[CODEGEN] Add a pass to prefetch operands of dot if applicable. (#105)

* update membar pass when data is double buffered

* Add instruction prefetch_s

* prefetch tests pass (except the 1 warp case)

* Fix the 1-warp bug

* Add back prefetch files

* Disable prefetch on a100

* Always add war barrier on sm>=80

lib/ir/builder.cc

@@ -381,6 +381,9 @@ value *builder::create_async_wait(int N) {
   return insert(async_wait_inst::create(ctx_, N));
 }
 
+value *builder::create_prefetch_s(value *arg, int inc) {
+  return insert(prefetch_s_inst::create(ctx_, arg, inc));
+}
 
 
 }

lib/ir/instructions.cc

@@ -832,6 +832,10 @@ async_wait_inst* async_wait_inst::create(context &ctx, int N, const std::string
   return new async_wait_inst(ctx, N, name, next);
 }
 
+// prefetch_s
+prefetch_s_inst *prefetch_s_inst::create(context &ctx, value *arg, int inc, const std::string &name, instruction *next) {
+  return new prefetch_s_inst(ctx, arg, inc, name, next);
+}
 
 //// nv_dynamic_program_idx
 //make_range_dyn::make_range_dyn(type *ty, const std::string &name, instruction *next)