Merge pull request #25 from echoedit/dev

Dev

1 files changed, 26 insertions, 4 deletions

diff --git a/libcuda/gpgpu_context.h b/libcuda/gpgpu_context.h
index 3c9f87c..45c5cdd 100644
--- a/libcuda/gpgpu_context.h
+++ b/libcuda/gpgpu_context.h
@@ -6,32 +6,51 @@
 #include "../src/gpgpusim_entrypoint.h"
 #include "../src/cuda-sim/cuda-sim.h"
 #include "../src/cuda-sim/cuda_device_runtime.h"
+#include "../src/cuda-sim/ptx-stats.h"
 
 class gpgpu_context {
     public:
 	gpgpu_context() {
 	    g_global_allfiles_symbol_table = NULL;
+	    sm_next_access_uid=0;
+	    warp_inst_sm_next_uid=0;
+	    operand_info_sm_next_uid = 1;
+	    kernel_info_m_next_uid = 1;
+	    g_num_ptx_inst_uid = 0;
+	    g_ptx_cta_info_uid = 1;
+	    symbol_sm_next_uid = 1;
+	    function_info_sm_next_uid = 1;
+	    debug_tensorcore = 0;
 	    api = new cuda_runtime_api(this);
 	    ptxinfo = new ptxinfo_data(this);
 	    ptx_parser = new ptx_recognizer(this);
 	    the_gpgpusim = new GPGPUsim_ctx(this);
 	    func_sim = new cuda_sim(this);
-#if (CUDART_VERSION >= 5000)
 	    device_runtime = new cuda_device_runtime(this);
-#endif
+	    stats = new ptx_stats(this);
 	}
 	// global list
 	symbol_table *g_global_allfiles_symbol_table;
 	const char *g_filename;
+	unsigned sm_next_access_uid;
+	unsigned warp_inst_sm_next_uid;
+	unsigned operand_info_sm_next_uid;//uid for operand_info
+	unsigned kernel_info_m_next_uid;//uid for kernel_info_t
+	unsigned g_num_ptx_inst_uid; //uid for ptx inst inside ptx_instruction
+	unsigned long long g_ptx_cta_info_uid;
+	unsigned symbol_sm_next_uid; //uid for symbol
+	unsigned function_info_sm_next_uid;
+	std::vector<ptx_instruction*> s_g_pc_to_insn; // a direct mapping from PC to instruction
+	bool debug_tensorcore;
+
 	// objects pointers for each file
 	cuda_runtime_api* api;
 	ptxinfo_data* ptxinfo;
 	ptx_recognizer* ptx_parser;
 	GPGPUsim_ctx* the_gpgpusim;
 	cuda_sim* func_sim;
-#if (CUDART_VERSION >= 5000)
 	cuda_device_runtime* device_runtime;
-#endif
+	ptx_stats* stats;
 	// member function list
 	void cuobjdumpParseBinary(unsigned int handle);
 	class symbol_table *gpgpu_ptx_sim_load_ptx_from_string( const char *p, unsigned source_num );
@@ -43,6 +62,9 @@ class gpgpu_context {
 	class gpgpu_sim *gpgpu_ptx_sim_init_perf();
 	struct _cuda_device_id *GPGPUSim_Init();
 	void ptx_reg_options(option_parser_t opp);
+	const ptx_instruction* pc_to_instruction(unsigned pc); 
+	const warp_inst_t *ptx_fetch_inst( address_type pc );
+	unsigned translate_pc_to_ptxlineno(unsigned pc);
 };
 gpgpu_context* GPGPU_Context();