Typed Array Pool

Allocate typed arrays from a typed array memory pool.

Installation

npm install @stdlib/array-pool

Usage

var typedarraypool = require( '@stdlib/array-pool' );

typedarraypool( [dtype] )

Returns an uninitialized typed array having a specified data type dtype.

var arr = typedarraypool();
// returns <Float64Array>[]

// ...

typedarraypool.free( arr );

The function recognizes the following data types:

• float64: double-precision floating-point numbers (IEEE 754)
• float32: single-precision floating-point numbers (IEEE 754)
• complex128: double-precision complex floating-point numbers
• complex64: single-precision complex floating-point numbers
• int32: 32-bit two's complement signed integers
• uint32: 32-bit unsigned integers
• int16: 16-bit two's complement signed integers
• uint16: 16-bit unsigned integers
• int8: 8-bit two's complement signed integers
• uint8: 8-bit unsigned integers
• uint8c: 8-bit unsigned integers clamped to 0-255

By default, the output typed array is float64. To specify an alternative data type, set the dtype parameter.

var arr = typedarraypool( 'int32' );
// returns <Int32Array>[]

// ...

typedarraypool.free( arr );

typedarraypool( length[, dtype] )

Returns an uninitialized typed array having a specified length from a typed array memory pool.

var arr1 = typedarraypool( 5 );
// returns <Float64Array>

var arr2 = typedarraypool( 5, 'uint8' );
// returns <Uint8Array>

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool( typedarray[, dtype] )

Returns a pooled typed array from another typed array.

var arr1 = typedarraypool( [ 5.0, -3.0, 2.0 ] );
// returns <Float64Array>[ 5.0, -3.0, 2.0 ]

var arr2 = typedarraypool( arr1 );
// returns <Float64Array>[ 5.0, -3.0, 2.0 ]

var arr3 = typedarraypool( arr1, 'int32' );
// returns <Int32Array>[ 5, -3, 2 ]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );
typedarraypool.free( arr3 );

typedarraypool( obj[, dtype] )

Returns a pooled typed array from an array-like object.

var arr1 = typedarraypool( [ 0.5, 0.5, 0.5 ] );
// returns <Float64Array>[ 0.5, 0.5, 0.5 ]

var arr2 = typedarraypool( [ 0.5, 0.5, 0.5 ], 'float32' );
// returns <Float32Array>[ 0.5, 0.5, 0.5 ]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.malloc( [dtype] )

Returns an uninitialized typed array having a specified data type dtype.

var arr1 = typedarraypool.malloc();
// returns <Float64Array>[]

var arr2 = typedarraypool.malloc( 'int32' );
// returns <Int32Array>[]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.malloc( length[, dtype] )

Returns an uninitialized typed array having a specified length from a typed array memory pool.

var arr1 = typedarraypool.malloc( 5 );
// returns <Float64Array>

var arr2 = typedarraypool.malloc( 5, 'uint8' );
// returns <Uint8Array>

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.malloc( typedarray[, dtype] )

Returns a pooled typed array from another typed array.

var arr1 = typedarraypool.malloc( [ 5.0, -3.0, 2.0 ] );
// returns <Float64Array>[ 5.0, -3.0, 2.0 ]

var arr2 = typedarraypool.malloc( arr1 );
// returns <Float64Array>[ 5.0, -3.0, 2.0 ]

var arr3 = typedarraypool.malloc( arr1, 'int32' );
// returns <Int32Array>[ 5, -3, 2 ]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );
typedarraypool.free( arr3 );

typedarraypool.malloc( obj[, dtype] )

Returns a pooled typed array from an array-like object.

var arr1 = typedarraypool.malloc( [ 0.5, 0.5, 0.5 ] );
// returns <Float64Array>[ 0.5, 0.5, 0.5 ]

var arr2 = typedarraypool.malloc( [ 0.5, 0.5, 0.5 ], 'float32' );
// returns <Float32Array>[ 0.5, 0.5, 0.5 ]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.calloc( [dtype] )

Returns a zero-initialized typed array having a specified data type dtype.

var arr1 = typedarraypool.calloc();
// returns <Float64Array>[]

var arr2 = typedarraypool.calloc( 'int32' );
// returns <Int32Array>[]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.calloc( length[, dtype] )

Returns a zero-initialized typed array having a specified length from a typed array memory pool.

var arr1 = typedarraypool.calloc( 5 );
// returns <Float64Array>[ 0.0, 0.0, 0.0, 0.0, 0.0 ]

var arr2 = typedarraypool.calloc( 5, 'uint8' );
// returns <Uint8Array>[ 0, 0, 0, 0, 0 ]

// ...

typedarraypool.free( arr1 );
typedarraypool.free( arr2 );

typedarraypool.free( buf )

Frees a typed array or typed array buffer for use in a future allocation.

var arr = typedarraypool( 10, 'float64' );
// returns <Float64Array>

// ...

// Free the allocated typed array for use in a future allocation:
typedarraypool.free( arr );

// Create another typed array:
arr = typedarraypool( 10, 'float64' );
// returns <Float64Array>

// ...

// Free the allocated typed array buffer for use in a future allocation:
typedarraypool.free( arr.buffer );

typedarraypool.clear()

Clears the typed array pool allowing garbage collection of previously allocated (and currently free) array buffers.

var arr = typedarraypool( 10, 'float64' );
// returns <Float64Array>

// ...

typedarraypool.free( arr );

// ...

// Clear all freed buffers:
typedarraypool.clear();

typedarraypool.highWaterMark

Read-only property returning the pool's high water mark (in bytes).

var limit = typedarraypool.highWaterMark;
// returns <number>

Once a high water mark is reached, typed array allocation fails.

typedarraypool.nbytes

Read-only property returning the total number of allocated bytes.

var arr = typedarraypool( 5, 'float64' );

var nbytes = typedarraypool.nbytes;
// returns <number>

The returned value is the total accumulated value. Hence, anytime a pool must allocate a new array buffer (i.e., more memory), the pool increments this value. The only time this value is decremented is when a pool is cleared. This behavior means that, while allocated buffers which are never freed may, in fact, be garbage collected, they continue to count against the high water mark limit. Accordingly, you should always free allocated buffers in order to prevent the pool from believing that non-freed buffers are continually in use.

typedarraypool.factory( [options] )

Creates a new typed array pool.

var pool = typedarraypool.factory();

var arr = pool( 5, 'float64' );
// returns <Float64Array>

// ...

pool.free( arr );

The method accepts the following options:

• highWaterMark: maximum total memory (in bytes) which can be allocated. Default: 2^53 bytes.

By default, the maximum total memory a pool may allocate is 2^53 bytes (approximately 1 petabyte, which, in practical terms, means a pool has unlimited capacity). To specify an alternative limit, set the highWaterMark option.

// Create a new typed array pool which can allocate up to 1MB:
var pool = typedarraypool.factory({
    'highWaterMark': 1e6
});

var arr = pool( 5, 'float64' );
// returns <Float64Array>

// ...

pool.free( arr );

Notes

• Uninitialized typed arrays may contain sensitive contents. If security is paramount (e.g., if freed typed arrays have been used to store sensitive contents), use calloc.
• An allocated typed array is guaranteed to have an underlying array buffer with at least N * w bytes, where N is the number of typed array elements and w is the number of bytes per element. Note, however, that the underlying array buffer is likely to have excess capacity. Thus, if you create many typed arrays which are held in memory and are not freed, you are likely to consume significantly more memory than if you had directly used typed array constructors. However, if you create many typed arrays which are rapidly discarded and of relatively large size, then using a typed array pool can offer significant performance advantages.

Examples

var randu = require( '@stdlib/random-base-randu' );
var typedarraypool = require( '@stdlib/array-pool' );

// Create a typed array pool which can allocate at most 1GB:
var typedarray = typedarraypool.factory({
    'highWaterMark': 1e9
});

// Inspect the pool:
console.log( 'Max bytes: %d', typedarray.highWaterMark );
console.log( 'nbytes: %d', typedarray.nbytes );

// Allocate an array for storing double-precision floating-point numbers:
var arr1 = typedarray( 5, 'float64' );
// returns <Float64Array>[ 0.0, 0.0, 0.0, 0.0, 0.0 ]

// Inspect the pool:
console.log( 'nbytes: %d', typedarray.nbytes );

// Fill the array...
var i;
for ( i = 0; i < arr1.length; i++ ) {
    arr1[ i ] = randu();
}

// Inspect array contents:
console.log( arr1 );

// Free the array:
typedarray.free( arr1 );

// Allocate another array similar to the previous one:
var arr2 = typedarray( 5, 'float64' );
// returns <Float64Array>

// Check that we have been returned a new typed array view:
console.log( arr2 === arr1 );
// => false

// Inspect array contents:
console.log( arr2 );

// Free the array:
typedarray.free( arr2 );

// Allocate an initialized array:
var arr3 = typedarray.calloc( 5, 'float64' );
// returns <Float64Array>[ 0.0, 0.0, 0.0, 0.0, 0.0 ]

// Inspect array contents:
console.log( arr3 );

// Free the array:
typedarray.free( arr3 );

// Clear the pool:
typedarray.clear();

// Inspect the pool:
console.log( 'nbytes: %d', typedarray.nbytes );

See Also

• @stdlib/array-typed: create a typed array.

Notice

This package is part of stdlib, a standard library for JavaScript and Node.js, with an emphasis on numerical and scientific computing. The library provides a collection of robust, high performance libraries for mathematics, statistics, streams, utilities, and more.

For more information on the project, filing bug reports and feature requests, and guidance on how to develop stdlib, see the main project repository.

Community

License

See LICENSE.

Copyright

Copyright © 2016-2024. The Stdlib Authors.