A thread executing BSMAD is now able to access information from all threads in its warp

1 files changed, 14 insertions, 0 deletions

diff --git a/src/cuda-sim/cuda-sim.cc b/src/cuda-sim/cuda-sim.cc
index e194a2a..059fbe2 100644
--- a/src/cuda-sim/cuda-sim.cc
+++ b/src/cuda-sim/cuda-sim.cc
@@ -849,8 +849,10 @@ void ptx_instruction::pre_decode()
 
    switch ( get_opcode() ) {
 #define OP_DEF(OP,FUNC,STR,DST,CLASSIFICATION) case OP: has_dst = (DST!=0); break;
+#define OP_W_DEF(OP,FUNC,STR,DST,CLASSIFICATION) case OP: has_dst = (DST!=0); break;
 #include "opcodes.def"
 #undef OP_DEF
+#undef OP_W_DEF
    default:
       printf( "Execution error: Invalid opcode (0x%x)\n", get_opcode() );
       break;
@@ -1242,10 +1244,22 @@ void ptx_thread_info::ptx_exec_inst( warp_inst_t &inst, unsigned lane_id)
          *((warp_inst_t*)pJ) = inst; // copy active mask information
          pI = pJ;
       }
+      /*const ptx_instruction **pA;
+      if( pI->get_opcode() == BSMAD_OP ) {
+    	  //pA = (const ptx_instruction**)malloc(get_core()->get_warp_size()*(sizeof(ptx_instruction*)));
+    	  pA = (const ptx_instruction**)malloc(8*(sizeof(ptx_instruction*)));
+    	  for (int i = 0; i < get_core()->get_warp_size() && inst.active(i); i++) {
+    		  //pA[i] = get_core()->get_thread_info()[inst.warp_id() * get_core()->get_warp_size() + i]->func_info()->get_instruction(pc+(i-lane_id)*(pI->inst_size()));
+    		  int tid = inst.warp_id() * get_core()->get_warp_size() + i;
+    		  pA[i] = get_core()->get_thread_info()[tid]->func_info()->get_instruction(pc);
+    	  }
+      }*/
       switch ( pI->get_opcode() ) {
 #define OP_DEF(OP,FUNC,STR,DST,CLASSIFICATION) case OP: FUNC(pI,this); op_classification = CLASSIFICATION; break;
+#define OP_W_DEF(OP,FUNC,STR,DST,CLASSIFICATION) case OP: FUNC(pI,get_core(),inst); op_classification = CLASSIFICATION; break;
 #include "opcodes.def"
 #undef OP_DEF
+#undef OP_W_DEF
       default: printf( "Execution error: Invalid opcode (0x%x)\n", pI->get_opcode() ); break;
       }
       delete pJ;