You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
// So long as iteration for each respective array always moves in the same direction (i.e., no mixed sign strides), we can leverage cache-optimal (i.e., normal) nested loops without resorting to blocked iteration...
375
375
obj->functions[ ndims ]( arrays, fcn );
376
376
return0;
377
377
}
378
378
// Fall-through to blocked iteration...
379
379
}
380
-
// At this point, we're either dealing with non-contiguous n-dimensional arrays and/or high dimensional n-dimensional arrays, so our only hope is that we can still perform blocked iteration...
380
+
// At this point, we're either dealing with non-contiguous n-dimensional arrays, high dimensional n-dimensional arrays, and/or arrays having differing memory layouts, so our only hope is that we can still perform blocked iteration...
381
381
382
382
// Determine whether we can perform blocked iteration...
0 commit comments