[RFC]: add C implementation for @stdlib/math/base/special/fast/uint32-log2

[performant23]

Description

This RFC proposes adding native C implementation for @stdlib/math/base/special/fast/uint32-log2:

uint32_t stdlib_base_fast_uint32_log2 (uint32_t x);

Related Issues

N/A

Questions

No.

Other

No.

Checklist

• I have read and understood the Code of Conduct.
• Searched for existing issues and pull requests.
• The issue name begins with RFC:.

[performant23]

Hello, had started working on this issue yesterday. Will open a PR once my last one #1918 is addressed. Thanks!

[performant23]

Also, just to clarify, there are variable declarations for constants and they're used later in lib/main.js as:

// 4294901760 => 0xFFFF0000 => 11111111111111110000000000000000
var B4 = 0xFFFF0000 >>> 0; // asm type annotation
.
.
.
if ( y & B4 ) {
do something;
}

Is it required that we do the same for the C implementation using // CONSTANTS // section:

const uint32_t B4 = 0xFFFF0000u; // asm type annotation

or can we use it directly in conditionals:

if (y & 0xFFFF0000u){
do something...
}

cc @kgryte, @Planeshifter

Presently I've implemented it using the // CONSTANTS // section for the declarations.

[performant23]

And also, since the implementation is relatively more complex and subtle, just thought it would be a good idea to flesh out the implementation in parts I have doubts about. I wanted to confirm about the use of minstd in our C implementation. Since the single function in JS is broken into multiple parts for C and directly addresses memory, how would generating the random numbers work basically?

So far, I have come up with this using the implementation and examples in those.

struct BasePRNGObject *obj = stdlib_base_random_minstd_shuffle_allocate(<seed>); 

uint64_t randNum;
int8_t status = obj->prng->next(obj, &randNum); (//We can later typecast randNum to int32_t.)
.
.
.

stdlib_base_random_minstd_shuffle_free(obj);

In this case, how would we handle failures in allocation and the failures in generating the random numbers?

[kgryte]

Re: variable declarations. Yes, prefer static const declarations over inlining.

Re: PRNGs. Not sure why this is necessary. Why do you need to use a stdlib PRNG for adding a C implementation?

[performant23]

Thanks for the clarifications, @kgryte! So currently, the PRNGs are used in benchmarking to generate random 32-bit unsigned integers. Thought we'd use their corresponding C exports for generation in benchmark/c/native/benchmark.c

[kgryte]

For random values, instead of rand_double (you can leave that function there!), just use rand() directly and cast to uint32_t.