whisper.rn 0.5.0-rc.9 → 0.5.1

package/cpp/ggml-backend-impl.h

@@ -8,7 +8,7 @@
 extern "C" {
 #endif
 
-    #define WSP_GGML_BACKEND_API_VERSION 1
+    #define WSP_GGML_BACKEND_API_VERSION 2
 
     //
     // Backend buffer type
@@ -114,6 +114,9 @@ extern "C" {
         void (*event_record)(wsp_ggml_backend_t backend, wsp_ggml_backend_event_t event);
         // wait for an event on on a different stream
         void (*event_wait)  (wsp_ggml_backend_t backend, wsp_ggml_backend_event_t event);
+
+        // (optional) sort/optimize the nodes in the graph
+        void                      (*graph_optimize)    (wsp_ggml_backend_t backend, struct wsp_ggml_cgraph * cgraph);
     };
 
     struct wsp_ggml_backend {

package/cpp/ggml-backend-reg.cpp

@@ -45,6 +45,14 @@
 #include "ggml-vulkan.h"
 #endif
 
+#ifdef WSP_GGML_USE_WEBGPU
+#include "ggml-webgpu.h"
+#endif
+
+#ifdef WSP_GGML_USE_ZDNN
+#include "ggml-zdnn.h"
+#endif
+
 #ifdef WSP_GGML_USE_OPENCL
 #include "ggml-opencl.h"
 #endif
@@ -61,10 +69,6 @@
 #include "ggml-cann.h"
 #endif
 
-#ifdef WSP_GGML_USE_KOMPUTE
-#include "ggml-kompute.h"
-#endif
-
 // disable C++17 deprecation warning for std::codecvt_utf8
 #if defined(__clang__)
 #    pragma clang diagnostic push
@@ -131,6 +135,10 @@ static void * dl_get_sym(dl_handle * handle, const char * name) {
     return p;
 }
 
+static const char * dl_error() {
+    return "";
+}
+
 #else
 
 using dl_handle = void;
@@ -151,6 +159,11 @@ static void * dl_get_sym(dl_handle * handle, const char * name) {
     return dlsym(handle, name);
 }
 
+static const char * dl_error() {
+    const char *rslt = dlerror();
+    return rslt != nullptr ? rslt : "";
+}
+
 #endif
 
 using dl_handle_ptr = std::unique_ptr<dl_handle, dl_handle_deleter>;
@@ -177,6 +190,12 @@ struct wsp_ggml_backend_registry {
 #ifdef WSP_GGML_USE_VULKAN
         register_backend(wsp_ggml_backend_vk_reg());
 #endif
+#ifdef WSP_GGML_USE_WEBGPU
+        register_backend(wsp_ggml_backend_webgpu_reg());
+#endif
+#ifdef WSP_GGML_USE_ZDNN
+        register_backend(wsp_ggml_backend_zdnn_reg());
+#endif
 #ifdef WSP_GGML_USE_OPENCL
         register_backend(wsp_ggml_backend_opencl_reg());
 #endif
@@ -189,9 +208,6 @@ struct wsp_ggml_backend_registry {
 #ifdef WSP_GGML_USE_RPC
         register_backend(wsp_ggml_backend_rpc_reg());
 #endif
-#ifdef WSP_GGML_USE_KOMPUTE
-        register_backend(wsp_ggml_backend_kompute_reg());
-#endif
 #ifdef WSP_GGML_USE_CPU
         register_backend(wsp_ggml_backend_cpu_reg());
 #endif
@@ -233,7 +249,7 @@ struct wsp_ggml_backend_registry {
         dl_handle_ptr handle { dl_load_library(path) };
         if (!handle) {
             if (!silent) {
-                WSP_GGML_LOG_ERROR("%s: failed to load %s\n", __func__, path_str(path).c_str());
+                WSP_GGML_LOG_ERROR("%s: failed to load %s: %s\n", __func__, path_str(path).c_str(), dl_error());
             }
             return nullptr;
         }
@@ -393,9 +409,8 @@ wsp_ggml_backend_t wsp_ggml_backend_init_by_type(enum wsp_ggml_backend_dev_type
 
 wsp_ggml_backend_t wsp_ggml_backend_init_best(void) {
     wsp_ggml_backend_dev_t dev = wsp_ggml_backend_dev_by_type(WSP_GGML_BACKEND_DEVICE_TYPE_GPU);
-    if (!dev) {
-        dev = wsp_ggml_backend_dev_by_type(WSP_GGML_BACKEND_DEVICE_TYPE_CPU);
-    }
+    dev = dev ? dev : wsp_ggml_backend_dev_by_type(WSP_GGML_BACKEND_DEVICE_TYPE_IGPU);
+    dev = dev ? dev : wsp_ggml_backend_dev_by_type(WSP_GGML_BACKEND_DEVICE_TYPE_CPU);
     if (!dev) {
         return nullptr;
     }
@@ -498,6 +513,9 @@ static wsp_ggml_backend_reg_t wsp_ggml_backend_load_best(const char * name, bool
 
     std::vector<fs::path> search_paths;
     if (user_search_path == nullptr) {
+#ifdef WSP_GGML_BACKEND_DIR
+        search_paths.push_back(fs::u8path(WSP_GGML_BACKEND_DIR));
+#endif
         // default search paths: executable directory, current directory
         search_paths.push_back(get_executable_path());
         search_paths.push_back(fs::current_path());
@@ -521,7 +539,7 @@ static wsp_ggml_backend_reg_t wsp_ggml_backend_load_best(const char * name, bool
                 if (filename.native().find(file_prefix) == 0 && ext == file_extension) {
                     dl_handle_ptr handle { dl_load_library(entry) };
                     if (!handle && !silent) {
-                        WSP_GGML_LOG_ERROR("%s: failed to load %s\n", __func__, path_str(entry.path()).c_str());
+                        WSP_GGML_LOG_ERROR("%s: failed to load %s: %s\n", __func__, path_str(entry.path()).c_str(), dl_error());
                     }
                     if (handle) {
                         auto score_fn = (wsp_ggml_backend_score_t) dl_get_sym(handle.get(), "wsp_ggml_backend_score");
@@ -575,7 +593,6 @@ void wsp_ggml_backend_load_all_from_path(const char * dir_path) {
     wsp_ggml_backend_load_best("cann", silent, dir_path);
     wsp_ggml_backend_load_best("cuda", silent, dir_path);
     wsp_ggml_backend_load_best("hip", silent, dir_path);
-    wsp_ggml_backend_load_best("kompute", silent, dir_path);
     wsp_ggml_backend_load_best("metal", silent, dir_path);
     wsp_ggml_backend_load_best("rpc", silent, dir_path);
     wsp_ggml_backend_load_best("sycl", silent, dir_path);