[harsh-thakkar7]

Resolves #16510

Description

What is the purpose of this pull request?

Fixes subnormal rounding in @stdlib/number/float16/base/to-word at the 11-bit shift boundary. When shift === 11, the code previously took the wrong rounding branch and returned zero for values that should round to the smallest subnormal:

• toWord( 2.98038125038147e-8 ) returned 0x0000; now returns 0x0001.
• toWord( -2.98038125038147e-8 ) returned 0x8000; now returns 0x8001.

The fix changes the branch condition from shift < 11 to shift <= 11, so the round bit is taken from the correct bit of the 11-bit mantissa. Values requiring shift > 11 already return zero earlier via the f16Exponent < -10 check.

Regression tests cover the positive and negative rounding cases, the exact 2^-25 tie, and values just below 2^-24.

Related Issues

Does this pull request have any related issues?

• resolves [BUG]: number/float16/base/to-word mis-rounds subnormals at the 11-bit shift boundary #16510

Questions

Any questions for reviewers of this pull request?

No.

Other

Any other information relevant to this pull request?

The C/native path is unaffected: it returns the bits of an already encoded stdlib_float16_t. The JS test suite passes (5,013 tests), and the result was cross-checked against native Float16Array: 0 mismatches over a 50,000-value sweep around 2^-25 (previously ~25,007 mismatches).

Checklist

Please ensure that the following tasks are completed before submitting this pull request.

• Read, understood, and followed the contributing guidelines.

---

@stdlib-js/reviewers

[stdlib-bot]

Coverage Report

Package	Statements	Branches	Functions	Lines
number/float16/base/to-word	$\\color{red}264/282$ $\\color{green}+0.00\\%$	$\\color{red}28/33$ $\\color{green}+0.00\\%$	$\\color{green}2/2$ $\\color{green}+0.00\\%$	$\\color{red}264/282$ $\\color{green}+0.00\\%$

The above coverage report was generated for the changes in this PR.