[Bug gprofng/30089] New: [display text] Incorrect thread count shown on aarch64

[ruud.vanderpas at oracle dot com]

https://sourceware.org/bugzilla/show_bug.cgi?id=30089

            Bug ID: 30089
           Summary: [display text] Incorrect thread count shown on aarch64
           Product: binutils
           Version: 2.40
            Status: NEW
          Severity: normal
          Priority: P2
         Component: gprofng
          Assignee: vladimir.mezentsev at oracle dot com
          Reporter: ruud.vanderpas at oracle dot com
  Target Milestone: ---

As the sample output below shows, the number of threads shown with "gprofng
display text" is incorrect on aarch64.

This tested was conducted on an Oracle Linux 9 ("OL9") aarch64 system with
multiple cores.

The sample output below shows that the performance improves when increasing the
number of threads from 1 to 2.

Then the gprofng experiment is conducted using 2 threads. The printed elapsed
time of about 0.6 seconds confirms that both threads are used in the
computation.

But the output of "gprofng display text" shows the information for one thread
only.

====================  LOG of the session =====================
$ cat run.res
+ exe=../bindir/mxv-pthreads
+ ../bindir/mxv-pthreads -t 1 -v
Verbose mode enabled
Allocated data structures
Timings for the initialization phase
  elapsed time init_data = 0.066946
  cpu     time init_data = 0.066852
  system  time init_data = 0.008289
Initialized matrix and vectors
Defined workload distribution
Assigned work to threads
Thread 0 has been created
Timings for the computational phase
  elapsed time mxv = 1.253865
  cpu     time mxv = 1.251356
  system  time mxv = 0.000000
Matrix vector multiplication has completed
Verify correctness of result
Error check passed
mxv: error check passed - rows = 2000 columns = 3000 threads = 1
+ ../bindir/mxv-pthreads -t 2 -v
Verbose mode enabled
Allocated data structures
Timings for the initialization phase
  elapsed time init_data = 0.067124
  cpu     time init_data = 0.067025
  system  time init_data = 0.004770
Initialized matrix and vectors
Defined workload distribution
Assigned work to threads
Thread 0 has been created
Thread 1 has been created
Timings for the computational phase
  elapsed time mxv = 0.613867
  cpu     time mxv = 1.222008
  system  time mxv = 0.003657
Matrix vector multiplication has completed
Verify correctness of result
Error check passed
mxv: error check passed - rows = 2000 columns = 3000 threads = 2
+ gprofng --version
GNU gprofng binutils version 2.40.50
Copyright (C) 2023 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
+ gprofng collect app -O mxv.2.thr.er ../bindir/mxv-pthreads -t 2 -v
CPU clock frequency: 25000000
Creating experiment directory mxv.2.thr.er (Process ID: 1147696) ...
Verbose mode enabled
Allocated data structures
Timings for the initialization phase
  elapsed time init_data = 0.067561
  cpu     time init_data = 0.067397
  system  time init_data = 0.014307
Initialized matrix and vectors
Defined workload distribution
Assigned work to threads
Thread 0 has been created
Thread 1 has been created
Timings for the computational phase
  elapsed time mxv = 0.618479
  cpu     time mxv = 1.230675
  system  time mxv = 0.000000
Matrix vector multiplication has completed
Verify correctness of result
Error check passed
mxv: error check passed - rows = 2000 columns = 3000 threads = 2
+ gprofng display text -threads ./mxv.2.thr.er
Objects sorted by metric: Exclusive Total CPU Time

Excl. Total    Name
CPU
 sec.      %
0.060 100.00   <Total>
0.060 100.00   Process 1, Thread 1

$

-- 
You are receiving this mail because:
You are on the CC list for the bug.