ggml_tensor : use 1 bit per flag

ggml.c

@@ -4592,9 +4592,9 @@ struct ggml_tensor * ggml_new_tensor_impl(
         /*.op           =*/ GGML_OP_NONE,
         /*.op_params    =*/ {0},
         /*.is_param     =*/ false,
+        /*.visited      =*/ false,
         /*.grad         =*/ NULL,
         /*.src          =*/ { NULL },
-        /*.visited      =*/ false,
         /*.perf_runs    =*/ 0,
         /*.perf_cycles  =*/ 0,
         /*.perf_time_us =*/ 0,

ggml.h

@@ -422,13 +422,12 @@ extern "C" {
         // op params - allocated as int32_t for alignment
         int32_t op_params[GGML_MAX_OP_PARAMS / sizeof(uint32_t)];
 
-        bool is_param;
+        uint32_t is_param:1;
+        uint32_t visited:1;  // used to build graphs
 
         struct ggml_tensor * grad;
         struct ggml_tensor * src[GGML_MAX_SRC];
 
-        bool visited;   // used to build graphs
-
         // performance
         int     perf_runs;
         int64_t perf_cycles;