mirror of
https://github.com/vortexgpgpu/vortex.git
synced 2025-04-23 21:39:10 -04:00
extending csv trace gen to support multiple traces
This commit is contained in:
Blaise Tine
parent
5c73243a4e
commit
e33b62f6fc
1 changed files with 174 additions and 154 deletions
328
ci/trace_csv.py
328
ci/trace_csv.py
|
|
@ -26,7 +26,7 @@ def parse_args():
|
|||
parser.add_argument('log', help='Input log file')
|
||||
return parser.parse_args()
|
||||
|
||||
def parse_simx(log_filename):
|
||||
def parse_simx(log_lines):
|
||||
pc_pattern = r"PC=(0x[0-9a-fA-F]+)"
|
||||
instr_pattern = r"Instr (0x[0-9a-fA-F]+):"
|
||||
opcode_pattern = r"Instr 0x[0-9a-fA-F]+: ([0-9a-zA-Z_\.]+)"
|
||||
|
|
@ -37,32 +37,31 @@ def parse_simx(log_filename):
|
|||
destination_pattern = r"Dest Reg: (.+)"
|
||||
uuid_pattern = r"#(\d+)"
|
||||
entries = []
|
||||
with open(log_filename, 'r') as log_file:
|
||||
instr_data = None
|
||||
for lineno, line in enumerate(log_file, start=1):
|
||||
try:
|
||||
if line.startswith("DEBUG Fetch:"):
|
||||
if instr_data:
|
||||
entries.append(instr_data)
|
||||
instr_data = {}
|
||||
instr_data["lineno"] = lineno
|
||||
instr_data["PC"] = re.search(pc_pattern, line).group(1)
|
||||
instr_data["core_id"] = re.search(core_id_pattern, line).group(1)
|
||||
instr_data["warp_id"] = re.search(warp_id_pattern, line).group(1)
|
||||
instr_data["tmask"] = re.search(tmask_pattern, line).group(1)
|
||||
instr_data["uuid"] = re.search(uuid_pattern, line).group(1)
|
||||
elif line.startswith("DEBUG Instr"):
|
||||
instr_data["instr"] = re.search(instr_pattern, line).group(1)
|
||||
instr_data["opcode"] = re.search(opcode_pattern, line).group(1)
|
||||
elif line.startswith("DEBUG Src"):
|
||||
src_reg = re.search(operands_pattern, line).group(1)
|
||||
instr_data["operands"] = (instr_data["operands"] + ', ' + src_reg) if 'operands' in instr_data else src_reg
|
||||
elif line.startswith("DEBUG Dest"):
|
||||
instr_data["destination"] = re.search(destination_pattern, line).group(1)
|
||||
except Exception as e:
|
||||
print("Error at line {}: {}".format(lineno, e))
|
||||
if instr_data:
|
||||
entries.append(instr_data)
|
||||
instr_data = None
|
||||
for lineno, line in enumerate(log_lines, start=1):
|
||||
try:
|
||||
if line.startswith("DEBUG Fetch:"):
|
||||
if instr_data:
|
||||
entries.append(instr_data)
|
||||
instr_data = {}
|
||||
instr_data["lineno"] = lineno
|
||||
instr_data["PC"] = re.search(pc_pattern, line).group(1)
|
||||
instr_data["core_id"] = re.search(core_id_pattern, line).group(1)
|
||||
instr_data["warp_id"] = re.search(warp_id_pattern, line).group(1)
|
||||
instr_data["tmask"] = re.search(tmask_pattern, line).group(1)
|
||||
instr_data["uuid"] = re.search(uuid_pattern, line).group(1)
|
||||
elif line.startswith("DEBUG Instr"):
|
||||
instr_data["instr"] = re.search(instr_pattern, line).group(1)
|
||||
instr_data["opcode"] = re.search(opcode_pattern, line).group(1)
|
||||
elif line.startswith("DEBUG Src"):
|
||||
src_reg = re.search(operands_pattern, line).group(1)
|
||||
instr_data["operands"] = (instr_data["operands"] + ', ' + src_reg) if 'operands' in instr_data else src_reg
|
||||
elif line.startswith("DEBUG Dest"):
|
||||
instr_data["destination"] = re.search(destination_pattern, line).group(1)
|
||||
except Exception as e:
|
||||
print("Error at line {}: {}".format(lineno, e))
|
||||
if instr_data:
|
||||
entries.append(instr_data)
|
||||
return entries
|
||||
|
||||
def reverse_binary(bin_str):
|
||||
|
|
@ -95,7 +94,7 @@ def append_value(text, reg, value, tmask_arr, sep):
|
|||
text += "}"
|
||||
return text, sep
|
||||
|
||||
def parse_rtlsim(log_filename):
|
||||
def parse_rtlsim(log_lines):
|
||||
config_pattern = r"CONFIGS: num_threads=(\d+), num_warps=(\d+), num_cores=(\d+), num_clusters=(\d+), socket_size=(\d+), local_mem_base=(\d+), num_barriers=(\d+)"
|
||||
line_pattern = r"\d+: cluster(\d+)-socket(\d+)-core(\d+)-(decode|issue|commit)"
|
||||
pc_pattern = r"PC=(0x[0-9a-fA-F]+)"
|
||||
|
|
@ -117,145 +116,166 @@ def parse_rtlsim(log_filename):
|
|||
eop_pattern = r"eop=(\d)"
|
||||
uuid_pattern = r"#(\d+)"
|
||||
entries = []
|
||||
with open(log_filename, 'r') as log_file:
|
||||
instr_data = {}
|
||||
num_threads = 0
|
||||
num_warps = 0
|
||||
num_cores = 0
|
||||
num_clusters = 0
|
||||
socket_size = 0
|
||||
local_mem_base = 0
|
||||
num_barriers = 0
|
||||
num_sockets = 0
|
||||
for lineno, line in enumerate(log_file, start=1):
|
||||
try:
|
||||
config_match = re.search(config_pattern, line)
|
||||
if config_match:
|
||||
num_threads = int(config_match.group(1))
|
||||
num_warps = int(config_match.group(2))
|
||||
num_cores = int(config_match.group(3))
|
||||
num_clusters = int(config_match.group(4))
|
||||
socket_size = int(config_match.group(5))
|
||||
local_mem_base = int(config_match.group(6))
|
||||
num_barriers = int(config_match.group(7))
|
||||
num_sockets = (num_cores + socket_size - 1) // socket_size
|
||||
continue
|
||||
line_match = re.search(line_pattern, line)
|
||||
if line_match:
|
||||
PC = re.search(pc_pattern, line).group(1)
|
||||
warp_id = re.search(warp_id_pattern, line).group(1)
|
||||
tmask = re.search(tmask_pattern, line).group(1)
|
||||
uuid = re.search(uuid_pattern, line).group(1)
|
||||
cluster_id = line_match.group(1)
|
||||
socket_id = line_match.group(2)
|
||||
core_id = line_match.group(3)
|
||||
stage = line_match.group(4)
|
||||
if stage == "decode":
|
||||
trace = {}
|
||||
trace["uuid"] = uuid
|
||||
trace["PC"] = PC
|
||||
trace["core_id"] = ((((cluster_id * num_sockets) + socket_id) * socket_size) + core_id)
|
||||
trace["warp_id"] = warp_id
|
||||
trace["tmask"] = reverse_binary(tmask)
|
||||
trace["instr"] = re.search(instr_pattern, line).group(1)
|
||||
trace["opcode"] = re.search(op_pattern, line).group(1)
|
||||
trace["opds"] = bin_to_array(re.search(opds_pattern, line).group(1))
|
||||
trace["rd"] = re.search(rd_pattern, line).group(1)
|
||||
trace["rs1"] = re.search(rs1_pattern, line).group(1)
|
||||
trace["rs2"] = re.search(rs2_pattern, line).group(1)
|
||||
trace["rs3"] = re.search(rs3_pattern, line).group(1)
|
||||
instr_data = {}
|
||||
num_threads = 0
|
||||
num_warps = 0
|
||||
num_cores = 0
|
||||
num_clusters = 0
|
||||
socket_size = 0
|
||||
local_mem_base = 0
|
||||
num_barriers = 0
|
||||
num_sockets = 0
|
||||
for lineno, line in enumerate(log_lines, start=1):
|
||||
try:
|
||||
config_match = re.search(config_pattern, line)
|
||||
if config_match:
|
||||
num_threads = int(config_match.group(1))
|
||||
num_warps = int(config_match.group(2))
|
||||
num_cores = int(config_match.group(3))
|
||||
num_clusters = int(config_match.group(4))
|
||||
socket_size = int(config_match.group(5))
|
||||
local_mem_base = int(config_match.group(6))
|
||||
num_barriers = int(config_match.group(7))
|
||||
num_sockets = (num_cores + socket_size - 1) // socket_size
|
||||
continue
|
||||
line_match = re.search(line_pattern, line)
|
||||
if line_match:
|
||||
PC = re.search(pc_pattern, line).group(1)
|
||||
warp_id = re.search(warp_id_pattern, line).group(1)
|
||||
tmask = re.search(tmask_pattern, line).group(1)
|
||||
uuid = re.search(uuid_pattern, line).group(1)
|
||||
cluster_id = line_match.group(1)
|
||||
socket_id = line_match.group(2)
|
||||
core_id = line_match.group(3)
|
||||
stage = line_match.group(4)
|
||||
if stage == "decode":
|
||||
trace = {}
|
||||
trace["uuid"] = uuid
|
||||
trace["PC"] = PC
|
||||
trace["core_id"] = ((((cluster_id * num_sockets) + socket_id) * socket_size) + core_id)
|
||||
trace["warp_id"] = warp_id
|
||||
trace["tmask"] = reverse_binary(tmask)
|
||||
trace["instr"] = re.search(instr_pattern, line).group(1)
|
||||
trace["opcode"] = re.search(op_pattern, line).group(1)
|
||||
trace["opds"] = bin_to_array(re.search(opds_pattern, line).group(1))
|
||||
trace["rd"] = re.search(rd_pattern, line).group(1)
|
||||
trace["rs1"] = re.search(rs1_pattern, line).group(1)
|
||||
trace["rs2"] = re.search(rs2_pattern, line).group(1)
|
||||
trace["rs3"] = re.search(rs3_pattern, line).group(1)
|
||||
instr_data[uuid] = trace
|
||||
elif stage == "issue":
|
||||
if uuid in instr_data:
|
||||
trace = instr_data[uuid]
|
||||
trace["lineno"] = lineno
|
||||
opds = trace["opds"]
|
||||
if opds[1]:
|
||||
trace["rs1_data"] = re.search(rs1_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if opds[2]:
|
||||
trace["rs2_data"] = re.search(rs2_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if opds[3]:
|
||||
trace["rs3_data"] = re.search(rs3_data_pattern, line).group(1).split(', ')[::-1]
|
||||
trace["issued"] = True
|
||||
instr_data[uuid] = trace
|
||||
elif stage == "issue":
|
||||
if uuid in instr_data:
|
||||
trace = instr_data[uuid]
|
||||
trace["lineno"] = lineno
|
||||
elif stage == "commit":
|
||||
if uuid in instr_data:
|
||||
trace = instr_data[uuid]
|
||||
if "issued" in trace:
|
||||
opds = trace["opds"]
|
||||
if opds[1]:
|
||||
trace["rs1_data"] = re.search(rs1_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if opds[2]:
|
||||
trace["rs2_data"] = re.search(rs2_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if opds[3]:
|
||||
trace["rs3_data"] = re.search(rs3_data_pattern, line).group(1).split(', ')[::-1]
|
||||
trace["issued"] = True
|
||||
dst_tmask_arr = bin_to_array(tmask)[::-1]
|
||||
wb = re.search(wb_pattern, line).group(1) == "1"
|
||||
if wb:
|
||||
rd_data = re.search(rd_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if 'rd_data' in trace:
|
||||
merged_rd_data = trace['rd_data']
|
||||
for i in range(len(dst_tmask_arr)):
|
||||
if dst_tmask_arr[i] == 1:
|
||||
merged_rd_data[i] = rd_data[i]
|
||||
trace['rd_data'] = merged_rd_data
|
||||
else:
|
||||
trace['rd_data'] = rd_data
|
||||
instr_data[uuid] = trace
|
||||
elif stage == "commit":
|
||||
if uuid in instr_data:
|
||||
trace = instr_data[uuid]
|
||||
if "issued" in trace:
|
||||
opds = trace["opds"]
|
||||
dst_tmask_arr = bin_to_array(tmask)[::-1]
|
||||
wb = re.search(wb_pattern, line).group(1) == "1"
|
||||
eop = re.search(eop_pattern, line).group(1) == "1"
|
||||
if eop:
|
||||
tmask_arr = bin_to_array(trace["tmask"])
|
||||
destination = ''
|
||||
if wb:
|
||||
rd_data = re.search(rd_data_pattern, line).group(1).split(', ')[::-1]
|
||||
if 'rd_data' in trace:
|
||||
merged_rd_data = trace['rd_data']
|
||||
for i in range(len(dst_tmask_arr)):
|
||||
if dst_tmask_arr[i] == 1:
|
||||
merged_rd_data[i] = rd_data[i]
|
||||
trace['rd_data'] = merged_rd_data
|
||||
else:
|
||||
trace['rd_data'] = rd_data
|
||||
instr_data[uuid] = trace
|
||||
eop = re.search(eop_pattern, line).group(1) == "1"
|
||||
if eop:
|
||||
tmask_arr = bin_to_array(trace["tmask"])
|
||||
destination = ''
|
||||
if wb:
|
||||
destination, sep = append_value(destination, trace["rd"], trace['rd_data'], tmask_arr, False)
|
||||
del trace['rd_data']
|
||||
trace["destination"] = destination
|
||||
operands = ''
|
||||
sep = False
|
||||
if opds[1]:
|
||||
operands, sep = append_value(operands, trace["rs1"], trace["rs1_data"], tmask_arr, sep)
|
||||
del trace["rs1_data"]
|
||||
if opds[2]:
|
||||
operands, sep = append_value(operands, trace["rs2"], trace["rs2_data"], tmask_arr, sep)
|
||||
del trace["rs2_data"]
|
||||
if opds[3]:
|
||||
operands, sep = append_value(operands, trace["rs3"], trace["rs3_data"], tmask_arr, sep)
|
||||
del trace["rs3_data"]
|
||||
trace["operands"] = operands
|
||||
del trace["opds"]
|
||||
del trace["rd"]
|
||||
del trace["rs1"]
|
||||
del trace["rs2"]
|
||||
del trace["rs3"]
|
||||
del trace["issued"]
|
||||
del instr_data[uuid]
|
||||
entries.append(trace)
|
||||
except Exception as e:
|
||||
print("Error at line {}: {}".format(lineno, e))
|
||||
destination, sep = append_value(destination, trace["rd"], trace['rd_data'], tmask_arr, False)
|
||||
del trace['rd_data']
|
||||
trace["destination"] = destination
|
||||
operands = ''
|
||||
sep = False
|
||||
if opds[1]:
|
||||
operands, sep = append_value(operands, trace["rs1"], trace["rs1_data"], tmask_arr, sep)
|
||||
del trace["rs1_data"]
|
||||
if opds[2]:
|
||||
operands, sep = append_value(operands, trace["rs2"], trace["rs2_data"], tmask_arr, sep)
|
||||
del trace["rs2_data"]
|
||||
if opds[3]:
|
||||
operands, sep = append_value(operands, trace["rs3"], trace["rs3_data"], tmask_arr, sep)
|
||||
del trace["rs3_data"]
|
||||
trace["operands"] = operands
|
||||
del trace["opds"]
|
||||
del trace["rd"]
|
||||
del trace["rs1"]
|
||||
del trace["rs2"]
|
||||
del trace["rs3"]
|
||||
del trace["issued"]
|
||||
del instr_data[uuid]
|
||||
entries.append(trace)
|
||||
except Exception as e:
|
||||
print("Error at line {}: {}".format(lineno, e))
|
||||
return entries
|
||||
|
||||
def write_csv(log_filename, csv_filename, log_type):
|
||||
entries = None
|
||||
|
||||
# parse log file
|
||||
if log_type == "rtlsim":
|
||||
entries = parse_rtlsim(log_filename)
|
||||
elif log_type == "simx":
|
||||
entries = parse_simx(log_filename)
|
||||
else:
|
||||
print('Error: invalid log type')
|
||||
sys.exit()
|
||||
|
||||
# sort entries by uuid
|
||||
entries.sort(key=lambda x: (int(x['uuid'])))
|
||||
for entry in entries:
|
||||
del entry['lineno']
|
||||
|
||||
# write to CSV
|
||||
def write_csv(sublogs, csv_filename, log_type):
|
||||
with open(csv_filename, 'w', newline='') as csv_file:
|
||||
fieldnames = ["uuid", "PC", "opcode", "instr", "core_id", "warp_id", "tmask", "destination", "operands"]
|
||||
writer = csv.DictWriter(csv_file, fieldnames=fieldnames)
|
||||
writer.writeheader()
|
||||
for entry in entries:
|
||||
writer.writerow(entry)
|
||||
|
||||
for sublog in sublogs:
|
||||
entries = None
|
||||
|
||||
# parse sublog
|
||||
if log_type == "rtlsim":
|
||||
entries = parse_rtlsim(sublog)
|
||||
elif log_type == "simx":
|
||||
entries = parse_simx(sublog)
|
||||
else:
|
||||
print('Error: invalid log type')
|
||||
sys.exit()
|
||||
|
||||
# sort entries by uuid
|
||||
entries.sort(key=lambda x: (int(x['uuid'])))
|
||||
for entry in entries:
|
||||
del entry['lineno']
|
||||
|
||||
for entry in entries:
|
||||
writer.writerow(entry)
|
||||
|
||||
def split_log_file(log_filename):
|
||||
with open(log_filename, 'r') as log_file:
|
||||
log_lines = log_file.readlines()
|
||||
|
||||
sublogs = []
|
||||
current_sublog = None
|
||||
|
||||
for line in log_lines:
|
||||
if line.startswith("[VXDRV] START"):
|
||||
if current_sublog is not None:
|
||||
sublogs.append(current_sublog)
|
||||
current_sublog = [line]
|
||||
elif current_sublog is not None:
|
||||
current_sublog.append(line)
|
||||
|
||||
if current_sublog is not None:
|
||||
sublogs.append(current_sublog)
|
||||
|
||||
return sublogs
|
||||
|
||||
def main():
|
||||
args = parse_args()
|
||||
write_csv(args.log, args.csv, args.type)
|
||||
sublogs = split_log_file(args.log)
|
||||
write_csv(sublogs, args.csv, args.type)
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue