feat!: add support for stacks in blas/sdot

BREAKING CHANGE: return an ndarray, rather than a scalar

This commit changes the return value from a scalar to an ndarray. Previously, the function only operated on one-dimensional ndarrays and returned a scalar value. Now, the function operates on ndarrays of arbitrary rank and always returns an ndarray. In order to migrate, for one-dimensional input ndarrays, users should call `out.get()` in order to retrieve the scalar dot product.

PR-URL: #2895
Co-authored-by: Athan Reines <[email protected]>
Reviewed-by: Athan Reines <[email protected]>

Author: aman-095

lib/node_modules/@stdlib/blas/ddot/lib/main.js

@@ -41,7 +41,7 @@ var format = require( '@stdlib/string/format' );
 *
 * @param {ndarrayLike} x - first input array
 * @param {ndarrayLike} y - second input array
-* @param {NegativeInteger} [dim] - dimension for which to compute the dot product
+* @param {NegativeInteger} [dim=-1] - dimension for which to compute the dot product
 * @throws {TypeError} first argument must be a ndarray containing double-precision floating-point numbers
 * @throws {TypeError} first argument must have at least one dimension
 * @throws {TypeError} second argument must be a ndarray containing double-precision floating-point numbers

lib/node_modules/@stdlib/blas/ddot/package.json

@@ -14,7 +14,6 @@
     }
   ],
   "main": "./lib",
-  "browser": "./lib/main.js",
   "directories": {
     "benchmark": "./benchmark",
     "doc": "./docs",

lib/node_modules/@stdlib/blas/sdot/README.md

@@ -33,7 +33,7 @@ The [dot product][dot-product] (or scalar product) is defined as
 ```
 
 <!-- <div class="equation" align="center" data-raw-text="\mathbf{x}\cdot\mathbf{y} = \sum_{i=0}^{N-1} x_i y_i = x_0 y_0 + x_1 y_1 + \ldots + x_{N-1} y_{N-1}" data-equation="eq:dot_product">
-    <img src="https://cdn.jsdelivr.net/gh/stdlib-js/stdlib@03fff24f5a7ba807a292f08cfef75ed0748e40de/lib/node_modules/@stdlib/blas/sdot/docs/img/equation_dot_product.svg" alt="Dot product definition.">
+    <img src="https://cdn.jsdelivr.net/gh/stdlib-js/stdlib@d0afc603cdda35b11d5bd1633dd4dddb0d59e117/lib/node_modules/@stdlib/blas/sdot/docs/img/equation_dot_product.svg" alt="Dot product definition.">
     <br>
 </div> -->
 
@@ -51,9 +51,9 @@ The [dot product][dot-product] (or scalar product) is defined as
 var sdot = require( '@stdlib/blas/sdot' );
 ```
 
-#### sdot( x, y )
+#### sdot( x, y\[, dim] )
 
-Calculates the dot product of vectors `x` and `y`.
+Calculates the dot product of two single-precision floating-point vectors `x` and `y`.
 
 ```javascript
 var Float32Array = require( '@stdlib/array/float32' );
@@ -63,25 +63,38 @@ var x = array( new Float32Array( [ 4.0, 2.0, -3.0, 5.0, -1.0 ] ) );
 var y = array( new Float32Array( [ 2.0, 6.0, -1.0, -4.0, 8.0 ] ) );
 
 var z = sdot( x, y );
+// returns <ndarray>
+
+var v = z.get();
 // returns -5.0
 ```
 
 The function has the following parameters:
 
--   **x**: a 1-dimensional [`ndarray`][@stdlib/ndarray/array] whose underlying data type is `float32`.
--   **y**: a 1-dimensional [`ndarray`][@stdlib/ndarray/array] whose underlying data type is `float32`.
+-   **x**: a non-zero-dimensional [`ndarray`][@stdlib/ndarray/ctor] whose underlying data type is `float32`. Must be [broadcast-compatible][@stdlib/ndarray/base/broadcast-shapes] with `y`.
+-   **y**: a non-zero-dimensional [`ndarray`][@stdlib/ndarray/ctor] whose underlying data type is `float32`. Must be [broadcast-compatible][@stdlib/ndarray/base/broadcast-shapes] with `x`.
+-   **dim**: dimension for which to compute the dot product. Must be a negative integer. Negative indices are resolved relative to the last array dimension, with the last dimension corresponding to `-1`. Default: `-1`.
 
-If provided empty vectors, the function returns `0.0`.
+If provided at least one input [`ndarray`][@stdlib/ndarray/ctor] having more than one dimension, the input [`ndarrays`][@stdlib/ndarray/ctor] are [broadcasted][@stdlib/ndarray/base/broadcast-shapes] to a common shape. For multi-dimensional input [`ndarrays`][@stdlib/ndarray/ctor], the function performs batched computation, such that the function computes the dot product for each pair of vectors in `x` and `y` according to the specified dimension index.
 
 ```javascript
 var Float32Array = require( '@stdlib/array/float32' );
 var array = require( '@stdlib/ndarray/array' );
 
-var x = array( new Float32Array() );
-var y = array( new Float32Array() );
+var opts = {
+    'shape': [ 2, 3 ]
+};
+var x = array( new Float32Array( [ 4.0, 2.0, -3.0, 5.0, -1.0, 3.0 ] ), opts );
+var y = array( new Float32Array( [ 2.0, 6.0, -1.0, -4.0, 8.0, 2.0 ] ), opts );
 
 var z = sdot( x, y );
-// returns 0.0
+// returns <ndarray>
+
+var v1 = z.get( 0 );
+// returns 23.0
+
+var v2 = z.get( 1 );
+// returns -22.0
 ```
 
 </section>
@@ -92,6 +105,11 @@ var z = sdot( x, y );
 
 ## Notes
 
+-   The size of the contracted dimension must be the same for both input [`ndarrays`][@stdlib/ndarray/ctor].
+-   The function resolves the dimension index for which to compute the dot product **before** broadcasting.
+-   Negative indices are resolved relative to the last [`ndarray`][@stdlib/ndarray/ctor] dimension, with the last dimension corresponding to `-1`.
+-   The output [`ndarray`][@stdlib/ndarray/ctor] has the same data type as the input [`ndarrays`][@stdlib/ndarray/ctor] and has a shape which is determined by broadcasting and excludes the contracted dimension.
+-   If provided empty vectors, the dot product is `0`.
 -   `sdot()` provides a higher-level interface to the [BLAS][blas] level 1 function [`sdot`][@stdlib/blas/base/sdot].
 
 </section>
@@ -105,27 +123,27 @@ var z = sdot( x, y );
 <!-- eslint no-undef: "error" -->
 
 ```javascript
-var discreteUniform = require( '@stdlib/random/base/discrete-uniform' );
-var Float32Array = require( '@stdlib/array/float32' );
+var discreteUniform = require( '@stdlib/random/array/discrete-uniform' );
+var ndarray2array = require( '@stdlib/ndarray/to-array' );
 var array = require( '@stdlib/ndarray/array' );
 var sdot = require( '@stdlib/blas/sdot' );
 
-var x = array( new Float32Array( 10 ) );
-var y = array( new Float32Array( 10 ) );
+var opts = {
+    'dtype': 'float32'
+};
 
-var rand1 = discreteUniform.factory( 0, 100 );
-var rand2 = discreteUniform.factory( 0, 10 );
+var x = array( discreteUniform( 10, 0, 100, opts ), {
+    'shape': [ 5, 2 ]
+});
+console.log( ndarray2array( x ) );
 
-var i;
-for ( i = 0; i < x.length; i++ ) {
-    x.set( i, rand1() );
-    y.set( i, rand2() );
-}
-console.log( x.toString() );
-console.log( y.toString() );
+var y = array( discreteUniform( 10, 0, 10, opts ), {
+    'shape': x.shape
+});
+console.log( ndarray2array( y ) );
 
-var z = sdot( x, y );
-console.log( z );
+var z = sdot( x, y, -1 );
+console.log( ndarray2array( z ) );
 ```
 
 </section>
@@ -136,14 +154,6 @@ console.log( z );
 
 <section class="related">
 
-* * *
-
-## See Also
-
--   <span class="package-name">[`@stdlib/blas/base/sdot`][@stdlib/blas/base/sdot]</span><span class="delimiter">: </span><span class="description">calculate the dot product of two single-precision floating-point vectors.</span>
--   <span class="package-name">[`@stdlib/blas/ddot`][@stdlib/blas/ddot]</span><span class="delimiter">: </span><span class="description">calculate the dot product of two double-precision floating-point vectors.</span>
--   <span class="package-name">[`@stdlib/blas/gdot`][@stdlib/blas/gdot]</span><span class="delimiter">: </span><span class="description">calculate the dot product of two vectors.</span>
-
 </section>
 
 <!-- /.related -->
@@ -156,18 +166,12 @@ console.log( z );
 
 [blas]: http://www.netlib.org/blas
 
-[@stdlib/ndarray/array]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/ndarray/array
+[@stdlib/ndarray/ctor]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/ndarray/ctor
 
-<!-- <related-links> -->
+[@stdlib/ndarray/base/broadcast-shapes]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/ndarray/base/broadcast-shapes
 
 [@stdlib/blas/base/sdot]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/blas/base/sdot
 
-[@stdlib/blas/ddot]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/blas/ddot
-
-[@stdlib/blas/gdot]: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/blas/gdot
-
-<!-- </related-links> -->
-
 </section>
 
 <!-- /.links -->

lib/node_modules/@stdlib/blas/sdot/benchmark/benchmark.js

@@ -21,15 +21,21 @@
 // MODULES //
 
 var bench = require( '@stdlib/bench' );
-var randu = require( '@stdlib/random/base/randu' );
 var isnan = require( '@stdlib/math/base/assert/is-nan' );
 var pow = require( '@stdlib/math/base/special/pow' );
-var Float32Array = require( '@stdlib/array/float32' );
+var uniform = require( '@stdlib/random/array/uniform' );
 var array = require( '@stdlib/ndarray/array' );
 var pkg = require( './../package.json' ).name;
 var sdot = require( './../lib/main.js' );
 
 
+// VARIABLES //
+
+var opts = {
+	'dtype': 'float32'
+};
+
+
 // FUNCTIONS //
 
 /**
@@ -40,34 +46,29 @@ var sdot = require( './../lib/main.js' );
 * @returns {Function} benchmark function
 */
 function createBenchmark( len ) {
-	var x;
-	var y;
-	var i;
-
-	x = new Float32Array( len );
-	y = new Float32Array( len );
-	for ( i = 0; i < len; i++ ) {
-		x[ i ] = ( randu()*10.0 ) - 20.0;
-		y[ i ] = ( randu()*10.0 ) - 20.0;
-	}
-	x = array( x );
-	y = array( y );
-
+	var x = array( uniform( len, -100.0, 100.0, opts ) );
+	var y = array( uniform( len, -100.0, 100.0, opts ) );
 	return benchmark;
 
+	/**
+	* Benchmark function.
+	*
+	* @private
+	* @param {Benchmark} b - benchmark instance
+	*/
 	function benchmark( b ) {
 		var d;
 		var i;
 
 		b.tic();
 		for ( i = 0; i < b.iterations; i++ ) {
 			d = sdot( x, y );
-			if ( isnan( d ) ) {
+			if ( isnan( d.get() ) ) {
 				b.fail( 'should not return NaN' );
 			}
 		}
 		b.toc();
-		if ( isnan( d ) ) {
+		if ( isnan( d.get() ) ) {
 			b.fail( 'should not return NaN' );
 		}
 		b.pass( 'benchmark finished' );
@@ -96,7 +97,7 @@ function main() {
 	for ( i = min; i <= max; i++ ) {
 		len = pow( 10, i );
 		f = createBenchmark( len );
-		bench( pkg+':len='+len, f );
+		bench( pkg+'::vectors:len='+len, f );
 	}
 }
 

lib/node_modules/@stdlib/blas/sdot/benchmark/benchmark.stack.js

@@ -0,0 +1,122 @@
+/**
+* @license Apache-2.0
+*
+* Copyright (c) 2020 The Stdlib Authors.
+*
+* Licensed under the Apache License, Version 2.0 (the "License");
+* you may not use this file except in compliance with the License.
+* You may obtain a copy of the License at
+*
+*    http://www.apache.org/licenses/LICENSE-2.0
+*
+* Unless required by applicable law or agreed to in writing, software
+* distributed under the License is distributed on an "AS IS" BASIS,
+* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+* See the License for the specific language governing permissions and
+* limitations under the License.
+*/
+
+'use strict';
+
+// MODULES //
+
+var bench = require( '@stdlib/bench' );
+var isnan = require( '@stdlib/math/base/assert/is-nan' );
+var pow = require( '@stdlib/math/base/special/pow' );
+var uniform = require( '@stdlib/random/array/uniform' );
+var numel = require( '@stdlib/ndarray/base/numel' );
+var array = require( '@stdlib/ndarray/array' );
+var pkg = require( './../package.json' ).name;
+var sdot = require( './../lib/main.js' );
+
+
+// VARIABLES //
+
+var OPTS = {
+	'dtype': 'float32'
+};
+
+
+// FUNCTIONS //
+
+/**
+* Creates a benchmark function.
+*
+* @private
+* @param {PositiveIntegerArray} shape - array shape
+* @returns {Function} benchmark function
+*/
+function createBenchmark( shape ) {
+	var x;
+	var y;
+	var N;
+	var o;
+
+	N = numel( shape );
+	o = {
+		'shape': shape
+	};
+	x = array( uniform( N, -100.0, 100.0, OPTS ), o );
+	y = array( uniform( N, -100.0, 100.0, OPTS ), o );
+
+	return benchmark;
+
+	/**
+	* Benchmark function.
+	*
+	* @private
+	* @param {Benchmark} b - benchmark instance
+	*/
+	function benchmark( b ) {
+		var d;
+		var i;
+
+		b.tic();
+		for ( i = 0; i < b.iterations; i++ ) {
+			d = sdot( x, y );
+			if ( isnan( d.iget( 0 ) ) ) {
+				b.fail( 'should not return NaN' );
+			}
+		}
+		b.toc();
+		if ( isnan( d.iget( 0 ) ) ) {
+			b.fail( 'should not return NaN' );
+		}
+		b.pass( 'benchmark finished' );
+		b.end();
+	}
+}
+
+
+// MAIN //
+
+/**
+* Main execution sequence.
+*
+* @private
+*/
+function main() {
+	var shape;
+	var min;
+	var max;
+	var N;
+	var f;
+	var i;
+
+	min = 1; // 10^min
+	max = 6; // 10^max
+
+	for ( i = min; i <= max; i++ ) {
+		N = pow( 10, i );
+
+		shape = [ 2, N/2 ];
+		f = createBenchmark( shape );
+		bench( pkg+'::stacks:size='+N+',ndims='+shape.length+',shape=('+shape.join( ',' )+')', f );
+
+		shape = [ N/2, 2 ];
+		f = createBenchmark( shape );
+		bench( pkg+'::stacks:size='+N+',ndims='+shape.length+',shape=('+shape.join( ',' )+')', f );
+	}
+}
+
+main();