Compiler Remark Injection

thicket/tests/conftest.py

@@ -188,6 +188,30 @@ def rajaperf_seq_O3_1M_cali(data_dir, tmpdir):
     return [os.path.join(str(tmpdir), f) for f in cali_files]
 
 
+@pytest.fixture
+def dane_rajaperf_seq_O3_8M_cali(data_dir, tmpdir):
+    """All trials of for Base Sequential optimization level O3 and problem size 8388608."""
+    cali_files = sorted(
+        glob(
+            f"{data_dir}/rajaperf/dane/clang14.0.6_8388608/O3/**/Base_Seq-default.cali",
+            recursive=True,
+        )
+    )
+    for cf in cali_files:
+        shutil.copy(cf, str(tmpdir))
+    return [os.path.join(str(tmpdir), f) for f in cali_files]
+
+
+@pytest.fixture
+def dane_rajaperf_seq_O3_8M_remark_cali(data_dir, tmpdir):
+    """Remark caliper file for Base Sequential optimization level O3 and problem size 8388608."""
+    cali_file = f"{data_dir}/rajaperf/dane/clang14.0.6_8388608/O3/remark_data/remark_output.cali"
+
+    shutil.copy(cali_file, str(tmpdir))
+
+    return os.path.join(str(tmpdir), cali_file)
+
+
 @pytest.fixture
 def rajaperf_unique_tunings(data_dir, tmpdir):
     """1 trial of each unique tuning in the dataset"""

thicket/tests/test_remark_injestion.py

@@ -0,0 +1,47 @@
+# Copyright 2022 Lawrence Livermore National Security, LLC and other
+# Thicket Project Developers. See the top-level LICENSE file for details.
+#
+# SPDX-License-Identifier: MIT
+
+import pytest
+
+import thicket as th
+
+
+def test_add_remark_data(
+    dane_rajaperf_seq_O3_8M_cali,
+    dane_rajaperf_seq_O3_8M_remark_cali,
+    thicket_axis_columns,
+):
+    _, _, combined_th = thicket_axis_columns
+
+    th_ens = th.Thicket.from_caliperreader(
+        dane_rajaperf_seq_O3_8M_cali,
+        disable_tqdm=True,
+    )
+
+    with pytest.raises(
+        ValueError, match="Parameter 'caliper_file' must be of type string."
+    ):
+        th_ens.add_remark_data(1)
+
+    with pytest.raises(
+        ValueError, match="Cannot run 'run_remark_data' on columnar joined thickets."
+    ):
+        combined_th.add_remark_data(dane_rajaperf_seq_O3_8M_remark_cali)
+
+    th_ens.add_remark_data(dane_rajaperf_seq_O3_8M_remark_cali)
+
+    assert sorted(th_ens.dataframe.index.get_level_values(0).unique()) == sorted(
+        th_ens.statsframe.dataframe.index.values
+    )
+
+    assert "remark_columns" in th_ens.metadata
+
+    remark_columns = th_ens.metadata["remark_columns"].iloc[0]
+
+    assert isinstance(remark_columns, list)
+
+    assert len(set(th_ens.dataframe.columns).intersection(remark_columns)) == len(
+        remark_columns
+    )

thicket/thicket.py

@@ -6,6 +6,7 @@
 import collections
 import copy
 import os
+import re
 import pickle
 import sys
 import json
@@ -622,6 +623,99 @@ def from_json(json_thicket):
         # make and return thicket?
         return th
 
+    def add_remark_data(
+        self,
+        caliper_file,
+    ):
+        if not isinstance(caliper_file, str):
+            raise ValueError("Parameter 'caliper_file' must be of type string.")
+
+        if self.dataframe.columns.nlevels != 1:
+            raise ValueError(
+                "Cannot run 'run_remark_data' on columnar joined thickets."
+            )
+
+        def process_remarks(remark_data_thicket):
+
+            static_remark_columns = set()
+            dynamic_remark_columns = set()
+
+            # TODO: Change this so that you use row directly
+            for index, row in remark_data_thicket.dataframe.iterrows():
+
+                num_nodes = len(remark_data_thicket.graph)
+
+                static_freq_dict = {}
+                dynamic_freq_dict = {}
+
+                node = index[0]
+
+                remark_data_thicket.dataframe.loc[node, "Remark_Attribute"].iloc[0] = re.findall(r"'([^']*)'", remark_data_thicket.dataframe.loc[node, "Remark_Attribute"].iloc[0])
+
+                for parent in node.parents:
+                    # TODO: Delete, might be unnecessary???
+                    if isinstance(remark_data_thicket.dataframe.loc[parent, "Remark_Attribute"].iloc[0], str):
+                        remark_data_thicket.dataframe.loc[parent, "Remark_Attribute"].iloc[0] = re.findall(r"'([^']*)'", remark_data_thicket.dataframe.loc[parent, "Remark_Attribute"].iloc[0])
+                    for value in remark_data_thicket.dataframe.loc[parent, "Remark_Attribute"].iloc[0]:
+                        try:
+                            remark_data_thicket.dataframe.loc[node, "Remark_Attribute"].iloc[0].remove(value)
+                            # print(f"Removed value: {value} for region: {node.name}")
+                        except Exception:
+                            pass
+
+                static_remarks = list(set(remark_data_thicket.dataframe.loc[node, "Remark_Attribute"].iloc[0]))
+                dynamic_remarks = list(remark_data_thicket.dataframe.loc[node, "Remark_Attribute"].iloc[0])
+
+                static_remarks = ["~".join(i.split("~")[:3]) for i in static_remarks]
+                dynamic_remarks = ["~".join(i.split("~")[:3]) for i in dynamic_remarks]
+
+                unique_remarks = list(set(static_remarks))
+
+                for remark in static_remarks:
+                    static_remark_name = f"{remark}~static"
+                    static_remark_columns.add(static_remark_name)
+                    if static_remark_name not in static_freq_dict.keys():
+                        static_freq_dict[static_remark_name] = 1
+                        continue
+
+                    static_freq_dict[static_remark_name] += 1
+
+                for remark in dynamic_remarks:
+                    dynamic_remark_name = f"{remark}~dynamic"
+                    dynamic_remark_columns.add(dynamic_remark_name)
+                    if dynamic_remark_name not in dynamic_freq_dict.keys():
+                        dynamic_freq_dict[dynamic_remark_name] = 1
+                        continue
+                    dynamic_freq_dict[dynamic_remark_name] += 1
+
+                remark_data_thicket.dataframe.loc[node, static_freq_dict.keys()] = list(static_freq_dict.values())
+                remark_data_thicket.dataframe.loc[node, dynamic_freq_dict.keys()] = list(dynamic_freq_dict.values())
+
+            return list(static_remark_columns), list(dynamic_remark_columns)
+
+        remark_data_thicket = self.from_caliperreader(caliper_file, string_attributes="Remark_Attribute")
+
+        static_remark_columns, dynamic_remark_columns = process_remarks(remark_data_thicket)
+
+        remark_data_thicket.dataframe = remark_data_thicket.dataframe.reindex(
+            remark_data_thicket.dataframe.index.repeat(len(self.profile))
+        )
+
+        remark_data_thicket.dataframe.index = self.dataframe.index
+
+        self.dataframe[static_remark_columns] = remark_data_thicket.dataframe[
+            static_remark_columns
+        ]
+
+        self.dataframe[dynamic_remark_columns] = remark_data_thicket.dataframe[
+            dynamic_remark_columns
+        ]
+
+        self.metadata["remark_columns_static"] = [static_remark_columns] * len(self.profile)
+        self.metadata["remark_columns_dynamic"] = [dynamic_remark_columns] * len(self.profile)
+
+        self.dataframe.sort_index(inplace=True)
+
     def add_ncu(
         self,
         ncu_report_mapping,