Detect browser support for multithreading

.gitignore

@@ -36,6 +36,6 @@ cert.pem
 key.pem
 
 # build
-dist/
+/dist/
 build/
 node_modules/

README.md

@@ -14,8 +14,7 @@ https://tangledgroup.github.io/llama-cpp-wasm/
 ```bash
 git clone https://github.com/tangledgroup/llama-cpp-wasm.git
 cd llama-cpp-wasm
-./build-single-thread.sh
-./build-multi-thread.sh
+./build.sh
 ```
 
 Once build is complete you can find `llama.cpp` built in `dist/llama-st` and `dist/llama-mt` directory.

build-multi-thread.sh

@@ -40,8 +40,8 @@ if [ -d $LLAMA_CPP_WASM_DIST_LLAMA_DIR ]; then
 fi
 
 mkdir -p $LLAMA_CPP_WASM_DIST_LLAMA_DIR
-cp -rv src/llama/* $LLAMA_CPP_WASM_DIST_LLAMA_DIR
+cp -rv src/llama/* $LLAMA_CPP_WASM_DIST_DIR
 cp $LLAMA_CPP_BUILD_DIR/bin/main.* $LLAMA_CPP_WASM_DIST_LLAMA_DIR
 
-rm -rf docs/llama-mt
-cp -rv $LLAMA_CPP_WASM_DIST_LLAMA_DIR docs/
+rm -rf docs/$LLAMA_CPP_WASM_DIST_DIR
+cp -rv $LLAMA_CPP_WASM_DIST_DIR docs/

build-single-thread.sh

@@ -40,8 +40,8 @@ if [ -d $LLAMA_CPP_WASM_DIST_LLAMA_DIR ]; then
 fi
 
 mkdir -p $LLAMA_CPP_WASM_DIST_LLAMA_DIR
-cp -rv src/llama/* $LLAMA_CPP_WASM_DIST_LLAMA_DIR
+cp -rv src/llama/* $LLAMA_CPP_WASM_DIST_DIR
 cp $LLAMA_CPP_BUILD_DIR/bin/main.* $LLAMA_CPP_WASM_DIST_LLAMA_DIR
 
-rm -rf docs/llama-st
-cp -rv $LLAMA_CPP_WASM_DIST_LLAMA_DIR docs/
+rm -rf docs/$LLAMA_CPP_WASM_DIST_DIR
+cp -rv $LLAMA_CPP_WASM_DIST_DIR docs/

docs/llama-mt/main-worker.js → docs/dist/main-worker.js

@@ -1,6 +1,5 @@
 import { action } from "./actions.js";
-import { loadBinaryResource } from "./utility.js";
-import Module from "./main.js";
+import { isMultiThreadSupported, loadBinaryResource } from "./utility.js";
 
 // WASM Module
 let module;
@@ -60,6 +59,10 @@ const initWorker = async (modelPath) => {
         }],
     };
 
+    const { default: Module } = await (isMultiThreadSupported()
+        ? import("./llama-mt/main.js")
+        : import("./llama-st/main.js"));
+
     module = await Module(emscrModule);
 
     const initCallback = (bytes) => {
@@ -103,7 +106,7 @@ const run_main = (
         "--prompt", prompt.toString(),
     ];
 
-    if (!!globalThis.SharedArrayBuffer) {
+    if (isMultiThreadSupported()) {
         args.push("--threads");
         args.push((navigator.hardwareConcurrency).toString());
     }

docs/llama-st/utility.js → docs/dist/utility.js

@@ -40,3 +40,29 @@ export async function loadBinaryResource(url, callback) {
 
     req.send(null);
 }
+
+export function isMultiThreadSupported() {
+    // If 'SharedArrayBuffer' is not available, WebAssembly threads will not work.
+    if (typeof SharedArrayBuffer === "undefined") {
+        return false;
+    }
+
+    try {
+        // Test for transferability of SABs (for browsers. needed for Firefox)
+        // https://groups.google.com/forum/#!msg/mozilla.dev.platform/IHkBZlHETpA/dwsMNchWEQAJ
+        if (typeof MessageChannel !== "undefined") {
+            new MessageChannel().port1.postMessage(new SharedArrayBuffer(1));
+        }
+
+        // Test for WebAssembly threads capability (for both browsers and Node.js)
+        // This typed array is a WebAssembly program containing threaded instructions.
+        return WebAssembly.validate(
+            new Uint8Array([
+                0, 97, 115, 109, 1, 0, 0, 0, 1, 4, 1, 96, 0, 0, 3, 2, 1, 0, 5,
+                4, 1, 3, 1, 1, 10, 11, 1, 9, 0, 65, 0, 254, 16, 2, 0, 26, 11,
+            ])
+        );
+    } catch (e) {
+        return false;
+    }
+}

docs/example-multi-thread.js → docs/example.js

@@ -1,4 +1,4 @@
-import { LlamaCpp } from "./llama-mt/llama.js";
+import { LlamaCpp } from "./dist/llama.js";
 
 let app;
 const buttonRun = document.querySelector("#run");