Press n or j to go to the next uncovered block, b, p or k for the previous block.
| 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 | 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x 1x | /**
* @license Apache-2.0
*
* Copyright (c) 2025 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.
*/
/* eslint-disable max-depth, max-len, max-statements */
'use strict';
// MODULES //
var loopOrder = require( '@stdlib/ndarray/base/binary-loop-interchange-order' );
var blockSize = require( '@stdlib/ndarray/base/binary-tiling-block-size' );
var takeIndexed = require( '@stdlib/array/base/take-indexed' );
var copyIndexed = require( '@stdlib/array/base/copy-indexed' );
var zeros = require( '@stdlib/array/base/zeros' );
var incrementOffsets = require( './increment_offsets.js' );
var setViewOffsets = require( './set_view_offsets.js' );
var offsets = require( './offsets.js' );
// MAIN //
/**
* Performs a reduction over a list of specified dimensions in two input ndarrays and assigns results to a provided output ndarray via loop blocking.
*
* @private
* @param {Function} fcn - wrapper for a one-dimensional strided array reduction function
* @param {Array<Object>} arrays - ndarrays
* @param {Array<Object>} views - initialized ndarray-like objects representing sub-array views
* @param {IntegerArray} strideX - loop dimension strides for the first input ndarray
* @param {IntegerArray} strideY - loop dimension strides for the second input ndarray
* @param {Function} strategy1 - first input ndarray reshape strategy
* @param {Function} strategy2 - second input ndarray reshape strategy
* @param {Options} opts - function options
* @returns {void}
*
* @example
* var toAccessorArray = require( '@stdlib/array/base/to-accessor-array' );
* var accessors = require( '@stdlib/array/base/accessors' );
* var Float64Array = require( '@stdlib/array/float64' );
* var ndarray2array = require( '@stdlib/ndarray/base/to-array' );
* var gdot = require( '@stdlib/blas/base/ndarray/gdot' );
*
* // Create data buffers:
* var xbuf = toAccessorArray( new Float64Array( [ 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0 ] ) );
* var ybuf = toAccessorArray( new Float64Array( [ 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0 ] ) );
* var zbuf = toAccessorArray( new Float64Array( [ 0.0, 0.0, 0.0 ] ) );
*
* // Define the array shapes:
* var xsh = [ 1, 1, 1, 1, 1, 3, 2, 2 ];
* var ysh = [ 1, 1, 1, 1, 1, 3, 2, 2 ];
* var zsh = [ 1, 1, 1, 1, 1, 3 ];
*
* // Define the array strides:
* var sx = [ 12, 12, 12, 12, 12, 4, 2, 1 ];
* var sy = [ 12, 12, 12, 12, 12, 4, 2, 1 ];
* var sz = [ 3, 3, 3, 3, 3, 1 ];
*
* // Define the index offsets:
* var ox = 0;
* var oy = 0;
* var oz = 0;
*
* // Create input ndarray-like objects:
* var x = {
* 'dtype': 'float64',
* 'data': xbuf,
* 'shape': xsh,
* 'strides': sx,
* 'offset': ox,
* 'order': 'row-major',
* 'accessors': accessors( xbuf ).accessors
* };
* var y = {
* 'dtype': 'float64',
* 'data': ybuf,
* 'shape': ysh,
* 'strides': sy,
* 'offset': oy,
* 'order': 'row-major',
* 'accessors': accessors( ybuf ).accessors
* };
*
* // Create an output ndarray-like object:
* var z = {
* 'dtype': 'float64',
* 'data': zbuf,
* 'shape': zsh,
* 'strides': sz,
* 'offset': oz,
* 'order': 'row-major',
* 'accessors': accessors( zbuf ).accessors
* };
*
* // Initialize ndarray-like objects representing sub-array views:
* var views = [
* {
* 'dtype': x.dtype,
* 'data': x.data,
* 'shape': [ 2, 2 ],
* 'strides': [ 2, 1 ],
* 'offset': x.offset,
* 'order': x.order
* },
* {
* 'dtype': y.dtype,
* 'data': y.data,
* 'shape': [ 2, 2 ],
* 'strides': [ 2, 1 ],
* 'offset': y.offset,
* 'order': y.order
* }
* ];
*
* // Define a reshape strategy:
* function strategy( x ) {
* return {
* 'dtype': x.dtype,
* 'data': x.data,
* 'shape': [ 4 ],
* 'strides': [ 1 ],
* 'offset': x.offset,
* 'order': x.order
* };
* }
*
* // Perform a reduction:
* blockedbinary6d( gdot, [ x, y, z ], views, [ 12, 12, 12, 12, 12, 4 ], [ 12, 12, 12, 12, 12, 4 ], strategy, strategy, {} );
*
* var arr = ndarray2array( z.data, z.shape, z.strides, z.offset, z.order );
* // returns [ [ [ [ [ [ 30.0, 174.0, 446.0 ] ] ] ] ] ]
*/
function blockedbinary6d( fcn, arrays, views, strideX, strideY, strategy1, strategy2, opts ) {
var bsize;
var zbuf;
var set;
var dv0;
var dv1;
var dv2;
var dv3;
var dv4;
var dv5;
var ov1;
var ov2;
var ov3;
var ov4;
var ov5;
var sh;
var s0;
var s1;
var s2;
var s3;
var s4;
var s5;
var sv;
var ov;
var iv;
var i0;
var i1;
var i2;
var i3;
var i4;
var i5;
var j0;
var j1;
var j2;
var j3;
var j4;
var j5;
var N;
var o;
var x;
var y;
var z;
var v;
var k;
// Note on variable naming convention: S#, dv#, i#, j# where # corresponds to the loop number, with `0` being the innermost loop...
N = arrays.length;
x = arrays[ 0 ];
y = arrays[ 1 ];
z = arrays[ 2 ];
// Resolve the loop interchange order:
o = loopOrder( z.shape, strideX, strideY, z.strides );
sh = o.sh;
sv = [ o.sx, o.sy, o.sz ];
for ( k = 3; k < N; k++ ) {
sv.push( takeIndexed( arrays[k].strides, o.idx ) );
}
// Determine the block size:
bsize = blockSize( x.dtype, y.dtype, z.dtype );
// Resolve a list of pointers to the first indexed elements in the respective ndarrays:
ov = offsets( arrays );
// Cache a reference to the output ndarray buffer:
zbuf = z.data;
// Cache accessors:
set = z.accessors[1];
// Cache offset increments for the innermost loop...
dv0 = [];
for ( k = 0; k < N; k++ ) {
dv0.push( sv[k][0] );
}
// Initialize loop variables...
ov1 = zeros( N );
ov2 = zeros( N );
ov3 = zeros( N );
ov4 = zeros( N );
ov5 = zeros( N );
dv1 = zeros( N );
dv2 = zeros( N );
dv3 = zeros( N );
dv4 = zeros( N );
dv5 = zeros( N );
iv = zeros( N );
// Shallow copy the list of views to an internal array so that we can update with reshaped views without impacting the original list of views:
v = copyIndexed( views );
// Iterate over blocks...
for ( j5 = sh[5]; j5 > 0; ) {
if ( j5 < bsize ) {
s5 = j5;
j5 = 0;
} else {
s5 = bsize;
j5 -= bsize;
}
for ( k = 0; k < N; k++ ) {
ov5[ k ] = ov[k] + ( j5*sv[k][5] );
}
for ( j4 = sh[4]; j4 > 0; ) {
if ( j4 < bsize ) {
s4 = j4;
j4 = 0;
} else {
s4 = bsize;
j4 -= bsize;
}
for ( k = 0; k < N; k++ ) {
dv5 = sv[k][5] - ( s4*sv[k][4] );
ov4[ k ] = ov5[k] + ( j4*sv[k][4] );
}
for ( j3 = sh[3]; j3 > 0; ) {
if ( j3 < bsize ) {
s3 = j3;
j3 = 0;
} else {
s3 = bsize;
j3 -= bsize;
}
for ( k = 0; k < N; k++ ) {
dv4[ k ] = sv[k][4] - ( s3*sv[k][3] );
ov3[ k ] = ov4[k] + ( j3*sv[k][3] );
}
for ( j2 = sh[2]; j2 > 0; ) {
if ( j2 < bsize ) {
s2 = j2;
j2 = 0;
} else {
s2 = bsize;
j2 -= bsize;
}
for ( k = 0; k < N; k++ ) {
dv3[ k ] = sv[k][3] - ( s2*sv[k][2] );
ov2[ k ] = ov3[k] + ( j2*sv[k][2] );
}
for ( j1 = sh[1]; j1 > 0; ) {
if ( j1 < bsize ) {
s1 = j1;
j1 = 0;
} else {
s1 = bsize;
j1 -= bsize;
}
for ( k = 0; k < N; k++ ) {
dv2[ k ] = sv[k][2] - ( s1*sv[k][1] );
ov1[ k ] = ov2[k] + ( j1*sv[k][1] );
}
for ( j0 = sh[0]; j0 > 0; ) {
if ( j0 < bsize ) {
s0 = j0;
j0 = 0;
} else {
s0 = bsize;
j0 -= bsize;
}
// Compute index offsets and loop offset increments for the first ndarray elements in the current block...
for ( k = 0; k < N; k++ ) {
iv[ k ] = ov1[k] + ( j0*sv[k][0] );
dv1[ k ] = sv[k][1] - ( s0*sv[k][0] );
}
// Iterate over the non-reduced ndarray dimensions...
for ( i5 = 0; i5 < s5; i5++ ) {
for ( i4 = 0; i4 < s4; i4++ ) {
for ( i3 = 0; i3 < s3; i3++ ) {
for ( i2 = 0; i2 < s2; i2++ ) {
for ( i1 = 0; i1 < s1; i1++ ) {
for ( i0 = 0; i0 < s0; i0++ ) {
setViewOffsets( views, iv );
v[ 0 ] = strategy1( views[ 0 ] );
v[ 1 ] = strategy2( views[ 1 ] );
set( zbuf, iv[ 2 ], fcn( v, opts ) );
incrementOffsets( iv, dv0 );
}
incrementOffsets( iv, dv1 );
}
incrementOffsets( iv, dv2 );
}
incrementOffsets( iv, dv3 );
}
incrementOffsets( iv, dv4 );
}
incrementOffsets( iv, dv5 );
}
}
}
}
}
}
}
}
// EXPORTS //
module.exports = blockedbinary6d;
|