Merge pull request #28 from jwang323/cdp_clean

Initial support of CUDA Dynamic Parallelism on GPGPUSim

1 files changed, 16 insertions, 1 deletions

diff --git a/src/cuda-sim/ptx_ir.h b/src/cuda-sim/ptx_ir.h
index 601a13d..55b01fd 100644
--- a/src/cuda-sim/ptx_ir.h
+++ b/src/cuda-sim/ptx_ir.h
@@ -291,6 +291,7 @@ private:
 
    std::list<operand_info> m_initializer;
    static unsigned sm_next_uid;
+   
 };
 
 class symbol_table {
@@ -330,6 +331,11 @@ public:
    iterator const_iterator_end() { return m_consts.end();}
 
    void dump();
+   
+   //Jin: handle instruction group for cdp
+   symbol_table* start_inst_group();
+   symbol_table* end_inst_group();
+
 private:
    unsigned m_reg_allocator;
    unsigned m_shared_next;
@@ -347,6 +353,10 @@ private:
    std::list<symbol*> m_consts;
    std::map<std::string,function_info*> m_function_info_lookup;
    std::map<std::string,symbol_table*> m_function_symtab_lookup;
+   
+   //Jin: handle instruction group for cdp
+   unsigned m_inst_group_id;
+   std::map<std::string,symbol_table*> m_inst_group_symtab;
 };
 
 class operand_info {
@@ -690,7 +700,7 @@ public:
    }
 
    bool is_immediate_address() const {
-	   return   m_immediate_address;
+       return   m_immediate_address;
    }
 
    bool is_literal() const { return m_type == int_t ||
@@ -1200,6 +1210,8 @@ public:
    {
       return m_args.size();
    }
+   unsigned get_args_aligned_size();
+
    const symbol* get_arg( unsigned n ) const
    {
       assert( n < m_args.size() );
@@ -1288,6 +1300,9 @@ private:
 
    static std::vector<ptx_instruction*> s_g_pc_to_insn; // a direct mapping from PC to instruction
    static unsigned sm_next_uid;
+
+   //parameter size for device kernels
+   int m_args_aligned_size;
 };
 
 class arg_buffer_t {