From 2f6fbd7c8c2897be4ae4ba7e3e61b49e5bfbf7d1 Mon Sep 17 00:00:00 2001
From: Kelvin Choi <kelvin.choi@intel.com>
Date: Wed, 15 Mar 2023 20:56:01 +0900
Subject: [PATCH] [GPU] Need to exclude fused mem_dep from shape_infer_dep

---
 .../src/graph/include/program_node.h          |  4 +++-
 .../intel_gpu/src/graph/primitive_inst.cpp    |  6 +++++-
 .../intel_gpu/src/graph/program_node.cpp      | 20 ++++++++++++++++++-
 3 files changed, 27 insertions(+), 3 deletions(-)

diff --git a/src/plugins/intel_gpu/src/graph/include/program_node.h b/src/plugins/intel_gpu/src/graph/include/program_node.h
index 1f72b49bd8b5d1..845936ed18d0f8 100644
--- a/src/plugins/intel_gpu/src/graph/include/program_node.h
+++ b/src/plugins/intel_gpu/src/graph/include/program_node.h
@@ -99,7 +99,9 @@ struct program_node {
         return false;
     }
 
-    std::map<size_t, memory::ptr> get_const_memory_deps() const;
+    bool is_fused_dep(size_t dep_idx) const;
+
+    std::map<size_t, memory::ptr> get_const_memory_deps(bool exclude_fused_dep = false) const;
 
     virtual std::unique_ptr<kernel_impl_params> get_kernel_impl_params() const {
         return get_kernel_impl_params(get_input_layouts(), output_layouts);
diff --git a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
index 29b97f400be713..ad1e2e61b5b3e5 100644
--- a/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
+++ b/src/plugins/intel_gpu/src/graph/primitive_inst.cpp
@@ -165,7 +165,7 @@ void primitive_inst::update_shape() {
         set_shape_change();
 
     // Even though the predecessors' shapes are not changed, the output shape might be udpated by the mem_dep
-    auto memory_deps = _node->get_const_memory_deps();
+    auto memory_deps = _node->get_const_memory_deps(true);
     for (auto& i : _node->get_shape_infer_dependencies()) {
         if (memory_deps.count(i) > 0) {
             continue;
@@ -190,6 +190,10 @@ void primitive_inst::update_shape() {
         }
         auto& dep = _node->get_dependency(i);
         auto dep_id = dep.id();
+        // exclude fused node from memory_deps
+        if (_node->is_fused_dep(i)) {
+            break;
+        }
         // Events may be not created for in-order queue, so take them for OOO queue only
         if (_network.has_event(dep.id()) && queue_type == QueueTypes::out_of_order) {
             dependencies_events.push_back(_network.get_primitive_event(dep_id));
diff --git a/src/plugins/intel_gpu/src/graph/program_node.cpp b/src/plugins/intel_gpu/src/graph/program_node.cpp
index 140562e8d9354d..7f7508116ad28d 100644
--- a/src/plugins/intel_gpu/src/graph/program_node.cpp
+++ b/src/plugins/intel_gpu/src/graph/program_node.cpp
@@ -378,13 +378,31 @@ bool program_node::has_padded_dependency() const {
     });
 }
 
-std::map<size_t, memory::ptr> program_node::get_const_memory_deps() const {
+bool program_node::is_fused_dep(size_t dep_idx) const {
+    for (auto fused : get_fused_primitives()) {
+        if (dep_idx >= fused.dep_start_idx) {
+            return true;
+        }
+    }
+
+    return false;
+}
+
+std::map<size_t, memory::ptr> program_node::get_const_memory_deps(bool exclude_fused_dep) const {
     std::map<size_t, memory::ptr> mem_deps;
     for (auto& i : get_shape_infer_dependencies()) {
         // Some primitives may have flexible count of deps (e.g. reshape), thus allow skipping some deps
         if (i >= get_dependencies().size())
             continue;
 
+        // exclude fused dependency
+        if (exclude_fused_dep) {
+            if (is_fused_dep(i)) {
+                continue;
+            }
+        }
+
+        // constant type only
         auto& dep = get_dependency(i);
         if (dep.is_type<data>()) {
             mem_deps.insert({i, dep.as<data>().get_attached_memory_ptr()});