[CANN] Support eager execution mode under ACL graph compilation

Add support for running operators in eager mode while ACL graph
compilation is enabled. This allows bypassing graph execution
and directly submitting ops, which is useful for debugging and
reducing graph build overhead in certain scenarios.

Signed-off-by: noemotiovon <[email protected]>

Author: noemotiovon

docs/backend/CANN.md

@@ -314,3 +314,7 @@ Controls automatic cleanup of the memory pool. This option is only effective whe
 
 Converting the matmul weight format from ND to NZ can significantly improve performance on the 310I DUO NPU.
 
+### GGML_CANN_EAGER_MODE
+
+Enabling eager execution mode will bypass ACL graph execution and submit operators directly.  
+This is useful for debugging or scenarios where graph building overhead is undesirable.

ggml/src/ggml-cann/common.h

@@ -398,6 +398,7 @@ struct ggml_backend_cann_context {
 #endif
     cann_task_queue task_queue;
     bool async_mode;
+    bool eager_mode; // not use acl graph
     // Rope Cache
     ggml_cann_rope_cache rope_cache;
     // Constant Pool
@@ -419,6 +420,10 @@ struct ggml_backend_cann_context {
         async_mode = parse_bool(get_env("GGML_CANN_ASYNC_MODE").value_or(""));
         GGML_LOG_INFO("%s: device %d async operator submission is %s\n", __func__,
             device, async_mode ? "ON" : "OFF");
+        
+        eager_mode = parse_bool(get_env("GGML_CANN_EAGER_MODE").value_or(""));
+        GGML_LOG_INFO("%s: device %d eager execution mode is %s (acl graph disabled)\n",
+              __func__, device, eager_mode ? "ON" : "OFF");
     }
 
     /**

ggml/src/ggml-cann/ggml-cann.cpp

@@ -2252,6 +2252,10 @@ static enum ggml_status ggml_backend_cann_graph_compute(
     bool use_cann_graph = true;
     bool cann_graph_update_required = false;
 
+    if (cann_ctx->eager_mode) {
+        use_cann_graph = false;
+    }
+
     if (use_cann_graph) {
         if (cann_ctx->cann_graph == nullptr) {
             cann_ctx->cann_graph.reset(new ggml_cann_graph());