feat: add lapack/base/dptts2
#2625
Conversation
Pranavchiku
added
Feature
Signed-off-by: Pranav <[email protected]>
Signed-off-by: Pranav <[email protected]>
Signed-off-by: Pranav <[email protected]>
kgryte
added
the
Needs Changes
Signed-off-by: Pranav <[email protected]>
Pranavchiku
added
Needs Review
|
I think this PR can now have a final review. |
|
|
||
| b.tic(); | ||
| for ( i = 0; i < b.iterations; i++ ) { | ||
| B = uniform( len * len, -10.0, 10.0, options ); |
There was a problem hiding this comment.
Do not allocate new arrays in benchmarks. Results will be tainted with non-relevant operations.
|
|
||
| b.tic(); | ||
| for ( i = 0; i < b.iterations; i++ ) { | ||
| B = uniform( len * len, -10.0, 10.0, options ); |
| B[ ob + ( ( N - 1 ) * strideB1 ) ] /= D[ offsetD + ( ( N - 1 ) * strideD ) ]; | ||
| se = N - 2; | ||
| for ( i = N - 2; i >= 0; i-- ) { | ||
| B[ ob + ( i * strideB1 ) ] = ( B[ ob + ( i * strideB1 ) ] / D[ offsetD + ( i * strideD ) ] ) - ( E[ offsetE + se ] * B[ ob + ( ( i + 1) * strideB1 ) ] ); |
There was a problem hiding this comment.
You have repeated computation of index offsets. Refactor and simplify.
| expected = new Float64Array( [ 11.0, 38.0, -5.0, -18.0, 2.0, 6.0 ] ); | ||
| out = dptts2( 'row-major', 3, 2, D, E, B, 2 ); | ||
| t.strictEqual( out, B, 'returns expected value' ); | ||
| isApprox( t, B, expected, 1.0e-7 ); |
There was a problem hiding this comment.
Why 1.0e-7? This tolerance does not make sense to me. That effectively makes the tol value 0 in isApprox.
| t.end(); | ||
| }); | ||
|
|
||
| tape( 'the function solves a tridiagonal system of the form `A * X = B` using `L * D * L^T` factorization of `A` (row-major, offsetD=3, offsetE=4, offsetB=6)', function test( t ) { |
There was a problem hiding this comment.
Parameterizing with explicit values does not mean much to the casual reader. What is this test (and the ones which follow) indicated to show?
| t.end(); | ||
| }); | ||
|
|
||
| tape( 'the function solves a tridiagonal system of the form `A * X = B` using `L * D * L^T` factorization of `A` (row-major, offsetD=3, strideD=2, offsetE=4, offsetB=6)', function test( t ) { |
There was a problem hiding this comment.
Here, you are testing that the function supports a D stride. Update the description accordingly.
| t.end(); | ||
| }); | ||
|
|
||
| tape( 'the function solves a tridiagonal system of the form `A * X = B` using `L * D * L^T` factorization of `A` (row-major, strideB1=-2, strideB2=-1)', function test( t ) { |
There was a problem hiding this comment.
Here, you are showing that the function supports negative strides for B. Update the description accordingly. Etc.
| t.end(); | ||
| }); | ||
|
|
||
| tape( 'the function solves a tridiagonal system of the form `A * X = B` using `L * D * L^T` factorization of `A` (row-major, strideB1=-3, strideB2=-2)', function test( t ) { |
There was a problem hiding this comment.
All your tests are row-major. You need to also systematically test behavior for column-major.
kgryte
added
Needs Changes
Towards #2464
Description
This pull request adds initial implementation for
lapack/base/dptts2.Related Issues
No.
Questions
No.
Other
No.
Checklist
TODO:
strideB1,strideB2.@stdlib-js/reviewers