metal : print node names for debugging (#17882)

PR #17091 set the VERSION of various libraries to 0.0.abcd, where abcd
is the LLAMA_BUILD_NUMBER. That build number is too large to fit in the
Mach-O 'current version' field's 'micro' part, which only goes up to
255. This just sets the Mach-O current version to 0 to get it building
properly again.

Fixes #17258.

ggml/src/ggml-alloc.c

@@ -25,6 +25,7 @@ static bool ggml_is_view(const struct ggml_tensor * t) {
 // ops that return true for this function must not use restrict pointers for their backend implementations
 bool ggml_op_can_inplace(enum ggml_op op) {
     switch (op) {
+        case GGML_OP_FILL:
         case GGML_OP_SCALE:
         case GGML_OP_DIAG_MASK_ZERO:
         case GGML_OP_DIAG_MASK_INF:

ggml/src/ggml-cuda/fill.cu

@@ -4,7 +4,7 @@
 #define CUDA_FILL_BLOCK_SIZE 256
 
 template <typename T>
-static __global__ void fill_kernel(T * __restrict__ dst, const int64_t k, const T value) {
+static __global__ void fill_kernel(T * dst, const int64_t k, const T value) {
     const int64_t i = (int64_t)blockDim.x * blockIdx.x + threadIdx.x;
     if (i >= k) {
         return;

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -221,7 +221,7 @@ static int ggml_metal_op_encode_impl(ggml_metal_op_t ctx, int idx) {
         }
 
         if (ctx->debug_graph > 0) {
-            GGML_LOG_DEBUG("%s: node[%5d] - %-12s %s\n", __func__, idx, ggml_op_name(node->op), is_concurrent ? "(concurrent)" : "");
+            GGML_LOG_DEBUG("%s: node[%5d] - %-12s %-12s %s\n", __func__, idx, ggml_op_name(node->op), ggml_get_name(node), is_concurrent ? "(concurrent)" : "");
         }
         if (ctx->debug_graph > 1) {
             GGML_TENSOR_LOCALS( int64_t, ne0, node->src[0], ne);

src/CMakeLists.txt

@@ -139,6 +139,7 @@ add_library(llama
 set_target_properties(llama PROPERTIES
     VERSION ${LLAMA_INSTALL_VERSION}
     SOVERSION 0
+    MACHO_CURRENT_VERSION 0 # keep macOS linker from seeing oversized version number
 )
 
 target_include_directories(llama PRIVATE .)

tools/mtmd/CMakeLists.txt

@@ -16,6 +16,7 @@ add_library(mtmd
 set_target_properties(mtmd PROPERTIES
     VERSION ${LLAMA_INSTALL_VERSION}
     SOVERSION 0
+    MACHO_CURRENT_VERSION 0 # keep macOS linker from seeing oversized version number
 )
 
 target_link_libraries     (mtmd PUBLIC ggml llama)