Added patch to llvm & works on loading mir files

WORKSPACE

@@ -175,6 +175,8 @@ http_archive(
     sha256 = LLVM_SHA256,
     strip_prefix = "llvm-project-" + LLVM_COMMIT,
     urls = ["https://github.com/llvm/llvm-project/archive/{commit}.zip".format(commit = LLVM_COMMIT)],
+    patches = ["//:mir_parser.patch"],  # Hack to make the MIR parser work
+    patch_args = ["-p1"],
 )
 
 load(

gematria/datasets/bhive_importer_test.cc

@@ -234,5 +234,10 @@ TEST_F(BHiveImporterTest, MIRDatasetBasicTest) {
               IsOk());
 }
 
+TEST_F(BHiveImporterTest, MIRDatasetTest2) {
+  EXPECT_THAT(x86_bhive_importer_->LoadMIRModule("/u9/z277zhu/research/gematria/mir_input/output/native_test.mir"),
+              IsOk());
+}
+
 }  // namespace
 }  // namespace gematria

gematria/datasets/python/BUILD

@@ -53,3 +53,14 @@ gematria_py_binary(
         "//gematria/utils/python:pybind11_abseil_status",
     ],
 )
+
+gematria_py_binary(
+    name = "import_from_mir",
+    srcs = ["import_from_mir.py"],
+    deps = [
+        ":bhive_importer",
+        "//gematria/llvm/python:canonicalizer",
+        "//gematria/llvm/python:llvm_architecture_support",
+        "//gematria/utils/python:pybind11_abseil_status",
+    ],
+)

gematria/datasets/python/import_from_mir.py

@@ -126,9 +126,11 @@ def main(argv: Sequence[str]) -> None:
     for filename in os.listdir(_INPUT_DIR.value):
         if filename.endswith(".mir"):
             mir_file = os.path.join(_INPUT_DIR.value, filename)
-            perf_file = os.path.join(_INPUT_DIR.value, filename.replace(".mir", ".perf"))
+            print("mir file is " + mir_file)
+            perf_file = os.path.join(_INPUT_DIR.value, filename.replace(".mir", ".ll.perf"))
             # load the MIR file
-            module = importer.loadMIRModule(mir_file)
+            module = importer.LoadMIRModule(mir_file)
+            logging.info('Procssing %s file', mir_file)
             # iterate over each line in the corresponding .perf file
             with tf.io.gfile.GFile(perf_file, 'r') as bhive_csv_file:
                 for line in bhive_csv_file:

mir_parser.patch

@@ -0,0 +1,22 @@
+diff --git a/llvm/lib/MC/MCContext.cpp b/llvm/lib/MC/MCContext.cpp
+index a1dde3e1fecb..88d8b11180e1 100644
+--- a/llvm/lib/MC/MCContext.cpp
++++ b/llvm/lib/MC/MCContext.cpp
+@@ -201,8 +201,15 @@ MCInst *MCContext::createMCInst() {
+ MCSymbol *MCContext::getOrCreateSymbol(const Twine &Name) {
+   SmallString<128> NameSV;
+   StringRef NameRef = Name.toStringRef(NameSV);
+-
+-  assert(!NameRef.empty() && "Normal symbols cannot be unnamed!");
++  static int cnt = 0;
++  if (NameRef.empty()){
++    // Append cnt to the end of NameRef
++    std::string ModifiedName = ("INVALID" + std::to_string(cnt));
++    cnt++; // Increment cnt
++
++    // Convert ModifiedName back to StringRef
++    NameRef = StringRef(ModifiedName);
++  }
+
+   MCSymbol *&Sym = Symbols[NameRef];
+   if (!Sym)