Skip to content

Commit 3f83dcc

Browse files
onewbiekhammad45
onewbiek
and
hammad45
authored
Fix DXT stack trace integration (#15)
* Fix DXT conflict with Recorder * Fix DXT merge conflict * Cut redundant code * Bug Fix in DXT Analysis --------- Co-authored-by: hammad45 <[email protected]>
1 parent 0ab1ded commit 3f83dcc

File tree

2 files changed

+152
-53
lines changed

2 files changed

+152
-53
lines changed

‎drishti/handlers/handle_darshan.py‎

Lines changed: 107 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -81,7 +81,10 @@ def handler():
8181

8282
information = darshanll.log_get_job(log)
8383

84-
log_version = information['metadata']['lib_ver']
84+
if 'log_ver' in information:
85+
log_version = information['log_ver']
86+
else:
87+
log_version = information['metadata']['lib_ver']
8588
library_version = darshanll.get_lib_version()
8689

8790
# Make sure log format is of the same version
@@ -144,6 +147,100 @@ def handler():
144147
df_mpiio = None
145148

146149
total_size_mpiio = 0
150+
151+
dxt_posix = None
152+
dxt_posix_read_data = None
153+
dxt_posix_write_data = None
154+
dxt_mpiio = None
155+
156+
df_lustre = None
157+
if "LUSTRE" in report.records:
158+
df_lustre = report.records['LUSTRE'].to_df()
159+
160+
if args.backtrace:
161+
if "DXT_POSIX" in report.records:
162+
dxt_posix = report.records["DXT_POSIX"].to_df()
163+
dxt_posix = pd.DataFrame(dxt_posix)
164+
if "address_line_mapping" not in dxt_posix:
165+
args.backtrace = False
166+
else:
167+
read_id = []
168+
read_rank = []
169+
read_length = []
170+
read_offsets = []
171+
read_end_time = []
172+
read_start_time = []
173+
read_operation = []
174+
175+
write_id = []
176+
write_rank = []
177+
write_length = []
178+
write_offsets = []
179+
write_end_time = []
180+
write_start_time = []
181+
write_operation = []
182+
183+
for r in zip(dxt_posix['rank'], dxt_posix['read_segments'], dxt_posix['write_segments'], dxt_posix['id']):
184+
if not r[1].empty:
185+
read_id.append([r[3]] * len((r[1]['length'].to_list())))
186+
read_rank.append([r[0]] * len((r[1]['length'].to_list())))
187+
read_length.append(r[1]['length'].to_list())
188+
read_end_time.append(r[1]['end_time'].to_list())
189+
read_start_time.append(r[1]['start_time'].to_list())
190+
read_operation.append(['read'] * len((r[1]['length'].to_list())))
191+
read_offsets.append(r[1]['offset'].to_list())
192+
193+
if not r[2].empty:
194+
write_id.append([r[3]] * len((r[2]['length'].to_list())))
195+
write_rank.append([r[0]] * len((r[2]['length'].to_list())))
196+
write_length.append(r[2]['length'].to_list())
197+
write_end_time.append(r[2]['end_time'].to_list())
198+
write_start_time.append(r[2]['start_time'].to_list())
199+
write_operation.append(['write'] * len((r[2]['length'].to_list())))
200+
write_offsets.append(r[2]['offset'].to_list())
201+
202+
read_id = [element for nestedlist in read_id for element in nestedlist]
203+
read_rank = [element for nestedlist in read_rank for element in nestedlist]
204+
read_length = [element for nestedlist in read_length for element in nestedlist]
205+
read_offsets = [element for nestedlist in read_offsets for element in nestedlist]
206+
read_end_time = [element for nestedlist in read_end_time for element in nestedlist]
207+
read_operation = [element for nestedlist in read_operation for element in nestedlist]
208+
read_start_time = [element for nestedlist in read_start_time for element in nestedlist]
209+
210+
write_id = [element for nestedlist in write_id for element in nestedlist]
211+
write_rank = [element for nestedlist in write_rank for element in nestedlist]
212+
write_length = [element for nestedlist in write_length for element in nestedlist]
213+
write_offsets = [element for nestedlist in write_offsets for element in nestedlist]
214+
write_end_time = [element for nestedlist in write_end_time for element in nestedlist]
215+
write_operation = [element for nestedlist in write_operation for element in nestedlist]
216+
write_start_time = [element for nestedlist in write_start_time for element in nestedlist]
217+
218+
dxt_posix_read_data = pd.DataFrame(
219+
{
220+
'id': read_id,
221+
'rank': read_rank,
222+
'length': read_length,
223+
'end_time': read_end_time,
224+
'start_time': read_start_time,
225+
'operation': read_operation,
226+
'offsets': read_offsets,
227+
})
228+
229+
dxt_posix_write_data = pd.DataFrame(
230+
{
231+
'id': write_id,
232+
'rank': write_rank,
233+
'length': write_length,
234+
'end_time': write_end_time,
235+
'start_time': write_start_time,
236+
'operation': write_operation,
237+
'offsets': write_offsets,
238+
})
239+
240+
if "DXT_MPIIO" in report.records:
241+
dxt_mpiio = report.records["DXT_MPIIO"].to_df()
242+
dxt_mpiio = pd.DataFrame(dxt_mpiio)
243+
147244

148245
# Since POSIX will capture both POSIX-only accesses and those comming from MPI-IO, we can subtract those
149246
if total_size_posix > 0 and total_size_posix >= total_size_mpiio:
@@ -262,7 +359,7 @@ def handler():
262359
detected_files.columns = ['id', 'total_reads', 'total_writes']
263360
detected_files.loc[:, 'id'] = detected_files.loc[:, 'id'].astype(str)
264361

265-
check_small_operation(total_reads, total_reads_small, total_writes, total_writes_small, detected_files, modules, file_map, df_posix)
362+
check_small_operation(total_reads, total_reads_small, total_writes, total_writes_small, detected_files, modules, file_map, dxt_posix, dxt_posix_read_data, dxt_posix_write_data)
266363

267364
#########################################################################################################################################################################
268365

@@ -271,7 +368,7 @@ def handler():
271368
total_mem_not_aligned = df['counters']['POSIX_MEM_NOT_ALIGNED'].sum()
272369
total_file_not_aligned = df['counters']['POSIX_FILE_NOT_ALIGNED'].sum()
273370

274-
check_misaligned(total_operations, total_mem_not_aligned, total_file_not_aligned, modules)
371+
check_misaligned(total_operations, total_mem_not_aligned, total_file_not_aligned, modules, file_map, df_lustre, dxt_posix, dxt_posix_read_data)
275372

276373
#########################################################################################################################################################################
277374

@@ -280,7 +377,7 @@ def handler():
280377
max_read_offset = df['counters']['POSIX_MAX_BYTE_READ'].max()
281378
max_write_offset = df['counters']['POSIX_MAX_BYTE_WRITTEN'].max()
282379

283-
check_traffic(max_read_offset, total_read_size, max_write_offset, total_written_size)
380+
check_traffic(max_read_offset, total_read_size, max_write_offset, total_written_size, dxt_posix, dxt_posix_read_data, dxt_posix_write_data)
284381

285382
#########################################################################################################################################################################
286383

@@ -305,7 +402,7 @@ def handler():
305402
write_random = total_writes - write_consecutive - write_sequential
306403
#print('WRITE Random: {} ({:.2f}%)'.format(write_random, write_random / total_writes * 100))
307404

308-
check_random_operation(read_consecutive, read_sequential, read_random, total_reads, write_consecutive, write_sequential, write_random, total_writes)
405+
check_random_operation(read_consecutive, read_sequential, read_random, total_reads, write_consecutive, write_sequential, write_random, total_writes, dxt_posix, dxt_posix_read_data, dxt_posix_write_data)
309406

310407
#########################################################################################################################################################################
311408

@@ -385,7 +482,7 @@ def handler():
385482

386483
column_names = ['id', 'data_imbalance']
387484
detected_files = pd.DataFrame(detected_files, columns=column_names)
388-
check_shared_data_imblance(stragglers_count, detected_files, file_map)
485+
check_shared_data_imblance(stragglers_count, detected_files, file_map, dxt_posix, dxt_posix_read_data, dxt_posix_write_data)
389486

390487
# POSIX_F_FASTEST_RANK_TIME
391488
# POSIX_F_SLOWEST_RANK_TIME
@@ -442,7 +539,7 @@ def handler():
442539

443540
column_names = ['id', 'write_imbalance']
444541
detected_files = pd.DataFrame(detected_files, columns=column_names)
445-
check_individual_write_imbalance(imbalance_count, detected_files, file_map)
542+
check_individual_write_imbalance(imbalance_count, detected_files, file_map, dxt_posix, dxt_posix_write_data)
446543

447544
imbalance_count = 0
448545

@@ -458,7 +555,7 @@ def handler():
458555

459556
column_names = ['id', 'read_imbalance']
460557
detected_files = pd.DataFrame(detected_files, columns=column_names)
461-
check_individual_read_imbalance(imbalance_count, detected_files, file_map)
558+
check_individual_read_imbalance(imbalance_count, detected_files, file_map, dxt_posix, dxt_posix_read_data)
462559

463560
#########################################################################################################################################################################
464561

@@ -493,7 +590,7 @@ def handler():
493590
column_names = ['id', 'absolute_indep_reads', 'percent_indep_reads']
494591
detected_files = pd.DataFrame(detected_files, columns=column_names)
495592

496-
check_mpi_collective_read_operation(mpiio_coll_reads, mpiio_indep_reads, total_mpiio_read_operations, detected_files, file_map)
593+
check_mpi_collective_read_operation(mpiio_coll_reads, mpiio_indep_reads, total_mpiio_read_operations, detected_files, file_map, dxt_mpiio)
497594

498595
df_mpiio_collective_writes = df_mpiio['counters'] #.loc[(df_mpiio['counters']['MPIIO_COLL_WRITES'] > 0)]
499596

@@ -518,7 +615,7 @@ def handler():
518615
column_names = ['id', 'absolute_indep_writes', 'percent_indep_writes']
519616
detected_files = pd.DataFrame(detected_files, columns=column_names)
520617

521-
check_mpi_collective_write_operation(mpiio_coll_writes, mpiio_indep_writes, total_mpiio_write_operations, detected_files, file_map)
618+
check_mpi_collective_write_operation(mpiio_coll_writes, mpiio_indep_writes, total_mpiio_write_operations, detected_files, file_map, dxt_mpiio)
522619

523620
#########################################################################################################################################################################
524621

0 commit comments

Comments
 (0)