Changes between Version 41 and Version 42 of Public/User_Guide/OmpSs-2

Timestamp:

Jun 14, 2019, 4:52:08 PM (5 years ago)

Author:

Pedro Martinez-Ferror

Comment:

—

Public/User_Guide/OmpSs-2

@@ -185 +185 @@
 `ssh -X USERNAME@deep.fz-juelich.de`
 
-and from there go to the ''multisaxpy'' folder
+and from there open the ''multisaxpy''
 
 `cd /work/cdeep/USERNAME/multisaxpy`
@@ -216 +216 @@
 }}}
 
-Now you should be able to clean, build and execute this benchmark consisting of 7 implementations via the command `make`:
+Now you should be able to clean, build and execute this benchmark via the command `make`:
 {{{
 $ make clean
@@ -260 +260 @@
 }}}
 
+== Override the Number of Threads Used ==
+
+Let's have a closer look at the third implementation ''03.multisaxpy_task'' which took 0.648536 seconds to finish using 48 threads.
+
+A full CM node features 48 threads (0-47) divided in two sockets: 0-11,24-35 for the first socket and 12-23,36-47 for the second socket.  **Notice that they are indeed not consecutive! **
+
+We can change the threads used by !OmpSs-2 with the Linux command `taskset`.  For example, the command to run this binary with 24 threads interleaved between the two sockets would be:
+
+`taskset -c 0-23 ./03.multisaxpy_task 16777216 8192 100`
+
+Similarly, to run this benchmark using all the 24 threads of the second socket use the following command:
+
+`taskset -c 12-23,36-47 ./03.multisaxpy_task 16777216 8192 100`
+
+You can also try to run this example with only 12 threads of the first socket:
+
+`taskset -c 0-11 ./03.multisaxpy_task 16777216 8192 100`
+
+or 12 threads interleaved between the two sockets:
+
+`taskset -c 0-5,12-17 ./03.multisaxpy_task 16777216 8192 100`
+
+Changing the number of threads assigned to !OmpSs-2 affects the performance of the application:
+{{{
+$ ./03.multisaxpy_task 16777216 8192 100
+size: 16777216, bs: 8192, iterations: 100, time: 0.653537, performance: 2.56714
+$ taskset -c 0-23 ./03.multisaxpy_task 16777216 8192 100
+size: 16777216, bs: 8192, iterations: 100, time: 0.686265, performance: 2.44471
+$ taskset -c 12-23,36-47 ./03.multisaxpy_task 16777216 8192 100
+size: 16777216, bs: 8192, iterations: 100, time: 0.650363, performance: 2.57967
+$ taskset -c 0-11 ./03.multisaxpy_task 16777216 8192 100
+size: 16777216, bs: 8192, iterations: 100, time: 0.55417, performance: 3.02745
+$ taskset -c 0-5,12-17 ./03.multisaxpy_task 16777216 8192 100
+size: 16777216, bs: 8192, iterations: 100, time: 0.705859, performance: 2.37685
+}}}
+
+
+
 == References ==