This contains my bachelors thesis and associated tex files, code snippets and maybe more. Topic: Data Movement in Heterogeneous Memories with Intel Data Streaming Accelerator
You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

92 lines
3.3 KiB

import os
import json
import pandas as pd
from itertools import chain
import seaborn as sns
import matplotlib.pyplot as plt
runid = "Run ID"
x_label = "Thread Count"
y_label = "Throughput in GiB/s LogScale"
var_label = "Thread Counts"
thread_counts = ["1t", "2t", "4t", "8t", "12t"]
thread_counts_nice = ["1 Thread", "2 Threads", "4 Threads", "8 Threads", "12 Threads"]
engine_counts = ["1mib-1e", "1mib-4e", "1gib-1e", "1gib-4e"]
engine_counts_nice = ["1 E/WQ and Tasksize 1 MiB", "4 E/WQ and Tasksize 1 MiB", "1 E/WQ and Tasksize 1 GiB", "4 E/WQ and Tasksize 1 GiB"]
title = "Per-Thread Throughput - 120 Copy Operations split on Threads Intra-Node on DDR with Size 1 MiB"
index = [runid, x_label, var_label]
data = []
def calc_throughput(size_bytes,time_ns):
time_seconds = time_ns * 1e-9
size_gib = size_bytes / (1024 ** 3)
throughput_gibs = size_gib / time_seconds
return throughput_gibs
def index_from_element(value,array):
for (idx,val) in enumerate(array):
if val == value: return idx
return 0
def load_time_mesurements(file_path):
with open(file_path, 'r') as file:
data = json.load(file)
count = data["count"]
iterations = data["list"][0]["task"]["iterations"]
# work queue size is 120 which is split over all available threads
# therefore we divide the result by 120/n_threads to get the per-element speed
return {
"total" : sum([x / (iterations * (120 / count)) for x in list(chain(*[data["list"][i]["report"]["time"]["total"] for i in range(count)]))]),
"combined" : [x / (120 / count) for x in list(chain(*[data["list"][i]["report"]["time"]["combined"] for i in range(count)]))],
"submission" : [x / (120 / count) for x in list(chain(*[data["list"][i]["report"]["time"]["submission"] for i in range(count)]))],
"completion" : [x / (120 / count) for x in list(chain(*[data["list"][i]["report"]["time"]["completion"] for i in range(count)]))]
}
def process_file_to_dataset(file_path, engine_label, thread_count):
engine_index = index_from_element(engine_label,engine_counts)
engine_nice = engine_counts_nice[engine_index]
threadc_index = index_from_element(thread_count, thread_counts)
thread_count_nice = thread_counts_nice[threadc_index]
data_size = 0
if engine_label in ["1gib-1e", "1gib-4e"]:
data_size = 1024*1024*1024
else:
data_size = 1024*1024
try:
time = load_time_mesurements(file_path)["total"]
run_idx = 0
for t in time:
data.append({ runid : run_idx, x_label: thread_count_nice, var_label : engine_nice, y_label : calc_throughput(data_size, t)})
run_idx = run_idx + 1
except FileNotFoundError:
return
def main():
folder_path = "benchmark-results/"
for engine_label in engine_counts:
for thread_count in thread_counts:
file = os.path.join(folder_path, f"mtsubmit-{thread_count}-{engine_label}.json")
process_file_to_dataset(file, engine_label, thread_count)
df = pd.DataFrame(data)
df.set_index(index, inplace=True)
sns.barplot(x=x_label, y=y_label, hue=var_label, data=df, palette="rocket", errorbar="sd")
plt.title(title)
plt.savefig(os.path.join(folder_path, "plot-perf-mtsubmit.png"), bbox_inches='tight')
plt.show()
if __name__ == "__main__":
main()