test: add tests for IEEE 754-2019 compliance

[anandkaranubc]

Progresses #365

Description

What is the purpose of this pull request?

This pull request:

• Adds all the tests for testing special cases of log*

Related Issues

Does this pull request have any related issues?

This pull request:

• progresses Add tests for IEEE 754-2019 compliance #365

Questions

Any questions for reviewers of this pull request?

No.

Other

Any other information relevant to this pull request? This may include screenshots, references, and/or implementation notes.

No.

Checklist

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

• Read, understood, and followed the contributing guidelines.

---

@stdlib-js/reviewers

[anandkaranubc]

All checked!

[Planeshifter]

@anandkaranubc Per failing CI checks, looks like our log10 might not behave as expected for -0. Would you mind taking a look, please?

[anandkaranubc]

Haha, was in the middle of writing the reason :)

Oops, there is a bug. log2 and log10 don't return NINF when x = -0.0. But it works in C.

Mainly because of no type casting on hx in JS.

stdlib/lib/node_modules/@stdlib/math/base/special/log2/lib/main.js

Lines 118 to 126 in 7966727

	toWords.assign( x, WORDS, 1, 0 );
	hx = WORDS[ 0 ];
	lx = WORDS[ 1 ];
	k = 0|0; // asm type annotation
	if ( hx < HIGH_MIN_NORMAL_EXP ) {
	// Case: x < 2**-1022
	if ( ( (hx&ABS_MASK) | lx ) === 0 ) {
	return NINF;
	}

hx still remains a "huge" positive number here and isn't less than HIGH_MIN_NORMAL_EXP

But in C,

Typecasting it back again in int32_t makes it negative again.

stdlib/lib/node_modules/@stdlib/math/base/special/log2/src/main.c

Lines 71 to 79 in 7966727

	xc = x;
	stdlib_base_float64_to_words( xc, &hx, &lx );
	ihx = (int32_t)hx;
	k = 0;
	if ( ihx < HIGH_MIN_NORMAL_EXP ) {
	// Case: x < 2**-1022
	if ( ( ( ihx & STDLIB_CONSTANT_FLOAT64_HIGH_WORD_ABS_MASK ) | lx ) == 0 ) {
	return STDLIB_CONSTANT_FLOAT64_NINF;
	}

cc: @Planeshifter @kgryte

[kgryte]

@anandkaranubc For L22, if the condition assumes ihx, then we can modify to

if ( (hx|0) < HIGH_MIN_NORMAL_EXP ) { 
 	// Case: x < 2**-1022 
 	if ( ( (hx&ABS_MASK) | lx ) === 0 ) { 
 		return NINF; 
 	} 

[anandkaranubc]

@kgryte Thanks for the reply!

Can I do hx = WORDS[ 0 ] | 0; instead?

[kgryte]

@anandkaranubc Actually, are we sure the issue is with hx here? Given

stdlib/lib/node_modules/@stdlib/math/base/special/log2/lib/main.js

Line 115 in 7966727

if ( isnan( x ) || x < 0.0 ) {

x will be positive once going to toWords. Hence, the sign-bit of the high word will be zero. In which case, hx and hx|0 should be the same value.

[anandkaranubc]

/stdlib merge

[kgryte]

@anandkaranubc Ah, I see. The special case is -0. In which case, yes, you are right. Based on the code, it should be reasonable to do

hx = WORDS[ 0 ] | 0; // asm type annotation

[stdlib-bot]

Coverage Report

Package	Statements	Branches	Functions	Lines
math/base/special/ln	$\color{red}370/382$ $\color{green}+96.86\%$	$\color{red}25/28$ $\color{green}+89.29\%$	$\color{green}4/4$ $\color{green}+100.00\%$	$\color{red}370/382$ $\color{green}+96.86\%$
math/base/special/lnf	$\color{red}372/376$ $\color{green}+98.94\%$	$\color{red}30/32$ $\color{green}+93.75\%$	$\color{green}4/4$ $\color{green}+100.00\%$	$\color{red}372/376$ $\color{green}+98.94\%$
math/base/special/log	$\color{green}171/171$ $\color{green}+100.00\%$	$\color{green}5/5$ $\color{green}+100.00\%$	$\color{green}2/2$ $\color{green}+100.00\%$	$\color{green}171/171$ $\color{green}+100.00\%$
math/base/special/log10	$\color{green}307/307$ $\color{green}+100.00\%$	$\color{green}16/16$ $\color{green}+100.00\%$	$\color{green}2/2$ $\color{green}+100.00\%$	$\color{green}307/307$ $\color{green}+100.00\%$
math/base/special/log2	$\color{green}290/290$ $\color{green}+100.00\%$	$\color{green}16/16$ $\color{green}+100.00\%$	$\color{green}2/2$ $\color{green}+100.00\%$	$\color{green}290/290$ $\color{green}+100.00\%$

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

[anandkaranubc]

I also updated this line accordingly to be consistent with upstream (and C) implementation.

Reference

[anandkaranubc]

Our C implementation:

stdlib/lib/node_modules/@stdlib/math/base/special/log10/src/main.c

Lines 104 to 106 in 864fc4f

	xc *= TWO54;
	stdlib_base_float64_get_high_word( xc, &hx );
	ihx = (int32_t)hx;

[anandkaranubc]

The failing test is unrelated to this PR.

[kgryte]

@anandkaranubc I'll fix that up on develop now.

[kgryte]

/stdlib merge

[kgryte]

LGTM