Improve graph build time

ggml.c

@@ -15742,6 +15742,34 @@ static void ggml_compute_backward(struct ggml_context * ctx, struct ggml_tensor
     }
 }
 
+static_assert(GGML_GRAPH_HASHTABLE_SIZE > GGML_MAX_NODES * 2, "GGML_GRAPH_HT_SIZE is too small");
+
+static size_t hash(void * p) {
+    return (size_t)p % GGML_GRAPH_HASHTABLE_SIZE;
+}
+
+static bool hash_insert(void * hash_table[], void * p) {
+    size_t h = hash(p);
+
+    // linear probing
+    size_t i = h;
+    while (hash_table[i] != NULL && hash_table[i] != p) {
+        i = (i + 1) % GGML_GRAPH_HASHTABLE_SIZE;
+        if (i == h) {
+            // hash table is full
+            GGML_ASSERT(false);
+        }
+    }
+
+    if (hash_table[i] == p) {
+        return true;
+    }
+
+    // insert
+    hash_table[i] = p;
+    return false;
+}
+
 static void ggml_visit_parents(struct ggml_cgraph * cgraph, struct ggml_tensor * node) {
     if (node->grad == NULL) {
         // this usually happens when we generate intermediate nodes from constants in the backward pass
@@ -15752,16 +15780,8 @@ static void ggml_visit_parents(struct ggml_cgraph * cgraph, struct ggml_tensor *
     }
 
     // check if already visited
-    for (int i = 0; i < cgraph->n_nodes; i++) {
-        if (cgraph->nodes[i] == node) {
-            return;
-        }
-    }
-
-    for (int i = 0; i < cgraph->n_leafs; i++) {
-        if (cgraph->leafs[i] == node) {
-            return;
-        }
+    if (hash_insert(cgraph->visited_hash_table, node)) {
+        return;
     }
 
     for (int i = 0; i < GGML_MAX_SRC; ++i) {
@@ -15824,6 +15844,7 @@ struct ggml_cgraph ggml_build_forward(struct ggml_tensor * tensor) {
         /*.nodes        =*/ { NULL },
         /*.grads        =*/ { NULL },
         /*.leafs        =*/ { NULL },
+        /*.hash_table   =*/ { NULL },
         /*.perf_runs    =*/ 0,
         /*.perf_cycles  =*/ 0,
         /*.perf_time_us =*/ 0,
@@ -15865,7 +15886,7 @@ struct ggml_cgraph ggml_build_backward(struct ggml_context * ctx, struct ggml_cg
 
         if (node->is_param) {
             GGML_PRINT_DEBUG("%s: found root node %p\n", __func__, (void *) node);
-            ggml_build_forward_impl(&result, node->grad, true);
+            ggml_build_forward_expand(&result, node->grad);
         }
     }
 

ggml.h

@@ -438,7 +438,7 @@ extern "C" {
 
         void * extra; // extra things e.g. for ggml-cuda.cu
 
-        char padding[8];
+        char padding[4];
     };
 
     static const size_t GGML_TENSOR_SIZE = sizeof(struct ggml_tensor);
@@ -459,6 +459,11 @@ extern "C" {
         void * abort_callback_data;
     };
 
+    // next prime after GGML_MAX_NODES
+    // #define GGML_GRAPH_HASHTABLE_SIZE 4099
+    // next prime after GGML_MAX_NODES * 2 (nodes + leafs)
+    #define GGML_GRAPH_HASHTABLE_SIZE 8273
+
     // computation graph
     struct ggml_cgraph {
         int n_nodes;
@@ -468,6 +473,8 @@ extern "C" {
         struct ggml_tensor * grads[GGML_MAX_NODES];
         struct ggml_tensor * leafs[GGML_MAX_NODES];
 
+        void * visited_hash_table[GGML_GRAPH_HASHTABLE_SIZE];
+
         // performance
         int     perf_runs;
         int64_t perf_cycles;

llama.cpp

@@ -1704,6 +1704,8 @@ static bool llama_eval_internal(
     // run the computation
     ggml_build_forward_expand(&gf, cur);
 
+    // fprintf(stderr, "graph build time: %.3f ms (%d nodes, %d leafs)\n", (ggml_time_us() - t_start_us)/1000.0, gf.n_nodes, gf.n_leafs);
+
 #if GGML_USE_MPI
     ggml_mpi_graph_compute_pre(lctx.ctx_mpi, &gf, n_layer);
 #endif