[RFC]: Refactor random number generation in JS benchmarks for stats/base/dists/truncated-normal

[anandkaranubc]

Description

This RFC proposes improving random number generation in JS benchmarks for stats/base/dists/truncated-normal.

Context: At present, in the remaining packages, random number generation in JS benchmarks occurs inside the benchmarking loop. Since random number generation is an expensive operation, it should be moved out of the benchmarking loops and initialized beforehand to avoid interfering with the results.

When adding support, the following tasks should be completed:

1. Move random number generation out of the benchmarking loops and initialize it before the benchmarks.

2. Ensure that the generated random values use the same range as the existing values to maintain consistency.

3. Use uniform and discreteUniform from @stdlib/random/base/uniform and @stdlib/random/base/discrete-uniform instead of randu expressions.

   • For example, replace ( randu() * 10.0 + EPS ) with uniform( EPS, 10.0 ).
   • Replace ceil( randu() * 10.0 + EPS ) with discreteUniform( 1, 10 ).

To provide a concrete example of what a PR implementing the desired changes should contain, see #4837 and #4955, which provide examples for using both uniform and discreteUniform.

Related Issues

#4837, #4955

Questions

No.

Other

Once the implementation is complete, you should be able to run the following make commands:

Build Native Add-on

NODE_ADDONS_PATTERN="@stdlib/stats/base/dists/truncated-normal/*" make install-node-addons

Run JavaScript Benchmarks

make benchmark-javascript-files FILES="$(pwd)/lib/node_modules/@stdlib/stats/base/dists/truncated-normal/*/benchmark/benchmark.js"

Run JavaScript Native Benchmarks

make benchmark-javascript-files FILES="$(pwd)/lib/node_modules/@stdlib/stats/base/dists/truncated-normal/*/benchmark/benchmark.native.js"

Note: If running benchmarks results in an error, it is likely due to the random number generators producing values where the functions are not defined. To resolve this, check the relevant benchmark files, adjust the input value ranges to ensure they remain within the domain where the functions are valid, and then rerun the benchmarks. Reference: #4955

Checklist

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

[ekambains]

Hi @anandkaranubc , I am unable to find the cdf/benchmark/benchmark.js directory for truncated-normal folder. Can you please point it out, so I can work on this issue?

[anandkaranubc]

Hi @anandkaranubc , I am unable to find the cdf/benchmark/benchmark.js directory for truncated-normal folder. Can you please point it out, so I can work on this issue?

That's correct. It hasn't been implemented yet.

[Krishna-Sharma-g]

hello @anandkaranubc i wanted to ask that can i work on the implementation of its inner functions so for that will i have to create another issue or i can do in this only so basically for the implementation of that we have to write total 9 files with the benchmarks, examples , src and lib

cdf (Cumulative Distribution Function) - shown above
pdf (Probability Density Function)
quantile (Inverse CDF)
mean
median
mode
variance
stdev (Standard Deviation)
logpdf (Log Probability Density Function)
and in them we have to give the docs src lib and all in them so can i do it and can you please guide me in this

[anandkaranubc]

Hi @Krishna-Sharma-g, are you referring to this current issue or the issues related to the C implementations? This issue only involves refactoring random number generation in JavaScript benchmarks and is not related to any C code.

[Krishna-Sharma-g]

no @anandkaranubc i have to implement this so for that do i have to create another issue for that

[anandkaranubc]

no @anandkaranubc i have to implement this so for that do i have to create another issue for that

If you want to create a PR targeting this issue, feel free to proceed. Just make sure to review the reference PRs first.

[saurabhraghuvanshii]

is anyone done it ?
if not I can .

[anandkaranubc]

Closing this issue as it does not apply to stats/base/dists/truncated-normal. There are no benchmarks in the namespace yet. Sorry :(