diff --git a/evaluate.py b/evaluate.py
new file mode 100755
index 000000000..ff4c3aa0a
--- /dev/null
+++ b/evaluate.py
@@ -0,0 +1,301 @@
+#!/usr/bin/python3
+
+import math
+import os
+import re
+import subprocess
+import sys
+import time
+import yaml
+
+import pprint
+
+gray = (186, 189, 182)
+
+def executeCommand(command, stdin = None, cwd = None):
+	with subprocess.Popen(command, stdout = subprocess.PIPE, stderr = subprocess.PIPE, stdin = (subprocess.PIPE if stdin != None else None), cwd = cwd) as process:
+		stdout, stderr = process.communicate(input = (stdin.encode("utf-8") if stdin != None else None))
+		exitCode = process.returncode
+
+	return stdout.decode("utf-8"), stderr.decode("utf-8"), exitCode
+
+def git(command, cwd, enforce = False):
+	stdout, stderr, exitCode = executeCommand(["git"] + command, cwd = cwd)
+
+	if exitCode != 0:
+		print(stderr, file = sys.stderr)
+
+		if enforce:
+			raise RuntimeError("git error")
+
+def initRepo(config):
+	dataDir = config["storage"]["local"]
+
+	# clone repo if not existing
+	if not os.path.isdir(config["storage"]["local"]):
+		git(["clone", config["storage"]["remote"], dataDir], None, enforce = True)
+
+	# fetch origin
+	git(["fetch"], cwd = dataDir)
+
+	# pull all branches
+	for key, branch in config["storage"]["branches"].items():
+		git(["checkout", branch], cwd = dataDir, enforce = True)
+		git(["pull"], cwd = dataDir)
+
+def readBenchmarkConfig(config):
+	initRepo(config)
+
+	dataDir = config["storage"]["local"]
+
+	# checkout config branch
+	git(["checkout", config["storage"]["branches"]["config"]], cwd = dataDir, enforce = True)
+
+	# read instance list
+	instancesFile = os.path.join(config["storage"]["local"], "instances.yml")
+
+	with open(instancesFile, "r") as stream:
+		instances = yaml.load(stream, Loader=yaml.CLoader)
+
+	# read configurations to test
+	configurationsFile = os.path.join(config["storage"]["local"], "configurations.yml")
+
+	with open(configurationsFile, "r") as stream:
+		configurations = yaml.load(stream, Loader=yaml.CLoader)
+
+	# flatten lists of options
+	for configuration in configurations["configurations"]:
+		configuration["options"] = [item for sublist in configuration["options"] for item in sublist]
+
+	return {"configurations": configurations, "instances": instances}
+
+def outputFilenames(configuration, instance, config):
+	instanceID = instance["ipc"] + "_" + instance["domain"] + "_" + str(instance["instance"])
+	outputFile = os.path.join(configuration["id"], instanceID + ".out")
+	errorFile = os.path.join(configuration["id"], instanceID + ".err")
+	environmentFile = os.path.join(configuration["id"], instanceID + ".env")
+
+	return {"outputFile": outputFile, "errorFile": errorFile, "environmentFile": environmentFile}
+
+def jobKey(configuration, instance):
+	return (configuration["id"], instance["ipc"], instance["domain"], instance["instance"])
+
+def instanceKey(instance):
+	return (instance["ipc"], instance["domain"], instance["instance"])
+
+def addResult(results, configuration, instance, result):
+	if not configuration["id"] in results:
+		results[configuration["id"]] = {}
+
+	results[configuration["id"]][instanceKey(instance)] = result
+
+def result(results, configuration, instance):
+	return results[configuration["id"]][instanceKey(instance)]
+
+def mix(color1, color2, t):
+	return (color1[0] * (1 - t) + color2[0] * t, color1[1] * (1 - t) + color2[1] * t, color1[2] * (1 - t) + color2[2] * t)
+
+def resultColor(result, config):
+	if result <= 0:
+		return colors[0]
+	elif result >= config["limits"]["time"]:
+		return colors[-1]
+
+	normalizedResult = (result / config["limits"]["time"]) ** 0.2
+	normalizedResult *= (len(colors) - 1)
+
+	c0 = min(math.floor(normalizedResult), len(colors) - 1)
+	t = normalizedResult - c0
+
+	if t <= 0:
+		return colors[c0]
+	elif t >= 1:
+		return colors[c0 + 1]
+
+	return mix(colors[c0], colors[c0 + 1], t)
+
+def collectResults(config):
+	benchmarkConfig = readBenchmarkConfig(config)
+
+	dataDir = config["storage"]["local"]
+
+	# checkout results branch
+	git(["checkout", config["storage"]["branches"]["results"]], cwd = dataDir, enforce = True)
+
+	configurations = benchmarkConfig["configurations"]["configurations"]
+	instances = benchmarkConfig["instances"]
+
+	results = {}
+
+	for instanceSetName, instanceSet in instances.items():
+		for instance in instanceSet:
+			for configuration in configurations:
+				filenames = outputFilenames(configuration, instance, config)
+				outputFile = os.path.join(config["storage"]["local"], filenames["outputFile"])
+				errorFile = os.path.join(config["storage"]["local"], filenames["errorFile"])
+				environmentFile = os.path.join(config["storage"]["local"], filenames["environmentFile"])
+
+				if not os.path.exists(outputFile) or not os.path.exists(errorFile) or not os.path.exists(environmentFile):
+					addResult(results, configuration, instance, None)
+					continue
+
+				with open(errorFile, "r") as errorOutput:
+					errors = errorOutput.read()
+
+				finishedRE = re.compile("^FINISHED CPU", re.M)
+				runtimeRE = re.compile("<time name=\"ALL\">(.*)</time>", re.M)
+				timeoutRE = re.compile("^TIMEOUT CPU", re.M)
+				memoutRE = re.compile("^MEM CPU", re.M)
+				exitCodeRE = re.compile("^# exit code: (\d+)$", re.M)
+
+				finished = finishedRE.search(errors)
+				runtime = runtimeRE.search(errors)
+				timeout = timeoutRE.search(errors)
+				memout = memoutRE.search(errors)
+				exitCode = exitCodeRE.search(errors)
+
+				if exitCode and int(exitCode.group(1)) != 0:
+					text = "error"
+					color = None
+				elif finished:
+					value = float(runtime.group(1)) / 1000
+					text = str(value)
+					color = (value / config["limits"]["time"]) ** 0.2
+				elif timeout:
+					text = "> " + str(config["limits"]["time"])
+					color = 1.0
+				elif memout:
+					text = "> " + str(config["limits"]["memory"] / 1000000) + " GB"
+					color = 1.0
+
+				result = {"text": text, "color": color}
+
+				addResult(results, configuration, instance, result)
+
+	return configurations, instances, results
+
+def aggregateResults(configurations, instanceSetID, instanceSet, instances, results):
+	aggregatedResults = {}
+
+	for instance in instanceSet:
+		ipcDomain = (instance["ipc"], instance["domain"])
+
+		if not ipcDomain in aggregatedResults:
+			aggregatedResults[ipcDomain] = {}
+
+		for configuration in configurations:
+			if not instanceSetID in configuration["instanceSets"]:
+				continue
+
+		if not configuration["id"] in aggregatedResults[ipcDomain]:
+			aggregatedResults[ipcDomain][configuration["id"]] = {"total": 0, "average": None, "results": []}
+
+		r = result(results, configuration, instance)
+
+		value = 900.0
+
+		try:
+			value = float(r["text"])
+			aggregatedResults[ipcDomain][configuration["id"]]["total"] += 1
+		except:
+			pass
+
+		aggregatedResults[ipcDomain][configuration["id"]]["results"].append(value)
+
+	for ipcDomain, results in aggregatedResults.items():
+		for configurationKey, configurationResults in aggregatedResults.items()
+			configurationResults["average"] = sum(configurationResults["results"]) / len(configurationResults["results"])
+
+	return aggregatedResults
+
+def requiresInstance(configuration, instance, instances):
+	for requiredInstanceSet in configuration["instanceSets"]:
+		if not requiredInstanceSet in instances:
+			raise RuntimeError("undefined instance set “" + requiredInstanceSet + "”")
+
+		if instance in instances[requiredInstanceSet]:
+			return True
+
+	return False
+
+def renderResultsTable(configurations, instanceSetID, instanceSet, instances, results):
+	print("<h2>" + instanceSetID + "</h2><table><thead><tr><th>IPC</th><th>domain</th><th>instance</th>")
+
+	for configuration in configurations:
+		if not instanceSetID in configuration["instanceSets"]:
+			continue
+
+		print("<th><div title=\"" + str(configuration["options"]) + "\">" + configuration["id"] + "</div></th>")
+
+	print("<tbody>")
+
+	for instance in instanceSet:
+		print("<tr><td class=\"col-header\">" + instance["ipc"] + "</td><td class=\"col-header\">" + instance["domain"] + "</td><td class=\"col-header\">" + str(instance["instance"]) + "</td>")
+
+		for configuration in configurations:
+			if not instanceSetID in configuration["instanceSets"]:
+				continue
+
+			r = result(results, configuration, instance)
+
+			if r and r["text"] != "error":
+				print("<td class=\"result result-" + str(int(r["color"] * 100)) + "\">")
+				print(r["text"])
+			elif r and r["text"] == "error":
+				print("<td class=\"error\">")
+				print(r["text"])
+			else:
+				print("<td class=\"tbd\">")
+
+			print("</td>")
+
+		print ("</tr>")
+
+	print("</tbody>")
+
+	print("</tr></thead></table>")
+
+def renderAggregatedResultsTable(type, configurations, instanceSetID, instanceSet, instances, results):
+	aggregatedResults = aggregateResults(configurations, instanceSetID, instanceSet, instances, results)
+
+	if not aggregatedResults:
+		print("<!-- error -->")
+		return
+
+	print("<h2>" + instanceSetID + " (" + type + ")</h2><table><thead><tr><th>IPC</th><th>domain</th>")
+
+	for configurationKey, configurationResults in aggregatedResults.items():
+		print("<th><div title=\"" + str(configurations[configurationKey]["options"]) + "\">" + configurations[configurationKey]["id"] + "</div></th>")
+
+	print("<tbody>")
+
+	for ipcDomain, results in aggregatedResults.items():
+		print("<tr><td class=\"col-header\">" + ipcDomain[0] + "</td><td class=\"col-header\">" + ipcDomain[1] + "</td>")
+
+		for configurationKey, configurationResults in aggregatedResults.items():
+	
+
+		r = results[type]
+
+		print("<td class=\"result\">" + str(r) + "</td></tr>")
+
+	print("</tbody>")
+
+	print("</tr></thead></table>")
+
+def main():
+	with open("config.yml", "r") as stream:
+		config = yaml.load(stream, Loader=yaml.CLoader)
+
+	configurations, instances, results = collectResults(config)
+
+	print("<!DOCTYPE html><html lang=\"en\"><head><title>TPLP benchmark results</title><meta charset=\"UTF-8\"><link rel=\"stylesheet\" href=\"style.css?v=2\" type=\"text/css\"></head><body><main><h1>TPLP Benchmark Results</h1><div class=\"footnote\">last updated at " + time.strftime("%Y-%m-%d %H:%M:%S %z") + "</div>")
+
+	for instanceSetID, instanceSet in instances.items():
+		renderResultsTable(configurations, instanceSetID, instanceSet, instances, results)
+		renderAggregatedResultsTable("total", configurations, instanceSetID, instanceSet, instances, results)
+		renderAggregatedResultsTable("average", configurations, instanceSetID, instanceSet, instances, results)
+
+	print("</main></body></html>")
+
+main()