# here comes the actual decomposition algorithm which returns just an array of arrays/integers and a array of integers with the cut position within the tree
lutstruc, cutHeights = resultTree.doNaiveDecomp(k, weight_dic_int, setOfLdMy)
print "\nChosen %s-LUT structure:" %k
print lutstruc
else:
# smarter decomposition with the help of a gain for each level here:
cutHeights = resultTree.doSmartDecomp(k, weight_dic_int, setOfLdMy)
print "\nCut positions:"
print cutHeights
ultimativeArray = bdd.encodeCutNodes(resultTree,cutHeights)
outputCore = bdd.getBLIF(ultimativeArray, resultTree)
#------write BLIF file ------------------------------------------
print "\n... creating BLIF file"
outputName = f.name.split('/')[-1].split('.')[0]+ '_%s_feasible' %k
outputContent = ".model " + outputName + "\n.inputs"
for inVar in range(1,inputs+1):
outputContent += " x%s" %inVar
def mainProcedure(inputfile, k, actSifting, plotGraph):
f = open(inputfile, "r")
content = f.readlines()
f.close()
print inputfile
# ------- splitting header / boundset / freeset ---------------------
content2 = []
equations = []
for line in content:
content2.append(line.split(" "))
for line in content2:
if line[0] == ".i":
inputs = int(line[1][:-1])
if line[0] == ".o":
outputs = int(line[1][:-1])
if line[0][0] != "." and len(line) >= 2:
equations.append([line[0], line[1][:-1]])
# ------- create minterms from blif equations ---------------------
outputs = 1 # comment if all outputs shall computed !!!!!
maxtermArray = []
variableOrderArray = []
for i in range(outputs):
maxtermArray.append(bf.Maxterm(i)) # i is index, increment for each output
for output in range(outputs): # for each output
equations_ONset = [] # only ON set equations
equations_NumberOfCares = [] # gives a number for each line, which indicates how many 0s or 1s are included
for line in equations:
if line[1][output] == "1": # select the line with an 1 at the end to create the minterm
equations_ONset.append(line[0])
tempCareCounter = 0
for position in range(inputs):
if line[0][position] == "0" or line[0][position] == "1":
tempCareCounter += 1
equations_NumberOfCares.append(float(tempCareCounter))
# ------- sort inputs depending on variable weights ----------------
weight_dic = {} # dictionary in the form of {'x1': 0.7337, 'x2': 0.1234, ... }
numberOfLines = float(len(equations_ONset))
# all values to zero
for i in range(inputs):
weight_dic["x" + str(i + 1)] = 0
# update weigth for each variable
for i in range(inputs):
for index, line in enumerate(equations_ONset):
if line[i] == "0" or line[i] == "1":
weight_dic["x" + str(i + 1)] += 1.0 / equations_NumberOfCares[index] * 1.0 / numberOfLines
# sort weight_dic
weight_dic = sorted(
weight_dic.items(), key=itemgetter(1)
) # notice that dictionary becomes a list of tuples now
# weight list becomes a list of variable indeces in sorted order [9, 3, ...]
weight_dic_int = []
for var in weight_dic:
weight_dic_int.append(int(var[0][1:]))
# lowest value at the beginning -> must be reversed
weight_dic_int.reverse()
# add the initial order to an array
variableOrderArray.append(weight_dic_int)
# finally build the maxterm
for line in equations_ONset:
maxtermArray[output].addMinterm(bf.buildMinterm(line))
# -------- building tree -----------------------------------------
print "\n... creating tree",
resultTree = bdd.doShannon(maxtermArray[0], 1, inputs, weight_dic_int) # must be done for every output !!!!
print "\n\n... creating QRBDD"
resultTree.makeQRBDD()
print "\n... updating level"
bdd.updateLevel(resultTree)
if actSifting == 1:
print "\n\n... sifting tree"
bdd.doSifting(resultTree)
# update the variable order after sifting:
weight_dic_int = bdd.getVariableOrder(resultTree, [])
print "\nNumber of Nodes:", bdd.countNodes(resultTree)
# --------create LUT structure ------------------------------------
# create ld(my) for every level (actually it is not ld(my), because every ld(my)=0 becomes ld(my)=1)
print "\n... calculating 'my' for each level"
setOfLdMy = []
for levels in range(1, inputs):
ldmy = int(math.ceil(math.log(bdd.getMy(resultTree, levels + 1), 2)))
# ldmy get modify if ldmy=0 -> necessary for doNaiveDecomp function
if ldmy == 0:
ldmy = 1
setOfLdMy.append(ldmy)
# naive or smart decomposition?
print "\n... creating LUT structure"
# here comes the actual decomposition algorithm which returns just an array of arrays/integers and a array of integers with the cut position within the tree
lutstruc, cutHeights = resultTree.doNaiveDecomp(k, weight_dic_int, setOfLdMy)
print "\nCut positions:"
print cutHeights
ultimativeArray = bdd.encodeCutNodes(resultTree, cutHeights)
outputCore = bdd.getBLIF(ultimativeArray, resultTree)
# ------write BLIF file ------------------------------------------
print "\n... creating BLIF file"
outputName = f.name.split("/")[-1].split(".")[0] + "_%s_feasible" % k
outputContent = ".model " + outputName + "\n.inputs"
for inVar in range(1, inputs + 1):
outputContent += " x%s" % inVar
outputContent += "\n.outputs y\n\n" + outputCore
print "\n########## BLIF #########\n\n" + outputContent
print "#########################"
outputFile = open(outputName + ".blif", "w")
outputFile.write(outputContent)
outputFile.close()
# ------- PLOT --------------------------------------------------
if plotGraph == 1:
print "\n... plotting tree"
resultTree.dotPrint2()
print "\nDone. Result saved in " + outputName + ".\n"
exit(1)