Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Perf converter update #289

Merged
merged 3 commits into from
Feb 21, 2025
Merged

Perf converter update #289

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
f2af4f6
Standardized how metric events get processed for translation
calebbiggers Feb 14, 2025
95d380b
Added check to ensure input/output dirs exist
calebbiggers Feb 14, 2025
96ac76e
Added flag for outputting TMA metrics
calebbiggers Feb 14, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions scripts/config/replacements_config.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,12 +14,14 @@
"PERF_METRICS.MEMORY_BOUND":"topdown\\-mem\\-bound",
"TOPDOWN.SLOTS:PERF_METRICS":"slots",
"TOPDOWN.SLOTS:PERCORE":"slots",
"SYSTEM_TSC_FREQ":"#SYSTEM_TSC_FREQ",
"SLOTS":"slots",
"SOCKET_COUNT":"#num_packages",
"HYPERTHREADING_ON":"#SMT_on",
"SMT_ON":"#SMT_on",
"CORES_PER_SOCKET":"#num_cores / #num_packages",
"DURATIONTIMEINMILLISECONDS":"( duration_time * 1000 )",
"DURATIONTIMEINSECONDS":"duration_time",
"DURATION_TIME":"duration_time",
"UNC_IIO_PAYLOAD_BYTES_IN.MEM_READ.PART0":"UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART0",
"UNC_IIO_PAYLOAD_BYTES_IN.MEM_READ.PART1":"UNC_IIO_DATA_REQ_OF_CPU.MEM_READ.PART1",
Expand Down
236 changes: 116 additions & 120 deletions scripts/perf_format_converter.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -38,20 +38,32 @@

def main():
# Get file pointers from args
arg_input_file = get_args()
arg_input_file, args_tma = get_args()

# Check that intput/output dirs exists
ensure_directories()

# Check for input file arg
if arg_input_file:
# If input file given, convert just input file
convert_file(arg_input_file)
convert_file(arg_input_file, args_tma)
else:
# If no input file, convert all files in input dir
glob = Path(FILE_PATH, INPUT_DIR_PATH).glob("*.json")
for file in glob:
convert_file(file)


def convert_file(file_path):
convert_file(file, args_tma)

def ensure_directories():
# Check that intput/output dirs exists
try:
if not Path(FILE_PATH, INPUT_DIR_PATH).exists():
Path(FILE_PATH, INPUT_DIR_PATH).mkdir(parents=True, exist_ok=True)
if not Path(FILE_PATH, OUTPUT_DIR_PATH).exists():
Path(FILE_PATH, OUTPUT_DIR_PATH).mkdir(parents=True, exist_ok=True)
except IOError as e:
sys.exit(f"[ERROR] - Error setting up inpur/output dirs {str(e)}. Exiting")

def convert_file(file_path, output_tma):
"""
Takes a standard json file and outputs a converted perf file

Expand All @@ -75,7 +87,7 @@ def convert_file(file_path):
format_converter.populate_issue_dict()

# Convert the dictionary to list of Perf format metric objects
perf_metrics = format_converter.convert_to_perf_metrics(platform)
perf_metrics = format_converter.convert_to_perf_metrics(platform, output_tma)
if not perf_metrics:
return

Expand All @@ -98,11 +110,13 @@ def get_args():
# Arguments
parser.add_argument("-i", "--finput", type=Path,
help="Path of input json file", required=False)

parser.add_argument("-t", "--tma", type=bool,
help="Output TMA metrics [true/false]", required=False)

# Get arguments
args = parser.parse_args()

return args.finput
return args.finput, args.tma


def get_output_file(path):
Expand Down Expand Up @@ -144,10 +158,44 @@ def isNum(string):
return False

def fixPercentages(string):
"""
Takes an inputted string containing a percentage value in % format, and
changed it to decimal format.
eg. 60 -> 0.6

@param string: string containing percentages to fix
@returns: string with fixed percentages
"""
splits = string.split(" ")
fixed = [str(float(split) / 100) if isNum(split) else split for split in splits]
return " ".join(fixed)

def fixSpacing(string):
"""
Takes an inputted formula as a string and fixes the spacing in the formula.
eg. (a / b) -> ( a / b )

@param string: formula to fix
@returns: fixed formula
"""
fixed = string

# Fix instances of " (x"
regex = r"(^|[\s])(\()([^\s])"
match = re.search(regex, fixed)
while(match):
fixed = fixed.replace(match.group(0), f" {match.group(2)} {match.group(3)}")
match = re.search(regex, fixed)

# Fix instances of "x) "
regex = r"([^\s])(\))($|[\s])"
match = re.search(regex, fixed)
while(match):
fixed = fixed.replace(match.group(0), f"{match.group(1)} {match.group(2)} ")
match = re.search(regex, fixed)

return fixed.strip()

class PerfFormatConverter:
"""
Perf Format Converter class. Used to convert the json files. Contains all
Expand Down Expand Up @@ -238,7 +286,7 @@ def populate_issue_dict(self):
else:
self.issue_dict[issue].append(self.translate_metric_name(metric))

def convert_to_perf_metrics(self, platform):
def convert_to_perf_metrics(self, platform, output_tma):
"""
Converts the json dictionary read into the script to a list of
metric objects in PERF format.
Expand All @@ -250,6 +298,11 @@ def convert_to_perf_metrics(self, platform):

try:
for metric in self.input_data["Metrics"]:
# Check if outputting TMA metrics
if not output_tma:
if "TMA" in metric["Category"]:
continue

# Add new metric object for each metric dictionary
new_metric = Metric(
public_description=self.get_public_description(metric),
Expand All @@ -274,60 +327,54 @@ def get_expression(self, metric, platform):
@param metric: metric data as a dictionary
@returns: string containing un-aliased expression
"""
# TMA metric
if "TMA" in metric["Category"]:
if "BaseFormula" in metric and metric["BaseFormula"] != "":
expression_list = [a.strip() for a in metric["BaseFormula"].split(" ") if a != ""]
for i, term in enumerate(expression_list):
if (term not in OPERATORS) and (not isNum(term)) and (not term.startswith("*")):
# Term is not an operator or a numeric value
if "tma_" not in term:
# Translate any event names
expression_list[i] = self.translate_metric_event(term, platform)

# Combine into formula
expression = " ".join(expression_list).strip()

# Add slots to metrics that have topdown events but not slots
if "topdown" in expression and "slots" not in expression:
expression = "( " + expression + " ) + ( 0 * slots )"

return expression.replace("TXL", "TxL")
try:
# TMA metric
if "TMA" in metric["Category"]:
if "BaseFormula" in metric and metric["BaseFormula"] != "":
base_formula = metric["BaseFormula"]
else:
print("Error: TMA metric without base formula found")
# Non TMA metric
else:
print("Error: TMA metric without base formula found")
# Non TMA metric
else:
try:
# Seperate operators with spaces
aliased_formula = fixSpacing(metric["Formula"])

# Get formula and events for conversion
base_formula = metric["Formula"].replace("DURATIONTIMEINSECONDS", "duration_time")
if base_formula.startswith("100 *") and metric["UnitOfMeasure"] == "percent":
base_formula = base_formula.replace("100 *", "")
if aliased_formula.startswith("100 *") and metric["UnitOfMeasure"] == "percent":
aliased_formula = aliased_formula.replace("100 *", "")
events = metric["Events"]
constants = metric["Constants"]

# Replace event/const aliases with names
expression = base_formula.lower()
base_formula = aliased_formula.lower()
for event in events:
reg = r"((?<=[\s+\-*\/\(\)])|(?<=^))({})((?=[\s+\-*\/\(\)\[])|(?=$))".format(event["Alias"].lower())
expression = re.sub(reg,
self.translate_metric_event(event["Name"], platform),
expression)
base_formula = re.sub(reg, event["Name"], base_formula)
for const in constants:
reg = r"((?<=[\s+\-*\/\(\)])|(?<=^))({})((?=[\s+\-*\/\(\)])|(?=$))".format(const["Alias"].lower())
expression = re.sub(reg,
pad(self.translate_metric_constant(const["Name"], metric)),
expression)

# Add slots to metrics that have topdown events but not slots
if any(event["Name"] for event in events if "PERF_METRICS" in event["Name"]):
if not any(event["Name"] for event in events if "SLOTS" in event["Name"]):
expression = "( " + expression + " ) + ( 0 * slots )"

except KeyError as error:
sys.exit("Error in input JSON format during get_expressions(): " + str(error) + ". Exiting")
base_formula = re.sub(reg, pad(const["Name"]), base_formula)

# Take base formula and translate all events
expression_list = [a.strip() for a in base_formula.split(" ") if a != ""]
for i, term in enumerate(expression_list):
if (term not in OPERATORS) and (not isNum(term)) and (not term.startswith("*")):
# Term is not an operator or a numeric value
if "tma_" not in term:
# Translate any event names
expression_list[i] = self.translate_metric_event(term, metric, platform)

# Combine into formula
expression = " ".join(expression_list).strip()

# Remove any extra spaces in expression
# Add slots to metrics that have topdown events but not slots
if "topdown" in expression and "slots" not in expression:
expression = "( " + expression + " ) + ( 0 * slots )"

return re.sub(r"[\s]{2,}", " ", expression.strip()).replace("TXL", "TxL")

except KeyError as error:
sys.exit("Error in input JSON format during get_expressions(): " + str(error) + ". Exiting")


def get_public_description(self, metric):
"""
Expand Down Expand Up @@ -413,7 +460,7 @@ def translate_metric_name(self, metric):
return "tma_" + metric["MetricName"].replace(" ", "_").lower()
return metric["MetricName"]

def translate_metric_event(self, event_name, platform):
def translate_metric_event(self, event_name, metric, platform):
"""
Replaces the event name with a replacement found in the metric
association replacements json file. (An "association" is either an event
Expand Down Expand Up @@ -443,10 +490,18 @@ def translate_metric_event(self, event_name, platform):
prefix = "cbox"

# Check if association has 1:1 replacement
for replacement in self.metric_assoc_replacement_dict:
if re.match(replacement, event_name.upper()):
return self.metric_assoc_replacement_dict[replacement.upper()]
if event_name in self.metric_assoc_replacement_dict:
return self.metric_assoc_replacement_dict[event_name]
elif event_name.upper() in self.metric_assoc_replacement_dict:
return self.metric_assoc_replacement_dict[event_name.upper()]

# Check for source_count() constant
for replacement in self.metric_source_event_dict:
if re.match(replacement, event_name):
for event in metric["Events"]:
if re.match(self.metric_source_event_dict[replacement], event["Name"]) and ":" not in event["Name"]:
return "source_count(" + event["Name"].split(":")[0] + ")"

# Check for ignored events
if event_name.lower() == "tsc":
return event_name.upper()
Expand Down Expand Up @@ -512,70 +567,11 @@ def translate_event_option(self, full_option, is_core_event):

if translated_option is None or translated_option == "None":
return None
return f"{translated_option}\\=0x{value}"

# def translate_event_options(self, split, event_info):
# """
# Takes info about an event with options and translates the options
# into a perf compatible format

# @param split: list of options as strings
# @param event_info: info on how to translate options
# @returns: string containing translated event
# """
# translation = event_info["unit"] + "@" + split[0]
# for option in split[1:]:
# if "=" in option:
# split = [s.lower() for s in option.split("=")]
# if split[0] in event_info["translations"]:
# if "x" in split[1] or "X" in split[1]:
# translation += "\\," + event_info["translations"][split[0]] + "\\=" + split[1]
# else:
# translation += "\\," + event_info["translations"][split[0]] + "\\=" + (int(split[1]) * event_info["scale"])
# elif "0x" in option:
# split = option.split("0x")
# if split[0] in event_info["translations"]:
# translation += "\\," + event_info["translations"][split[0]] + "\\=" + "0x" + split[1]
# else:
# print(f"ERROR - {split[0]} not in event translations...")
# elif "0X" in option:
# split = option.split("0X")
# if split[0].lower() in event_info["translations"]:
# translation += "\\," + event_info["translations"][split[0].lower()] + "\\=" + "0x" + split[1]
# else:
# print(f"ERROR - {split[0]} not in event translations...")
# else:
# match = re.match(r"([a-zA-Z]+)([\d]+)", option.lower())
# if match:
# if match[1] in event_info["translations"]:
# translation += "\\,"+ event_info["translations"][match[1]] + "\\=" + "0x" + match[2]
# return translation + "@"


def translate_metric_constant(self, constant_name, metric):
"""
Replaces the constant name with a replacement found in the metric
association replacements json file. Also handles the source_count()
formatting for specific constants using the config file.

@param constant_name: string containing constant name
@param metric: metric data as a dictionary
@returns: string containing un-aliased expression
"""
# Check if association has replacement

for replacement in self.metric_assoc_replacement_dict:
if re.match(replacement, constant_name):
return self.metric_assoc_replacement_dict[replacement]

for replacement in self.metric_source_event_dict:
# source_count() formatting
if re.match(replacement, constant_name):
for event in metric["Events"]:
if re.match(self.metric_source_event_dict[replacement], event["Name"]) and ":" not in event["Name"]:
return "source_count(" + event["Name"].split(":")[0] + ")"

return "#" + constant_name

if "0x" not in value:
value = f"0x{value}"

return f"{translated_option}\\={value}"

def serialize_output(self, perf_metrics, output_fp):
"""
Expand Down