diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/README.md b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/README.md index 0409c0431cc4..099e037756ed 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/README.md +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/README.md @@ -36,7 +36,7 @@ limitations under the License. var ssumpw = require( '@stdlib/blas/ext/base/ssumpw' ); ``` -#### ssumpw( N, x, stride ) +#### ssumpw( N, x, strideX ) Computes the sum of single-precision floating-point strided array elements using pairwise summation. @@ -44,9 +44,8 @@ Computes the sum of single-precision floating-point strided array elements using var Float32Array = require( '@stdlib/array/float32' ); var x = new Float32Array( [ 1.0, -2.0, 2.0 ] ); -var N = x.length; -var v = ssumpw( N, x, 1 ); +var v = ssumpw( x.length, x, 1 ); // returns 1.0 ``` @@ -54,9 +53,9 @@ The function has the following parameters: - **N**: number of indexed elements. - **x**: input [`Float32Array`][@stdlib/array/float32]. -- **stride**: index increment for `x`. +- **strideX**: stride length for `x`. -The `N` and stride parameters determine which elements in the strided array are accessed at runtime. For example, to compute the sum of every other element in `x`, +The `N` and stride parameters determine which elements in the strided array are accessed at runtime. For example, to compute the sum of every other element: ```javascript var Float32Array = require( '@stdlib/array/float32' ); @@ -81,7 +80,7 @@ var v = ssumpw( 4, x1, 2 ); // returns 5.0 ``` -#### ssumpw.ndarray( N, x, stride, offset ) +#### ssumpw.ndarray( N, x, strideX, offsetX ) Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. @@ -89,17 +88,16 @@ Computes the sum of single-precision floating-point strided array elements using var Float32Array = require( '@stdlib/array/float32' ); var x = new Float32Array( [ 1.0, -2.0, 2.0 ] ); -var N = x.length; -var v = ssumpw.ndarray( N, x, 1, 0 ); +var v = ssumpw.ndarray( x.length, x, 1, 0 ); // returns 1.0 ``` The function has the following additional parameters: -- **offset**: starting index for `x`. +- **offsetX**: starting index for `x`. -While [`typed array`][mdn-typed-array] views mandate a view offset based on the underlying `buffer`, the `offset` parameter supports indexing semantics based on a starting index. For example, to calculate the sum of every other value in `x` starting from the second value +While [`typed array`][mdn-typed-array] views mandate a view offset based on the underlying buffer, the offset parameter supports indexing semantics based on a starting index. For example, to calculate the sum of every other element starting from the second element: ```javascript var Float32Array = require( '@stdlib/array/float32' ); @@ -147,6 +145,123 @@ console.log( v ); + + +* * * + +
+ +## C APIs + + + +
+ +
+ + + + + +
+ +### Usage + +```c +#include "stdlib/blas/ext/base/ssumpw.h" +``` + +#### stdlib_strided_ssumpw( N, \*X, strideX ) + +Computes the sum of single-precision floating-point strided array elements using pairwise summation. + +```c +const float x[] = { 1.0f, -2.0f, 2.0f }; + +double v = stdlib_strided_ssumpw( 3, x, 1 ); +// returns 1.0 +``` + +The function accepts the following arguments: + +- **N**: `[in] CBLAS_INT` number of indexed elements. +- **X**: `[in] float*` input array. +- **strideX**: `[in] CBLAS_INT` stride length for `X`. + +```c +double stdlib_strided_ssumpw( const CBLAS_INT N, const float *X, const CBLAS_INT strideX ); +``` + +#### stdlib_strided_ssumpw_ndarray( N, \*X, strideX, offsetX ) + +Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. + +```c +const float x[] = { 1.0f, -2.0f, 2.0f }; + +double v = stdlib_strided_ssumpw_ndarray( 3, x, 1, 0 ); +// returns 1.0 +``` + +The function accepts the following arguments: + +- **N**: `[in] CBLAS_INT` number of indexed elements. +- **X**: `[in] float*` input array. +- **strideX**: `[in] CBLAS_INT` stride length for `X`. +- **offsetX**: `[in] CBLAS_INT` starting index for `X`. + +```c +double stdlib_strided_ssumpw_ndarray( const CBLAS_INT N, const float *X, const CBLAS_INT strideX, const CBLAS_INT offsetX ); +``` + +
+ + + + + +
+ +
+ + + + + +
+ +### Examples + +```c +#include "stdlib/blas/ext/base/ssumpw.h" +#include + +int main( void ) { + // Create a strided array: + const float x[] = { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f }; + + // Specify the number of elements: + const int N = 4; + + // Specify the stride length: + const int strideX = 2; + + // Compute the sum: + float v = stdlib_strided_ssumpw( N, x, strideX ); + + // Print the result: + printf( "sum: %f\n", v ); +} +``` + +
+ + + +
+ + + * * *
diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/benchmark/c/benchmark.length.c b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/benchmark/c/benchmark.length.c index b16f3996f92e..43deee0705a7 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/benchmark/c/benchmark.length.c +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/benchmark/c/benchmark.length.c @@ -94,7 +94,7 @@ static float rand_float( void ) { * @param len array length * @return elapsed time in seconds */ -static double benchmark( int iterations, int len ) { +static double benchmark1( int iterations, int len ) { double elapsed; float x[ len ]; float v; @@ -107,6 +107,7 @@ static double benchmark( int iterations, int len ) { v = 0.0f; t = tic(); for ( i = 0; i < iterations; i++ ) { + // cppcheck-suppress uninitvar v = stdlib_strided_ssumpw( len, x, 1 ); if ( v != v ) { printf( "should not return NaN\n" ); @@ -120,6 +121,40 @@ static double benchmark( int iterations, int len ) { return elapsed; } +/** +* Runs a benchmark. +* +* @param iterations number of iterations +* @param len array length +* @return elapsed time in seconds +*/ +static double benchmark2( int iterations, int len ) { + double elapsed; + float x[ len ]; + double v; + double t; + int i; + + for ( i = 0; i < len; i++ ) { + x[ i ] = ( rand_float()*20000.0f ) - 10000.0f; + } + v = 0.0f; + t = tic(); + for ( i = 0; i < iterations; i++ ) { + // cppcheck-suppress uninitvar + v = stdlib_strided_ssumpw_ndarray( len, x, 1, 0 ); + if ( v != v ) { + printf( "should not return NaN\n" ); + break; + } + } + elapsed = tic() - t; + if ( v != v ) { + printf( "should not return NaN\n" ); + } + return elapsed; +} + /** * Main execution sequence. */ @@ -142,7 +177,18 @@ int main( void ) { for ( j = 0; j < REPEATS; j++ ) { count += 1; printf( "# c::%s:len=%d\n", NAME, len ); - elapsed = benchmark( iter, len ); + elapsed = benchmark1( iter, len ); + print_results( iter, elapsed ); + printf( "ok %d benchmark finished\n", count ); + } + } + for ( i = MIN; i <= MAX; i++ ) { + len = pow( 10, i ); + iter = ITERATIONS / pow( 10, i-1 ); + for ( j = 0; j < REPEATS; j++ ) { + count += 1; + printf( "# c::%s:ndarray:len=%d\n", NAME, len ); + elapsed = benchmark2( iter, len ); print_results( iter, elapsed ); printf( "ok %d benchmark finished\n", count ); } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/repl.txt b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/repl.txt index 124979eb58c8..7a9163d7aed9 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/repl.txt +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/repl.txt @@ -1,10 +1,10 @@ -{{alias}}( N, x, stride ) +{{alias}}( N, x, strideX ) Computes the sum of single-precision floating-point strided array elements using pairwise summation. - The `N` and `stride` parameters determine which elements in the strided - array are accessed at runtime. + The `N` and stride parameters determine which elements in the strided array + are accessed at runtime. Indexing is relative to the first index. To introduce an offset, use a typed array view. @@ -19,8 +19,8 @@ x: Float32Array Input array. - stride: integer - Index increment. + strideX: integer + Stride length. Returns ------- @@ -34,7 +34,7 @@ > {{alias}}( x.length, x, 1 ) 1.0 - // Using `N` and `stride` parameters: + // Using `N` and stride parameters: > x = new {{alias:@stdlib/array/float32}}( [ -2.0, 1.0, 1.0, -5.0, 2.0, -1.0 ] ); > {{alias}}( 3, x, 2 ) 1.0 @@ -46,12 +46,12 @@ -1.0 -{{alias}}.ndarray( N, x, stride, offset ) +{{alias}}.ndarray( N, x, strideX, offsetX ) Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. While typed array views mandate a view offset based on the underlying - buffer, the `offset` parameter supports indexing semantics based on a + buffer, the offset parameter supports indexing semantics based on a starting index. Parameters @@ -62,10 +62,10 @@ x: Float32Array Input array. - stride: integer - Index increment. + strideX: integer + Stride length. - offset: integer + offsetX: integer Starting index. Returns diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/types/index.d.ts b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/types/index.d.ts index 02c827ce1925..e902e3055afa 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/types/index.d.ts +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/docs/types/index.d.ts @@ -27,7 +27,7 @@ interface Routine { * * @param N - number of indexed elements * @param x - input array - * @param stride - stride length + * @param strideX - stride length * @returns sum * * @example @@ -38,15 +38,15 @@ interface Routine { * var v = ssumpw( x.length, x, 1 ); * // returns 1.0 */ - ( N: number, x: Float32Array, stride: number ): number; + ( N: number, x: Float32Array, strideX: number ): number; /** * Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. * * @param N - number of indexed elements * @param x - input array - * @param stride - stride length - * @param offset - starting index + * @param strideX - stride length + * @param offsetX - starting index * @returns sum * * @example @@ -57,7 +57,7 @@ interface Routine { * var v = ssumpw.ndarray( x.length, x, 1, 0 ); * // returns 1.0 */ - ndarray( N: number, x: Float32Array, stride: number, offset: number ): number; + ndarray( N: number, x: Float32Array, strideX: number, offsetX: number ): number; } /** @@ -65,7 +65,7 @@ interface Routine { * * @param N - number of indexed elements * @param x - input array -* @param stride - stride length +* @param strideX - stride length * @returns sum * * @example diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/examples/c/example.c b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/examples/c/example.c index 60e6e6fad586..47874abf1740 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/examples/c/example.c +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/examples/c/example.c @@ -17,21 +17,20 @@ */ #include "stdlib/blas/ext/base/ssumpw.h" -#include #include int main( void ) { // Create a strided array: - const float x[] = { 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0 }; + const float x[] = { 1.0f, 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f }; // Specify the number of elements: - const int64_t N = 4; + const int N = 4; // Specify the stride length: - const int64_t stride = 2; + const int strideX = 2; // Compute the sum: - float v = stdlib_strided_ssumpw( N, x, stride ); + float v = stdlib_strided_ssumpw( N, x, strideX ); // Print the result: printf( "sum: %f\n", v ); diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/include/stdlib/blas/ext/base/ssumpw.h b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/include/stdlib/blas/ext/base/ssumpw.h index 9f6f0c4a599f..3f44488af3da 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/include/stdlib/blas/ext/base/ssumpw.h +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/include/stdlib/blas/ext/base/ssumpw.h @@ -19,7 +19,7 @@ #ifndef STDLIB_BLAS_EXT_BASE_SSUMPW_H #define STDLIB_BLAS_EXT_BASE_SSUMPW_H -#include +#include "stdlib/blas/base/shared.h" /* * If C++, prevent name mangling so that the compiler emits a binary file having undecorated names, thus mirroring the behavior of a C compiler. @@ -31,7 +31,12 @@ extern "C" { /** * Computes the sum of single-precision floating-point strided array elements using pairwise summation. */ -float stdlib_strided_ssumpw( const int64_t N, const float *X, const int64_t stride ); +double API_SUFFIX(stdlib_strided_ssumpw)( const CBLAS_INT N, const float *X, const CBLAS_INT strideX ); + +/** +* Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. +*/ +double API_SUFFIX(stdlib_strided_ssumpw_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT strideX, const CBLAS_INT offsetX ); #ifdef __cplusplus } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/index.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/index.js index 1fff659ec3b6..fb578cab4308 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/index.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/index.js @@ -28,9 +28,8 @@ * var ssumpw = require( '@stdlib/blas/ext/base/ssumpw' ); * * var x = new Float32Array( [ 1.0, -2.0, 2.0 ] ); -* var N = x.length; * -* var v = ssumpw( N, x, 1 ); +* var v = ssumpw( x.length, x, 1 ); * // returns 1.0 * * @example diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.js index 8626313fcd9a..d73d98f00de9 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.js @@ -22,6 +22,7 @@ var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); var floor = require( '@stdlib/math/base/special/floor' ); +var isnan = require( '@stdlib/math/base/assert/is-nan' ); // VARIABLES // @@ -45,8 +46,8 @@ var BLOCKSIZE = 128; * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {integer} stride - stride length -* @param {NonNegativeInteger} offset - starting index +* @param {integer} strideX - stride length +* @param {NonNegativeInteger} offsetX - starting index * @returns {number} sum * * @example @@ -57,7 +58,7 @@ var BLOCKSIZE = 128; * var v = ssumpw( 4, x, 2, 1 ); * // returns 5.0 */ -function ssumpw( N, x, stride, offset ) { +function ssumpw( N, x, strideX, offsetX ) { var ix; var s0; var s1; @@ -75,42 +76,45 @@ function ssumpw( N, x, stride, offset ) { if ( N <= 0 ) { return 0.0; } - if ( N === 1 || stride === 0 ) { - return x[ offset ]; + ix = offsetX; + if ( strideX === 0 ) { + if ( isnan( x[ ix ] ) ) { + return 0.0; + } + return N * x[ ix ]; } - ix = offset; if ( N < 8 ) { // Use simple summation... s = 0.0; for ( i = 0; i < N; i++ ) { s = float64ToFloat32( s + x[ ix ] ); - ix += stride; + ix += strideX; } return s; } if ( N <= BLOCKSIZE ) { // Sum a block with 8 accumulators (by loop unrolling, we lower the effective blocksize to 16)... s0 = x[ ix ]; - s1 = x[ ix+stride ]; - s2 = x[ ix+(2*stride) ]; - s3 = x[ ix+(3*stride) ]; - s4 = x[ ix+(4*stride) ]; - s5 = x[ ix+(5*stride) ]; - s6 = x[ ix+(6*stride) ]; - s7 = x[ ix+(7*stride) ]; - ix += 8 * stride; + s1 = x[ ix+strideX ]; + s2 = x[ ix+(2*strideX) ]; + s3 = x[ ix+(3*strideX) ]; + s4 = x[ ix+(4*strideX) ]; + s5 = x[ ix+(5*strideX) ]; + s6 = x[ ix+(6*strideX) ]; + s7 = x[ ix+(7*strideX) ]; + ix += 8 * strideX; M = N % 8; for ( i = 8; i < N-M; i += 8 ) { s0 = float64ToFloat32( s0 + x[ ix ] ); - s1 = float64ToFloat32( s1 + x[ ix+stride ] ); - s2 = float64ToFloat32( s2 + x[ ix+(2*stride) ] ); - s3 = float64ToFloat32( s3 + x[ ix+(3*stride) ] ); - s4 = float64ToFloat32( s4 + x[ ix+(4*stride) ] ); - s5 = float64ToFloat32( s5 + x[ ix+(5*stride) ] ); - s6 = float64ToFloat32( s6 + x[ ix+(6*stride) ] ); - s7 = float64ToFloat32( s7 + x[ ix+(7*stride) ] ); - ix += 8 * stride; + s1 = float64ToFloat32( s1 + x[ ix+strideX ] ); + s2 = float64ToFloat32( s2 + x[ ix+(2*strideX) ] ); + s3 = float64ToFloat32( s3 + x[ ix+(3*strideX) ] ); + s4 = float64ToFloat32( s4 + x[ ix+(4*strideX) ] ); + s5 = float64ToFloat32( s5 + x[ ix+(5*strideX) ] ); + s6 = float64ToFloat32( s6 + x[ ix+(6*strideX) ] ); + s7 = float64ToFloat32( s7 + x[ ix+(7*strideX) ] ); + ix += 8 * strideX; } // Pairwise sum the accumulators: s = float64ToFloat32( float64ToFloat32( float64ToFloat32(s0+s1) + float64ToFloat32(s2+s3) ) + float64ToFloat32( float64ToFloat32(s4+s5) + float64ToFloat32(s6+s7) ) ); // eslint-disable-line max-len @@ -118,14 +122,14 @@ function ssumpw( N, x, stride, offset ) { // Clean-up loop... for ( i; i < N; i++ ) { s = float64ToFloat32( s + x[ ix ] ); - ix += stride; + ix += strideX; } return s; } // Recurse by dividing by two, but avoiding non-multiples of unroll factor... n = floor( N/2 ); n -= n % 8; - return float64ToFloat32( ssumpw( n, x, stride, ix ) + ssumpw( N-n, x, stride, ix+(n*stride) ) ); // eslint-disable-line max-len + return float64ToFloat32( ssumpw( n, x, strideX, ix ) + ssumpw( N-n, x, strideX, ix+(n*strideX) ) ); // eslint-disable-line max-len } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.native.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.native.js index e6bfcf630720..f206e28dec82 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.native.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ndarray.native.js @@ -20,9 +20,7 @@ // MODULES // -var minViewBufferIndex = require( '@stdlib/strided/base/min-view-buffer-index' ); -var offsetView = require( '@stdlib/strided/base/offset-view' ); -var addon = require( './ssumpw.native.js' ); +var addon = require( './../src/addon.node' ); // MAIN // @@ -32,8 +30,8 @@ var addon = require( './ssumpw.native.js' ); * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {integer} stride - stride length -* @param {NonNegativeInteger} offset - starting index +* @param {integer} strideX - stride length +* @param {NonNegativeInteger} offsetX - starting index * @returns {number} sum * * @example @@ -44,11 +42,8 @@ var addon = require( './ssumpw.native.js' ); * var v = ssumpw( 4, x, 2, 1 ); * // returns 5.0 */ -function ssumpw( N, x, stride, offset ) { - var view; - offset = minViewBufferIndex( N, stride, offset ); - view = offsetView( x, offset ); - return addon( N, view, stride ); +function ssumpw( N, x, strideX, offsetX ) { + return addon.ndarray( N, x, strideX, offsetX ); } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.js index 50eefb89e3c7..bf5308c132be 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.js @@ -20,8 +20,8 @@ // MODULES // -var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); -var sum = require( './ndarray.js' ); +var stride2offset = require( '@stdlib/strided/base/stride2offset' ); +var ndarray = require( './ndarray.js' ); // MAIN // @@ -39,44 +39,19 @@ var sum = require( './ndarray.js' ); * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {integer} stride - stride length +* @param {integer} strideX - stride length * @returns {number} sum * * @example * var Float32Array = require( '@stdlib/array/float32' ); * * var x = new Float32Array( [ 1.0, -2.0, 2.0 ] ); -* var N = x.length; * -* var v = ssumpw( N, x, 1 ); +* var v = ssumpw( x.length, x, 1 ); * // returns 1.0 */ -function ssumpw( N, x, stride ) { - var ix; - var s; - var i; - - if ( N <= 0 ) { - return 0.0; - } - if ( N === 1 || stride === 0 ) { - return x[ 0 ]; - } - if ( stride < 0 ) { - ix = (1-N) * stride; - } else { - ix = 0; - } - if ( N < 8 ) { - // Use simple summation... - s = 0.0; - for ( i = 0; i < N; i++ ) { - s = float64ToFloat32( s + x[ ix ] ); - ix += stride; - } - return s; - } - return sum( N, x, stride, ix ); +function ssumpw( N, x, strideX ) { + return ndarray( N, x, strideX, stride2offset( N, strideX ) ); } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.native.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.native.js index e52820181c60..88fdadc0cefe 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.native.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/lib/ssumpw.native.js @@ -30,20 +30,19 @@ var addon = require( './../src/addon.node' ); * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {integer} stride - stride length +* @param {integer} strideX - stride length * @returns {number} sum * * @example * var Float32Array = require( '@stdlib/array/float32' ); * * var x = new Float32Array( [ 1.0, -2.0, 2.0 ] ); -* var N = x.length; * -* var v = ssumpw( N, x, 1 ); +* var v = ssumpw( x.length, x, 1 ); * // returns 1.0 */ -function ssumpw( N, x, stride ) { - return addon( N, x, stride ); +function ssumpw( N, x, strideX ) { + return addon( N, x, strideX ); } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/manifest.json b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/manifest.json index 92cb2fbd64fb..2e4581f2d856 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/manifest.json +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/manifest.json @@ -28,49 +28,55 @@ { "task": "build", "src": [ - "./src/ssumpw.c" + "./src/main.c" ], "include": [ "./include" ], - "libraries": [ - "-lm" - ], + "libraries": [], "libpath": [], "dependencies": [ "@stdlib/napi/export", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", - "@stdlib/napi/argv-strided-float32array" + "@stdlib/napi/argv-strided-float32array", + "@stdlib/napi/create-double", + "@stdlib/math/base/assert/is-nanf", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/stride2offset" ] }, { "task": "benchmark", "src": [ - "./src/ssumpw.c" + "./src/main.c" ], "include": [ "./include" ], - "libraries": [ - "-lm" - ], + "libraries": [], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/math/base/assert/is-nanf", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/stride2offset" + ] }, { "task": "examples", "src": [ - "./src/ssumpw.c" + "./src/main.c" ], "include": [ "./include" ], - "libraries": [ - "-lm" - ], + "libraries": [], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/math/base/assert/is-nanf", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/stride2offset" + ] } ] } diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/addon.c b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/addon.c index 48352d37510f..e0868a246b27 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/addon.c +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/addon.c @@ -17,12 +17,13 @@ */ #include "stdlib/blas/ext/base/ssumpw.h" +#include "stdlib/blas/base/shared.h" #include "stdlib/napi/export.h" #include "stdlib/napi/argv.h" #include "stdlib/napi/argv_int64.h" #include "stdlib/napi/argv_strided_float32array.h" +#include "stdlib/napi/create_double.h" #include -#include /** * Receives JavaScript callback invocation data. @@ -34,14 +35,27 @@ static napi_value addon( napi_env env, napi_callback_info info ) { STDLIB_NAPI_ARGV( env, info, argv, argc, 3 ); STDLIB_NAPI_ARGV_INT64( env, N, argv, 0 ); - STDLIB_NAPI_ARGV_INT64( env, stride, argv, 2 ); - STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, stride, argv, 1 ); - - napi_value v; - napi_status status = napi_create_double( env, (double)stdlib_strided_ssumpw( N, X, stride ), &v ); - assert( status == napi_ok ); + STDLIB_NAPI_ARGV_INT64( env, strideX, argv, 2 ); + STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, strideX, argv, 1 ); + STDLIB_NAPI_CREATE_DOUBLE( env, API_SUFFIX(stdlib_strided_ssumpw)( N, X, strideX ), v ) + return v; +} +/** +* Receives JavaScript callback invocation data. +* +* @param env environment under which the function is invoked +* @param info callback data +* @return Node-API value +*/ +static napi_value addon_method( napi_env env, napi_callback_info info ) { + STDLIB_NAPI_ARGV( env, info, argv, argc, 4 ); + STDLIB_NAPI_ARGV_INT64( env, N, argv, 0 ); + STDLIB_NAPI_ARGV_INT64( env, strideX, argv, 2 ); + STDLIB_NAPI_ARGV_INT64( env, offsetX, argv, 3 ); + STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, strideX, argv, 1 ); + STDLIB_NAPI_CREATE_DOUBLE( env, API_SUFFIX(stdlib_strided_ssumpw_ndarray)( N, X, strideX, offsetX ), v ) return v; } -STDLIB_NAPI_MODULE_EXPORT_FCN( addon ) +STDLIB_NAPI_MODULE_EXPORT_FCN_WITH_METHOD( addon, "ndarray", addon_method ); diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/main.c b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/main.c new file mode 100644 index 000000000000..910cc7a804d2 --- /dev/null +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/main.c @@ -0,0 +1,135 @@ +/** +* @license Apache-2.0 +* +* Copyright (c) 2024 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. +*/ + +#include "stdlib/blas/ext/base/ssumpw.h" +#include "stdlib/math/base/assert/is_nanf.h" +#include "stdlib/blas/base/shared.h" +#include "stdlib/strided/base/stride2offset.h" + +/** +* Computes the sum of single-precision floating-point strided array elements using pairwise summation. +* +* ## Method +* +* - This implementation uses pairwise summation, which accrues rounding error `O(log2 N)` instead of `O(N)`. The recursion depth is also `O(log2 N)`. +* +* ## References +* +* - Higham, Nicholas J. 1993. "The Accuracy of Floating Point Summation." _SIAM Journal on Scientific Computing_ 14 (4): 783–99. doi:[10.1137/0914050](https://doi.org/10.1137/0914050). +* +* @param N number of indexed elements +* @param X input array +* @param strideX stride length +* @return output value +*/ +double API_SUFFIX(stdlib_strided_ssumpw)( const CBLAS_INT N, const float *X, const CBLAS_INT strideX ) { + CBLAS_INT ox = stdlib_strided_stride2offset( N, strideX ); + API_SUFFIX(stdlib_strided_ssumpw_ndarray)( N, X, strideX, ox ); +} + +/** +* Computes the sum of single-precision floating-point strided array elements using pairwise summation and alternative indexing semantics. +* +* ## Method +* +* - This implementation uses pairwise summation, which accrues rounding error `O(log2 N)` instead of `O(N)`. The recursion depth is also `O(log2 N)`. +* +* ## References +* +* - Higham, Nicholas J. 1993. "The Accuracy of Floating Point Summation." _SIAM Journal on Scientific Computing_ 14 (4): 783–99. doi:[10.1137/0914050](https://doi.org/10.1137/0914050). +* +* @param N number of indexed elements +* @param X input array +* @param strideX stride length +* @param offsetX starting index +* @return output value +*/ +double API_SUFFIX(stdlib_strided_ssumpw_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT strideX, const CBLAS_INT offsetX ) { + CBLAS_INT ix; + CBLAS_INT M; + CBLAS_INT n; + CBLAS_INT i; + double sum; + double s0; + double s1; + double s2; + double s3; + double s4; + double s5; + double s6; + double s7; + + if ( N <= 0 ) { + return 0.0f; + } + ix = offsetX; + if ( strideX == 0 ) { + if ( stdlib_base_is_nanf( X[ ix ] ) ) { + return 0.0f; + } + return N * X[ ix ]; + } + if ( N < 8 ) { + // Use simple summation... + sum = 0.0f; + for ( i = 0; i < N; i++ ) { + sum += X[ ix ]; + ix += strideX; + } + return sum; + } + // Blocksize for pairwise summation: 128 (NOTE: decreasing the blocksize decreases rounding error as more pairs are summed, but also decreases performance. Because the inner loop is unrolled eight times, the blocksize is effectively `16`.) + if ( N <= 128 ) { + // Sum a block with 8 accumulators (by loop unrolling, we lower the effective blocksize to 16)... + s0 = X[ ix ]; + s1 = X[ ix+strideX ]; + s2 = X[ ix+(2*strideX) ]; + s3 = X[ ix+(3*strideX) ]; + s4 = X[ ix+(4*strideX) ]; + s5 = X[ ix+(5*strideX) ]; + s6 = X[ ix+(6*strideX) ]; + s7 = X[ ix+(7*strideX) ]; + ix += 8 * strideX; + + M = N % 8; + for ( i = 8; i < N-M; i += 8 ) { + s0 += X[ ix ]; + s1 += X[ ix+strideX ]; + s2 += X[ ix+(2*strideX) ]; + s3 += X[ ix+(3*strideX) ]; + s4 += X[ ix+(4*strideX) ]; + s5 += X[ ix+(5*strideX) ]; + s6 += X[ ix+(6*strideX) ]; + s7 += X[ ix+(7*strideX) ]; + ix += 8 * strideX; + } + // Pairwise sum the accumulators: + sum = ( (s0+s1) + (s2+s3) ) + ( (s4+s5) + (s6+s7) ); + + // Clean-up loop... + for (; i < N; i++ ) { + sum += X[ ix ]; + ix += strideX; + } + return sum; + } + // Recurse by dividing by two, but avoiding non-multiples of unroll factor... + n = N / 2; + n -= n % 8; + return stdlib_strided_ssumpw_ndarray( n, X, strideX, ix ) + stdlib_strided_ssumpw_ndarray( N-n, X, strideX, ix+(n*strideX) ); +} diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/ssumpw.c b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/ssumpw.c deleted file mode 100644 index 0fcdb8965649..000000000000 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/src/ssumpw.c +++ /dev/null @@ -1,121 +0,0 @@ -/** -* @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. -*/ - -#include "stdlib/blas/ext/base/ssumpw.h" -#include - -/** -* Computes the sum of single-precision floating-point strided array elements using pairwise summation. -* -* ## Method -* -* - This implementation uses pairwise summation, which accrues rounding error `O(log2 N)` instead of `O(N)`. The recursion depth is also `O(log2 N)`. -* -* ## References -* -* - Higham, Nicholas J. 1993. "The Accuracy of Floating Point Summation." _SIAM Journal on Scientific Computing_ 14 (4): 783–99. doi:[10.1137/0914050](https://doi.org/10.1137/0914050). -* -* @param N number of indexed elements -* @param X input array -* @param stride stride length -* @return output value -*/ -float stdlib_strided_ssumpw( const int64_t N, const float *X, const int64_t stride ) { - float *xp1; - float *xp2; - int64_t ix; - int64_t M; - int64_t n; - int64_t i; - float sum; - float s0; - float s1; - float s2; - float s3; - float s4; - float s5; - float s6; - float s7; - - if ( N <= 0 ) { - return 0.0f; - } - if ( N == 1 || stride == 0 ) { - return X[ 0 ]; - } - if ( stride < 0 ) { - ix = (1-N) * stride; - } else { - ix = 0; - } - if ( N < 8 ) { - // Use simple summation... - sum = 0.0f; - for ( i = 0; i < N; i++ ) { - sum += X[ ix ]; - ix += stride; - } - return sum; - } - // Blocksize for pairwise summation: 128 (NOTE: decreasing the blocksize decreases rounding error as more pairs are summed, but also decreases performance. Because the inner loop is unrolled eight times, the blocksize is effectively `16`.) - if ( N <= 128 ) { - // Sum a block with 8 accumulators (by loop unrolling, we lower the effective blocksize to 16)... - s0 = X[ ix ]; - s1 = X[ ix+stride ]; - s2 = X[ ix+(2*stride) ]; - s3 = X[ ix+(3*stride) ]; - s4 = X[ ix+(4*stride) ]; - s5 = X[ ix+(5*stride) ]; - s6 = X[ ix+(6*stride) ]; - s7 = X[ ix+(7*stride) ]; - ix += 8 * stride; - - M = N % 8; - for ( i = 8; i < N-M; i += 8 ) { - s0 += X[ ix ]; - s1 += X[ ix+stride ]; - s2 += X[ ix+(2*stride) ]; - s3 += X[ ix+(3*stride) ]; - s4 += X[ ix+(4*stride) ]; - s5 += X[ ix+(5*stride) ]; - s6 += X[ ix+(6*stride) ]; - s7 += X[ ix+(7*stride) ]; - ix += 8 * stride; - } - // Pairwise sum the accumulators: - sum = ((s0+s1) + (s2+s3)) + ((s4+s5) + (s6+s7)); - - // Clean-up loop... - for (; i < N; i++ ) { - sum += X[ ix ]; - ix += stride; - } - return sum; - } - // Recurse by dividing by two, but avoiding non-multiples of unroll factor... - n = N / 2; - n -= n % 8; - if ( stride < 0 ) { - xp1 = (float *)X + ( (n-N)*stride ); - xp2 = (float *)X; - } else { - xp1 = (float *)X; - xp2 = (float *)X + ( n*stride ); - } - return stdlib_strided_ssumpw( n, xp1, stride ) + stdlib_strided_ssumpw( N-n, xp2, stride ); -} diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.js index e9c183089d57..7f66952466c4 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.js @@ -156,14 +156,26 @@ tape( 'the function supports a negative `stride` parameter', function test( t ) t.end(); }); -tape( 'if provided a `stride` parameter equal to `0`, the function returns the first indexed element', function test( t ) { +tape( 'if provided a `stride` parameter equal to `0`, the function returns the sum of the first element repeated N times', function test( t ) { var x; var v; x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); v = ssumpw( x.length, x, 0, 0 ); - t.strictEqual( v, 1.0, 'returns expected value' ); + t.strictEqual( v, 5.0, 'returns expected value' ); + + t.end(); +}); + +tape( 'if provided a `stride` parameter equal to `0` and the first element is `NaN`, the function returns 0.0', function test( t ) { + var x; + var v; + + x = new Float32Array( [ NaN, -2.0, -4.0, 5.0, 3.0 ] ); + + v = ssumpw( x.length, x, 0, 0 ); + t.strictEqual( v, 0.0, 'returns expected value' ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.native.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.native.js index 99493e472db6..32adc1f11529 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.native.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ndarray.native.js @@ -165,14 +165,26 @@ tape( 'the function supports a negative `stride` parameter', opts, function test t.end(); }); -tape( 'if provided a `stride` parameter equal to `0`, the function returns the first indexed element', opts, function test( t ) { +tape( 'if provided a `stride` parameter equal to `0`, the function returns the sum of the first element repeated N times', opts, function test( t ) { var x; var v; x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); v = ssumpw( x.length, x, 0, 0 ); - t.strictEqual( v, 1.0, 'returns expected value' ); + t.strictEqual( v, 5.0, 'returns expected value' ); + + t.end(); +}); + +tape( 'if provided a `stride` parameter equal to `0` and the first element is `NaN`, the function returns 0.0', opts, function test( t ) { + var x; + var v; + + x = new Float32Array( [ NaN, -2.0, -4.0, 5.0, 3.0 ] ); + + v = ssumpw( x.length, x, 0, 0 ); + t.strictEqual( v, 0.0, 'returns expected value' ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.js index ea6438f6a75a..0eb782742a30 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.js @@ -156,14 +156,26 @@ tape( 'the function supports a negative `stride` parameter', function test( t ) t.end(); }); -tape( 'if provided a `stride` parameter equal to `0`, the function returns the first element', function test( t ) { +tape( 'if provided a `stride` parameter equal to `0`, the function returns the sum of the first element repeated N times', function test( t ) { var x; var v; x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); v = ssumpw( x.length, x, 0 ); - t.strictEqual( v, 1.0, 'returns expected value' ); + t.strictEqual( v, 5.0, 'returns expected value' ); + + t.end(); +}); + +tape( 'if provided a `stride` parameter equal to `0` and the first element is `NaN`, the function returns 0.0', function test( t ) { + var x; + var v; + + x = new Float32Array( [ NaN, -2.0, -4.0, 5.0, 3.0 ] ); + + v = ssumpw( x.length, x, 0 ); + t.strictEqual( v, 0.0, 'returns expected value' ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.native.js b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.native.js index f8c3bac2d293..182dba95473b 100644 --- a/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.native.js +++ b/lib/node_modules/@stdlib/blas/ext/base/ssumpw/test/test.ssumpw.native.js @@ -247,14 +247,26 @@ tape( 'the function supports a negative `stride` parameter', opts, function test t.end(); }); -tape( 'if provided a `stride` parameter equal to `0`, the function returns the first element', opts, function test( t ) { +tape( 'if provided a `stride` parameter equal to `0`, the function returns the sum of the first element repeated N times', opts, function test( t ) { var x; var v; x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); v = ssumpw( x.length, x, 0 ); - t.strictEqual( v, 1.0, 'returns expected value' ); + t.strictEqual( v, 5.0, 'returns expected value' ); + + t.end(); +}); + +tape( 'if provided a `stride` parameter equal to `0` and the first element is `NaN`, the function returns 0.0', opts, function test( t ) { + var x; + var v; + + x = new Float32Array( [ NaN, -2.0, -4.0, 5.0, 3.0 ] ); + + v = ssumpw( x.length, x, 0 ); + t.strictEqual( v, 0.0, 'returns expected value' ); t.end(); });