rpc: fix register position

Signed-off-by: thxCode <[email protected]>

Author: thxCode

common/arg.cpp

@@ -403,7 +403,7 @@ static void add_rpc_devices(std::string servers) {
     for (const auto & server : rpc_servers) {
         ggml_backend_dev_t dev = ggml_backend_rpc_add_device_fn(server.c_str());
         if (dev) {
-            ggml_backend_device_register(dev);
+            ggml_backend_device_register(dev, true);
         } else {
             throw std::invalid_argument("failed to register RPC device");
         }

ggml/include/ggml-backend.h

@@ -203,7 +203,7 @@ extern "C" {
     // Backend registry
     //
 
-    GGML_API void ggml_backend_device_register(ggml_backend_dev_t device);
+    GGML_API void ggml_backend_device_register(ggml_backend_dev_t device, bool front);
 
     // Backend (reg) enumeration
     GGML_API size_t             ggml_backend_reg_count(void);

ggml/src/ggml-backend-reg.cpp

@@ -210,11 +210,15 @@ struct ggml_backend_registry {
         }
     }
 
-    void register_device(ggml_backend_dev_t device) {
+    void register_device(ggml_backend_dev_t device, bool front = false) {
 #ifndef NDEBUG
         GGML_LOG_DEBUG("%s: registered device %s (%s)\n", __func__, ggml_backend_dev_name(device), ggml_backend_dev_description(device));
 #endif
-        devices.push_back(device);
+        if (front) {
+            devices.insert(devices.begin(), device);
+        } else {
+            devices.push_back(device);
+        }
     }
 
     ggml_backend_reg_t load_backend(const std::wstring & path, bool silent) {
@@ -298,8 +302,8 @@ void ggml_backend_register(ggml_backend_reg_t reg) {
     get_reg().register_backend(reg);
 }
 
-void ggml_backend_device_register(ggml_backend_dev_t device) {
-    get_reg().register_device(device);
+void ggml_backend_device_register(ggml_backend_dev_t device, bool front) {
+    get_reg().register_device(device, front);
 }
 
 // Backend (reg) enumeration

src/llama-model.cpp

@@ -1303,10 +1303,12 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
     const int act_gpu_layers = devices.empty() ? 0 : std::min(n_gpu_layers, (int)n_layer + 1);
     auto get_layer_buft_list = [&](int il) -> llama_model::impl::layer_dev {
         if (il < i_gpu_start || (il - i_gpu_start) >= act_gpu_layers) {
+            LLAMA_LOG_DEBUG("load_tensors: layer %3d assigned to device %s\n", il, ggml_backend_dev_name(cpu_dev));
             return {cpu_dev, &pimpl->cpu_buft_list};
         }
         const int layer_gpu = std::upper_bound(splits.begin(), splits.begin() + n_devices(), float(il - i_gpu_start)/act_gpu_layers) - splits.begin();
         auto * dev = devices.at(layer_gpu);
+        LLAMA_LOG_DEBUG("load_tensors: layer %3d assigned to device %s\n", il, ggml_backend_dev_name(dev));
         return {dev, &pimpl->gpu_buft_list.at(dev)};
     };