Merge branch 'fixes-to-config' into 'main' (80db3825) · Commits · ExaDigiT / sim-raps

main.py

+6 −2

Original line number	Diff line number	Diff line
		@@ -76,7 +76,7 @@ if args.seed:
		np.random.seed(SEED)

		if args.cooling:
		cooling_model = ThermoFluidsModel(FMU_PATH)
		cooling_model = ThermoFluidsModel(**config)
		cooling_model.initialize()
		args.layout = "layout2"

		@@ -99,7 +99,11 @@ else:
		flops_manager = FLOPSManager(**config)
		layout_manager = LayoutManager(args.layout, args.debug, **config)
		args_dict['config'] = config
		sc = Scheduler(power_manager, flops_manager, layout_manager, cooling_model, **args_dict)
		sc = Scheduler(
		power_manager = power_manager, flops_manager = flops_manager, layout_manager = layout_manager,
		cooling_model = cooling_model,
		**args_dict,
		)

		if args.replay:

raps/cooling.py

+13 −13

Original line number	Diff line number	Diff line
		@@ -73,7 +73,7 @@ class ThermoFluidsModel:
		cleanup():
		Cleans up the extracted FMU directory, ensuring no temporary files are left behind.
		"""
		def __init__(self, FMU_PATH):
		def __init__(self, **config):
		"""
		Constructs all the necessary attributes for the ThermoFluidsModel object.

		@@ -82,7 +82,7 @@ class ThermoFluidsModel:
		FMU_PATH : str
		The file path to the FMU file.
		"""
		self.FMU_PATH = FMU_PATH
		self.config = config
		self.fmu_history = []
		self.inputs = None
		self.outputs = None
		@@ -102,8 +102,8 @@ class ThermoFluidsModel:
		print('Initializing FMU...')

		# Unzip the FMU file and get the unzip directory
		self.unzipdir = extract(self.FMU_PATH)
		model_description = read_model_description(self.FMU_PATH)
		self.unzipdir = extract(self.config['FMU_PATH'])
		model_description = read_model_description(self.config['FMU_PATH'])

		# Add to list of variable names
		var_model = []
		@@ -139,12 +139,12 @@ class ThermoFluidsModel:
		"""
		# Dynamically generate the power inputs
		runtime_values = {
		f"simulator_1_datacenter_1_computeBlock_{i+1}_cabinet_1_sources_Q_flow_total": cdu_power[i] * COOLING_EFFICIENCY / RACKS_PER_CDU
		for i in range(NUM_CDUS)
		f"simulator_1_datacenter_1_computeBlock_{i+1}_cabinet_1_sources_Q_flow_total": cdu_power[i] * self.config['COOLING_EFFICIENCY'] / self.config['RACKS_PER_CDU']
		for i in range(self.config['NUM_CDUS'])
		}

		# Default temperature is from the config
		temperature = WET_BULB_TEMP
		temperature = self.config['WET_BULB_TEMP']

		# If replay mode is on and weather data is available
		if sc.replay and self.weather and self.weather.start is not None and self.weather.has_coords:
		@@ -153,10 +153,10 @@ class ThermoFluidsModel:
		target_datetime = self.weather.start + delta

		# Get temperature from weather data
		temperature = self.weather.get_temperature(target_datetime) or WET_BULB_TEMP
		temperature = self.weather.get_temperature(target_datetime) or self.config['WET_BULB_TEMP']

		# Set the temperature value
		runtime_values[TEMPERATURE_KEY] = temperature
		runtime_values[self.config['TEMPERATURE_KEY']] = temperature

		return runtime_values

		@@ -226,14 +226,14 @@ class ThermoFluidsModel:
		return np.array(value_in_kw) * 1e3 if value_in_kw is not None else 0.0

		# Convert values from kW to Watts using the utility function
		W_HTWPs = convert_to_watts(cooling_output.get(W_HTWPs_KEY))
		W_CTWPs = convert_to_watts(cooling_output.get(W_CTWPs_KEY))
		W_CTs = convert_to_watts(cooling_output.get(W_CTs_KEY))
		W_HTWPs = convert_to_watts(cooling_output.get(self.config['W_HTWPs_KEY']))
		W_CTWPs = convert_to_watts(cooling_output.get(self.config['W_CTWPs_KEY']))
		W_CTs = convert_to_watts(cooling_output.get(self.config['W_CTs_KEY']))

		# Get the sum of the work done by all CDU pumps
		W_CDUPs = sum(
		convert_to_watts(cooling_output.get(f'simulator[1].datacenter[1].computeBlock[{idx+1}].cdu[1].summary.W_flow_CDUP_kW'))
		for idx in range(NUM_CDUS)
		for idx in range(self.config['NUM_CDUS'])
		)

		# Sum all values in the cooling_input dictionary

raps/dataloaders/frontier.py

+10 −10

Original line number	Diff line number	Diff line
		@@ -148,7 +148,7 @@ def load_data_from_df(jobs_df: pd.DataFrame, jobprofile_df: pd.DataFrame, **kwar
		return jobs


		def xname_to_index(xname: str):
		def xname_to_index(xname: str, config: dict):
		"""
		Converts an xname string to an index value based on system configuration.

		@@ -167,11 +167,11 @@ def xname_to_index(xname: str):
		if row == 6:
		col -= 9
		rack_index = row * 12 + col
		node_index = chassis * BLADES_PER_CHASSIS * NODES_PER_BLADE + slot * NODES_PER_BLADE + node
		return rack_index * SC_SHAPE[2] + node_index
		node_index = chassis * config['BLADES_PER_CHASSIS'] * config['NODES_PER_BLADE'] + slot * config['NODES_PER_BLADE'] + node
		return rack_index * config['SC_SHAPE'][2] + node_index


		def index_to_xname(index: int):
		def index_to_xname(index: int, config: dict):
		"""
		Converts an index value back to an xname string based on system configuration.

		@@ -185,17 +185,17 @@ def index_to_xname(index: int):
		str
		The xname string corresponding to the index.
		"""
		rack_index = index // SC_SHAPE[2]
		node_index = index % SC_SHAPE[2]
		rack_index = index // config['SC_SHAPE'][2]
		node_index = index % config['SC_SHAPE'][2]

		row = rack_index // 12
		col = rack_index % 12
		if row == 6:
		col += 9

		chassis = node_index // (BLADES_PER_CHASSIS * NODES_PER_BLADE)
		remaining = node_index % (BLADES_PER_CHASSIS * NODES_PER_BLADE)
		slot = remaining // NODES_PER_BLADE
		node = remaining % NODES_PER_BLADE
		chassis = node_index // (config['BLADES_PER_CHASSIS'] * config['NODES_PER_BLADE'])
		remaining = node_index % (config['BLADES_PER_CHASSIS'] * config['NODES_PER_BLADE'])
		slot = remaining // config['NODES_PER_BLADE']
		node = remaining % config['NODES_PER_BLADE']

		return f"x2{row}{col:02}c{chassis}s{slot}b{node}"

raps/power.py

+1 −1

Original line number	Diff line number	Diff line
		@@ -161,7 +161,7 @@ class PowerManager:
		- down_nodes: Nodes that are currently down.
		- down_rack: Rack number of down nodes.
		"""
		def __init__(self, power_func, **config):
		def __init__(self, power_func=compute_node_power, **config):
		"""
		Initialize the PowerManager object.

raps/scheduler.py

+2 −2

Original line number	Diff line number	Diff line
		@@ -77,8 +77,8 @@ def get_utilization(trace, time_quanta_index):

		class Scheduler:
		"""Job scheduler and simulation manager."""
		def __init__(self, power_manager, flops_manager, layout_manager, cooling_model=None, **kwargs):
		self.config = kwargs.get('config')
		def __init__(self, , power_manager, flops_manager, layout_manager, cooling_model=None, config, *kwargs):
		self.config = config
		self.down_nodes = summarize_ranges(self.config['DOWN_NODES'])
		self.available_nodes = list(set(range(self.config['TOTAL_NODES'])) - set(self.config['DOWN_NODES']))
		self.num_free_nodes = len(self.available_nodes)