format code

gpucachesim/benchmarks.py

@@ -132,7 +132,9 @@
 
 BENCH_TARGET_INDEX_COLS = ["target", "benchmark"]
 
-PREVIEW_COLS = list(BENCH_TARGET_INDEX_COLS + ["input_id"] + INDEX_COLS + SIMULATE_INPUT_COLS)
+PREVIEW_COLS = list(
+    BENCH_TARGET_INDEX_COLS + ["input_id"] + INDEX_COLS + SIMULATE_INPUT_COLS
+)
 
 RATE_COLUMNS = [
     "l2_hit_rate",
@@ -235,7 +237,9 @@ def _map_dtype(dtype: str) -> str:
 assert len(missing_dtypes) == 0, "missing dtypes for {}".format(missing_dtypes)
 
 
-CATEGORICAL_COLS = set([col for col, dtype in SPECIAL_DTYPES.items() if dtype == "category"])
+CATEGORICAL_COLS = set(
+    [col for col, dtype in SPECIAL_DTYPES.items() if dtype == "category"]
+)
 
 
 def benchmark_name_human_readable(name: str) -> str:
@@ -461,10 +465,18 @@ def construct_playground_simulate_target_config(self, node):
 BenchmarkLoader.add_constructor("!Profile", construct_profile_target_config)
 BenchmarkLoader.add_constructor("!Trace", construct_trace_target_config)
 BenchmarkLoader.add_constructor("!Simulate", construct_simulate_target_config)
-BenchmarkLoader.add_constructor("!ExecDrivenSimulate", construct_exec_driven_simulate_target_config)
-BenchmarkLoader.add_constructor("!AccelsimSimulate", construct_accelsim_simulate_target_config)
-BenchmarkLoader.add_constructor("!AccelsimTrace", construct_accelsim_trace_target_config)
-BenchmarkLoader.add_constructor("!PlaygroundSimulate", construct_playground_simulate_target_config)
+BenchmarkLoader.add_constructor(
+    "!ExecDrivenSimulate", construct_exec_driven_simulate_target_config
+)
+BenchmarkLoader.add_constructor(
+    "!AccelsimSimulate", construct_accelsim_simulate_target_config
+)
+BenchmarkLoader.add_constructor(
+    "!AccelsimTrace", construct_accelsim_trace_target_config
+)
+BenchmarkLoader.add_constructor(
+    "!PlaygroundSimulate", construct_playground_simulate_target_config
+)
 
 
 class Benchmarks:
@@ -508,18 +520,24 @@ def main():
 
 
 @main.command()
-@click.option("--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config")
+@click.option(
+    "--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config"
+)
 def count_bench_configs(path):
     print("loading", path)
     b = Benchmarks(path)
     benches = b.benchmarks[Target.Simulate.value]
 
-    total_bench_configs = sum([len(bench_configs) for bench_configs in benches.values()])
+    total_bench_configs = sum(
+        [len(bench_configs) for bench_configs in benches.values()]
+    )
     print("total bench configs: {}".format(total_bench_configs))
 
 
 @main.command()
-@click.option("--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config")
+@click.option(
+    "--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config"
+)
 @click.option("--baseline", type=bool, default=True, help="Baseline configurations")
 def table(path, baseline):
     print("loading", path)
@@ -550,13 +568,17 @@ def is_baseline(config):
             return not baseline or all(
                 [
                     config["values"].get("memory_only") in [False, None],
-                    config["values"].get("num_clusters") in [int(BASELINE["num_clusters"]), None],
-                    config["values"].get("cores_per_cluster") in [int(BASELINE["cores_per_cluster"]), None],
+                    config["values"].get("num_clusters")
+                    in [int(BASELINE["num_clusters"]), None],
+                    config["values"].get("cores_per_cluster")
+                    in [int(BASELINE["cores_per_cluster"]), None],
                     config["values"].get("mode") in ["serial", None],
                 ]
             )
 
-        baseline_bench_configs = [config for config in bench_configs if is_baseline(config)]
+        baseline_bench_configs = [
+            config for config in bench_configs if is_baseline(config)
+        ]
 
         print(bench_name)
 
@@ -615,7 +637,9 @@ def is_baseline(config):
 
 
 @main.command()
-@click.option("--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config")
+@click.option(
+    "--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config"
+)
 def list(path):
     print("loading", path)
     b = Benchmarks(path)
@@ -625,7 +649,9 @@ def list(path):
 
 
 @main.command()
-@click.option("--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config")
+@click.option(
+    "--path", default=DEFAULT_BENCH_FILE, help="Path to materialized benchmark config"
+)
 def fix(path):
     print("loading", path)
     b = Benchmarks(path)
@@ -695,7 +721,9 @@ def fix(path):
             pass
 
         try:
-            if (result_dir / "Simulate").is_dir() and (result_dir / "simulate").is_dir():
+            if (result_dir / "Simulate").is_dir() and (
+                result_dir / "simulate"
+            ).is_dir():
                 # Simulate is newer
                 shutil.rmtree(result_dir / "simulate")
             os.rename(result_dir / "Simulate", result_dir / "simulate")

gpucachesim/microbench/pchase.py

@@ -2322,14 +2322,19 @@ def find_cache_set_mapping(
     print(combined.shape)
 
     combined_binary = combined.copy()
-    combined_binary["offset"] = combined_binary["offset"].apply(lambda x: np.binary_repr(x, width=num_sets_log2))
-    combined_binary["set"] = combined_binary["set"].apply(lambda x: np.binary_repr(x, width=num_sets_log2))
+    combined_binary["offset"] = combined_binary["offset"].apply(
+        lambda x: np.binary_repr(x, width=num_sets_log2)
+    )
+    combined_binary["set"] = combined_binary["set"].apply(
+        lambda x: np.binary_repr(x, width=num_sets_log2)
+    )
     highest_virt_addr = combined_binary["virt_addr"].max()
-    combined_binary["virt_addr"] = combined_binary["virt_addr"].apply(lambda x: np.binary_repr(x, width=int(np.ceil(np.log2(highest_virt_addr))) + 1))
+    combined_binary["virt_addr"] = combined_binary["virt_addr"].apply(
+        lambda x: np.binary_repr(x, width=int(np.ceil(np.log2(highest_virt_addr))) + 1)
+    )
     print(combined_binary.head(n=10))
     print(combined_binary.shape)
 
-
     print(compute_set_probability(combined))
 
     # return

gpucachesim/plot/__init__.py

@@ -133,6 +133,8 @@ def human_format_thousands(num, round_to=2, variable_precision=False):
             ["", "K", "M", "G", "T", "P"][magnitude],
         )
     return "{}{}".format(
-        round_to_precision_str(num, round_to=round_to, variable_precision=variable_precision),
+        round_to_precision_str(
+            num, round_to=round_to, variable_precision=variable_precision
+        ),
         ["", "K", "M", "G", "T", "P"][magnitude],
     )

gpucachesim/plot/equations.py

@@ -139,7 +139,9 @@ def equations():
         print(color("wrote {}".format(pdf_output_path), fg="cyan"))
 
         png_output_path = (plot.EQUATIONS_DIR / "png" / name).with_suffix(".png")
-        utils.convert_to_png(input_path=pdf_output_path, output_path=png_output_path, density=600)
+        utils.convert_to_png(
+            input_path=pdf_output_path, output_path=png_output_path, density=600
+        )
         print(color("wrote {}".format(png_output_path), fg="cyan"))
 
     pass

gpucachesim/stats/__init__.py

@@ -205,7 +205,6 @@ def different_cols(df):
     return [col for col in df.columns if len(df[col].value_counts()) > 1]
 
 
-
 @main.command(name="speed-table")
 @click.option("-p", "--path", help="Path to materialized benchmark config")
 @click.option("-b", "--bench", "bench_name", help="Benchmark name")
@@ -224,9 +223,13 @@ def different_cols(df):
 def run_speed_table(bench_name, path, nsight, verbose, include_mean_time, png):
     profiler = "nsight" if nsight else "nvprof"
     selected_df = load_stats(bench_name=bench_name, profiler=profiler, path=path)
-    gpucachesim.stats.speed_table.speed_table(selected_df, bench_name, include_mean_time=include_mean_time, verbose=verbose, png=png)
-
-
+    gpucachesim.stats.speed_table.speed_table(
+        selected_df,
+        bench_name,
+        include_mean_time=include_mean_time,
+        verbose=verbose,
+        png=png,
+    )
 
 
 @main.command(name="result-table")
@@ -235,7 +238,11 @@ def run_speed_table(bench_name, path, nsight, verbose, include_mean_time, png):
 @click.option("--bench", "bench_name", help="Benchmark name")
 @click.option("--metric", "metric", type=str, help="metric")
 @click.option(
-    "--combined-only", "combined_only", type=bool, is_flag=True, help="only output combined metrics"
+    "--combined-only",
+    "combined_only",
+    type=bool,
+    is_flag=True,
+    help="only output combined metrics",
 )
 @click.option("--nsight", "nsight", type=bool, is_flag=True, help="use nsight")
 @click.option(
@@ -245,7 +252,15 @@ def run_speed_table(bench_name, path, nsight, verbose, include_mean_time, png):
 def run_result_table(path, bench_name, metric, combined_only, nsight, verbose, png):
     profiler = "nsight" if nsight else "nvprof"
     selected_df = load_stats(bench_name=bench_name, profiler=profiler, path=path)
-    gpucachesim.stats.result_table.result_table(selected_df, bench_name=bench_name, metrics=[metric], combined_only=combined_only, verbose=verbose, png=png)
+    gpucachesim.stats.result_table.result_table(
+        selected_df,
+        bench_name=bench_name,
+        metrics=[metric],
+        combined_only=combined_only,
+        verbose=verbose,
+        png=png,
+    )
+
 
 @main.command(name="all-result-table")
 # @click.pass_context
@@ -263,20 +278,34 @@ def all_result_table(path, bench_name, metric, nsight, verbose, png):
 
     all_benches = sorted(list(selected_df["benchmark"].unique()))
     if metric is None:
-        metrics = ["dramreads", "dramwrites", "l2accesses", "l2dhitrate", "l1accesses", "l1dhitrate", "cycles"]
+        metrics = [
+            "dramreads",
+            "dramwrites",
+            "l2accesses",
+            "l2dhitrate",
+            "l1accesses",
+            "l1dhitrate",
+            "cycles",
+        ]
     else:
         metrics = [metric]
 
     options = dict(verbose=verbose, png=png)
 
     for bench_name in all_benches:
-        gpucachesim.stats.result_table.result_table(selected_df.copy(), bench_name=bench_name, metrics=metrics, **options)
+        gpucachesim.stats.result_table.result_table(
+            selected_df.copy(), bench_name=bench_name, metrics=metrics, **options
+        )
 
     for combined_only in [True, False]:
-        gpucachesim.stats.result_table.result_table(selected_df.copy(), bench_name=None, metrics=metrics, combined_only=combined_only, **options)
-
+        gpucachesim.stats.result_table.result_table(
+            selected_df.copy(),
+            bench_name=None,
+            metrics=metrics,
+            combined_only=combined_only,
+            **options,
+        )
 
-
 
 @main.command(name="parallel-table")
 # @click.pass_context
@@ -309,7 +338,13 @@ def run_parallel_table(bench_name, path, nsight, scale_clusters, large, verbose,
     profiler = "nsight" if nsight else "nvprof"
     selected_df = load_stats(bench_name=bench_name, profiler=profiler, path=path)
     gpucachesim.stats.parallel_table.parallel_table(
-        selected_df, bench_name=bench_name, scale_clusters=scale_clusters, large=large, verbose=verbose, png=png)
+        selected_df,
+        bench_name=bench_name,
+        scale_clusters=scale_clusters,
+        large=large,
+        verbose=verbose,
+        png=png,
+    )
 
 
 @main.command(name="all-parallel-table")
@@ -318,8 +353,7 @@ def run_parallel_table(bench_name, path, nsight, scale_clusters, large, verbose,
 @click.option("--png", "png", type=bool, is_flag=True, help="convert to png")
 def run_all_parallel_table(path, nsight, png):
     profiler = "nsight" if nsight else "nvprof"
-    selected_df = load_stats(
-        bench_name=None, profiler=profiler, path=path)
+    selected_df = load_stats(bench_name=None, profiler=profiler, path=path)
 
     bench_names = sorted(list(selected_df["benchmark"].unique()))
     configs = list(itertools.product([True, False], [True, False]))
@@ -329,17 +363,23 @@ def run_all_parallel_table(path, nsight, png):
     options = dict(batch=True, verbose=False, png=png)
 
     done = 0
-    for (scale_clusters, large) in configs:
+    for scale_clusters, large in configs:
         print("========= {:>4}/{:<4} =======".format(done, total))
         gpucachesim.stats.parallel_table.parallel_table(
-            selected_df, bench_name=None, scale_clusters=scale_clusters, large=large, **options)
+            selected_df,
+            bench_name=None,
+            scale_clusters=scale_clusters,
+            large=large,
+            **options,
+        )
 
         # for all benchmarks
         for bench_name in bench_names:
             mask = selected_df["benchmark"] == bench_name
             bench_df = selected_df[mask].copy()
             gpucachesim.stats.parallel_table.parallel_table(
-                bench_df, bench_name=bench_name,
+                bench_df,
+                bench_name=bench_name,
                 scale_clusters=scale_clusters,
                 **options,
             )