ggml : fix some compile warnings

ggml-cuda.cu

@@ -7962,6 +7962,15 @@ bool ggml_cuda_compute_forward(struct ggml_compute_params * params, struct ggml_
         return false;
     }
 
+    if (tensor->op == GGML_OP_MUL_MAT) {
+        if (tensor->src[0]->ne[3] != tensor->src[1]->ne[3]) {
+#ifndef NDEBUG
+            fprintf(stderr, "%s: cannot compute %s: src0->ne[3] = %d, src1->ne[3] = %d - fallback to CPU\n", __func__, tensor->name, tensor->src[0]->ne[3], tensor->src[1]->ne[3]);
+#endif
+            return false;
+        }
+    }
+
     switch (tensor->op) {
         case GGML_OP_REPEAT:
             func = ggml_cuda_repeat;

ggml-metal.m

@@ -128,7 +128,7 @@ struct ggml_metal_context {
 // MSL code
 // TODO: move the contents here when ready
 //       for now it is easier to work in a separate file
-static NSString * const msl_library_source = @"see metal.metal";
+//static NSString * const msl_library_source = @"see metal.metal";
 
 // Here to assist with NSBundle Path Hack
 @interface GGMLMetalClass : NSObject
@@ -144,7 +144,8 @@ void ggml_metal_log_set_callback(ggml_log_callback log_callback, void * user_dat
     ggml_metal_log_user_data = user_data;
 }
 
-static void ggml_metal_log(enum ggml_log_level level, const char* format, ...){
+GGML_ATTRIBUTE_FORMAT(2, 3)
+static void ggml_metal_log(enum ggml_log_level level, const char * format, ...){
     if (ggml_metal_log_callback != NULL) {
         va_list args;
         va_start(args, format);
@@ -339,7 +340,7 @@ struct ggml_metal_context * ggml_metal_init(int n_cb) {
     // https://developer.apple.com/metal/Metal-Feature-Set-Tables.pdf
     for (int i = MTLGPUFamilyApple1 + 20; i >= MTLGPUFamilyApple1; --i) {
         if ([ctx->device supportsFamily:i]) {
-            GGML_METAL_LOG_INFO("%s: GPU family: MTLGPUFamilyApple%d (%d)\n", __func__, i - MTLGPUFamilyApple1 + 1, i);
+            GGML_METAL_LOG_INFO("%s: GPU family: MTLGPUFamilyApple%d (%d)\n", __func__, i - (int) MTLGPUFamilyApple1 + 1, i);
             break;
         }
     }
@@ -583,7 +584,7 @@ bool ggml_metal_add_buffer(
                 ctx->device.recommendedMaxWorkingSetSize / 1024.0 / 1024.0);
 
         if (ctx->device.currentAllocatedSize > ctx->device.recommendedMaxWorkingSetSize) {
-            GGML_METAL_LOG_WARN(", warning: current allocated size is greater than the recommended max working set size\n", __func__);
+            GGML_METAL_LOG_WARN("%s: warning: current allocated size is greater than the recommended max working set size\n", __func__);
         } else {
             GGML_METAL_LOG_INFO("\n");
         }

whisper.cpp

@@ -118,7 +118,7 @@ static void byteswap_tensor(ggml_tensor * tensor) {
 //
 
 WHISPER_ATTRIBUTE_FORMAT(2, 3)
-static void whisper_log_internal        (ggml_log_level level, const char* format, ...);
+static void whisper_log_internal        (ggml_log_level level, const char * format, ...);
 static void whisper_log_callback_default(ggml_log_level level, const char * text, void * user_data);
 
 #define WHISPER_LOG_INFO(...)  whisper_log_internal(GGML_LOG_LEVEL_INFO , __VA_ARGS__)
@@ -5848,27 +5848,21 @@ void whisper_log_set(ggml_log_callback log_callback, void * user_data) {
     g_state.log_callback_user_data = user_data;
 }
 
-static void whisper_log_internal_v(ggml_log_level level, const char * format, va_list args) {
-    va_list args_copy;
-    va_copy(args_copy, args);
-    char buffer[128];
-    int len = vsnprintf(buffer, 128, format, args);
-    if (len < 128) {
+GGML_ATTRIBUTE_FORMAT(2, 3)
+static void whisper_log_internal(ggml_log_level level, const char * format, ...) {
+    va_list args;
+    va_start(args, format);
+    char buffer[1024];
+    int len = vsnprintf(buffer, 1024, format, args);
+    if (len < 1024) {
         g_state.log_callback(level, buffer, g_state.log_callback_user_data);
     } else {
         char* buffer2 = new char[len+1];
-        vsnprintf(buffer2, len+1, format, args_copy);
+        vsnprintf(buffer2, len+1, format, args);
         buffer2[len] = 0;
         g_state.log_callback(level, buffer2, g_state.log_callback_user_data);
         delete[] buffer2;
     }
-    va_end(args_copy);
-}
-
-static void whisper_log_internal(ggml_log_level level, const char * format, ...) {
-    va_list args;
-    va_start(args, format);
-    whisper_log_internal_v(level, format, args);
     va_end(args);
 }