Add benchmark history updates

Author: ianayl

devops/scripts/benchmarks/history.py

@@ -1,19 +1,19 @@
-# Copyright (C) 2024 Intel Corporation
+# Copyright (C) 2024-2025 Intel Corporation
 # Part of the Unified-Runtime Project, under the Apache License v2.0 with LLVM Exceptions.
 # See LICENSE.TXT
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
 import os
 import json
 from pathlib import Path
+import socket
 from utils.result import Result, BenchmarkRun
 from options import Compare, options
 from datetime import datetime, timezone
 from utils.utils import run
 
 
 class BenchmarkHistory:
-    benchmark_run_index_max = 0
     runs = []
 
     def __init__(self, dir):
@@ -35,42 +35,55 @@ def load(self, n: int):
         # Get all JSON files in the results directory
         benchmark_files = list(results_dir.glob("*.json"))
 
-        # Extract index numbers and sort files by index number
-        def extract_index(file_path: Path) -> int:
+        # Extract timestamp and sort files by it
+        def extract_timestamp(file_path: Path) -> str:
             try:
-                return int(file_path.stem.split("_")[0])
-            except (IndexError, ValueError):
-                return -1
+                return file_path.stem.split("_")[-1]
+            except IndexError:
+                return ""
 
-        benchmark_files = [
-            file for file in benchmark_files if extract_index(file) != -1
-        ]
-        benchmark_files.sort(key=extract_index)
+        benchmark_files.sort(key=extract_timestamp, reverse=True)
 
         # Load the first n benchmark files
         benchmark_runs = []
-        for file_path in benchmark_files[n::-1]:
+        for file_path in benchmark_files[:n]:
             benchmark_run = self.load_result(file_path)
             if benchmark_run:
                 benchmark_runs.append(benchmark_run)
 
-        if benchmark_files:
-            self.benchmark_run_index_max = extract_index(benchmark_files[-1])
-
         self.runs = benchmark_runs
 
     def create_run(self, name: str, results: list[Result]) -> BenchmarkRun:
         try:
-            result = run("git rev-parse --short HEAD")
+            script_dir = os.path.dirname(os.path.abspath(__file__))
+            result = run("git rev-parse --short HEAD", cwd=script_dir)
             git_hash = result.stdout.decode().strip()
+
+            # Get the GitHub repo URL from git remote
+            remote_result = run("git remote get-url origin", cwd=script_dir)
+            remote_url = remote_result.stdout.decode().strip()
+
+            # Convert SSH or HTTPS URL to owner/repo format
+            if remote_url.startswith("[email protected]:"):
+                # SSH format: [email protected]:owner/repo.git
+                github_repo = remote_url.split("[email protected]:")[1].rstrip(".git")
+            elif remote_url.startswith("https://github.com/"):
+                # HTTPS format: https://github.com/owner/repo.git
+                github_repo = remote_url.split("https://github.com/")[1].rstrip(".git")
+            else:
+                github_repo = None
+
         except:
             git_hash = "unknown"
+            github_repo = None
 
         return BenchmarkRun(
             name=name,
             git_hash=git_hash,
+            github_repo=github_repo,
             date=datetime.now(tz=timezone.utc),
             results=results,
+            hostname=socket.gethostname(),
         )
 
     def save(self, save_name, results: list[Result], to_file=True):
@@ -84,12 +97,9 @@ def save(self, save_name, results: list[Result], to_file=True):
         results_dir = Path(os.path.join(self.dir, "results"))
         os.makedirs(results_dir, exist_ok=True)
 
-        self.benchmark_run_index_max += 1
-        file_path = Path(
-            os.path.join(
-                results_dir, f"{self.benchmark_run_index_max}_{save_name}.json"
-            )
-        )
+        # Use formatted timestamp for the filename
+        timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+        file_path = Path(os.path.join(results_dir, f"{save_name}_{timestamp}.json"))
         with file_path.open("w") as file:
             json.dump(serialized, file, indent=4)
         print(f"Benchmark results saved to {file_path}")
@@ -120,6 +130,7 @@ def compute_average(self, data: list[BenchmarkRun]):
             name=first_run.name,
             git_hash="average",
             date=first_run.date,  # should this be different?
+            hostname=first_run.hostname,
         )
 
         return average_benchmark_run

devops/scripts/benchmarks/main.py

@@ -253,7 +253,10 @@ def main(directory, additional_env_vars, save_name, compare_names, filter):
     if not options.dry_run:
         chart_data = {this_name: results}
 
-    history = BenchmarkHistory(directory)
+    results_dir = directory
+    if options.custom_results_dir:
+        results_dir = Path(options.custom_results_dir)
+    history = BenchmarkHistory(results_dir)
     # limit how many files we load.
     # should this be configurable?
     history.load(1000)
@@ -445,7 +448,12 @@ def validate_and_parse_env_args(env_args):
         help="The name of the results which should be used as a baseline for metrics calculation",
         default=options.current_run_name,
     )
-
+    parser.add_argument(
+        "--results-dir",
+        type=str,
+        help="Specify a custom results directory",
+        default=options.custom_results_dir,
+    )
     parser.add_argument(
         "--build-jobs",
         type=int,
@@ -476,6 +484,7 @@ def validate_and_parse_env_args(env_args):
     options.iterations_stddev = args.iterations_stddev
     options.build_igc = args.build_igc
     options.current_run_name = args.relative_perf
+    options.custom_results_dir = args.results_dir
     options.build_jobs = args.build_jobs
 
     if args.build_igc and args.compute_runtime is None:

devops/scripts/benchmarks/options.py

@@ -42,6 +42,7 @@ class Options:
     compute_runtime_tag: str = "25.05.32567.12"
     build_igc: bool = False
     current_run_name: str = "This PR"
+    custom_results_dir = None
     build_jobs: int = multiprocessing.cpu_count()
 
 options = Options()