[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
2021-05-13 10:42:18 +08:00
parent 147675923e
commit 967e629c0c
14 changed files with 408 additions and 42 deletions
--- a/lib/driver/module.cc
+++ b/lib/driver/module.cc
@@ -209,7 +209,8 @@ static std::map<int, int> vptx = {
  {10020, 65},
  {11000, 70},
  {11010, 71},
-  {11020, 72}
+  {11020, 72},
+  {11030, 73},
 };

 std::string cu_module::compile_llvm_module(llvm::Module* module, driver::device* device) {