refactored more functions and fixed bugs

Author: khoing0810

llvm_ir_dataset_utils/tools/opt_analysis_tools.py

@@ -290,13 +290,7 @@ def parse_pass_analysis_exec(
         else:
             data = read_data(output_file_path, bitcode_file, opt).split("\n")
 
-        if relative:  # include both relative and absolute
-            result = {
-                "pass-exec": None,
-                "analysis-exec": None,
-            }
-        else:
-            result = {"pass-exec": None, "analysis-exec": None}
+        result = {"pass-exec": None, "analysis-exec": None}
 
         line_start = find_start_line(data)
         if line_start is None:
@@ -334,7 +328,7 @@ def parse_pass_analysis_exec(
 
 def read_data(file_path: str, bitcode_file: bool, opt: str):
     command = [
-        "/p/lustre1/khoidng/LLVM/build/bin/opt",
+        "opt",
         "-" + opt,
         "--stats",
         "--disable-output",
@@ -359,13 +353,12 @@ def read_data(file_path: str, bitcode_file: bool, opt: str):
 
 def read_data_bc(bitcode_module, opt: str):
     command = [
-        "/p/lustre1/khoidng/LLVM/build/bin/opt",
+        "opt",
         "-" + opt,
         "--stats",
         "--disable-output",
         "--time-passes",
-    ]  # TODO: replace the hardcoded opt path with something more flexible
-
+    ]
     with subprocess.Popen(
         command,
         stdout=subprocess.PIPE,
@@ -394,7 +387,6 @@ def sampling(
     bitcode_file: bool = False,
     opt: str = "O3",
 ):
-    # available_pass_list = import_pass_from_file('./opt_passes.txt')
     wall_time = {k: [] for k in OPT_O3_PASS_LIST}
     passes = set(OPT_O3_PASS_LIST)
     files = listdir(dir_path)
@@ -459,6 +451,19 @@ def sampling_csv(
     return result
 
 
+"""
+source_dir: source directory of bitcode files
+fp: name for output csv file
+nsamples: number of files to sample
+ncols: number of columns for output file table
+relative (deprecated, don't use): output whether data table is relative time or absolute time
+col_labels: list of labels for the output data table (recommend: []abs_time, rel_time, pass])
+bitcode_file: whether type of files in source_dir bitcode file
+opt: optimization pipeline
+data_type: output transformation pass ('pass-exec') or analysis pass ('analysis-exec') data
+"""
+
+
 def sample_then_export_csv(
     source_dir: str,
     fp: str,
@@ -475,27 +480,6 @@ def sample_then_export_csv(
     return export_to_csv(o, fp, ncols, col_labels)
 
 
-def export_pass_name(pass_collection: list[str], fp, append=True):
-    # append_cond = lambda append: 'a' if append else 'w'
-    def append_cond(append):
-        if append:
-            return "a"
-        return "w"
-
-    with open(fp, append_cond) as f:
-        for i in pass_collection:
-            f.write(i)
-            f.write("\n")
-    return 0
-
-
-def import_pass_from_file(fp, delimeter="\n"):
-    data = None
-    with open(fp, "r") as f:
-        data = f.read().split(delimeter)
-    return data
-
-
 def export_to_json(data: Union[str, list[float]], fn: str = "", indent: int = 2):
     out = json.dumps(data, indent=indent)
     name = fn if fn != "" else "json_file.json"
@@ -659,7 +643,7 @@ def cat_encode(data: Union[str, list[float]]):
     return (result, encoding)
 
 
-def download_bitcode(target_dir: str, languages: list[str], n: int = -1):
+def download_bitcode(target_dir: str, languages: list[str], n: int = -1, random=False):
     if not isinstance(languages, list):
         print("languages arg has to be list type")
         return 1
@@ -685,6 +669,7 @@ def download_bitcode(target_dir: str, languages: list[str], n: int = -1):
     print("converted to python iterable")
     counter = 0
     spin_text = "Downloading bitcode files"
+
     with yaspin(text=spin_text) as sp:
         while (n != -1 and counter < n) or (row is not None and n == -1):
             if row["language"] in lang_set:

llvm_ir_dataset_utils/tools/outlier_analysis.py

@@ -3,7 +3,7 @@
 import os
 from os import listdir
 from os.path import join
-from typing import List, Union, Callable
+from typing import List
 import pandas as pd
 import parallelbar
 import ray
@@ -17,6 +17,13 @@
 
 num_cpus = psutil.cpu_count(logical=False)
 
+opt_load_args = [
+    "opt",
+    "-load",
+    "RemoveFunctionBodyPass/build/libRemoveFunctionBody.so",
+    "-load-pass-plugin=RemoveFunctionBodyPass/build/libRemoveFunctionBody.so",
+    "-passes=remove-fn-body",
+]
 
 """
     inputs:
@@ -30,9 +37,9 @@ def remove_fn(i: int, src: str, dst: str):
     if not (os.path.isfile(src) or os.path.exists(os.path.dirname(dst))):
         print("invalid file path in either src or dst argument")
         return None
-    command = [
-        "/p/lustre1/khoidng/LLVM/build/bin/opt",  # TODO: replace this
-        f"-passes=remove-fn-body<i={i}>",
+    command = opt_load_args + [
+        "-index",
+        f"{i}",
         src,
         "-o",
         dst,
@@ -55,15 +62,16 @@ def remove_fn(i: int, src: str, dst: str):
 
 
 def remove_fn_bc(i: int, bc):
-    if i == -1:
-        print("Currently not supporting i == -1")
+    if i < 0:
+        print("No negative index!")
         return None
     try:
         with subprocess.Popen(
-            [
-                "/p/lustre1/khoidng/LLVM/build/bin/opt",
-                f"-passes=remove-fn-body<i={i}>",
-            ],  # TODO: replace this
+            opt_load_args
+            + [
+                "-index",
+                f"{i}",
+            ],
             stdout=subprocess.PIPE,
             stderr=subprocess.STDOUT,
             stdin=subprocess.PIPE,
@@ -85,9 +93,10 @@ def get_n_functions(file_path: str):
         bc = f.read()
     try:
         with subprocess.Popen(
-            [
-                "/p/lustre1/khoidng/LLVM/build/bin/opt",  # TODO: replace this
-                "-passes=remove-fn-body<i=-1>",
+            opt_load_args
+            + [
+                "-index",
+                "-1",
                 "--disable-output",
             ],
             stdout=subprocess.PIPE,
@@ -108,9 +117,10 @@ def get_n_functions(file_path: str):
 
 def get_ir(bitcode_module):
     with subprocess.Popen(
-        [
-            "/p/lustre1/khoidng/LLVM/build/bin/opt",  # TODO: replace this
-            "-passes=remove-fn-body<i=-1>",
+        opt_load_args
+        + [
+            "-index",
+            "-1",
             "-S",
         ],
         stdout=subprocess.PIPE,
@@ -121,9 +131,12 @@ def get_ir(bitcode_module):
 
 
 """
-- outlier_check_fn: must be a callable function. Requires at least time analysis data and ref_data as arguments.
-                    must return True if outlier, and False otherwise.
-- Return: number of functions removed, total number of functions, average fraction of passes being outliers if ith function is removed
+Non-optimized version of get outliers.
+Inputs:
+- file_path: path to bitcode file
+- opt: O1, O2, O3, Oz,... optimization options (case-sensitive)
+- outlier_threshold: ratio of number of outlier passes over total number of passes 
+- quantile: percentile of passes given runtime data to be considered outlier for every pass
 """
 
 
@@ -136,9 +149,11 @@ def get_outliers(file_path: str, opt: str, outlier_threshold=1, quantile=0.95):
     os.makedirs(os.path.dirname(tmp_path), exist_ok=True)
 
     ref_data = pd.read_csv(
-        f"{opt.lower()}_cpp.csv"
+        f"pass_runtime/transformations/{opt.lower()}_cpp.csv"
     )  # TODO: filename needs to be abstracted away!
-
+    groups = ref_data.groupby("pass").quantile(
+        q=quantile
+    )  # TODO: put this outside of the loop
     bc = None
     n_removed = 0
     src_path = file_path
@@ -155,11 +170,6 @@ def get_outliers(file_path: str, opt: str, outlier_threshold=1, quantile=0.95):
         # this ignores the src_path and only takes the bitcode module
         time_analysis_tmp = parse_pass_analysis_exec(src_path, True, True, opt, tmp)
 
-        # pd.DataFrame consisting of quantiles grouped by passes
-        groups = ref_data.groupby("pass").quantile(
-            q=quantile
-        )  # TODO: put this outside of the loop
-
         # if removed function still results in module being in outlier range,
         # delete that function because it doesn't affect the outliers
         # (minimize functions in module such that outliers are preserved)
@@ -201,8 +211,6 @@ def check_outliers(time, pass_name, ref_data, time_col="fraction_total_time"):
 def is_outlier(
     data,
     ref_data: pd.DataFrame,
-    quantile=0.95,
-    pass_col="pass",
     time_col="fraction_total_time",
     threshold=1,
 ):
@@ -252,7 +260,7 @@ def preserve_outliers_dir(dir_path: str, opt: str, outlier_threshold=1):
 outlier functions extracted.
 
 Inputs:
-- file_path: path to bitcode module file.
+- file_path: absolute path to bitcode module file.
 - opt: optimization pipeline in {O1,O2,O3,Oz}.
 - pass_name: pass name in which outlier threshold used to extract
   outlier functions.
@@ -287,7 +295,9 @@ def get_outliers_pass_specific(
     os.makedirs(os.path.dirname(tmp_path), exist_ok=True)
 
     if ref_data is None:
-        ref_data = pd.read_csv(f"{opt.lower()}_cpp.csv")  # TODO: optimize this away
+        ref_data = pd.read_csv(
+            f"pass_runtime/transformations/{opt.lower()}_cpp.csv"
+        )  # TODO: abstract this away. currently hard coding transformations pass data
 
     bc = None
     n_removed = 0
@@ -388,7 +398,9 @@ def preserve_outliers_dir_pass_specific(
     if ref_data is not None:
         df = ref_data
     else:
-        df = pd.read_csv(f"{opt.lower()}_cpp.csv")  # TODO: change this
+        df = pd.read_csv(
+            f"pass_runtime/transformations/{opt.lower()}_cpp.csv"
+        )  # TODO: change this
     data = []
     s = time.time()
     results = [
@@ -461,27 +473,6 @@ def check_correctness(
             / n_files
         )
 
-    # for i in range(len(passes)):
-    #     # print(f"{p}", flush=True)
-    #     # tmp_dir_name = f"tmp_{p.split('<')[0]}"
-    #     # print(f"in directory {tmp_dir_name}", flush=True)
-    #     # result = ray.get(
-    #     #     preserve_outliers_dir_pass_specific.remote(
-    #     #         dir_path, opt, p, quantile, fp_list=fp, dst=tmp_dir_name
-    #     #     )
-    #     # )
-    #     result_dict = Counter(results[i])
-    #     data[passes[i]] = (
-    #         1
-    #         - (
-    #             result_dict[(0, 0)]
-    #             + result_dict[(-1, -1)]
-    #             + result_dict[(-2, -2)]
-    #             + result_dict[(-3, -3)]
-    #         )
-    #         / n_files
-    #     )
-
     return data