Benchmark: ArrayBuffer to base64 String vs. superhuman

HTML Preparation code:


<script src='https://cdn.jsdelivr.net/gh/superhuman/fast64@master/index.js'></script>

Script Preparation code:

var ab = new ArrayBuffer(70*1024);
var uint8 = new Uint8Array(ab);
for (var i = 0; i < ab.byteLength; i++) {
    uint8[i] = i % 256;
}
var expected = window.btoa(String.fromCharCode.apply(null, uint8));

function byteArrayToString(bytes) {
  var CHUNK_SIZE = 8*1024;
  if (bytes.length <= CHUNK_SIZE)
    return String.fromCharCode.apply(null, bytes);
  var str = '';
  for (var i = 0; i < bytes.length; i += CHUNK_SIZE)
    str += String.fromCharCode.apply(null, bytes.slice(i, i+CHUNK_SIZE));
  return str;
}

function validate(res){
  if (res !== expected)
      throw "expected=" + expected + " but got " + res;
}


function uint6ToB64(nUint6) {
  return nUint6 < 26 ?
      nUint6 + 65
    : nUint6 < 52 ?
      nUint6 + 71
    : nUint6 < 62 ?
      nUint6 - 4
    : nUint6 === 62 ?
      43
    : nUint6 === 63 ?
      47
    :
      65;
};

function bytesToBase64(aBytes) {
  var eqLen = (3 - (aBytes.length % 3)) % 3, sB64Enc = "";

  for (var nMod3, nLen = aBytes.length, nUint24 = 0, nIdx = 0; nIdx < nLen; nIdx++) {
    nMod3 = nIdx % 3;
    nUint24 |= aBytes[nIdx] << (16 >>> nMod3 & 24);
    if (nMod3 === 2 || aBytes.length - nIdx === 1) {
      sB64Enc += String.fromCharCode(uint6ToB64(nUint24 >>> 18 & 63), uint6ToB64(nUint24 >>> 12 & 63), uint6ToB64(nUint24 >>> 6 & 63), uint6ToB64(nUint24 & 63));
      nUint24 = 0;
    }
  }

  return  eqLen === 0 ?
            sB64Enc
          :
            sB64Enc.substring(0, sB64Enc.length - eqLen) + (eqLen === 1 ? "=" : "==");
};

Tests:

String.fromCharCode.apply
var bin = String.fromCharCode.apply(null, uint8); var res = window.btoa(bin) validate(res)
byteArrayToString
var bin = byteArrayToString(uint8); var res = window.btoa(bin) validate(res)
superhuman
var res = B64.encode(uint8); validate(res)

Rendered benchmark preparation results:

Suite status: <idle, ready to run>

Previous results

Test case name	Result
String.fromCharCode.apply
byteArrayToString
superhuman

Fastest: N/A

Slowest: N/A

Latest run results:

Run details: (Test run date: one month ago)

User agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/134.0.0.0 Safari/537.36

Browser/OS: Chrome 134 on Mac OS X 10.15.7

View result in a separate tab

Test name	Executions per second
String.fromCharCode.apply	4356.4 Ops/sec
byteArrayToString	4538.0 Ops/sec
superhuman	9407.4 Ops/sec

Autogenerated LLM Summary (model llama3.2:3b, generated 6 months ago):

LLMs can make mistakes. Check important info.

Let's break down what is being tested in the provided JSON benchmark.

**Benchmark Overview**

The benchmark compares three different approaches to convert an ArrayBuffer (a binary data type) to a base64 string:

1. `window.btoa(String.fromCharCode.apply(null, uint8))` (using `String.fromCharCode`)
2. `byteArrayToString(uint8)` and then passing it to `window.btoa()` (using a custom function `byteArrayToString`)
3. `B64.encode(uint8)` from the `superhuman` library

**Options Compared**

The three options are compared in terms of execution time, which is measured as `ExecutionsPerSecond`. The benchmark measures how many times each approach can be executed per second on a specific hardware configuration.

**Pros and Cons of Each Approach**

1. **Using `window.btoa(String.fromCharCode.apply(null, uint8))`**
	* Pros: Simple and straightforward, uses built-in JavaScript functions.
	* Cons: May not be optimized for performance, may have overhead due to string creation.
2. **Using `byteArrayToString(uint8)` and then passing it to `window.btoa()`**
	* Pros: Allows for custom optimization of the conversion process, can handle larger ArrayBuffer sizes more efficiently.
	* Cons: Requires implementing a custom function (`byteArrayToString`) which may add complexity.
3. **Using `B64.encode(uint8)` from `superhuman` library**
	* Pros: Optimized for performance, uses a dedicated library that is likely to be highly optimized.
	* Cons: Requires including an external library, may have dependencies or versioning issues.

**Library and its Purpose**

The `superhuman` library provides a high-performance base64 encoding function (`B64.encode`) that is specifically designed to optimize the encoding process. The library is written in C++ and uses SIMD instructions for optimal performance.

**Special JS Feature/Syntax**

The benchmark does not explicitly mention any special JavaScript features or syntax, but it does use the `apply()` method to call a function with an array of arguments. This is a common pattern in JavaScript for invoking functions with multiple arguments.

**Other Alternatives**

Other alternatives for base64 encoding exist, such as using a library like `base64-js` or implementing a custom solution. However, these alternatives may not be optimized for performance like the `superhuman` library.

Related benchmarks:

ArrayBuffer to base64 String vs. superhuman (version: 0)

Comparing performance of: String.fromCharCode.apply vs byteArrayToString vs superhuman

Created: 4 years ago by: Guest

Jump to the latest result

String.fromCharCode.apply

byteArrayToString

superhuman

Suite status: <idle, ready to run>

Experimental features:

Fastest: N/A

Slowest: N/A

Autogenerated LLM Summary (model llama3.2:3b, generated 6 months ago):