feat: add `math/base/special/cinvf` #3103

aayush0325 · 2024-11-12T08:35:51Z

Resolves a part of #649

Description

What is the purpose of this pull request?

This pull request:

resolves part of [RFC]: Add C implementations to base special math functions (tracking issue) #649

Related Issues

Does this pull request have any related issues?

This pull request:

resolves a part of [RFC]: Add C implementations to base special math functions (tracking issue) #649

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.

Need help with overflow case.

Checklist

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

Read, understood, and followed the contributing guidelines.

@stdlib-js/reviewers

aayush0325 · 2024-11-12T09:19:31Z

tape( 'the function may overflow', function test( t ) {
	var v;

	v = cinvf( new Complex64( 5.0e-324, 5.0e-324 ) );
	t.strictEqual( real( v ), PINF, 'real component is +infinity' );
	t.strictEqual( imag( v ), NINF, 'imaginary component is -infinity' );

	v = cinvf( new Complex64( -5.0e-324, 5.0e-324 ) );
	t.strictEqual( real( v ), NINF, 'real component is -infinity' );
	t.strictEqual( imag( v ), NINF, 'imaginary component is -infinity' );

	v = cinvf( new Complex64( -5.0e-324, -5.0e-324 ) );
	t.strictEqual( real( v ), NINF, 'real component is -infinity' );
	t.strictEqual( imag( v ), PINF, 'imaginary component is +infinity' );

	v = cinvf( new Complex64( 5.0e-324, -5.0e-324 ) );
	t.strictEqual( real( v ), PINF, 'real component is +infinity' );
	t.strictEqual( imag( v ), PINF, 'imaginary component is +infinity' );

	v = cinvf( new Complex64( 0.0, 5.0e-324 ) );
	t.strictEqual( real( v ), 0.0, 'real component is 0' );
	t.strictEqual( imag( v ), NINF, 'imaginary component is -infinity' );

	v = cinvf( new Complex64( 0.0, -5.0e-324 ) );
	t.strictEqual( real( v ), 0.0, 'real component is 0' );
	t.strictEqual( imag( v ), PINF, 'imaginary component is +infinity' );

	v = cinvf( new Complex64( 5.0e-324, 0.0 ) );
	t.strictEqual( real( v ), PINF, 'real component is +infinity' );
	t.strictEqual( imag( v ), 0.0, 'imaginary component is 0' );

	v = cinvf( new Complex64( -5.0e-324, 0.0 ) );
	t.strictEqual( real( v ), NINF, 'real component is -infinity' );
	t.strictEqual( imag( v ), 0.0, 'imaginary component is 0' );

	t.end();
});

Good Afternoon, @gunjjoshi can you please help me out in adding the overflow testcase in test.js and test.native.js as it is not there in this PR right now, i tried various values for this but the complex number real( v ) imag( v ) turns out to be NaN NaN and not Infinity or -Infinity as expected. how did you arrive at the number 5.0e-324?

gunjjoshi · 2024-11-12T14:01:16Z

@aayush0325 Tried the following test on your branch, and it works:

tape( 'the function may produce very large values for very small inputs', function test( t ) {
    var v;

    v = cinvf( new Complex64( FLOAT32_SMALLEST_NORMAL, FLOAT32_SMALLEST_NORMAL ) );
    t.ok( real( v ) > 1e37, 'real component is very large' );
    t.ok( imag( v ) < -1e37, 'imaginary component is very large and negative' );

    v = cinvf( new Complex64( -FLOAT32_SMALLEST_NORMAL, FLOAT32_SMALLEST_NORMAL ) );
    t.ok( real( v ) < -1e37, 'real component is very large and negative' );
    t.ok( imag( v ) < -1e37, 'imaginary component is very large and negative' );

    v = cinvf( new Complex64( -FLOAT32_SMALLEST_NORMAL, -FLOAT32_SMALLEST_NORMAL ) );
    t.ok( real( v ) < -1e37, 'real component is very large and negative' );
    t.ok( imag( v ) > 1e37, 'imaginary component is very large and positive' );

    v = cinvf( new Complex64( FLOAT32_SMALLEST_NORMAL, -FLOAT32_SMALLEST_NORMAL ) );
    t.ok( real( v ) > 1e37, 'real component is very large and positive' );
    t.ok( imag( v ) > 1e37, 'imaginary component is very large and positive' );

    v = cinvf( new Complex64( 0.0, FLOAT32_SMALLEST_NORMAL ) );
    t.strictEqual( real( v ), 0.0, 'real component is 0' );
    t.ok( imag( v ) < -1e37, 'imaginary component is very large and negative' );

    v = cinvf( new Complex64( 0.0, -FLOAT32_SMALLEST_NORMAL ) );
    t.strictEqual( real( v ), 0.0, 'real component is 0' );
    t.ok( imag( v ) > 1e37, 'imaginary component is very large and positive' );

    v = cinvf( new Complex64( FLOAT32_SMALLEST_NORMAL, 0.0 ) );
    t.ok( real( v ) > 1e37, 'real component is very large and positive' );
    t.strictEqual( imag( v ), 0.0, 'imaginary component is 0' );

    v = cinvf( new Complex64( -FLOAT32_SMALLEST_NORMAL, 0.0 ) );
    t.ok( real( v ) < -1e37, 'real component is very large and negative' );
    t.strictEqual( imag( v ), 0.0, 'imaginary component is 0' );

    t.end();
});

You can try this. I think there might be a loss of precision as we are handling single-precision floating-point numbers and as we are not explicitly returning PINF or NINF in our implementation, so, we are not comparing with exactly PINF and NINF here.

gunjjoshi · 2024-11-12T14:03:17Z

You can declare separate variables for tiny and huge, such as:

var tiny = -1e37;
var huge = 1e37;

and then use them in the tests.

aayush0325 · 2024-11-12T14:09:04Z

thanks @gunjjoshi! i'll push these changes soon, anything else to add? i'll make all the changes together then.

gunjjoshi · 2024-11-12T14:16:41Z

thanks @gunjjoshi! i'll push these changes soon, anything else to add? i'll make all the changes together then.

Nothing for now, once you've added these tests in both test.js and test.native.js, we can review it.

aayush0325 · 2024-11-12T15:33:05Z

updated the tests, kindly review @gunjjoshi.

lib/node_modules/@stdlib/math/base/special/cinvf/README.md

lib/node_modules/@stdlib/math/base/special/cinvf/benchmark/c/Makefile

lib/node_modules/@stdlib/math/base/special/cinvf/benchmark/c/native/Makefile

lib/node_modules/@stdlib/math/base/special/cinvf/docs/types/index.d.ts

lib/node_modules/@stdlib/math/base/special/cinvf/test/test.native.js

gunjjoshi · 2024-11-17T03:49:50Z

The failing tests might be originating from

stdlib/.github/workflows/run_tests_coverage.yml

Line 213 in 1b6e722

comment_id: botComment.id,

as we're not declaring botComment, if it is a bot comment.

cc: @Planeshifter

aayush0325 · 2024-11-17T04:17:06Z

I see, anything else I can do in this PR @gunjjoshi?

lib/node_modules/@stdlib/math/base/special/cinvf/test/test.js

lib/node_modules/@stdlib/math/base/special/cinvf/test/test.native.js

aayush0325 · 2024-11-18T02:15:04Z

The tests seem to be running fine locally for both test.js and test.native.js, not sure why there's an error here.

kgryte · 2024-11-18T04:05:27Z

/stdlib merge

kgryte · 2024-11-18T04:06:26Z

not sure why there's an error here

Test coverage build errors should be resolved now, given recent fixes by @Planeshifter to our CI.

aayush0325 · 2024-11-28T04:44:49Z

/stdlib merge

aayush0325 · 2024-12-07T19:32:03Z

@gunjjoshi kindly give this a review!

gunjjoshi · 2024-12-08T11:53:38Z

lib/node_modules/@stdlib/math/base/special/cinvf/README.md

+```javascript
+var Complex64 = require( '@stdlib/complex/float32/ctor' );
+var uniform = require( '@stdlib/random/base/uniform' );
+var cinvf = require( '@stdlib/math/base/special/cinvf' );
+
+var z1;
+var z2;
+var i;
+
+for ( i = 0; i < 100; i++ ) {
+    z1 = new Complex64( uniform( -50.0, 50.0 ), uniform( -50.0, 50.0 ) );
+    z2 = cinvf( z1 );
+    console.log( '1.0 / (%s) = %s', z1.toString(), z2.toString() );
+}
+```


@aayush0325 Can you modify this example to generate the example values outside the loop? You can use @stdlib/random/array/uniform here.

You can follow this in examples/index.js as well.

Signed-off-by: Gunj Joshi <[email protected]>

gunjjoshi · 2024-12-08T15:39:02Z

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

+	if ( ab >= LARGE_THRESHOLD ) {
+		re *= 0.5;
+		im *= 0.5;
+		s *= 0.5;


@aayush0325 Can you add a test both in test.js and test.native.js, to test this block? This block is not being reached in any test, according to the code coverage report.

@gunjjoshi should i increase the range from 1e+30 to 1e+38 in all fixtures so that the code reaches this block or should i just add 4 new fixtures for very large positive/negative imaginary and very large positive/negative real (which goes to the order of +- 1e+38). we may need to further increase the tolerance of tests as well i think.

You probably don't need to create fixtures (as in fixture files), but simply need to add a separate test block in which you hard code a few test values which exercise the branch. In that test block, you can specify a test tolerance independent of other test blocks.

I do recommend, however, validating what test values you use against a known reference implementation.

If it is easier to just create additional fixture files based on a reference implementation, that is also fine.

aayush0325 · 2024-12-11T08:29:17Z

good morning, i've added a test with hardcoded values where magnitude of the real and imaginary component exceeds LARGE_THRESHOLD so that the code reaches the mentioned block, i have taken these values from the numpy implementation using the following code:

import numpy as np

# Function to compute the inverse of a complex number using single-precision floating-point numbers
def compute_inverse_single_precision(real, imag):
    complex_number = np.complex64(real + imag * 1j)
    inverse = 1 / complex_number
    return inverse

# Given complex numbers
complex_numbers = [
    (2e+38, 200),
    (-2e+38, 200),
    (200, 2e+38),
    (200, -2e+38)
]

# Compute inverses
for real, imag in complex_numbers:
    inverse = compute_inverse_single_precision(real, imag)
    print(f"The inverse of {real} + {imag}I is {inverse}")

which gave me the output:

The inverse of 2e+38 + 200I is (4.999999675228202e-39-0j)
The inverse of -2e+38 + 200I is (-4.999999675228202e-39-0j)
The inverse of 200 + 2e+38I is -4.999999675228202e-39j
The inverse of 200 + -2e+38I is 4.999999675228202e-39j

if there's anything else that i should do then please let me know!

AadishJ · 2025-01-11T07:51:43Z

hey is this implementation complete, I was trying to do this one as I did not find it in the source code?

aayush0325 added 8 commits November 12, 2024 11:01

feat: basic setup

9df4fb1

feat: basic c files added

a574281

feat: basic js files added

9aee3b0

feat: tests added

f0e76ad

feat: docs added

1dc1612

feat: examples added

accab3f

feat: benchmarks added

89640dc

feat: julia benchmark updated

c67937b

stdlib-bot added the Math Issue or pull request specific to math functionality. label Nov 12, 2024

aayush0325 added 7 commits November 12, 2024 14:08

fix: example error in README

2c47328

feat: updated c code using stdlib_base_maxf

ccedb23

feat: updated c code using stdlib_base_maxf

193c5c7

feat: updated tolerance for C tests

8d92eff

feat: code using maxf

b933479

feat: code using fmaxf

f0bf2fe

feat: updated tests for c files

15e5d34

feat: updated test coverage

6a21435

gunjjoshi requested changes Nov 13, 2024

View reviewed changes

gunjjoshi added the Needs Changes Pull request which needs changes before being merged. label Nov 13, 2024

aayush0325 added 5 commits November 13, 2024 22:39

chore: stuff from code review

068956c

chore: stuff from code review

072e306

chore: stuff from code review

ed89c11

chore: stuff from code review

a054780

feat: updated code using stdlib_base_maxf

1d636a1

gunjjoshi reviewed Nov 17, 2024

View reviewed changes

lib/node_modules/@stdlib/math/base/special/cinvf/test/test.js Outdated Show resolved Hide resolved

lib/node_modules/@stdlib/math/base/special/cinvf/test/test.native.js Outdated Show resolved Hide resolved

chore: stuff from code review

e1e187a

aayush0325 requested a review from gunjjoshi November 18, 2024 02:16

Merge remote-tracking branch 'upstream/develop' into cinvf

6bbd6ee

kgryte added the Feature Issue or pull request for adding a new feature. label Nov 18, 2024

stdlib-bot added the bot: In Progress Pull request is currently awaiting automation. label Nov 28, 2024

Merge remote-tracking branch 'upstream/develop' into cinvf

f7979ec

stdlib-bot removed the bot: In Progress Pull request is currently awaiting automation. label Nov 28, 2024

gunjjoshi requested changes Dec 8, 2024

View reviewed changes

gunjjoshi and others added 6 commits December 8, 2024 17:36

docs: reorder variable declarations

dcfbf2a

Signed-off-by: Gunj Joshi <[email protected]>

docs: use correct function name

e70b3c1

Signed-off-by: Gunj Joshi <[email protected]>

docs: update output values

bd17893

Signed-off-by: Gunj Joshi <[email protected]>

test: update tests

3448159

Update package.json

61e9d32

Signed-off-by: Gunj Joshi <[email protected]>

feat: updated examples with precomputation

2b28783

aayush0325 requested a review from gunjjoshi December 8, 2024 13:24

stdlib-bot added the Needs Review A pull request which needs code review. label Dec 8, 2024

gunjjoshi reviewed Dec 8, 2024

View reviewed changes

feat: updated test coverage

5cf999a

aayush0325 requested a review from gunjjoshi December 11, 2024 08:29

kgryte removed the Needs Review A pull request which needs code review. label Dec 14, 2024

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat: add `math/base/special/cinvf` #3103

feat: add `math/base/special/cinvf` #3103

aayush0325 commented Nov 12, 2024 •

edited by gunjjoshi

Loading

aayush0325 commented Nov 12, 2024 •

edited

Loading

gunjjoshi commented Nov 12, 2024 •

edited

Loading

gunjjoshi commented Nov 12, 2024

aayush0325 commented Nov 12, 2024

gunjjoshi commented Nov 12, 2024

aayush0325 commented Nov 12, 2024

gunjjoshi commented Nov 17, 2024 •

edited

Loading

aayush0325 commented Nov 17, 2024

aayush0325 commented Nov 18, 2024

kgryte commented Nov 18, 2024

kgryte commented Nov 18, 2024

aayush0325 commented Nov 28, 2024

aayush0325 commented Dec 7, 2024

gunjjoshi Dec 8, 2024

gunjjoshi Dec 8, 2024

gunjjoshi Dec 8, 2024

aayush0325 Dec 9, 2024 •

edited

Loading

kgryte Dec 10, 2024

kgryte Dec 10, 2024

aayush0325 commented Dec 11, 2024 •

edited

Loading

AadishJ commented Jan 11, 2025

feat: add math/base/special/cinvf #3103

Are you sure you want to change the base?

feat: add math/base/special/cinvf #3103

Conversation

aayush0325 commented Nov 12, 2024 • edited by gunjjoshi Loading

Description

Related Issues

Questions

Other

Checklist

aayush0325 commented Nov 12, 2024 • edited Loading

gunjjoshi commented Nov 12, 2024 • edited Loading

gunjjoshi commented Nov 12, 2024

aayush0325 commented Nov 12, 2024

gunjjoshi commented Nov 12, 2024

aayush0325 commented Nov 12, 2024

gunjjoshi commented Nov 17, 2024 • edited Loading

aayush0325 commented Nov 17, 2024

aayush0325 commented Nov 18, 2024

kgryte commented Nov 18, 2024

kgryte commented Nov 18, 2024

aayush0325 commented Nov 28, 2024

aayush0325 commented Dec 7, 2024

gunjjoshi Dec 8, 2024

Choose a reason for hiding this comment

gunjjoshi Dec 8, 2024

Choose a reason for hiding this comment

gunjjoshi Dec 8, 2024

Choose a reason for hiding this comment

aayush0325 Dec 9, 2024 • edited Loading

Choose a reason for hiding this comment

kgryte Dec 10, 2024

Choose a reason for hiding this comment

kgryte Dec 10, 2024

Choose a reason for hiding this comment

aayush0325 commented Dec 11, 2024 • edited Loading

AadishJ commented Jan 11, 2025

feat: add `math/base/special/cinvf` #3103

feat: add `math/base/special/cinvf` #3103

aayush0325 commented Nov 12, 2024 •

edited by gunjjoshi

Loading

aayush0325 commented Nov 12, 2024 •

edited

Loading

gunjjoshi commented Nov 12, 2024 •

edited

Loading

gunjjoshi commented Nov 17, 2024 •

edited

Loading

aayush0325 Dec 9, 2024 •

edited

Loading

aayush0325 commented Dec 11, 2024 •

edited

Loading