whisper.wasm : do not block page while processing (close ggerganov#86)

stancx1 · Nov 25, 2022 · be16dfa · be16dfa
1 parent 0f619b5
commit be16dfa
Show file tree

Hide file tree

Showing 7 changed files with 47 additions and 9 deletions.
diff --git a/bindings/javascript/emscripten.cpp b/bindings/javascript/emscripten.cpp
@@ -6,10 +6,16 @@
 #include <vector>
 #include <thread>
 
+std::thread g_worker;
+
 std::vector<struct whisper_context *> g_contexts(4, nullptr);
 
 EMSCRIPTEN_BINDINGS(whisper) {
     emscripten::function("init", emscripten::optional_override([](const std::string & path_model) {
+        if (g_worker.joinable()) {
+            g_worker.join();
+        }
+
         for (size_t i = 0; i < g_contexts.size(); ++i) {
             if (g_contexts[i] == nullptr) {
                 g_contexts[i] = whisper_init(path_model.c_str());
@@ -25,6 +31,10 @@ EMSCRIPTEN_BINDINGS(whisper) {
     }));
 
     emscripten::function("free", emscripten::optional_override([](size_t index) {
+        if (g_worker.joinable()) {
+            g_worker.join();
+        }
+
         --index;
 
         if (index < g_contexts.size()) {
@@ -34,6 +44,10 @@ EMSCRIPTEN_BINDINGS(whisper) {
     }));
 
     emscripten::function("full_default", emscripten::optional_override([](size_t index, const emscripten::val & audio, const std::string & lang, bool translate) {
+        if (g_worker.joinable()) {
+            g_worker.join();
+        }
+
         --index;
 
         if (index >= g_contexts.size()) {
@@ -80,10 +94,15 @@ EMSCRIPTEN_BINDINGS(whisper) {
             printf("\n");
         }
 
-        int ret = whisper_full(g_contexts[index], params, pcmf32.data(), pcmf32.size());
-
-        whisper_print_timings(g_contexts[index]);
+        // run the worker
+        {
+            g_worker = std::thread([index, params, pcmf32 = std::move(pcmf32)]() {
+                whisper_reset_timings(g_contexts[index]);
+                whisper_full(g_contexts[index], params, pcmf32.data(), pcmf32.size());
+                whisper_print_timings(g_contexts[index]);
+            });
+        }
 
-        return ret;
+        return 0;
     }));
 }
diff --git a/bindings/javascript/whisper.js b/bindings/javascript/whisper.js
diff --git a/examples/command/command.cpp b/examples/command/command.cpp
@@ -109,9 +109,8 @@ void whisper_print_usage(int argc, char ** argv, const whisper_params & params)
 
 class audio_async {
 public:
-    audio_async(int len_ms) {
-        m_len_ms = len_ms;
-    }
+    audio_async(int len_ms);
+    ~audio_async();
 
     bool init(int capture_id, int sample_rate);
 
@@ -142,6 +141,16 @@ class audio_async {
     size_t             m_audio_len = 0;
 };
 
+audio_async::audio_async(int len_ms) {
+    m_len_ms = len_ms;
+}
+
+audio_async::~audio_async() {
+    if (m_dev_id_in) {
+        SDL_CloseAudioDevice(m_dev_id_in);
+    }
+}
+
 bool audio_async::init(int capture_id, int sample_rate) {
     SDL_LogSetPriority(SDL_LOG_CATEGORY_APPLICATION, SDL_LOG_PRIORITY_INFO);
 

diff --git a/examples/stream/stream.cpp b/examples/stream/stream.cpp
@@ -384,6 +384,10 @@ int main(int argc, char ** argv) {
         }
     }
 
+    if (g_dev_id_in >= 0) {
+        SDL_CloseAudioDevice(g_dev_id_in);
+    }
+
     whisper_print_timings(ctx);
     whisper_free(ctx);
 

diff --git a/examples/whisper.wasm/index-tmpl.html b/examples/whisper.wasm/index-tmpl.html
@@ -526,7 +526,6 @@
                 if (instance) {
                     printTextarea('');
                     printTextarea('js: processing - this might take a while ...');
-                    printTextarea('js: the page will be unresponsive until the processing is completed');
                     printTextarea('');
 
                     setTimeout(function() {

diff --git a/whisper.cpp b/whisper.cpp
@@ -2380,6 +2380,12 @@ void whisper_print_timings(struct whisper_context * ctx) {
     fprintf(stderr, "%s:    total time = %8.2f ms\n", __func__, (t_end_us - ctx->t_start_us)/1000.0f);
 }
 
+void whisper_reset_timings(struct whisper_context * ctx) {
+    ctx->t_sample_us = 0;
+    ctx->t_encode_us = 0;
+    ctx->t_decode_us = 0;
+}
+
 ////////////////////////////////////////////////////////////////////////////
 
 struct whisper_full_params whisper_full_default_params(enum whisper_sampling_strategy strategy) {

diff --git a/whisper.h b/whisper.h
@@ -167,6 +167,7 @@ extern "C" {
 
     // Performance information
     WHISPER_API void whisper_print_timings(struct whisper_context * ctx);
+    WHISPER_API void whisper_reset_timings(struct whisper_context * ctx);
 
     ////////////////////////////////////////////////////////////////////////////