From 69234e5c789afb4f534d72a694ff8552ce4ab4bd Mon Sep 17 00:00:00 2001 From: aman-095 Date: Thu, 19 Sep 2024 08:01:18 +0530 Subject: [PATCH 1/3] feat: add C ndarray implemenatation for snrm2 --- .../base/snrm2/benchmark/c/benchmark.length.c | 44 ++++++++++++++++++- .../blas/base/snrm2/examples/c/example.c | 6 +++ .../snrm2/include/stdlib/blas/base/snrm2.h | 9 +++- .../include/stdlib/blas/base/snrm2_cblas.h | 4 +- 4 files changed, 59 insertions(+), 4 deletions(-) diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/benchmark/c/benchmark.length.c b/lib/node_modules/@stdlib/blas/base/snrm2/benchmark/c/benchmark.length.c index c78af1732557..e8d870efdd63 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/benchmark/c/benchmark.length.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/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 z; @@ -120,6 +120,39 @@ 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 ]; + float z; + double t; + int i; + + for ( i = 0; i < len; i++ ) { + x[ i ] = ( rand_float() * 20000.0f ) - 10000.0f; + } + z = 0.0f; + t = tic(); + for ( i = 0; i < iterations; i++ ) { + z = c_snrm2_ndarray( len, x, 1, 0 ); + if ( z != z ) { + printf( "should not return NaN\n" ); + break; + } + } + elapsed = tic() - t; + if ( z != z ) { + printf( "should not return NaN\n" ); + } + return elapsed; +} + /** * Main execution sequence. */ @@ -142,7 +175,14 @@ 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 ( 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/base/snrm2/examples/c/example.c b/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c index 7bd544c00112..da3b0a8cb0f1 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c @@ -34,4 +34,10 @@ int main( void ) { // Print the result: printf( "L2-norm: %f\n", l2 ); + + // Compute the L2-norm: + l2 = c_snrm2_ndarray( N, x, strideX, 0 ); + + // Print the result: + printf( "L2-norm: %f\n", l2 ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2.h b/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2.h index 346152670b15..c8858765a011 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2.h +++ b/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2.h @@ -22,6 +22,8 @@ #ifndef SNRM2_H #define SNRM2_H +#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. */ @@ -32,7 +34,12 @@ extern "C" { /** * Computes the L2-norm of a single-precision floating-point vector. */ -float c_snrm2( const int N, const float *X, const int stride ); +float API_SUFFIX(c_snrm2)( const CBLAS_INT N, const float *X, const CBLAS_INT stride ); + +/** +* Computes the L2-norm of a single-precision floating-point vector using alternative indexing semantics. +*/ +float API_SUFFIX(c_snrm2_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT stride, const CBLAS_INT offset ); #ifdef __cplusplus } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2_cblas.h b/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2_cblas.h index ff9198e8bef7..d46693ae9ae7 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2_cblas.h +++ b/lib/node_modules/@stdlib/blas/base/snrm2/include/stdlib/blas/base/snrm2_cblas.h @@ -22,6 +22,8 @@ #ifndef SNRM2_CBLAS_H #define SNRM2_CBLAS_H +#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. */ @@ -32,7 +34,7 @@ extern "C" { /** * Computes the L2-norm of a single-precision floating-point vector. */ -float cblas_snrm2( const int N, const float *X, const int stride ); +float API_SUFFIX(cblas_snrm2)( const CBLAS_INT N, const float *X, const CBLAS_INT stride ); #ifdef __cplusplus } From 326e25e4b3c34b55b44f0b9cd098bd72007e21df Mon Sep 17 00:00:00 2001 From: aman-095 Date: Thu, 19 Sep 2024 17:55:17 +0530 Subject: [PATCH 2/3] feat: update Js implementation and add C ndarray implementation for snrm2 --- .../@stdlib/blas/base/snrm2/README.md | 123 ++++++++++++++++ .../@stdlib/blas/base/snrm2/docs/repl.txt | 2 +- .../@stdlib/blas/base/snrm2/lib/ndarray.js | 93 +++++++++--- .../blas/base/snrm2/lib/ndarray.native.js | 12 +- .../@stdlib/blas/base/snrm2/lib/snrm2.js | 38 +---- .../blas/base/snrm2/lib/snrm2.native.js | 2 +- .../@stdlib/blas/base/snrm2/manifest.json | 133 +++++++++++++++--- .../@stdlib/blas/base/snrm2/src/addon.c | 26 +++- .../@stdlib/blas/base/snrm2/src/snrm2.c | 32 +---- .../@stdlib/blas/base/snrm2/src/snrm2.f | 100 ++++++++++--- .../@stdlib/blas/base/snrm2/src/snrm2_cblas.c | 28 +++- .../@stdlib/blas/base/snrm2/src/snrm2_f.c | 20 ++- .../blas/base/snrm2/src/snrm2_ndarray.c | 112 +++++++++++++++ .../blas/base/snrm2/test/test.ndarray.js | 49 +++++-- .../base/snrm2/test/test.ndarray.native.js | 49 +++++-- .../blas/base/snrm2/test/test.snrm2.js | 64 +++++++-- .../blas/base/snrm2/test/test.snrm2.native.js | 64 +++++++-- 17 files changed, 765 insertions(+), 182 deletions(-) create mode 100644 lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_ndarray.c diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/README.md b/lib/node_modules/@stdlib/blas/base/snrm2/README.md index a88478bab488..7e545c23b273 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/README.md +++ b/lib/node_modules/@stdlib/blas/base/snrm2/README.md @@ -160,6 +160,129 @@ console.log( out ); + + +* * * + +
+ +## C APIs + + + +
+ +
+ + + + + +
+ +### Usage + +```c +#include "stdlib/blas/base/snrm2.h" +``` + +#### c_snrm2( N, \*X, stride ) + +Computes the L2-norm of a complex single-precision floating-point vector. + +```c +const float x[] = { 1.0f, -2.0f, 3.0f, -4.0f, 5.0f, -6.0f, 7.0f, -8.0f }; + +float norm = c_snrm2( 8, x, 1 ); +// returns 14.3 +``` + +The function accepts the following arguments: + +- **N**: `[in] CBLAS_INT` number of indexed elements. +- **X**: `[in] float*` input array. +- **stride**: `[in] CBLAS_INT` index increment for `X`. + +```c +float c_snrm2( const CBLAS_INT N, const float *X, const CBLAS_INT stride ); +``` + +#### c_snrm2_ndarray( N, \*X, stride, offset ) + +Computes the L2-norm of a complex single-precision floating-point vector using alternative indexing semantics. + +```c +const float x[] = { 1.0f, -2.0f, 3.0f, -4.0f, 5.0f, -6.0f, 7.0f, -8.0f }; + +float norm = c_snrm2_ndarray( 8, x, 1, 0 ); +// returns 14.3 +``` + +The function accepts the following arguments: + +- **N**: `[in] CBLAS_INT` number of indexed elements. +- **X**: `[in] float*` input array. +- **stride**: `[in] CBLAS_INT` index increment for `X`. +- **offset**: `[in] CBLAS_INT` starting index for `X`. + +```c +float c_snrm2_ndarray( const CBLAS_INT N, const float *X, const CBLAS_INT stride, const CBLAS_INT offset ); +``` + +
+ + + + + +
+ +
+ + + + + +
+ +### Examples + +```c +#include "stdlib/blas/base/snrm2.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 indexed elements: + const int N = 8; + + // Specify a stride: + const int strideX = 1; + + // Compute the L2-norm: + float l2 = c_snrm2( N, x, strideX ); + + // Print the result: + printf( "L2-norm: %f\n", l2 ); + + // Compute the L2-norm: + l2 = c_snrm2_ndarray( N, x, strideX, 0 ); + + // Print the result: + printf( "L2-norm: %f\n", l2 ); +} +``` + +
+ + + +
+ + +
diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/docs/repl.txt b/lib/node_modules/@stdlib/blas/base/snrm2/docs/repl.txt index 628744db5c76..016670f1b4d5 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/docs/repl.txt +++ b/lib/node_modules/@stdlib/blas/base/snrm2/docs/repl.txt @@ -8,7 +8,7 @@ Indexing is relative to the first index. To introduce an offset, use a typed array view. - If `N <= 0` or `stride <= 0`, the function returns `0`. + If `N <= 0` the function returns `0`. Parameters ---------- diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.js b/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.js index 08dedee49364..c9b704118679 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.js @@ -20,9 +20,20 @@ // MODULES // -var sqrtf = require( '@stdlib/math/base/special/sqrtf' ); +var FLOAT32_MAX = require( '@stdlib/constants/float32/max' ); +var f32 = require( '@stdlib/number/float64/base/to-float32' ); var absf = require( '@stdlib/math/base/special/absf' ); -var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); +var abs2f = require( '@stdlib/math/base/special/abs2f' ); +var sqrtf = require( '@stdlib/math/base/special/sqrtf' ); + + +// VARIABLES // + +// Blue's scaling constants: +var tsml = 1.08420217E-19; +var tbig = 4.50359963E+15; +var ssml = 3.77789319E+22; +var sbig = 1.32348898E-23; // MAIN // @@ -45,37 +56,79 @@ var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); * // returns 5.0 */ function snrm2( N, x, stride, offset ) { - var scale; - var ssq; + var notbig; + var sumsq; + var abig; + var amed; + var asml; + var ymax; + var ymin; + var scl; var ax; var ix; - var v; var i; if ( N <= 0 ) { return 0.0; } - if ( N === 1 ) { - return absf( x[ offset ] ); - } ix = offset; - scale = 0.0; - ssq = 1.0; + + // Initialize loop values for accumulation: + notbig = true; + + sumsq = 0.0; + abig = 0.0; + amed = 0.0; + asml = 0.0; + scl = 1.0; + + // Compute the sum of squares using 3 accumulators--`abig` (sum of squares scaled down to avoid overflow), `asml` (sum of squares scaled up to avoid underflow), `amed` (sum of squares that do not require scaling)--and thresholds and multipliers--`tbig` (values bigger than this are scaled down by `sbig`) and `tsml` (values smaller than this are scaled up by `ssml`)... for ( i = 0; i < N; i++ ) { - if ( x[ ix ] !== 0.0 ) { - ax = absf( x[ ix ] ); - if ( scale < ax ) { - v = float64ToFloat32( scale/ax ); - ssq = float64ToFloat32( 1.0 + float64ToFloat32( ssq * float64ToFloat32( v*v ) ) ); // eslint-disable-line max-len - scale = ax; - } else { - v = float64ToFloat32( ax/scale ); - ssq = float64ToFloat32( ssq + float64ToFloat32( v*v ) ); + ax = absf( x[ ix ] ); + if ( ax > tbig ) { + abig = f32( abig + abs2f( ax * sbig ) ); + notbig = false; + } else if ( ax < tsml ) { + if ( notbig ) { + asml = f32( asml + abs2f( ax * ssml ) ); } + } else { + amed = f32( amed + f32( ax * ax ) ); } ix += stride; } - return float64ToFloat32( scale * sqrtf( ssq ) ); + // Combine `abig` and `amed` or `amed` and `asml` if more than one accumulator was used... + if ( abig > 0.0 ) { + // Combine `abig` and `amed` if `abig` > 0... + if ( amed > 0.0 || ( amed > FLOAT32_MAX ) || ( amed !== amed ) ) { + abig = f32( abig + f32( f32( amed * sbig ) * sbig ) ); + } + scl = f32( 1.0 / sbig ); + sumsq = abig; + } else if ( asml > 0.0 ) { + // Combine `amed` and `asml` if `asml` > 0... + if ( amed > 0.0 || amed > FLOAT32_MAX || ( amed !== amed ) ) { + amed = sqrtf( amed ); + asml = f32( sqrtf( asml ) / ssml ); + if ( asml > amed ) { + ymin = amed; + ymax = asml; + } else { + ymin = asml; + ymax = amed; + } + scl = 1.0; + sumsq = f32( f32( ymax * ymax ) * f32( 1.0 + abs2f( ymin / ymax ) ) ); // eslint-disable-line max-len + } else { + scl = f32( 1.0 / ssml ); + sumsq = asml; + } + } else { + // All values are mid-range... + scl = 1.0; + sumsq = amed; + } + return f32( sqrtf( sumsq ) * scl ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.native.js b/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.native.js index 63f8461a2118..c4ff92365b82 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/lib/ndarray.native.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/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( './snrm2.native.js' ); +var addon = require( './../src/addon.node' ); // MAIN // @@ -45,13 +43,7 @@ var addon = require( './snrm2.native.js' ); * // returns 5.0 */ function snrm2( N, x, stride, offset ) { - var view; - offset = minViewBufferIndex( N, stride, offset ); - if ( stride < 0 ) { - stride *= -1; - } - view = offsetView( x, offset ); - return addon( N, view, stride ); + return addon.ndarray( N, x, stride, offset ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.js b/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.js index 7d6897f536af..d7b41c20f329 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.js @@ -20,9 +20,8 @@ // MODULES // -var sqrtf = require( '@stdlib/math/base/special/sqrtf' ); -var absf = require( '@stdlib/math/base/special/absf' ); -var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); +var stride2offset = require( '@stdlib/strided/base/stride2offset' ); +var ndarray = require( './ndarray.js' ); // MAIN // @@ -32,7 +31,7 @@ var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {PositiveInteger} stride - stride length +* @param {integer} stride - stride length * @returns {number} L2-norm * * @example @@ -44,35 +43,8 @@ var float64ToFloat32 = require( '@stdlib/number/float64/base/to-float32' ); * // returns 3.0 */ function snrm2( N, x, stride ) { - var scale; - var ssq; - var ax; - var v; - var i; - - if ( N <= 0 || stride <= 0 ) { - return 0.0; - } - if ( N === 1 ) { - return absf( x[ 0 ] ); - } - scale = 0.0; - ssq = 1.0; - N *= stride; - for ( i = 0; i < N; i += stride ) { - if ( x[ i ] !== 0.0 ) { - ax = absf( x[ i ] ); - if ( scale < ax ) { - v = float64ToFloat32( scale/ax ); - ssq = float64ToFloat32( 1.0 + float64ToFloat32( ssq * float64ToFloat32( v*v ) ) ); // eslint-disable-line max-len - scale = ax; - } else { - v = float64ToFloat32( ax/scale ); - ssq = float64ToFloat32( ssq + float64ToFloat32( v*v ) ); - } - } - } - return float64ToFloat32( scale * sqrtf( ssq ) ); + var ox = stride2offset( N, stride ); + return ndarray( N, x, stride, ox ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.native.js b/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.native.js index c9e5e1d32273..fc11fd8d0348 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.native.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/lib/snrm2.native.js @@ -30,7 +30,7 @@ var addon = require( './../src/addon.node' ); * * @param {PositiveInteger} N - number of indexed elements * @param {Float32Array} x - input array -* @param {PositiveInteger} stride - stride length +* @param {integer} stride - stride length * @returns {number} L2-norm * * @example diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/manifest.json b/lib/node_modules/@stdlib/blas/base/snrm2/manifest.json index f3f8fd9e4706..69d2d10b0745 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/manifest.json +++ b/lib/node_modules/@stdlib/blas/base/snrm2/manifest.json @@ -45,6 +45,8 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -57,7 +59,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -66,7 +69,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { "task": "examples", @@ -74,7 +84,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -83,7 +94,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { @@ -104,6 +122,8 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -126,7 +146,10 @@ "-lpthread" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { "task": "examples", @@ -144,7 +167,10 @@ "-lpthread" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { @@ -164,6 +190,8 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -176,7 +204,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -185,7 +214,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { "task": "examples", @@ -193,7 +229,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -202,7 +239,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { @@ -222,6 +266,8 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -243,7 +289,10 @@ "-lblas" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { "task": "examples", @@ -260,7 +309,10 @@ "-lblas" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { @@ -281,6 +333,8 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -303,7 +357,10 @@ "-lpthread" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { "task": "examples", @@ -321,7 +378,10 @@ "-lpthread" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/strided/base/min-view-buffer-index" + ] }, { @@ -330,7 +390,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -341,6 +402,12 @@ "libpath": [], "dependencies": [ "@stdlib/napi/export", + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset", "@stdlib/napi/argv", "@stdlib/napi/argv-int64", "@stdlib/napi/argv-strided-float32array", @@ -353,7 +420,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -362,7 +430,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { "task": "examples", @@ -370,7 +445,8 @@ "blas": "", "wasm": false, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -379,7 +455,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] }, { @@ -388,7 +471,8 @@ "blas": "", "wasm": true, "src": [ - "./src/snrm2.c" + "./src/snrm2.c", + "./src/snrm2_ndarray.c" ], "include": [ "./include" @@ -397,7 +481,14 @@ "-lm" ], "libpath": [], - "dependencies": [] + "dependencies": [ + "@stdlib/blas/base/shared", + "@stdlib/math/base/special/absf", + "@stdlib/math/base/special/abs2f", + "@stdlib/math/base/special/sqrtf", + "@stdlib/constants/float32/max", + "@stdlib/strided/base/stride2offset" + ] } ] } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/addon.c b/lib/node_modules/@stdlib/blas/base/snrm2/src/addon.c index 2d17d9c0df2c..8c4cd0254079 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/src/addon.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/addon.c @@ -17,6 +17,7 @@ */ #include "stdlib/blas/base/snrm2.h" +#include "stdlib/blas/base/shared.h" #include "stdlib/napi/export.h" #include "stdlib/napi/argv.h" #include "stdlib/napi/argv_int64.h" @@ -34,10 +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, strideX, argv, 2 ); - STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, strideX, argv, 1 ); - STDLIB_NAPI_CREATE_DOUBLE( env, (double)c_snrm2( N, X, strideX ), v ); + STDLIB_NAPI_ARGV_INT64( env, stride, argv, 2 ); + STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, stride, argv, 1 ); + STDLIB_NAPI_CREATE_DOUBLE( env, (double)API_SUFFIX(c_snrm2)( N, X, stride ), v ); return v; } -STDLIB_NAPI_MODULE_EXPORT_FCN( addon ) +/** +* 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, stride, argv, 2 ); + STDLIB_NAPI_ARGV_INT64( env, offset, argv, 3 ); + STDLIB_NAPI_ARGV_STRIDED_FLOAT32ARRAY( env, X, N, stride, argv, 1 ); + STDLIB_NAPI_CREATE_DOUBLE( env, (double)API_SUFFIX(c_snrm2_ndarray)( N, X, stride, offset ), v ); + return v; +} + +STDLIB_NAPI_MODULE_EXPORT_FCN_WITH_METHOD( addon, "ndarray", addon_method ) diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.c b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.c index e495f6e5ecda..087947a805fa 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.c @@ -17,7 +17,8 @@ */ #include "stdlib/blas/base/snrm2.h" -#include +#include "stdlib/blas/base/shared.h" +#include "stdlib/strided/base/stride2offset.h" /** * Computes the L2-norm of a single-precision floating-point vector. @@ -27,30 +28,7 @@ * @param stride stride length * @return output value */ -float c_snrm2( const int N, const float *X, const int stride ) { - float scale; - float ssq; - float ax; - int i; - - if ( N <= 0 || stride <= 0 ) { - return 0.0f; - } - if ( N == 1 ) { - return fabsf( X[ 0 ] ); - } - scale = 0.0f; - ssq = 1.0f; - for ( i = 0; i < N*stride; i += stride ) { - if ( X[ i ] != 0.0f ) { - ax = fabsf( X[ i ] ); - if ( scale < ax ) { - ssq = 1.0f + ( ssq * powf( scale/ax, 2 ) ); - scale = ax; - } else { - ssq += powf( ax/scale, 2 ); - } - } - } - return scale * sqrtf( ssq ); +float API_SUFFIX(c_snrm2)( const CBLAS_INT N, const float *X, const CBLAS_INT stride ) { + CBLAS_INT ox = stdlib_strided_stride2offset( N, stride ); + return API_SUFFIX(c_snrm2_ndarray)( N, X, stride, ox ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.f b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.f index 9ac87657dd03..acfd421411f9 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.f +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2.f @@ -52,6 +52,20 @@ !< real function snrm2( N, sx, stride ) implicit none + ! .. + ! Define a kind parameter for single precision: + integer, parameter :: wp = kind( 1.0e0 ) + ! .. + ! Define constants: + real( wp ), parameter :: zero = 0.0_wp + real( wp ), parameter :: one = 1.0_wp + real( wp ), parameter :: maxN = huge( 0.0_wp ) + ! .. + ! Blue's scaling constants: + real( wp ), parameter :: tsml = 1.08420217e-19_wp + real( wp ), parameter :: tbig = 4.50359963e15_wp + real( wp ), parameter :: ssml = 3.77789319e22_wp + real( wp ), parameter :: sbig = 1.32348898e-23_wp !.. ! Scalar arguments: integer :: N, stride @@ -60,35 +74,85 @@ real function snrm2( N, sx, stride ) real, intent(in) :: sx(*) ! .. ! Local scalars: - real :: ax, scale, ssq - integer :: i + integer :: i, ix + logical :: notbig + real( wp ) :: abig, amed, asml, ax, scl, sumsq, ymax, ymin ! .. ! Intrinsic functions: intrinsic abs, sqrt ! .. - snrm2 = 0.0e0 + snrm2 = zero ! .. - if ( N <= 0 .OR. stride <= 0 ) then + if ( N <= 0 ) then return end if !.. - if ( N == 1 ) then - snrm2 = abs( sx( 1 ) ) - return + scl = one + sumsq = zero + ! .. + ! Compute the sum of squares in 3 accumulators: + ! abig -- sums of squares scaled down to avoid overflow + ! asml -- sums of squares scaled up to avoid underflow + ! amed -- sums of squares that do not require scaling + ! Thresholds and multipliers: + ! tbig -- values bigger than this are scaled down by sbig + ! tsml -- values smaller than this are scaled up by ssml + ! .. + notbig = .true. + asml = zero + amed = zero + abig = zero + ix = 1 + if ( stride < 0 ) then + ix = 1 - ( N - 1 ) * stride end if ! .. - scale = 0.0e0 - ssq = 1.0e0 - do i = 1, 1+((N-1)*stride), stride - if ( sx( i ) /= 0.0e0 ) then - ax = abs( sx( i ) ) - if ( scale < ax ) then - ssq = 1.0e0 + ( ssq * (scale/ax)**2 ) - scale = ax - else - ssq = ssq + (ax/scale)**2 + do i = 1, N + ax = abs( sx( ix ) ) + if ( ax > tbig ) then + abig = abig + ( ax * sbig )**2 + notbig = .false. + else if ( ax < tsml ) then + if ( notbig ) then + asml = asml + ( ax * ssml )**2 end if + else + amed = amed + ax**2 end if + ix = ix + stride end do - snrm2 = scale * sqrt( ssq ) + ! .. + ! Combine abig and amed or amed and asml if more than one accumulator was used: + if ( abig > zero ) then + ! Combine abig and amed if abig > 0... + if ( ( amed > zero ) .or. ( amed > maxN ) .or. ( amed /= amed ) ) then + abig = abig + ( amed * sbig ) * sbig + end if + scl = one / sbig + sumsq = abig + else if ( asml > zero ) then + ! Combine amed and asml if asml > 0... + if ( ( amed > zero ) .or. ( amed > maxN ) .or. ( amed /= amed ) ) then + amed = sqrt( amed ) + asml = sqrt( asml ) / ssml + if ( asml > amed ) then + ymin = amed + ymax = asml + else + ymin = asml + ymax = amed + end if + scl = one + sumsq = ymax**2 * ( one + ( ymin / ymax )**2 ) + else + scl = one / ssml + sumsq = asml + end if + else + ! All values are mid-range... + scl = one + sumsq = amed + end if + snrm2 = scl * sqrt( sumsq ) + return end function snrm2 diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_cblas.c b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_cblas.c index 0d366ef4d066..3ac85ff844bc 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_cblas.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_cblas.c @@ -18,6 +18,8 @@ #include "stdlib/blas/base/snrm2.h" #include "stdlib/blas/base/snrm2_cblas.h" +#include "stdlib/blas/base/shared.h" +#include "stdlib/strided/base/min_view_buffer_index.h" /** * Computes the L2-norm of a single-precision floating-point vector. @@ -27,6 +29,28 @@ * @param stride stride length * @return output value */ -float c_snrm2( const int N, const float *X, const int stride ) { - return cblas_snrm2( N, X, stride ); +float API_SUFFIX(c_snrm2)( const CBLAS_INT N, const float *X, const CBLAS_INT stride ) { + CBLAS_INT sx = stride; + if ( sx < 0 ) { + sx = -sx; + } + return API_SUFFIX(cblas_snrm2)( N, X, sx ); +} + +/** +* Computes the L2-norm of a single-precision floating-point vector using alternative indexing semantics. +* +* @param N number of indexed elements +* @param X input array +* @param stride stride length +* @param offset starting index +* @return output value +*/ +float API_SUFFIX(c_snrm2_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT stride, const CBLAS_INT offset ) { + CBLAS_INT sx = stride; + if ( sx < 0 ) { + sx = -sx; + } + X += stdlib_strided_min_view_buffer_index( N, stride, offset ); // adjust array pointer + return API_SUFFIX(cblas_snrm2)( N, X, sx ); } diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_f.c b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_f.c index 197df29c15eb..d2231ad9f517 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_f.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_f.c @@ -18,6 +18,8 @@ #include "stdlib/blas/base/snrm2.h" #include "stdlib/blas/base/snrm2_fortran.h" +#include "stdlib/blas/base/shared.h" +#include "stdlib/strided/base/min_view_buffer_index.h" /** * Computes the L2-norm of a single-precision floating-point vector. @@ -27,8 +29,24 @@ * @param stride stride length * @return output value */ -float c_snrm2( const int N, const float *X, const int stride ) { +float API_SUFFIX(c_snrm2)( const CBLAS_INT N, const float *X, const CBLAS_INT stride ) { float nrm2; snrm2sub( &N, X, &stride, &nrm2 ); return nrm2; } + +/** +* Computes the L2-norm of a single-precision floating-point vector using alternative indexing semantics. +* +* @param N number of indexed elements +* @param X input array +* @param stride stride length +* @param offset starting index +* @return output value +*/ +float API_SUFFIX(c_snrm2_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT stride, const CBLAS_INT offset ) { + float nrm2; + X += stdlib_strided_min_view_buffer_index( N, stride, offset ); // adjust array pointer + snrm2sub( &N, X, &stride, &nrm2 ); + return nrm2; +} diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_ndarray.c b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_ndarray.c new file mode 100644 index 000000000000..f4e3e8b31b70 --- /dev/null +++ b/lib/node_modules/@stdlib/blas/base/snrm2/src/snrm2_ndarray.c @@ -0,0 +1,112 @@ +/** +* @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/base/snrm2.h" +#include "stdlib/blas/base/shared.h" +#include "stdlib/math/base/special/absf.h" +#include "stdlib/math/base/special/abs2f.h" +#include "stdlib/math/base/special/sqrtf.h" +#include "stdlib/constants/float32/max.h" +#include + +// Blue's scaling constants... +static const float tsml = 1.08420217E-19f; +static const float tbig = 4.50359963E+15f; +static const float ssml = 3.77789319E+22f; +static const float sbig = 1.32348898E-23f; + +/** +* Computes the L2-norm of a single-precision floating-point vector using alternative indexing semantics. +* +* @param N number of indexed elements +* @param X input array +* @param stride stride length +* @param offset starting index +* @return output value +*/ +float API_SUFFIX(c_snrm2_ndarray)( const CBLAS_INT N, const float *X, const CBLAS_INT stride, const CBLAS_INT offset ) { + CBLAS_INT ix; + bool notbig; + CBLAS_INT i; + float sumsq; + float abig; + float amed; + float asml; + float ymax; + float ymin; + float scl; + float ax; + + if ( N <= 0 ) { + return 0.0f; + } + ix = offset; + + // Compute the sum of squares using 3 accumulators--`abig` (sum of squares scaled down to avoid overflow), `asml` (sum of squares scaled up to avoid underflow), `amed` (sum of squares that do not require scaling)--and thresholds and multipliers--`tbig` (values bigger than this are scaled down by `sbig`) and `tsml` (values smaller than this are scaled up by `ssml`)... + notbig = true; + sumsq = 0.0f; + abig = 0.0f; + amed = 0.0f; + asml = 0.0f; + for ( i = 0; i < N; i++ ) { + ax = stdlib_base_absf( X[ ix ] ); + if ( ax > tbig ) { + abig += stdlib_base_abs2f( ax * sbig ); + notbig = false; + } else if ( ax < tsml ) { + if ( notbig ) { + asml += stdlib_base_abs2f( ax * ssml ); + } + } else { + amed += stdlib_base_abs2f( ax ); + } + ix += stride; + } + // Combine `abig` and `amed` or `amed` and `asml` if more than one accumulator was used... + if ( abig > 0.0f ) { + // Combine `abig` and `amed` if `abig` > 0... + if ( amed > 0.0f || ( amed > STDLIB_CONSTANT_FLOAT32_MAX ) || ( amed != amed ) ) { + abig += ( amed * sbig ) * sbig; + } + scl = 1.0f / sbig; + sumsq = abig; + } else if ( asml > 0.0f ) { + // Combine `amed` and `asml` if `asml` > 0... + if ( amed > 0.0f || amed > STDLIB_CONSTANT_FLOAT32_MAX || ( amed != amed ) ) { + amed = stdlib_base_sqrtf( amed ); + asml = stdlib_base_sqrtf( asml ) / ssml; + if ( asml > amed ) { + ymin = amed; + ymax = asml; + } else { + ymin = asml; + ymax = amed; + } + scl = 1.0f; + sumsq = stdlib_base_abs2f( ymax ) * ( 1.0 + stdlib_base_abs2f( ymin / ymax ) ); + } else { + scl = 1.0f / ssml; + sumsq = asml; + } + } else { + // All values are mid-range... + scl = 1.0f; + sumsq = amed; + } + return stdlib_base_sqrtf( sumsq ) * scl; +} diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.js b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.js index 2b1767680f24..a377dd57f1b6 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.js @@ -22,9 +22,40 @@ var tape = require( 'tape' ); var Float32Array = require( '@stdlib/array/float32' ); +var EPS = require( '@stdlib/constants/float32/eps' ); +var abs = require( '@stdlib/math/base/special/abs' ); var snrm2 = require( './../lib/ndarray.js' ); +// FUNCTIONS // + +/** +* Tests for element-wise approximate equality. +* +* @private +* @param {Object} t - test object +* @param {Collection} actual - actual values +* @param {Collection} expected - expected values +* @param {number} rtol - relative tolerance +*/ +function isApprox( t, actual, expected, rtol ) { + var delta; + var tol; + var i; + + t.strictEqual( actual.length, expected.length, 'returns expected value' ); + for ( i = 0; i < expected.length; i++ ) { + if ( actual[ i ] === expected[ i ] ) { + t.strictEqual( actual[ i ], expected[ i ], 'returns expected value' ); + } else { + delta = abs( actual[ i ] - expected[ i ] ); + tol = rtol * EPS * abs( expected[ i ] ); + t.ok( delta <= tol, 'within tolerance. actual: '+actual[ i ]+'. expected: '+expected[ i ]+'. delta: '+delta+'. tol: '+tol+'.' ); + } + } +} + + // TESTS // tape( 'main export is a function', function test( t ) { @@ -45,12 +76,12 @@ tape( 'the function calculates the L2-norm of a vector', function test( t ) { x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 0.0, 3.0 ] ); z = snrm2( x.length, x, 1, 0 ); - t.strictEqual( z, 7.41619873046875, 'returns expected value' ); + isApprox( t, z, 7.4, 2.0 ); x = new Float32Array( [ -4.0 ] ); z = snrm2( x.length, x, 1, 0 ); - t.strictEqual( z, 4.0, 'returns expected value' ); + isApprox( t, z, 4.0, 1.0 ); t.end(); }); @@ -89,7 +120,7 @@ tape( 'the function supports a `stride` parameter', function test( t ) { N = 4; z = snrm2( N, x, 2, 0 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); @@ -99,20 +130,20 @@ tape( 'the function supports a negative `stride` parameter', function test( t ) var z; x = new Float32Array([ - 1.0, // 0 + 1.0, // 3 2.0, - 2.0, // 1 + 2.0, // 2 -7.0, - -2.0, // 2 + -2.0, // 1 3.0, - 4.0, // 3 + 4.0, // 0 2.0 ]); N = 4; z = snrm2( N, x, -2, x.length-2 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); @@ -134,7 +165,7 @@ tape( 'the function supports an `offset` parameter', function test( t ) { N = 4; z = snrm2( N, x, 2, 1 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.native.js b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.native.js index ae5b8106efdd..b21dda7b9459 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.native.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.ndarray.native.js @@ -23,6 +23,8 @@ var resolve = require( 'path' ).resolve; var tape = require( 'tape' ); var Float32Array = require( '@stdlib/array/float32' ); +var EPS = require( '@stdlib/constants/float32/eps' ); +var abs = require( '@stdlib/math/base/special/abs' ); var tryRequire = require( '@stdlib/utils/try-require' ); @@ -34,6 +36,35 @@ var opts = { }; +// FUNCTIONS // + +/** +* Tests for element-wise approximate equality. +* +* @private +* @param {Object} t - test object +* @param {Collection} actual - actual values +* @param {Collection} expected - expected values +* @param {number} rtol - relative tolerance +*/ +function isApprox( t, actual, expected, rtol ) { + var delta; + var tol; + var i; + + t.strictEqual( actual.length, expected.length, 'returns expected value' ); + for ( i = 0; i < expected.length; i++ ) { + if ( actual[ i ] === expected[ i ] ) { + t.strictEqual( actual[ i ], expected[ i ], 'returns expected value' ); + } else { + delta = abs( actual[ i ] - expected[ i ] ); + tol = rtol * EPS * abs( expected[ i ] ); + t.ok( delta <= tol, 'within tolerance. actual: '+actual[ i ]+'. expected: '+expected[ i ]+'. delta: '+delta+'. tol: '+tol+'.' ); + } + } +} + + // TESTS // tape( 'main export is a function', opts, function test( t ) { @@ -54,12 +85,12 @@ tape( 'the function calculates the L2-norm of a vector', opts, function test( t x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 0.0, 3.0 ] ); z = snrm2( x.length, x, 1, 0 ); - t.strictEqual( z, 7.41619873046875, 'returns expected value' ); + isApprox( t, z, 7.4, 2.0 ); x = new Float32Array( [ -4.0 ] ); z = snrm2( x.length, x, 1, 0 ); - t.strictEqual( z, 4.0, 'returns expected value' ); + isApprox( t, z, 4.0, 1.0 ); t.end(); }); @@ -98,7 +129,7 @@ tape( 'the function supports a `stride` parameter', opts, function test( t ) { N = 4; z = snrm2( N, x, 2, 0 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); @@ -108,20 +139,20 @@ tape( 'the function supports a negative `stride` parameter', opts, function test var z; x = new Float32Array([ - 1.0, // 0 + 1.0, // 3 2.0, - 2.0, // 1 + 2.0, // 2 -7.0, - -2.0, // 2 + -2.0, // 1 3.0, - 4.0, // 3 + 4.0, // 0 2.0 ]); N = 4; z = snrm2( N, x, -2, x.length-2 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); @@ -143,7 +174,7 @@ tape( 'the function supports an `offset` parameter', opts, function test( t ) { N = 4; z = snrm2( N, x, 2, 1 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.js b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.js index 082648b5d10e..5cc4cd1ceab4 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.js @@ -22,9 +22,40 @@ var tape = require( 'tape' ); var Float32Array = require( '@stdlib/array/float32' ); +var EPS = require( '@stdlib/constants/float32/eps' ); +var abs = require( '@stdlib/math/base/special/abs' ); var snrm2 = require( './../lib/snrm2.js' ); +// FUNCTIONS // + +/** +* Tests for element-wise approximate equality. +* +* @private +* @param {Object} t - test object +* @param {Collection} actual - actual values +* @param {Collection} expected - expected values +* @param {number} rtol - relative tolerance +*/ +function isApprox( t, actual, expected, rtol ) { + var delta; + var tol; + var i; + + t.strictEqual( actual.length, expected.length, 'returns expected value' ); + for ( i = 0; i < expected.length; i++ ) { + if ( actual[ i ] === expected[ i ] ) { + t.strictEqual( actual[ i ], expected[ i ], 'returns expected value' ); + } else { + delta = abs( actual[ i ] - expected[ i ] ); + tol = rtol * EPS * abs( expected[ i ] ); + t.ok( delta <= tol, 'within tolerance. actual: '+actual[ i ]+'. expected: '+expected[ i ]+'. delta: '+delta+'. tol: '+tol+'.' ); + } + } +} + + // TESTS // tape( 'main export is a function', function test( t ) { @@ -42,15 +73,15 @@ tape( 'the function calculates the L2-norm of a vector', function test( t ) { var x; var z; - x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 0.0, 3.0 ] ); + x = new Float32Array( [ 2.0, -4.0, -5.0, 6.0, 0.0, 5.0 ] ); z = snrm2( x.length, x, 1 ); - t.strictEqual( z, 7.41619873046875, 'returns expected value' ); + isApprox( t, z, 10.3, 2.0 ); x = new Float32Array( [ -4.0 ] ); z = snrm2( x.length, x, 1 ); - t.strictEqual( z, 4.0, 'returns expected value' ); + isApprox( t, z, 4.0, 1.0 ); t.end(); }); @@ -89,22 +120,29 @@ tape( 'the function supports a `stride` parameter', function test( t ) { N = 4; z = snrm2( N, x, 2 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); -tape( 'if provided a `stride` parameter less than or equal to `0`, the function returns `0`', function test( t ) { +tape( 'the function supports specifying a negative stride', function test( t ) { + var expected; + var actual; var x; - var z; - x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); - - z = snrm2( x.length, x, 0 ); - t.strictEqual( z, 0.0, 'returns expected value' ); - - z = snrm2( x.length, x, -1 ); - t.strictEqual( z, 0.0, 'returns expected value' ); + x = new Float32Array([ + 0.3, // 3 + 5.0, + 0.5, // 2 + 6.0, + 0.4, // 1 + 8.0, + 0.4, // 0 + 9.0 + ]); + expected = 0.81; + actual = snrm2( 4, x, -2 ); + isApprox( t, actual, expected, 2.0 ); t.end(); }); diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.native.js b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.native.js index 0d31d7321b07..fa9bc7aa06d4 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.native.js +++ b/lib/node_modules/@stdlib/blas/base/snrm2/test/test.snrm2.native.js @@ -23,6 +23,8 @@ var resolve = require( 'path' ).resolve; var tape = require( 'tape' ); var Float32Array = require( '@stdlib/array/float32' ); +var EPS = require( '@stdlib/constants/float32/eps' ); +var abs = require( '@stdlib/math/base/special/abs' ); var tryRequire = require( '@stdlib/utils/try-require' ); @@ -34,6 +36,35 @@ var opts = { }; +// FUNCTIONS // + +/** +* Tests for element-wise approximate equality. +* +* @private +* @param {Object} t - test object +* @param {Collection} actual - actual values +* @param {Collection} expected - expected values +* @param {number} rtol - relative tolerance +*/ +function isApprox( t, actual, expected, rtol ) { + var delta; + var tol; + var i; + + t.strictEqual( actual.length, expected.length, 'returns expected value' ); + for ( i = 0; i < expected.length; i++ ) { + if ( actual[ i ] === expected[ i ] ) { + t.strictEqual( actual[ i ], expected[ i ], 'returns expected value' ); + } else { + delta = abs( actual[ i ] - expected[ i ] ); + tol = rtol * EPS * abs( expected[ i ] ); + t.ok( delta <= tol, 'within tolerance. actual: '+actual[ i ]+'. expected: '+expected[ i ]+'. delta: '+delta+'. tol: '+tol+'.' ); + } + } +} + + // TESTS // tape( 'main export is a function', opts, function test( t ) { @@ -54,12 +85,12 @@ tape( 'the function calculates the L2-norm of a vector', opts, function test( t x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 0.0, 3.0 ] ); z = snrm2( x.length, x, 1 ); - t.strictEqual( z, 7.41619873046875, 'returns expected value' ); + isApprox( t, z, 7.4, 2.0 ); x = new Float32Array( [ -4.0 ] ); z = snrm2( x.length, x, 1 ); - t.strictEqual( z, 4.0, 'returns expected value' ); + isApprox( t, z, 4.0, 1.0 ); t.end(); }); @@ -98,22 +129,29 @@ tape( 'the function supports a `stride` parameter', opts, function test( t ) { N = 4; z = snrm2( N, x, 2 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); -tape( 'if provided a `stride` parameter less than or equal to `0`, the function returns `0`', opts, function test( t ) { +tape( 'the function supports specifying a negative stride', opts, function test( t ) { + var expected; + var actual; var x; - var z; - x = new Float32Array( [ 1.0, -2.0, -4.0, 5.0, 3.0 ] ); - - z = snrm2( x.length, x, 0 ); - t.strictEqual( z, 0.0, 'returns expected value' ); - - z = snrm2( x.length, x, -1 ); - t.strictEqual( z, 0.0, 'returns expected value' ); + x = new Float32Array([ + 0.3, // 3 + 5.0, + 0.5, // 2 + 6.0, + 0.4, // 1 + 8.0, + 0.4, // 0 + 9.0 + ]); + expected = 0.81; + actual = snrm2( 4, x, -2 ); + isApprox( t, actual, expected, 2.0 ); t.end(); }); @@ -139,7 +177,7 @@ tape( 'the function supports view offsets', opts, function test( t ) { N = 4; z = snrm2( N, x1, 2 ); - t.strictEqual( z, 5.0, 'returns expected value' ); + isApprox( t, z, 5.0, 1.0 ); t.end(); }); From 3781bcda8b8c9006fac5b246ac2a2f752fd60191 Mon Sep 17 00:00:00 2001 From: aman-095 Date: Sat, 21 Sep 2024 09:10:31 +0530 Subject: [PATCH 3/3] docs: update descriptions in README --- .../@stdlib/blas/base/snrm2/README.md | 16 ++++++++-------- .../@stdlib/blas/base/snrm2/examples/c/example.c | 2 +- 2 files changed, 9 insertions(+), 9 deletions(-) diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/README.md b/lib/node_modules/@stdlib/blas/base/snrm2/README.md index 7e545c23b273..5fd7646b133f 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/README.md +++ b/lib/node_modules/@stdlib/blas/base/snrm2/README.md @@ -91,7 +91,7 @@ var z = snrm2( 4, x1, 2 ); // returns 5.0 ``` -If either `N` or `stride` is less than or equal to `0`, the function returns `0`. +If `N` is less than or equal to `0`, the function returns `0`. #### snrm2.ndarray( N, x, stride, offset ) @@ -191,10 +191,10 @@ console.log( out ); Computes the L2-norm of a complex single-precision floating-point vector. ```c -const float x[] = { 1.0f, -2.0f, 3.0f, -4.0f, 5.0f, -6.0f, 7.0f, -8.0f }; +const float x[] = { 1.0f, 2.0f, 2.0f, -7.0f, -2.0f, 3.0f, 4.0f, 2.0f }; -float norm = c_snrm2( 8, x, 1 ); -// returns 14.3 +float norm = c_snrm2( 4, x, 2 ); +// returns 5.0f ``` The function accepts the following arguments: @@ -212,10 +212,10 @@ float c_snrm2( const CBLAS_INT N, const float *X, const CBLAS_INT stride ); Computes the L2-norm of a complex single-precision floating-point vector using alternative indexing semantics. ```c -const float x[] = { 1.0f, -2.0f, 3.0f, -4.0f, 5.0f, -6.0f, 7.0f, -8.0f }; +const float x[] = { 1.0f, 2.0f, 2.0f, -7.0f, -2.0f, 3.0f, 4.0f, 2.0f }; -float norm = c_snrm2_ndarray( 8, x, 1, 0 ); -// returns 14.3 +float norm = c_snrm2_ndarray( 4, x, 2, 0 ); +// returns 5.0f ``` The function accepts the following arguments: @@ -268,7 +268,7 @@ int main( void ) { printf( "L2-norm: %f\n", l2 ); // Compute the L2-norm: - l2 = c_snrm2_ndarray( N, x, strideX, 0 ); + l2 = c_snrm2_ndarray( N, x, -strideX, 7 ); // Print the result: printf( "L2-norm: %f\n", l2 ); diff --git a/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c b/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c index da3b0a8cb0f1..d1a09f22b11b 100644 --- a/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c +++ b/lib/node_modules/@stdlib/blas/base/snrm2/examples/c/example.c @@ -36,7 +36,7 @@ int main( void ) { printf( "L2-norm: %f\n", l2 ); // Compute the L2-norm: - l2 = c_snrm2_ndarray( N, x, strideX, 0 ); + l2 = c_snrm2_ndarray( N, x, -strideX, 7 ); // Print the result: printf( "L2-norm: %f\n", l2 );