Aggregate all results to a pandas dataframe (#9)

kernels/streamer_matmul/Snakefile

@@ -1,5 +1,6 @@
 from util.snake.configs import get_snax_gemmx_config
 from quantized_matmul import main as get_quantized_matmul
+from aggregate import aggregate_results
 
 config = get_snax_gemmx_config(snax_mlir_path="../../snax-mlir")
 config["snaxoptflags"] = ",".join(
@@ -53,10 +54,17 @@ values = [16,32]
 # Rules
 rule all:
     input:
-        expand("{file}_{m}_{n}_{k}_traces.json", file=files, m=values, n=values, k=values),
-        "tiled_quantized_matmul_traces.json"
+        "tiled_quantized_matmul_traces.json",
+        "quantized_matmul_results.csv"
 
-#rule get_cycles:
+rule get_dataframe:
+    input:
+        expand("{file}_{m}_{n}_{k}_traces.json", file=files, m=values, n=values, k=values),
+    output:
+        "{file}_results.csv"
+    run:
+        aggregate_results(wildcards.file)
+
 
 rule generate_quantized_matmul:
     output:

kernels/streamer_matmul/aggregate.py

@@ -0,0 +1,25 @@
+import json
+import pandas as pd
+import re
+import glob
+
+def aggregate_results(filename: str):
+    FILENAME_PATTERN = filename + r"_(\d+)_(\d+)_(\d+)_traces\.json"
+    json_files = glob.glob(f"{filename}_*_traces.json")  # Adjust the path
+    data_list = []
+    for file in json_files:
+        match = re.search(FILENAME_PATTERN, file)
+        if match:
+            n, m, k = map(int, match.groups())  # Extract n, m, k as integers
+            with open(file, "r") as f:
+                json_data = json.load(f)
+                extracted_cycles = json_data[0][0][2]["cycles"]
+            data_list.append({"n": n, "m": m, "k": k, "cycles": extracted_cycles })
+    df = pd.DataFrame(data_list)
+    df.to_csv(f"{filename}_results.csv", index=False)
+    return df
+
+if __name__ == "__main__":
+    df = aggregate_results("quantized_matmul")
+    print(df)
+