add back const to finish, recode is_split_point and is_merge_point

src/include/migraphx/context.hpp

@@ -20,7 +20,7 @@ inline namespace MIGRAPHX_INLINE_NS {
 struct context
 {
     /// Wait for any tasks in the context to complete
-    void finish();
+    void finish() const;
 };
 
 #else
@@ -30,7 +30,7 @@ struct context
  *
  * struct context
  * {
- *      void finish() ;
+ *      void finish() const;
  * };
  *
  */
@@ -92,7 +92,7 @@ struct context
             return private_detail_te_get_handle().type();
     }
 
-    void finish()
+    void finish() const
     {
         assert((*this).private_detail_te_handle_mem_var);
         (*this).private_detail_te_get_handle().finish();
@@ -111,7 +111,7 @@ struct context
         virtual std::shared_ptr<private_detail_te_handle_base_type> clone() const = 0;
         virtual const std::type_info& type() const                                = 0;
 
-        virtual void finish() = 0;
+        virtual void finish() const = 0;
     };
 
     template <typename PrivateDetailTypeErasedT>
@@ -142,7 +142,7 @@ struct context
 
         const std::type_info& type() const override { return typeid(private_detail_te_value); }
 
-        void finish() override { private_detail_te_value.finish(); }
+        void finish() const override { private_detail_te_value.finish(); }
 
         PrivateDetailTypeErasedT private_detail_te_value;
     };

src/opt/dom_info.cpp

@@ -149,28 +149,34 @@ void dom_info::compute_dom(bool reversed)
 
 bool dom_info::is_split_point(instruction_ref ins)
 {
-    std::set<int> stream_set;
+    int stream = -1;
     for(auto&& arg : ins->outputs())
     {
         int arg_stream = arg->get_stream();
-        if(arg_stream >= 0)
-            stream_set.insert(arg_stream);
+        if(arg_stream < 0)
+            continue;
+        if((stream >= 0) && (arg_stream != stream))
+            return true;
+        stream = arg_stream;
     }
-    return (stream_set.size() > 1);
+    return false;
 }
 
 // Identify merge points.  A merge point has more than one
 // inputs that are executed in different streams.
 bool dom_info::is_merge_point(instruction_ref ins)
 {
-    std::set<int> stream_set;
+    int stream = -1;
     for(auto&& arg : ins->inputs())
     {
         int arg_stream = arg->get_stream();
-        if(arg_stream >= 0)
-            stream_set.insert(arg_stream);
+        if(arg_stream < 0)
+            continue;
+        if((stream >= 0) && (arg_stream != stream))
+            return true;
+        stream = arg_stream;
     }
-    return (stream_set.size() > 1);
+    return false;
 }
 
 //  Propagate split points through the graph and identify concurrent instructions.

src/targets/cpu/include/migraphx/cpu/context.hpp

@@ -9,7 +9,7 @@ namespace cpu {
 
 struct context
 {
-    void finish() {}
+    void finish() const {}
 };
 } // namespace cpu
 } // namespace MIGRAPHX_INLINE_NS

src/targets/gpu/include/migraphx/gpu/context.hpp

@@ -148,7 +148,7 @@ struct context
 {
     context(std::size_t n = 0) : current_device(std::make_shared<hip_device>(n)) {}
 
-    hip_device& get_current_device()
+    hip_device& get_current_device() const
     {
         assert(current_device != nullptr);
         return *current_device;
@@ -165,7 +165,7 @@ struct context
     void wait_event(int event) { get_current_device().wait_event(event); }
 
     std::vector<argument> literals{};
-    void finish()
+    void finish() const
     {
         get_current_device().stream_sync();
         gpu_sync();

tools/include/context.hpp

@@ -20,14 +20,14 @@ inline namespace MIGRAPHX_INLINE_NS {
 struct context
 {
     /// Wait for any tasks in the context to complete
-    void finish();
+    void finish() const;
 };
 
 #else
 
 <%
 interface('context',
-    virtual('finish', returns='void'),
+          virtual('finish', returns='void', const=True),
 )
 %>