feat: Implement hybrid MPI+OpenMP parallelization

- Enable -qopenmp in makefile.inc
- Add OpenMP directives to 4th order derivatives in diff_new.f90
- Update makefile_and_run.py to dynamic calculate OMP_NUM_THREADS based on 96 cores and remove hardcoded CPU binding

AMSS_NCKU_source/diff_new.f90

@@ -69,6 +69,7 @@
   fy = ZEO
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -151,6 +152,7 @@
 
   fx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -218,6 +220,7 @@
 
   fy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -282,6 +285,7 @@
 
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -371,6 +375,7 @@
   fxz = ZEO
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -469,6 +474,7 @@
 
   fxx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -531,6 +537,7 @@
 
   fyy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -594,6 +601,7 @@
 
   fzz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -657,6 +665,7 @@
 
   fxy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -719,6 +728,7 @@
 
   fxz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -780,6 +790,7 @@
 
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -866,6 +877,7 @@
   fxz = ZEO
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -997,6 +1009,7 @@
   fy = ZEO
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1151,6 +1164,7 @@
 
   fx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1227,6 +1241,7 @@
 
   fy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1297,6 +1312,7 @@
 
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1401,6 +1417,7 @@
   fxz = ZEO
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1576,6 +1593,7 @@
 
   fxx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1643,6 +1661,7 @@
 
   fyy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1712,6 +1731,7 @@
 
   fzz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1781,6 +1801,7 @@
 
   fxy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1851,6 +1872,7 @@
 
   fxz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -1919,6 +1941,7 @@
 
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2011,6 +2034,7 @@
   fy = ZEO
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2127,6 +2151,7 @@
 
   fx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2212,6 +2237,7 @@
 
   fy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2288,6 +2314,7 @@
 
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2406,6 +2433,7 @@
   fxz = ZEO
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2593,6 +2621,7 @@
 
   fxx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2665,6 +2694,7 @@
 
   fyy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2740,6 +2770,7 @@
 
   fzz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2815,6 +2846,7 @@
 
   fxy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2895,6 +2927,7 @@
 
   fxz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -2973,6 +3006,7 @@
 
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3080,6 +3114,7 @@
   fy = ZEO
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3216,6 +3251,7 @@
 
   fx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3311,6 +3347,7 @@
 
   fy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3395,6 +3432,7 @@
 
   fz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3530,6 +3568,7 @@
   fxz = ZEO
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3802,6 +3841,7 @@
 
   fxx = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3883,6 +3923,7 @@
 
   fyy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -3967,6 +4008,7 @@
   
   fzz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -4051,6 +4093,7 @@
 
   fxy = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -4153,6 +4196,7 @@
 
   fxz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1
@@ -4253,6 +4297,7 @@
 
   fyz = ZEO
 
+  !$omp parallel do collapse(3) schedule(static)
   do k=1,ex(3)-1
   do j=1,ex(2)-1
   do i=1,ex(1)-1

AMSS_NCKU_source/makefile.inc

@@ -15,16 +15,16 @@ LDLIBS  = -L${MKLROOT}/lib -lmkl_intel_lp64 -lmkl_sequential -lmkl_core -lifcore
 ## -xHost: Optimize for the host CPU architecture (Intel/AMD compatible)
 ## -fp-model fast=2: Aggressive floating-point optimizations
 ## -fma: Enable fused multiply-add instructions
-## Note: OpenMP has been disabled (-qopenmp removed) due to performance issues
-CXXAPPFLAGS  = -O3 -xHost -fp-model fast=2 -fma \
+## Note: OpenMP enabled for hybrid MPI+OpenMP
+CXXAPPFLAGS  = -O3 -xHost -fp-model fast=2 -fma -qopenmp \
                -Dfortran3 -Dnewc -I${MKLROOT}/include
-f90appflags  = -O3 -xHost -fp-model fast=2 -fma \
+f90appflags  = -O3 -xHost -fp-model fast=2 -fma -qopenmp \
                -fpp -I${MKLROOT}/include
 f90          = ifx
 f77          = ifx
 CXX          = icpx
 CC           = icx
-CLINKER      = mpiicpx 
+CLINKER      = mpiicpx -qopenmp 
 
 Cu = nvcc
 CUDA_LIB_PATH = -L/usr/lib/cuda/lib64 -I/usr/include -I/usr/lib/cuda/include

makefile_and_run.py

@@ -11,16 +11,14 @@
 import AMSS_NCKU_Input as input_data
 import subprocess
 
-## CPU core binding configuration using taskset
-## taskset ensures all child processes inherit the CPU affinity mask
-## This forces make and all compiler processes to use only nohz_full cores (4-55, 60-111)
-## Format: taskset -c 4-55,60-111 ensures processes only run on these cores
-NUMACTL_CPU_BIND = "taskset -c 4-55,60-111"
+## CPU core binding configuration
+## Removed hardcoded taskset to allow full utilization of 96 cores via MPI+OpenMP
+NUMACTL_CPU_BIND = ""
 
 ## Build parallelism configuration
 ## Use nohz_full cores (4-55, 60-111) for compilation: 52 + 52 = 104 cores
 ## Set make -j to utilize available cores for faster builds
-BUILD_JOBS = 104
+BUILD_JOBS = 96
 
 
 ##################################################################
@@ -37,7 +35,7 @@ def makefile_ABE():
     print( " Compiling the AMSS-NCKU executable file ABE/ABEGPU " ) 
     print(                                                        )
 
-    ## Build command with CPU binding to nohz_full cores
+    ## Build command
     if (input_data.GPU_Calculation == "no"):
         makefile_command  = f"{NUMACTL_CPU_BIND} make -j{BUILD_JOBS} ABE"
     elif (input_data.GPU_Calculation == "yes"):
@@ -78,7 +76,7 @@ def makefile_TwoPunctureABE():
     print( " Compiling the AMSS-NCKU executable file TwoPunctureABE " )
     print(                                                            )
     
-    ## Build command with CPU binding to nohz_full cores
+    ## Build command
     makefile_command = f"{NUMACTL_CPU_BIND} make -j{BUILD_JOBS} TwoPunctureABE"
 
     ## Execute the command with subprocess.Popen and stream output
@@ -113,13 +111,23 @@ def run_ABE():
     print( " Running the AMSS-NCKU executable file ABE/ABEGPU " ) 
     print(                                                      )
 
+    ## Calculate OMP_NUM_THREADS
+    ## User has 96 cores. Calculate threads per MPI process.
+    total_physical_cores = 96
+    omp_num_threads = total_physical_cores // input_data.MPI_processes
+    if omp_num_threads < 1:
+        omp_num_threads = 1
+    
+    print( f" Configuration: {input_data.MPI_processes} MPI processes, {omp_num_threads} OpenMP threads per process." )
+    print( f" Total cores utilized: {input_data.MPI_processes * omp_num_threads}" )
+
     ## Define the command to run; cast other values to strings as needed
     
     if (input_data.GPU_Calculation == "no"):
-        mpi_command         = NUMACTL_CPU_BIND + " mpirun -np " + str(input_data.MPI_processes) + " ./ABE"
+        mpi_command         = f"{NUMACTL_CPU_BIND} mpirun -genv OMP_NUM_THREADS {omp_num_threads} -np {input_data.MPI_processes} ./ABE"
         mpi_command_outfile = "ABE_out.log"
     elif (input_data.GPU_Calculation == "yes"):
-        mpi_command         = NUMACTL_CPU_BIND + " mpirun -np " + str(input_data.MPI_processes) + " ./ABEGPU"
+        mpi_command         = f"{NUMACTL_CPU_BIND} mpirun -genv OMP_NUM_THREADS {omp_num_threads} -np {input_data.MPI_processes} ./ABEGPU"
         mpi_command_outfile = "ABEGPU_out.log"
  
     ## Execute the MPI command and stream output