common : remove unused local variables (#16140)

These two local variables 'arg' and 'arg_prefix' have been overriden by:

  1. for (const auto & arg : opt.args)

  2. for (int i = 1; i < argc; i++) {
        const std::string arg_prefix = "--";

        std::string arg = argv[i];

CODEOWNERS

@@ -2,7 +2,10 @@
 
 /ci/ @ggerganov
 /.devops/*.Dockerfile @ngxson
-/tools/server/ @ngxson
+/tools/server/* @ngxson # no subdir
+/tools/server/webui/ @allozaur
+/tools/mtmd/ @ngxson
+/src/llama-chat.* @ngxson
 /ggml/src/ggml-cuda/fattn* @JohannesGaessler
 /ggml/src/ggml-cuda/mmq.* @JohannesGaessler
 /ggml/src/ggml-cuda/mmvq.* @JohannesGaessler

common/arg.cpp

@@ -1068,8 +1068,6 @@ static std::string get_all_kv_cache_types() {
 //
 
 static bool common_params_parse_ex(int argc, char ** argv, common_params_context & ctx_arg) {
-    std::string arg;
-    const std::string arg_prefix = "--";
     common_params & params = ctx_arg.params;
 
     std::unordered_map<std::string, common_arg *> arg_to_options;

ggml/src/ggml-impl.h

@@ -73,7 +73,7 @@ static inline int ggml_up(int n, int m) {
     return (n + m - 1) & ~(m - 1);
 }
 
-// TODO: move to ggml.h?
+// TODO: move to ggml.h? (won't be able to inline)
 static bool ggml_are_same_layout(const struct ggml_tensor * a, const struct ggml_tensor * b) {
     if (a->type != b->type) {
         return false;
@@ -89,6 +89,19 @@ static bool ggml_are_same_layout(const struct ggml_tensor * a, const struct ggml
     return true;
 }
 
+static bool ggml_op_is_empty(enum ggml_op op) {
+    switch (op) {
+        case GGML_OP_NONE:
+        case GGML_OP_RESHAPE:
+        case GGML_OP_TRANSPOSE:
+        case GGML_OP_VIEW:
+        case GGML_OP_PERMUTE:
+            return true;
+        default:
+            return false;
+    }
+}
+
 //
 // logging
 //
@@ -570,27 +583,27 @@ static inline bool ggml_node_has_n_uses(const struct ggml_cgraph * cgraph, int n
     return true;
 }
 
-// Returns true if nodes [i, i+ops.size()) are the sequence of ggml_ops in ops[]
+// Returns true if nodes with indices { node_idxs } are the sequence of ggml_ops in ops[]
 // and are fusable. Nodes are considered fusable according to this function if:
 // - all nodes except the last have only one use and are not views/outputs (see ggml_node_has_N_uses).
 // - all nodes except the last are a src of the following node.
 // - all nodes are the same shape.
 // TODO: Consider allowing GGML_OP_NONE nodes in between
-static inline bool ggml_can_fuse(const struct ggml_cgraph * cgraph, int node_idx, const enum ggml_op * ops, int num_ops) {
-    if (node_idx + num_ops > cgraph->n_nodes) {
-        return false;
-    }
-
+static inline bool ggml_can_fuse_ext(const struct ggml_cgraph * cgraph, const int * node_idxs, const enum ggml_op * ops, int num_ops) {
     for (int i = 0; i < num_ops; ++i) {
-        struct ggml_tensor * node = cgraph->nodes[node_idx + i];
+        if (node_idxs[i] >= cgraph->n_nodes) {
+            return false;
+        }
+
+        struct ggml_tensor * node = cgraph->nodes[node_idxs[i]];
         if (node->op != ops[i]) {
             return false;
         }
-        if (i < num_ops - 1 && !ggml_node_has_n_uses(cgraph, node_idx + i, 1)) {
+        if (i < num_ops - 1 && !ggml_node_has_n_uses(cgraph, node_idxs[i], 1)) {
             return false;
         }
         if (i > 0) {
-            struct ggml_tensor * prev = cgraph->nodes[node_idx + i - 1];
+            struct ggml_tensor * prev = cgraph->nodes[node_idxs[i - 1]];
             if (node->src[0] != prev && node->src[1] != prev) {
                 return false;
             }
@@ -602,6 +615,22 @@ static inline bool ggml_can_fuse(const struct ggml_cgraph * cgraph, int node_idx
     return true;
 }
 
+// same as above, for sequential indices starting at node_idx
+static inline bool ggml_can_fuse(const struct ggml_cgraph * cgraph, int node_idx, const enum ggml_op * ops, int num_ops) {
+    assert(num_ops < 32);
+
+    if (node_idx + num_ops > cgraph->n_nodes) {
+        return false;
+    }
+
+    int idxs[32];
+    for (int i = 0; i < num_ops; ++i) {
+        idxs[i] = node_idx + i;
+    }
+
+    return ggml_can_fuse_ext(cgraph, idxs, ops, num_ops);
+}
+
 #ifdef __cplusplus
 }
 #endif

ggml/src/ggml-metal/ggml-metal-common.cpp

@@ -184,20 +184,6 @@ bool ggml_mem_ranges_check(ggml_mem_ranges_t mrs, const ggml_tensor * tensor) {
     return ggml_mem_ranges_check_dst(mrs, tensor);
 }
 
-// TODO: move to ggml.h?
-static bool is_empty(ggml_op op) {
-    switch (op) {
-        case GGML_OP_NONE:
-        case GGML_OP_RESHAPE:
-        case GGML_OP_TRANSPOSE:
-        case GGML_OP_VIEW:
-        case GGML_OP_PERMUTE:
-            return true;
-        default:
-            return false;
-    }
-}
-
 struct node_info {
     ggml_tensor * node;
 
@@ -212,7 +198,7 @@ struct node_info {
     }
 
     bool is_empty() const {
-        return ::is_empty(node->op);
+        return ggml_op_is_empty(node->op);
     }
 
     void add_fused(ggml_tensor * t) {
@@ -289,7 +275,7 @@ static std::vector<int> ggml_metal_graph_optimize_reorder(const std::vector<node
             case GGML_OP_GET_ROWS:
                 return true;
             default:
-                return is_empty(op);
+                return ggml_op_is_empty(op);
         }
     };