of CI pipeline, we shall stick to deterministic random number generation.

Got it.
Then I am curious the purpose of this PR. The old implementation has already presented a deterministic random number generation. Maybe we should just close the original issue?

[Cuda-Chen]

After searching, I realize that "consistent random number sequence" means the RNG always produces the same random number sequence every time the test program is run. Is that what we are going to discuss about?

Yes that is exactly what I mentioned. So do we need to have a consistent random number sequence or maybe not?

For the purpose of CI pipeline, we shall stick to deterministic random number generation.

Got it. Then I am curious the purpose of this PR. The old implementation has already presented a deterministic random number generation. Maybe we should just close the original issue?

I agree. Just close the original issue with a comment mentioning about sticking to deterministic random number generation of CI pipelines.

[jserv]

Refine the subject per discussions.

[howjmay]

issue was closed

[Cuda-Chen]

As the rand() usually acts good for deterministic random number generation, I would like to close this PR if we do not need fast random number generation.

[jserv]

As the rand() usually acts good for deterministic random number generation, I would like to close this PR if we do not need fast random number generation.

The proposed RNG still makes sense for cross-platform consideration. In particular, this project supports GCC, Clang, and MSVC in combination with Linux, macOS, and Windows. Therefore, a fast and deterministic RNG can minimize potential issues while performing tests.

@Cuda-Chen, please refine the commit messages.