examples : command.wasm updates

CMakeLists.txt

@@ -38,8 +38,13 @@ if (EMSCRIPTEN)
 
     # TODO: without these, we get the following error:
     #       wasm-ld: error: --shared-memory is disallowed by whisper.cpp.o because it was not compiled with 'atomics' or 'bulk-memory' features.
-    set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS}   -pthread -s TOTAL_STACK=5242880")
-    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread -s TOTAL_STACK=5242880")
+    set(CMAKE_C_FLAGS   "${CMAKE_C_FLAGS}   -pthread")
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -pthread")
+
+    set(CMAKE_EXE_LINKER_FLAGS "${CMAKE_EXE_LINKER_FLAGS} -s TOTAL_STACK=5242880")
+    set(CMAKE_SHARED_LINKER_FLAGS "${CMAKE_SHARED_LINKER_FLAGS} -s TOTAL_STACK=5242880")
+
+    set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -Wno-deprecated")
 else()
     if (MINGW)
         set(BUILD_SHARED_LIBS_DEFAULT OFF)

examples/command.wasm/README.md

@@ -15,9 +15,18 @@ git clone https://github.com/ggerganov/whisper.cpp
 cd whisper.cpp
 mkdir build-em && cd build-em
 emcmake cmake ..
-make -j
+make -j libcommand
+```
+The example can then be started by running a local HTTP server:
+```console
+python3 examples/server.py
+```
+And then opening a browser to the following URL:
+http://localhost:8000/command.wasm/
 
-# copy the produced page to your HTTP path
+To run the example in a different server, you need to copy the following files
+to the server's HTTP path:
+```
 cp bin/command.wasm/*       /path/to/html/
 cp bin/libcommand.worker.js /path/to/html/
 ```

examples/common.cpp

@@ -248,14 +248,37 @@ std::map<std::string, int32_t> json_parse(const std::string & fname) {
 }
 
 std::string convert_to_utf8(const std::wstring & input) {
+#if defined(__clang__)
+    // disable C++17 deprecation warning for std::codecvt_utf8
+#    pragma clang diagnostic push
+#    pragma clang diagnostic ignored "-Wdeprecated-declarations"
+#endif
+
     std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
-    return converter.to_bytes(input);
-}
+    auto result = converter.to_bytes(input);
 
+#if defined(__clang__)
+#    pragma clang diagnostic pop
+#endif
+
+    return result;
+}
 
 std::wstring convert_to_wstring(const std::string & input) {
+#if defined(__clang__)
+    // disable C++17 deprecation warning for std::codecvt_utf8
+#    pragma clang diagnostic push
+#    pragma clang diagnostic ignored "-Wdeprecated-declarations"
+#endif
+
     std::wstring_convert<std::codecvt_utf8<wchar_t>> converter;
-    return converter.from_bytes(input);
+    auto result = converter.from_bytes(input);
+
+#if defined(__clang__)
+#    pragma clang diagnostic pop
+#endif
+
+    return result;
 }
 
 void gpt_split_words(std::string str, std::vector<std::string>& words) {

examples/server.py

@@ -0,0 +1,27 @@
+import http.server
+import socketserver
+import os
+from pathlib import Path
+
+SCRIPT_DIR = Path(__file__).parent.absolute()
+DIRECTORY = os.path.join(SCRIPT_DIR, "../build-em/bin")
+DIRECTORY = os.path.abspath(DIRECTORY)
+
+class CustomHTTPRequestHandler(http.server.SimpleHTTPRequestHandler):
+    def __init__(self, *args, **kwargs):
+        super().__init__(*args, directory=DIRECTORY, **kwargs)
+
+    def end_headers(self):
+        # Add required headers for SharedArrayBuffer
+        self.send_header("Cross-Origin-Opener-Policy", "same-origin")
+        self.send_header("Cross-Origin-Embedder-Policy", "require-corp")
+        super().end_headers()
+
+PORT = 8000
+
+with socketserver.TCPServer(("", PORT), CustomHTTPRequestHandler) as httpd:
+    print(f"Serving directory '{DIRECTORY}' at http://localhost:{PORT}")
+    try:
+        httpd.serve_forever()
+    except KeyboardInterrupt:
+        print("\nServer stopped.")

examples/stb_vorbis.c

@@ -1401,7 +1401,7 @@ static int set_file_offset(stb_vorbis *f, unsigned int loc)
    #endif
    f->eof = 0;
    if (USE_MEMORY(f)) {
-      if (f->stream_start + loc >= f->stream_end || f->stream_start + loc < f->stream_start) {
+      if (loc >= f->stream_len) {
          f->stream = f->stream_end;
          f->eof = 1;
          return 0;

ggml/src/ggml-cpu/ggml-cpu.c

@@ -901,24 +901,24 @@ inline static void __wasm_f16x4_store(ggml_fp16_t * p, v128_t x) {
 #define GGML_F16x4_FMA         GGML_F32x4_FMA
 #define GGML_F16x4_ADD         wasm_f32x4_add
 #define GGML_F16x4_MUL         wasm_f32x4_mul
-#define GGML_F16x4_REDUCE(res, x)                  \
-{                                                  \
-    int offset = GGML_F16_ARR >> 1;                \
-    for (int i = 0; i < offset; ++i) {             \
-        x[i] = wasm_f32x4_add(x[i], x[offset+i]);  \
-    }                                              \
-    offset >>= 1;                                  \
-    for (int i = 0; i < offset; ++i) {             \
-        x[i] = wasm_f32x4_add(x[i], x[offset+i]);  \
-    }                                              \
-    offset >>= 1;                                  \
-    for (int i = 0; i < offset; ++i) {             \
-        x[i] = wasm_f32x4_add(x[i], x[offset+i]);  \
-    }                                              \
-    res = wasm_f32x4_extract_lane(x[0], 0) +       \
-          wasm_f32x4_extract_lane(x[0], 1) +       \
-          wasm_f32x4_extract_lane(x[0], 2) +       \
-          wasm_f32x4_extract_lane(x[0], 3);        \
+#define GGML_F16x4_REDUCE(res, x)                           \
+{                                                           \
+    int offset = GGML_F16_ARR >> 1;                         \
+    for (int i = 0; i < offset; ++i) {                      \
+        x[i] = wasm_f32x4_add(x[i], x[offset+i]);           \
+    }                                                       \
+    offset >>= 1;                                           \
+    for (int i = 0; i < offset; ++i) {                      \
+        x[i] = wasm_f32x4_add(x[i], x[offset+i]);           \
+    }                                                       \
+    offset >>= 1;                                           \
+    for (int i = 0; i < offset; ++i) {                      \
+        x[i] = wasm_f32x4_add(x[i], x[offset+i]);           \
+    }                                                       \
+    res = (ggml_float) (wasm_f32x4_extract_lane(x[0], 0) +  \
+          wasm_f32x4_extract_lane(x[0], 1) +                \
+          wasm_f32x4_extract_lane(x[0], 2) +                \
+          wasm_f32x4_extract_lane(x[0], 3));                \
 }
 
 #define GGML_F16_VEC                GGML_F16x4