01_processor-core/02_simd-if

Hands-on: SIMD in loops with if statements

• Choose either C/C++ (cpp) or Fortran (fortran) samples. Both of them are fine, as well.

C/C++

How to compile and how to execute

1. Run a script of creating working space

• edit task.sh (modify --gname option).
• See objst in create_project.sh.
  • It indicates a list of the settings of compile options.
• Running create_project.sh, the executable file (run.x) is generated in Obj_*** directory.
  • This directory is also automatically generated.
  • When you desire to use the own compiler, set MAKE_DIR variable in create_project.sh from config to config.own.
• Example:

$ cd cpp
$ bash create_project.sh 
$ ls 
Obj_clang  Obj_clang.novec  

2. Run program

• A job script to execute the program is located in Obj_***/results.
• You can run the program either:

## To run as a batch job
$ cd Obj_clang/results
$ pjsub task.sh  
## Or, to run in an interactive job
$ cd Obj_clang/results
$ bash task.sh  

• Each of the samples in the Exercises will be completed within 1 minutes.
  • For safety, we set the upper limit of job elapsed time as 3 minutes.

Exercises A

• E1: Check the compiler reports (*.lst) about Makefile.clang or Makefile.clang.novec. Are the loops of the kernels (calc1, calc2, and calc2_mod) successfully vectorized?
• E2: Compare the performance (Elapsed time) in Obj_clang.novec to that in Obj_clang, to consider whether vectorization is effective even in the presence of if statements inside a loop.

Exercises B (advanced)

• E3: Set the second argument of run.x as 0 (Thus, the if statements in mykernel.cpp are always false). Then, check the elapsed time.
• E4: Test the case of cpp_mix-trad, in which the main part is compiled with Clang mode but a part of mykernel.cpp (renamed by optkernel.cpp) with Trad mode.

Note

• We summarize the main compiler options of each Makefile.

Makefile.clang         #Clang mode. Auto-vectorization with scalable vector length and interleaving are allowed. 
Makefile.clang.512     #Clang mode. Auto-vectorization with 512-bit is allowed.
Makefile.clang.novec   #Clang mode, without vectorization. 
Makefile.trad.nosimd   #Trad mode, without vectorization.
Makefile.trad.simd     #Trad mode. Auto-vectorization with -Ksimd=auto is allowed.
Makefile.trad.simd2    #Trad mode. Auto-vectorization with -Ksimd=2 is allowed.

Fortran

1. Run a script of creating working space

• edit task.sh (modify --gname option).
• See objst in create_project.sh.
  • It indicates a list of the settings of compile options.
• Running create_project.sh, the executable file (run.x) is generated in Obj_*** directory.
  • This directory is also automatically generated.
  • When you desire to use the own compiler, set MAKE_DIR variable in create_project.sh from config to config.own.
• Example:

$ cd fortran
$ bash create_project.sh 
$ ls 
Obj_nosimd  Obj_simd  Obj_simd2  

2. Run program

• A job script to execute the program is located in Obj_***/results.
• You can run the program either:

## To run as a batch job
$ cd Obj_simd/results
$ pjsub task.sh  
## Or, to run in an interactive job
$ cd Obj_simd/results
$ bash task.sh  

• Each of the samples in the Exercises will be completed within 1 minutes.
  • For safety, we set the upper limit of job elapsed time as 3 minutes.

Exercises A

• E1: Check the compiler reports (*.lst) about Makefile.simd and Makefile.simd2. Are the loops of the kernels (calc1, calc2, and calc2_mod) successfully vectorized?
• E2: Compare the performance (Elapsed time) in Obj_nosimd to that in Obj_simd or Obj_simd2, to consider whether vectorization is effective even in the presence of if statements inside a loop.

Exercises B (advanced)

• E3: Set the second argument of run.x as 0 (Thus, the if statements in mykernel.cpp are always false). Then, check the elapsed time.

Note

• We summarize the main compiler options of each Makefile.

Makefile.nosimd  #Auto-vectorization is not allowed. 
Makefile.simd    #Auto-vectorization with -Ksimd=auto is allowed.
Makefile.simd2   #Auto-vectorization with -Ksimd=2 is allowed.