def GenerateExactMRCFromTrace(exactTraceName):
"""
-------------------------------------------------------
This program creates an exact MRC curve using my histogram
class by parsing the output of a Parda trace analysis.
This allows us to compare the cache curves resulting
from my implementation of SHARDS against the actual cache curve.
-------------------------------------------------------
"""
FIRST_LINE = 2
PATH_TO_TRACE_DIR = os.path.normpath(os.path.join(os.getcwd(), ".."))
fp = open(os.path.join(PATH_TO_TRACE_DIR, "Traces", exactTraceName), "r", encoding = "utf-8")
actualMRCHistogram = Histogram()
for i in range(0,FIRST_LINE):
thisRecord = fp.readline().strip()
thisRecord = fp.readline().strip()
while thisRecord != "" and thisRecord[0].isdigit():
thisRecord = thisRecord.split()
actualMRCHistogram.AddBucket(int(thisRecord[0]), int(thisRecord[1]))
thisRecord = fp.readline().strip()
#Need to add the infinite stack depth indicated at the end of the file
thisRecord = fp.readline().split()
actualMRCHistogram.AddBucket(-1, int(thisRecord[1]))
return actualMRCHistogram
def ClassicLRUSHARDS(traceName):
"""
Since this is fixed size SHARDS, start by sampling every reference. The sampling rate
will be lowered accordingly as the SampleSet reaches maximum capacity.
"""
SAMPLE_RATE = 0.5
mySampleSet = SampleSet(S_MAX)
SHARDSHistogram = Histogram()
myDistanceTree = LRUTree()
i = 0
fp = open(os.path.join(PATH_TO_TRACE_DIR, "Traces",traceName), "r", encoding = "utf-8")
#Get the current time
t1 = clock()
thisReference = fp.readline().strip()
while thisReference != "":
i += 1
thisReference = thisReference[0:INDEX_OF_LAST_CHAR_IN_REF]
#We only sample those disk references which satisfy our sampling condition
if (hash(thisReference) % 100) < SAMPLE_RATE:
"""
Check if the disk reference is in our sample set
"""
#If no
if mySampleSet.FindElement(thisReference) == False:
#Insert the reference into the sample set
mySampleSet.InsertElement(thisReference, hash(thisReference))
#Insert the element into the distance tree
myDistanceTree.InsertElement(thisReference)
#A miss occurred ("infinite" stack depth), record it in the histogram
SHARDSHistogram.IncrementBucket(-1)
#If yes
else:
#Since the address is already in the sample set, it is also in the tree. Get the stack depth
stackDistanceOfThisReference = myDistanceTree.GetDistanceOfElement(thisReference)
#This reuse distance needs to be scaled before it is inserted into the histogram
rescaleFactor = SAMPLE_RATE/100
stackDistanceOfThisReference /= SAMPLE_RATE/100
stackDistanceOfThisReference = round(stackDistanceOfThisReference)
# Remove it from the stack and re-push it (since the stack distance of this element is now 1)
myDistanceTree.RemoveElement(thisReference)
myDistanceTree.InsertElement(thisReference)
# Update the histogram with the old stack depth of thisReference
if SHARDSHistogram.BucketInHistogram(stackDistanceOfThisReference):
SHARDSHistogram.IncrementBucket(stackDistanceOfThisReference)
else:
SHARDSHistogram.AddBucket(stackDistanceOfThisReference, 1)
if i % 10000 == 0:
print(i)
thisReference = fp.readline().strip()
#SHARDSHistogram.PrintDetailedInfo()
#Get the time again and calculate time elapsed
t2 = clock()
#print("Time elapsed: {0}".format(t2 - t1))
return SHARDSHistogram
def Worker(myWorkerId, fileOffset, lengthToProcess, globalProcessReferenceDict,
histogramList, fileName):
"""
Since this is fixed size SHARDS, start by sampling every reference. The sampling rate
will be lowered accordingly as the SampleSet reaches maximum capacity.
"""
mySampleSet = SampleSet(S_MAX)
SHARDSHistogram = Histogram()
myDistanceTree = LRUTree()
#Open the trace file
fp = open(os.path.join(PATH_TO_TRACE_DIR, "Traces", fileName),
"r",
encoding="utf-8")
#Position the file pointer accordingly
currentOffset = 0
thisReference = fp.readline().strip()
while thisReference != "" and currentOffset < fileOffset:
currentOffset += 1
thisReference = fp.readline().strip()
#Process the desired partition of the trace
currentOffset = 0
thisReference = fp.readline().strip()
while thisReference != "" and currentOffset < lengthToProcess:
thisReference = thisReference[0:INDEX_OF_LAST_CHAR_IN_REF]
#We only sample those disk references which satisfy our sampling condition
if (hash(thisReference) % 100) < SAMPLE_RATE:
"""
Check if the disk reference is in our sample set
"""
#If no, this means we have a LOCAL infinity
if mySampleSet.FindElement(thisReference) == False:
#We must process this local infinity to determine if it is indeed global
try:
lastProcessToAccess = globalProcessReferenceDict[
globalProcessReferenceDict]
if lastProcessToAccess < myWorkerId:
#crunch approximate depth
#approximateDepth = ((myWorkerId - lastProcessToAccess) * lengthToProcess) + currentOffset
approximateDepth = (
(myWorkerId - lastProcessToAccess - 1) *
lengthToProcess) + currentOffset + (
0.5 * lengthToProcess)
#This reuse distance needs to be scaled before it is inserted into the histogram
rescaleFactor = SAMPLE_RATE / 100
stackDistanceOfThisReference /= rescaleFactor
stackDistanceOfThisReference = round(
stackDistanceOfThisReference)
# Remove it from the stack and re-push it (since the stack distance of this element is now 1)
myDistanceTree.RemoveElement(thisReference)
myDistanceTree.InsertElement(thisReference)
# Update the histogram with the old stack depth of thisReference
if SHARDSHistogram.BucketInHistogram(
stackDistanceOfThisReference):
SHARDSHistogram.IncrementBucket(
stackDistanceOfThisReference)
else:
SHARDSHistogram.AddBucket(
stackDistanceOfThisReference, 1)
else:
#It is really a global infinity
#Insert the reference into the sample set
mySampleSet.InsertElement(thisReference,
hash(thisReference))
#Insert the element into the distance tree
myDistanceTree.InsertElement(thisReference)
#A miss occurred ("infinite" stack depth), record it in the histogram
SHARDSHistogram.IncrementBucket(-1)
except:
lastProcessToAccess = -1
#It really is a global infinity!
#It is really a global infinity
#Insert the reference into the sample set
mySampleSet.InsertElement(thisReference,
hash(thisReference))
#Insert the element into the distance tree
myDistanceTree.InsertElement(thisReference)
#A miss occurred ("infinite" stack depth), record it in the histogram
SHARDSHistogram.IncrementBucket(-1)
#If yes
else:
#Since the address is already in the sample set, it is also in the tree. Get the stack depth
stackDistanceOfThisReference = myDistanceTree.GetDistanceOfElement(
thisReference)
#This reuse distance needs to be scaled before it is inserted into the histogram
rescaleFactor = SAMPLE_RATE / 100
stackDistanceOfThisReference /= SAMPLE_RATE / 100
stackDistanceOfThisReference = round(
stackDistanceOfThisReference)
# Remove it from the stack and re-push it (since the stack distance of this element is now 1)
myDistanceTree.RemoveElement(thisReference)
myDistanceTree.InsertElement(thisReference)
# Update the histogram with the old stack depth of thisReference
if SHARDSHistogram.BucketInHistogram(
stackDistanceOfThisReference):
SHARDSHistogram.IncrementBucket(
stackDistanceOfThisReference)
else:
SHARDSHistogram.AddBucket(stackDistanceOfThisReference, 1)
#We need to update the global dictionary
globalProcessReferenceDict[thisReference] = myWorkerId
thisReference = fp.readline().strip()
currentOffset += 1
print("Thread {0}. Processed {1} references".format(
myWorkerId, currentOffset))
#Update the manager with the partial histogram
histogramList.append(SHARDSHistogram.GetBuckets())
return
