Official sqlite3 benchmark, compiled to WebAssembly

This is a Wasm (shell-only) build of SQLite's official speedtest1.c benchmark program. Quoting from https://sqlite.org/cpu.html:

This program strives to exercise the SQLite library in a way that is typical of real-world applications. Of course, every application is different, and so no test program can exactly mirror the behavior of all applications. The speedtest1.c program is updated from time to time as the SQLite developers' understanding of what constitutes "typical" usage evolves.

Build Instructions

See build.sh or just run it. See build.log for the last build time, used sources, and toolchain versions.

Prerequisites:

Emscripten: https://emscripten.org. Make sure emcc is on your PATH, e.g., by running source /path/to/emsdk_env.sh.
wasm-strip: Install WABT, e.g., via sudo apt install wabt or by compiling yourself from the source at https://github.com/WebAssembly/wabt.

Since this benchmark is meant for the developers of SQLite, it is not part of the official Wasm build of SQLite at https://sqlite.org/wasm/doc/trunk/index.md and also not contained in the "amalgamation" single-file source code of SQLite (see https://sqlite.org/amalgamation.html). So we need to build it ourselves from the full sources. You can download the full SQLite source tree from https://sqlite.org/download.html. See under "Alternative Source Code Formats" or search for "Snapshot of the complete (raw) source tree for SQLite".

Running in JS shells

The SQLite developers only maintain compatibility of the Wasm build for running inside browsers, so benchmark.js contains some polyfills to run it JavaScript shells (such as d8 (V8), js (SpiderMonkey), and jsc (JavaScriptCore)). Ideally, it should run just via $shell benchmark.js from the current directory.

To keep the shell runner and browser results consistent, the benchmark is configured to never use OPFS (Origin Private File System) as the underlying storage layer ("VFS" in SQLite), since that is not available in shells. It might thus show slightly different performance characteristics compared to the upstream speedtest1.html running with OPFS.

Running the upstream version of the benchmark in browsers

Start a webserver in build/ that serves with the correct headers (CORS/COOP/COEP) set for Wasm execution. E.g., this simple Python server will do:

#!/usr/bin/env python3
from http.server import HTTPServer, SimpleHTTPRequestHandler, test
import sys

class CORSRequestHandler (SimpleHTTPRequestHandler):
    def end_headers (self):
        self.send_header('Cross-Origin-Embedder-Policy', 'require-corp')
        self.send_header('Cross-Origin-Opener-Policy', 'same-origin')

        SimpleHTTPRequestHandler.end_headers(self)

if __name__ == '__main__':
    test(CORSRequestHandler, HTTPServer, port=int(sys.argv[1]) if len(sys.argv) > 1 else 8083)

Then browse to http://localhost:8083/speedtest1.html. (Note that, e.g., Chrome requires localhost, not an IP!) Because the computation happens on the main thread, it may hang for a little while but should show the console output afterwards.