From 4f25d438add8a42bf23f7aaa37c5c1394074fd99 Mon Sep 17 00:00:00 2001
From: Dominic Kempf <dominic.kempf@iwr.uni-heidelberg.de>
Date: Fri, 28 Sep 2018 14:46:26 +0200
Subject: [PATCH] Introduce an asynchronous minimization function
---
python/dune/perftool/sumfact/vectorization.py | 36 ++++++++++++++++---
1 file changed, 32 insertions(+), 4 deletions(-)
diff --git a/python/dune/perftool/sumfact/vectorization.py b/python/dune/perftool/sumfact/vectorization.py
index 58df262f..0b3c068e 100644
--- a/python/dune/perftool/sumfact/vectorization.py
+++ b/python/dune/perftool/sumfact/vectorization.py
@@ -30,6 +30,7 @@ import itertools as it
import loopy as lp
import numpy as np
import math
+import sys
@generator_factory(item_tags=("vecinfo", "dryrundata"), cache_key_generator=lambda o, n: o)
@@ -160,6 +161,33 @@ def stringify_vectorization_strategy(strategy):
return result
+def minimize(iterable, key=lambda x: x):
+ """ A minimization function that is capable of asynchronous
+ evaluation of the iterable if the python version supports it.
+ """
+ version = sys.version_info
+ if version.major == 3 and version.minor > 5:
+ import asyncio
+ from concurrent.futures import ThreadPoolExecutor
+
+ loop = asyncio.get_event_loop()
+ executor = ThreadPoolExecutor(max_workers=1)
+
+ @asyncio.coroutine
+ def key_coro(i):
+ return loop.run_in_executor(executor, key, i)
+
+ tasks = {}
+ for i in iterable:
+ tasks[i] = asyncio.async(key_coro(i), loop=loop)
+
+ loop.run_until_complete(asyncio.gather(*tasks.values()))
+
+ return min(tasks.items(), key=lambda t: t[1].result())[0]
+ else:
+ return min(iterable, key=key)
+
+
def short_stringify_vectorization_strategy(strategy):
""" A short string decribing the vectorization strategy. This is used
in costmodel validation plots to describe what a data point does
@@ -272,7 +300,7 @@ def level1_optimal_vectorization_strategy(sumfacts, width):
# Print the achieved cost and the target cost on the screen
set_form_option("vectorization_strategy", "model")
target = float(get_form_option("vectorization_target"))
- qp = min(optimal_strategies, key=lambda qp: abs(strategy_cost((qp, optimal_strategies[qp])) - target))
+ qp = minimize(optimal_strategies, key=lambda qp: abs(strategy_cost((qp, optimal_strategies[qp])) - target))
cost = strategy_cost((qp, optimal_strategies[qp]))
print("The target cost was: {}".format(target))
@@ -300,7 +328,7 @@ def level1_optimal_vectorization_strategy(sumfacts, width):
with open("mapping.csv", 'a') as f:
f.write(" ".join((identifier, str(cost), short_stringify_vectorization_strategy((qp, optimal_strategies[qp])))) + "\n")
else:
- qp = min(optimal_strategies, key=lambda qp: strategy_cost((qp, optimal_strategies[qp])))
+ qp = minimize(optimal_strategies, key=lambda qp: strategy_cost((qp, optimal_strategies[qp])))
return qp, optimal_strategies[qp]
@@ -316,8 +344,8 @@ def level2_optimal_vectorization_strategy(sumfacts, width, qp):
key_sumfacts = frozenset(sf for sf in sumfacts if sf.parallel_key == key)
# Minimize over all the opportunities for the subset given by the current key
- key_strategy = min(level2_optimal_vectorization_strategy_generator(key_sumfacts, width, qp),
- key=fixedqp_strategy_costfunction(qp))
+ key_strategy = minimize(level2_optimal_vectorization_strategy_generator(key_sumfacts, width, qp),
+ key=fixedqp_strategy_costfunction(qp))
sfdict = add_to_frozendict(sfdict, key_strategy)
return sfdict
--
GitLab