def organizedata(fileVars):
organizeFunction = {
'scalar':OrganizeScalar, # Scalar data
'impVec':nullOrganizedShader, # Implicit vector data for multiple units (used by Shader Core stats)
'stackbar':nullOrganizedStackedBar, # Stacked bars
'idxVec':nullOrganizedDram, # Vector data with index (used by DRAM stats)
'idx2DVec':nullOrganizedDramV2, # Vector data with 2D index (used by DRAM access stats)
'sparse':OrganizeSparse, # Vector data with 2D index (used by DRAM access stats)
'custom':0
}
data_type_char = {int:'I', float:'f'}
print "Organizing data into internal format..."
# Organize globalCycle in advance because it is used as a reference
if ('globalCycle' in fileVars):
statData = fileVars['globalCycle']
fileVars['globalCycle'].data = organizeFunction[statData.organize](statData.data, data_type_char[statData.datatype])
# Organize other stat data into internal format
for statName, statData in fileVars.iteritems():
if (statName != 'CFLOG' and statName != 'globalCycle' and statData.organize != 'custom'):
fileVars[statName].data = organizeFunction[statData.organize](statData.data, data_type_char[statData.datatype])
# Custom routines to organize stat data into internal format
if fileVars.has_key('averagemflatency'):
zeros = []
for count in range(len(fileVars['averagemflatency'].data),len(fileVars['globalCycle'].data)):
zeros.append(0)
fileVars['averagemflatency'].data = zeros + fileVars['averagemflatency'].data
if (skipCFLog == 0) and fileVars.has_key('CFLOG'):
ptxFile = CFLOGptxFile
statFile = CFLOGInsnInfoFile
print "PC Histogram to CUDA Src = %d" % convertCFLog2CUDAsrc
parseCFLOGCUDA = convertCFLog2CUDAsrc
if parseCFLOGCUDA == 1:
print "Obtaining PTX-to-CUDA Mapping from %s..." % ptxFile
map = lexyacctexteditor.ptxToCudaMapping(ptxFile.rstrip())
print "Obtaining Program Range from %s..." % statFile
maxStats = max(lexyacctexteditor.textEditorParseMe(statFile.rstrip()).keys())
if parseCFLOGCUDA == 1:
newMap = {}
for lines in map:
for ptxLines in map[lines]:
newMap[ptxLines] = lines
print " Total number of CUDA src lines = %s..." % len(newMap)
markForDel = []
for ptxLines in newMap:
if ptxLines > maxStats:
markForDel.append(ptxLines)
for lines in markForDel:
del newMap[lines]
print " Number of touched CUDA src lines = %s..." % len(newMap)
fileVars['CFLOGglobalPTX'] = vc.variable('',2,0)
fileVars['CFLOGglobalCUDA'] = vc.variable('',2,0)
count = 0
for iter in fileVars['CFLOG']:
print "Organizing data for %s" % iter
fileVars[iter + 'PTX'] = fileVars['CFLOG'][iter]
fileVars[iter + 'PTX'].data = CFLOGOrganizePTX(fileVars['CFLOG'][iter].data, fileVars['CFLOG'][iter].maxPC)
if parseCFLOGCUDA == 1:
fileVars[iter + 'CUDA'] = vc.variable('',2,0)
fileVars[iter + 'CUDA'].data = CFLOGOrganizeCuda(fileVars[iter + 'PTX'].data, newMap)
try:
if count == 0:
fileVars['CFLOGglobalPTX'] = fileVars[iter + 'PTX']
if parseCFLOGCUDA == 1:
fileVars['CFLOGglobalCUDA'] = fileVars[iter + 'CUDA']
else:
for rows in range(0, len(fileVars[iter + 'PTX'].data)):
for columns in range(0, len(fileVars[iter + 'PTX'].data[rows])):
fileVars['CFLOGglobalPTX'].data[rows][columns] += fileVars[iter + 'PTX'].data[rows][columns]
if parseCFLOGCUDA == 1:
for rows in range(0, len(fileVars[iter + 'CUDA'].data)):
for columns in range(0, len(fileVars[iter + 'CUDA'].data[rows])):
fileVars['CFLOGglobalCUDA'].data[rows][columns] += fileVars[iter + 'CUDA'].data[rows][columns]
except:
print "Error in generating globalCFLog data"
count += 1
del fileVars['CFLOG']
return fileVars
def organizedata(fileVars):
print "Organizing data into internal format...";
#Section 2.1 for adding a variable
if fileVars.has_key('shaderInsn'):
fileVars['shaderInsn'].data = nullOrganizedShader(fileVars['shaderInsn'].data)
if fileVars.has_key('shdrctacount'):
fileVars['shdrctacount'].data = nullOrganizedShader(fileVars['shdrctacount'].data)
if fileVars.has_key('cacheMissRate_globalL1_all'):
fileVars['cacheMissRate_globalL1_all'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_globalL1_all'].data)
if fileVars.has_key('cacheMissRate_textureL1_all'):
fileVars['cacheMissRate_textureL1_all'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_textureL1_all'].data)
if fileVars.has_key('cacheMissRate_constL1_all'):
fileVars['cacheMissRate_constL1_all'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_constL1_all'].data)
if fileVars.has_key('cacheMissRate_globalL1_noMgHt'):
fileVars['cacheMissRate_globalL1_noMgHt'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_globalL1_noMgHt'].data)
if fileVars.has_key('cacheMissRate_textureL1_noMgHt'):
fileVars['cacheMissRate_textureL1_noMgHt'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_textureL1_noMgHt'].data)
if fileVars.has_key('cacheMissRate_constL1_noMgHt'):
fileVars['cacheMissRate_constL1_noMgHt'].data = nullOrganizedShaderv2(fileVars['cacheMissRate_constL1_noMgHt'].data)
if fileVars.has_key('dram_writes_per_cycle'):
fileVars['dram_writes_per_cycle'].data = [0] + [float(x) for x in fileVars['dram_writes_per_cycle'].data]
if fileVars.has_key('dram_reads_per_cycle'):
fileVars['dram_reads_per_cycle'].data = [0] + [float(x) for x in fileVars['dram_reads_per_cycle'].data]
if fileVars.has_key('globalInsn'):
fileVars['globalInsn'].data = [0] + [int(x) for x in fileVars['globalInsn'].data]
if fileVars.has_key('globalCycle'):
fileVars['globalCycle'].data = [0] + [int(x) for x in fileVars['globalCycle'].data]
if fileVars.has_key('L1ReadMiss'):
fileVars['L1ReadMiss'].data = [0] + [int(x) for x in fileVars['L1ReadMiss'].data]
if fileVars.has_key('L1TextMiss'):
fileVars['L1TextMiss'].data = [0] + fileVars['L1TextMiss'].data
if fileVars.has_key('L1ConstMiss'):
fileVars['L1ConstMiss'].data = [0] + fileVars['L1ConstMiss'].data
if fileVars.has_key('shaderWarpDiv'):
fileVars['shaderWarpDiv'].data = nullOrganizedShader(fileVars['shaderWarpDiv'].data)
if fileVars.has_key('globalTotInsn'):
fileVars['globalTotInsn'].data = fileVars['globalTotInsn'].data
if fileVars.has_key('STmemlatdist'):
fileVars['STmemlatdist'].data = nullOrganizedShader(fileVars['STmemlatdist'].data)
if fileVars.has_key('LDmemlatdist'):
fileVars['LDmemlatdist'].data = nullOrganizedShader(fileVars['LDmemlatdist'].data)
if fileVars.has_key('WarpDivergenceBreakdown'):
fileVars['WarpDivergenceBreakdown'].data = nullOrganizedShader(fileVars['WarpDivergenceBreakdown'].data)
if fileVars.has_key('dramCMD'):
fileVars['dramCMD'].data = nullOrganizedDram(fileVars['dramCMD'].data)
if fileVars.has_key('dramNOP'):
fileVars['dramNOP'].data = nullOrganizedDram(fileVars['dramNOP'].data)
if fileVars.has_key('dramNACT'):
fileVars['dramNACT'].data = nullOrganizedDram(fileVars['dramNACT'].data)
if fileVars.has_key('dramNPRE'):
fileVars['dramNPRE'].data = nullOrganizedDram(fileVars['dramNPRE'].data)
if fileVars.has_key('dramNREQ'):
fileVars['dramNREQ'].data = nullOrganizedDram(fileVars['dramNREQ'].data)
if fileVars.has_key('dramAveMRQS'):
fileVars['dramAveMRQS'].data = nullOrganizedDram(fileVars['dramAveMRQS'].data)
if fileVars.has_key('dramUtil'):
fileVars['dramUtil'].data = nullOrganizedDram(fileVars['dramUtil'].data)
if fileVars.has_key('dramEff'):
fileVars['dramEff'].data = nullOrganizedDram(fileVars['dramEff'].data)
if fileVars.has_key('dramglobal_acc_r'):
fileVars['dramglobal_acc_r'].data = nullOrganizedDramV2(fileVars['dramglobal_acc_r'].data)
if fileVars.has_key('dramglobal_acc_w'):
fileVars['dramglobal_acc_w'].data = nullOrganizedDramV2(fileVars['dramglobal_acc_w'].data)
if fileVars.has_key('dramlocal_acc_r'):
fileVars['dramlocal_acc_r'].data = nullOrganizedDramV2(fileVars['dramlocal_acc_r'].data)
if fileVars.has_key('dramlocal_acc_w'):
fileVars['dramlocal_acc_w'].data = nullOrganizedDramV2(fileVars['dramlocal_acc_w'].data)
if fileVars.has_key('dramconst_acc_r'):
fileVars['dramconst_acc_r'].data = nullOrganizedDramV2(fileVars['dramconst_acc_r'].data)
if fileVars.has_key('dramtexture_acc_r'):
fileVars['dramtexture_acc_r'].data = nullOrganizedDramV2(fileVars['dramtexture_acc_r'].data)
if fileVars.has_key('globalCompletedThreads'):
fileVars['globalCompletedThreads'].data = [0] + fileVars['globalCompletedThreads'].data
if fileVars.has_key('globalSentWrites'):
fileVars['globalSentWrites'].data = [0] + fileVars['globalSentWrites'].data
if fileVars.has_key('globalProcessedWrites'):
fileVars['globalProcessedWrites'].data = [0] + fileVars['globalProcessedWrites'].data
if fileVars.has_key('averagemflatency'):
zeros = []
for count in range(len(fileVars['averagemflatency'].data),len(fileVars['globalCycle'].data)):
zeros.append(0)
fileVars['averagemflatency'].data = zeros + fileVars['averagemflatency'].data
if fileVars.has_key('gpu_stall_by_MSHRwb'):
fileVars['gpu_stall_by_MSHRwb'].data = [0] + fileVars['gpu_stall_by_MSHRwb'].data
if (skipCFLog == 0) and fileVars.has_key('CFLOG'):
loadfile = open('recentfiles.txt', 'r')
bool = 0
while loadfile:
line = loadfile.readline()
if not line: break
if '.ptx' in line:
ptxFile = line
bool += 1
if 'gpgpu_inst_stats' in line:
statFile = line
bool += 1
if bool == 2:
break
print "PC Histogram to CUDA Src = %d" % convertCFLog2CUDAsrc
parseCFLOGCUDA = convertCFLog2CUDAsrc
if parseCFLOGCUDA == 1:
map = lexyacctexteditor.ptxToCudaMapping(ptxFile.rstrip())
maxStats = max(lexyacctexteditor.textEditorParseMe(statFile.rstrip()).keys())
if parseCFLOGCUDA == 1:
newMap = {}
for lines in map:
for ptxLines in map[lines]:
newMap[ptxLines] = lines
markForDel = []
for ptxLines in newMap:
if ptxLines > maxStats:
markForDel.append(ptxLines)
for lines in markForDel:
del newMap[lines]
fileVars['CFLOGglobalPTX'] = vc.variable(2,0)
fileVars['CFLOGglobalCUDA'] = vc.variable(2,0)
count = 0
for iter in fileVars['CFLOG']:
print "Organizing data for %s" % iter
fileVars[iter + 'PTX'] = fileVars['CFLOG'][iter]
fileVars[iter + 'PTX'].data = CFLOGOrganizePTX(fileVars['CFLOG'][iter].data, fileVars['CFLOG'][iter].maxPC)
if parseCFLOGCUDA == 1:
fileVars[iter + 'CUDA'] = vc.variable(2,0)
fileVars[iter + 'CUDA'].data = CFLOGOrganizeCuda(fileVars[iter + 'PTX'].data, newMap)
try:
if count == 0:
fileVars['globalPTX'] = fileVars[iter + 'PTX']
if parseCFLOGCUDA == 1:
fileVars['globalCUDA'] = fileVars[iter + 'CUDA']
else:
for rows in range(0, len(fileVars[iter + 'PTX'].data)):
for columns in range(0, len(fileVars[iter + 'PTX'].data[rows])):
fileVars['globalPTX'].data[rows][columns] += fileVars[iter + 'PTX'].data[rows][columns]
if parseCFLOGCUDA == 1:
for rows in range(0, len(fileVars[iter + 'CUDA'].data)):
for columns in range(0, len(fileVars[iter + 'CUDA'].data[rows])):
fileVars['globalCUDA'].data[rows][columns] += fileVars[iter + 'CUDA'].data[rows][columns]
except:
print "Error in generating globalCFLog data"
count += 1
del fileVars['CFLOG']
return fileVars
for count in range(len(fileVars['averagemflatency'].data),len(fileVars['globalCycle'].data)):
zeros.append(0)
fileVars['averagemflatency'].data = zeros + fileVars['averagemflatency'].data
if (skipCFLog == 0) and fileVars.has_key('CFLOG'):
ptxFile = CFLOGptxFile
statFile = CFLOGInsnInfoFile
print "PC Histogram to CUDA Src = %d" % convertCFLog2CUDAsrc
parseCFLOGCUDA = convertCFLog2CUDAsrc
if parseCFLOGCUDA == 1:
print "Obtaining PTX-to-CUDA Mapping from %s..." % ptxFile
map = lexyacctexteditor.ptxToCudaMapping(ptxFile.rstrip())
print "Obtaining Program Range from %s..." % statFile
maxStats = max(lexyacctexteditor.textEditorParseMe(statFile.rstrip()).keys())
if parseCFLOGCUDA == 1:
newMap = {}
for lines in map:
for ptxLines in map[lines]:
newMap[ptxLines] = lines
print " Total number of CUDA src lines = %s..." % len(newMap)
markForDel = []
for ptxLines in newMap:
if ptxLines > maxStats:
markForDel.append(ptxLines)
for lines in markForDel:
del newMap[lines]
print " Number of touched CUDA src lines = %s..." % len(newMap)
def organizedata(fileVars):
organizeFunction = {
'scalar': OrganizeScalar, # Scalar data
'impVec':
nullOrganizedShader, # Implicit vector data for multiple units (used by Shader Core stats)
'stackbar': nullOrganizedStackedBar, # Stacked bars
'idxVec':
nullOrganizedDram, # Vector data with index (used by DRAM stats)
'idx2DVec':
nullOrganizedDramV2, # Vector data with 2D index (used by DRAM access stats)
'sparse':
OrganizeSparse, # Vector data with 2D index (used by DRAM access stats)
'custom': 0
}
data_type_char = {int: 'I', float: 'f'}
print "Organizing data into internal format..."
# Organize globalCycle in advance because it is used as a reference
if ('globalCycle' in fileVars):
statData = fileVars['globalCycle']
fileVars['globalCycle'].data = organizeFunction[statData.organize](
statData.data, data_type_char[statData.datatype])
# Organize other stat data into internal format
for statName, statData in fileVars.iteritems():
if (statName != 'CFLOG' and statName != 'globalCycle'
and statData.organize != 'custom'):
fileVars[statName].data = organizeFunction[statData.organize](
statData.data, data_type_char[statData.datatype])
# Custom routines to organize stat data into internal format
if fileVars.has_key('averagemflatency'):
zeros = []
for count in range(len(fileVars['averagemflatency'].data),
len(fileVars['globalCycle'].data)):
zeros.append(0)
fileVars['averagemflatency'].data = zeros + fileVars[
'averagemflatency'].data
if (skipCFLog == 0) and fileVars.has_key('CFLOG'):
ptxFile = CFLOGptxFile
statFile = CFLOGInsnInfoFile
print "PC Histogram to CUDA Src = %d" % convertCFLog2CUDAsrc
parseCFLOGCUDA = convertCFLog2CUDAsrc
if parseCFLOGCUDA == 1:
print "Obtaining PTX-to-CUDA Mapping from %s..." % ptxFile
map = lexyacctexteditor.ptxToCudaMapping(ptxFile.rstrip())
print "Obtaining Program Range from %s..." % statFile
maxStats = max(
lexyacctexteditor.textEditorParseMe(statFile.rstrip()).keys())
if parseCFLOGCUDA == 1:
newMap = {}
for lines in map:
for ptxLines in map[lines]:
newMap[ptxLines] = lines
markForDel = []
for ptxLines in newMap:
if ptxLines > maxStats:
markForDel.append(ptxLines)
for lines in markForDel:
del newMap[lines]
fileVars['CFLOGglobalPTX'] = vc.variable('', 2, 0)
fileVars['CFLOGglobalCUDA'] = vc.variable('', 2, 0)
count = 0
for iter in fileVars['CFLOG']:
print "Organizing data for %s" % iter
fileVars[iter + 'PTX'] = fileVars['CFLOG'][iter]
fileVars[iter + 'PTX'].data = CFLOGOrganizePTX(
fileVars['CFLOG'][iter].data, fileVars['CFLOG'][iter].maxPC)
if parseCFLOGCUDA == 1:
fileVars[iter + 'CUDA'] = vc.variable('', 2, 0)
fileVars[iter + 'CUDA'].data = CFLOGOrganizeCuda(
fileVars[iter + 'PTX'].data, newMap)
try:
if count == 0:
fileVars['globalPTX'] = fileVars[iter + 'PTX']
if parseCFLOGCUDA == 1:
fileVars['globalCUDA'] = fileVars[iter + 'CUDA']
else:
for rows in range(0, len(fileVars[iter + 'PTX'].data)):
for columns in range(
0, len(fileVars[iter + 'PTX'].data[rows])):
fileVars['globalPTX'].data[rows][
columns] += fileVars[iter +
'PTX'].data[rows][columns]
if parseCFLOGCUDA == 1:
for rows in range(0,
len(fileVars[iter + 'CUDA'].data)):
for columns in range(
0,
len(fileVars[iter + 'CUDA'].data[rows])):
fileVars['globalCUDA'].data[rows][
columns] += fileVars[
iter + 'CUDA'].data[rows][columns]
except:
print "Error in generating globalCFLog data"
count += 1
del fileVars['CFLOG']
return fileVars
