diff --git a/source/KKRnano/source/Makefile b/source/KKRnano/source/Makefile index 04592b9fb0cb1661d10ed765cdd3f2aa20b286b1..d4893402d597b067ef3b94e23d594ed235a7c2f6 100644 --- a/source/KKRnano/source/Makefile +++ b/source/KKRnano/source/Makefile @@ -70,7 +70,7 @@ ifeq ($(PLATFORM),ifort) FCFLAGS += -O0 -g -check bounds -traceback -debug all -check all PPFLAGS += -D USE_MTRADII -D USE_OLD_MESH -D NOLOGGING else ifeq ($(TYPE),voronoi_mesh) - FCFLAGS += -O2 -xHost + FCFLAGS += -O2 PPFLAGS += -D NOLOGGING -D USE_OLD_MESH else ifeq ($(TYPE),voronoi_mesh_debug) FCFLAGS += -O0 -g -check bounds -traceback -debug all -check all -fpe1 @@ -83,7 +83,7 @@ ifeq ($(PLATFORM),ifort) PPFLAGS += -D NOLOGGING -D USE_OLD_MESH -D PRINT_MTRADII else - FCFLAGS += -O2 -xHost + FCFLAGS += -O2 PPFLAGS += -D NOLOGGING endif endif @@ -116,7 +116,6 @@ endif # ================== JURECA ====================== -# optimal setup on one node seems to be 8MPI/3OMP for 64 atoms # The following software stages were tested and work: # module --force purge # module use /usr/local/software/jureca/OtherStages @@ -132,21 +131,21 @@ ifeq ($(PLATFORM),jureca) FCFLAGS += -module $(BUILDDIR) FCFLAGS += -I $(BUILDDIR) # PPFLAGS += -D TASKLOCAL_FILES - FCFLAGS += -g + FCFLAGS += -g -march=core-avx2 LDFLAGS += -g -L${MKLROOT}/lib/intel64 -lmkl_intel_lp64 -lmkl_intel_thread -lmkl_core -liomp5 -lpthread -lm -ldl ifeq ($(TYPE),debug_light) FCFLAGS += -O0 -g else ifeq ($(TYPE),debug) FCFLAGS += -O0 -g -check bounds -traceback -debug all -check all -fpe1 else ifeq ($(TYPE),mtradii) - FCFLAGS += -O2 -xHost + FCFLAGS += -O2 PPFLAGS += -D USE_MTRADII PPFLAGS += -D NOLOGGING -D USE_OLD_MESH else ifeq ($(TYPE),debug_mtradii) FCFLAGS += -O0 -g -check bounds -traceback -debug all -check all PPFLAGS += -D USE_MTRADII else ifeq ($(TYPE),voronoi_mesh) - FCFLAGS += -O2 -xHost + FCFLAGS += -O2 PPFLAGS += -D NOLOGGING -D USE_OLD_MESH else ifeq ($(TYPE),voronoi_mesh_debug) FCFLAGS += -O0 -g -check bounds -traceback -debug all -check all -fpe1 @@ -155,7 +154,7 @@ ifeq ($(PLATFORM),jureca) FCFLAGS += -O0 -g PPFLAGS += -D NOLOGGING -D USE_OLD_MESH else - FCFLAGS += -O2 -xHost + FCFLAGS += -O2 PPFLAGS += -D NOLOGGING endif endif @@ -204,14 +203,14 @@ ifeq ($(PLATFORM),jureca_knl) PPFLAGS += -D NOLOGGING endif endif -# ================== HAZEL HEN ====================== +# ================== HAWK =========================== # Usage of Intel compiler and Intel MKL seems to give best results: # module swap PrgEnv-cray PrgEnv-intel # module unload cray-libsci # Remarks: - Bad GGA performance -> Use LDA! # - Home directory slow -> Use Workspace (see HLRS wiki) # - If 'seg 0' error occurs, add 'module unload craype-hugepages16M' to job file -ifeq ($(PLATFORM),hazelhen) +ifeq ($(PLATFORM),hawk) FC = ftn FC90 = ftn # -module <path> specifies where to put .mod files @@ -565,5 +564,6 @@ ChebMeshData_mod.o: InputParams_mod.o RadialMeshData_mod.o Truncation_mod.o: Logging_mod.o Exceptions_mod.o TruncationZone_mod.o NonCollinearMagnetism_mod.o: RadialMeshData_mod.o ChebMeshData_mod.o read_formatted_shapefun_mod.o NonCollinearMagnetismData_mod.o: Exceptions_mod.o +vintras_new.o: SingleSiteHelpers_mod.o # DO NOT DELETE