added incount and outcount, addressed all of the pull request comments

3 files changed, 29 insertions, 9 deletions

diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index 86a1d45..23c5ad5 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -602,7 +602,7 @@ void ptx_instruction::set_opcode_and_latency()
 	 * [3] MAD
 	 * [4] DIV
 	 */
-	sscanf(opcode_latency_int, "%u,%u,%u,%u,%u,%u,%u",
+	sscanf(opcode_latency_int, "%u,%u,%u,%u,%u",
 			&int_latency[0],&int_latency[1],&int_latency[2],
 			&int_latency[3],&int_latency[4]);
 	sscanf(opcode_latency_fp, "%u,%u,%u,%u,%u",
@@ -850,12 +850,14 @@ void ptx_instruction::pre_decode()
 {
    pc = m_PC;
    isize = m_inst_size;
-   for(unsigned i=0; i<8; i++) {
+   for(unsigned i=0; i<MAX_OUTPUT_VALUES; i++) {
        out[i] = 0;
    }
-   for(unsigned i=0; i<24; i++) {
+   for(unsigned i=0; i<MAX_INPUT_VALUES; i++) {
        in[i] = 0;
    }
+   incount=0;
+   outcount=0;
    is_vectorin = 0;
    is_vectorout = 0;
    std::fill_n(arch_reg.src, MAX_REG_OPERANDS, -1);
@@ -960,6 +962,15 @@ void ptx_instruction::pre_decode()
          }
       }
    }
+   
+   //Setting number of input and output operands which is required for scoreboard check
+   for(int i=0;i<MAX_OUTPUT_VALUES;i++)
+	if(out[i]>0)
+		outcount++;   
+ 
+   for(int i=0;i<MAX_INPUT_VALUES;i++)
+	if(in[i]>0)
+		incount++;   
 
    // Get predicate
    if(has_pred()) {
@@ -1316,7 +1327,10 @@ void ptx_thread_info::ptx_exec_inst( warp_inst_t &inst, unsigned lane_id)
      
       if(((inst_opcode==MMA_OP||inst_opcode==MMA_LD_OP||inst_opcode==MMA_ST_OP))){
       		if(inst.active_count()!=MAX_WARP_SIZE)
-			while(1); 
+		{	
+			printf("Tensor Core operation are warp synchronous operation. All the threads needs to be active.");
+			assert(0);
+		}
       }
 
       if(((inst_opcode!=MMA_OP)&&(inst_opcode!=MMA_LD_OP)&&(inst_opcode!=MMA_ST_OP))||((inst_opcode==MMA_OP||inst_opcode==MMA_LD_OP||inst_opcode==MMA_ST_OP)&&(lane_id==0))){
diff --git a/src/cuda-sim/instructions.cc b/src/cuda-sim/instructions.cc
index 2677c40..06cad54 100644
--- a/src/cuda-sim/instructions.cc
+++ b/src/cuda-sim/instructions.cc
@@ -1726,7 +1726,7 @@ void mma_impl( const ptx_instruction *pI, core_t *core, warp_inst_t inst )
 	unsigned b_layout = pI->get_wmma_layout(1);
 	unsigned type = pI->get_type();
 	unsigned type2 = pI->get_type2();
-	int tid = inst.warp_id_func() * core->get_warp_size();
+	int tid = inst.warp_id() * core->get_warp_size();
 	const operand_info &dst = pI->operand_lookup(0);
 	
 	unsigned thread_group_index;
@@ -2962,7 +2962,7 @@ void mma_st_impl( const ptx_instruction *pI, core_t *core, warp_inst_t &inst )
    const operand_info &src  = pI->operand_lookup(1);
    const operand_info &src1 = pI->operand_lookup(0);
    const operand_info &src2 = pI->operand_lookup(2);
-   int tid = inst.warp_id_func()*core->get_warp_size();
+   int tid = inst.warp_id()*core->get_warp_size();
    unsigned type = pI->get_type();
    unsigned wmma_type = pI->get_wmma_type();
    unsigned wmma_layout = pI->get_wmma_layout(0);
@@ -3069,7 +3069,7 @@ void mma_ld_impl( const ptx_instruction *pI, core_t *core, warp_inst_t &inst )
    unsigned type = pI->get_type();
    unsigned wmma_type = pI->get_wmma_type();
    unsigned wmma_layout = pI->get_wmma_layout(0);
-   int tid = inst.warp_id_func()*core->get_warp_size();
+   int tid = inst.warp_id()*core->get_warp_size();
    int thrd,stride;
    ptx_thread_info *thread;
    _memory_op_t insn_memory_op = pI->has_memory_read() ? memory_load : memory_store;
@@ -4469,7 +4469,13 @@ void set_impl( const ptx_instruction *pI, ptx_thread_info *thread )
 void shfl_impl( const ptx_instruction *pI, core_t *core, warp_inst_t inst )
 {
 	unsigned i_type = pI->get_type();
-	int tid = inst.warp_id_func() * core->get_warp_size();
+  	int tid;
+
+  	if(core->get_gpu()->is_functional_sim())
+    		tid = inst.warp_id_func() * core->get_warp_size();
+  	else
+	 	tid = inst.warp_id() * core->get_warp_size();
+	
 	ptx_thread_info *thread = core->get_thread_info()[tid];
 	ptx_warp_info *warp_info = thread->m_warp_info;
 	int lane = warp_info->get_done_threads();
diff --git a/src/cuda-sim/ptx_parser.cc b/src/cuda-sim/ptx_parser.cc
index d5324d0..9671ab7 100644
--- a/src/cuda-sim/ptx_parser.cc
+++ b/src/cuda-sim/ptx_parser.cc
@@ -39,7 +39,7 @@ void set_ptx_warp_size(const struct core_config * warp_size)
    g_shader_core_config=warp_size;
 }
 
-static bool g_debug_ir_generation=true;
+static bool g_debug_ir_generation=false;
 const char *g_filename;
 unsigned g_max_regs_per_thread = 0;