Refactor code
This commit is contained in:
18
make/include_CLANG.mk
Normal file
18
make/include_CLANG.mk
Normal file
@@ -0,0 +1,18 @@
|
||||
CC = /opt/homebrew/Cellar/llvm/18.1.5/bin/clang
|
||||
LINKER = $(CC)
|
||||
|
||||
ANSI_CFLAGS = -ansi
|
||||
ANSI_CFLAGS += -std=c99
|
||||
ANSI_CFLAGS += -pedantic
|
||||
# ANSI_CFLAGS += -Wextra
|
||||
|
||||
CFLAGS = -Ofast -march=native $(ANSI_CFLAGS) -Xpreprocessor -fopenmp #-g
|
||||
#CFLAGS = -Ofast -march=core-avx2 $(ANSI_CFLAGS) #-Xpreprocessor -fopenmp -g
|
||||
#CFLAGS = -O3 -march=cascadelake $(ANSI_CFLAGS) #-Xpreprocessor -fopenmp -g
|
||||
#CFLAGS = -Ofast $(ANSI_CFLAGS) -g #-Xpreprocessor -fopenmp -g
|
||||
ASFLAGS = #-masm=intel
|
||||
LFLAGS =
|
||||
DEFINES = -D_GNU_SOURCE
|
||||
# MacOSX with Apple Silicon and homebrew
|
||||
INCLUDES = -I/opt/homebrew/Cellar/libomp/18.1.5/include/
|
||||
LIBS = -lm -L/opt/homebrew/Cellar/libomp/18.1.5/lib/ -lomp
|
||||
36
make/include_GCC.mk
Normal file
36
make/include_GCC.mk
Normal file
@@ -0,0 +1,36 @@
|
||||
CC = gcc
|
||||
LINKER = $(CC)
|
||||
|
||||
ANSI_CFLAGS = -ansi
|
||||
ANSI_CFLAGS += -std=c99
|
||||
ANSI_CFLAGS += -pedantic
|
||||
ANSI_CFLAGS += -Wextra
|
||||
|
||||
ifeq ($(ISA),AVX512)
|
||||
CFLAGS = -Ofast -mavx512f -mavx512vl -mavx512bw -mavx512dq -mavx512cd -ffast-math -funroll-loops # -fopenmp
|
||||
#CFLAGS = -O3 -march=cascadelake -ffast-math -funroll-loops # -fopenmp
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX2)
|
||||
#CFLAGS = -Ofast -march=native -mavx2 -ffast-math -funroll-loops # -fopenmp
|
||||
#CFLAGS = -O3 -march=znver1 -ffast-math -funroll-loops # -fopenmp
|
||||
#CFLAGS = -Ofast -mavx2 -ffast-math -funroll-loops # -fopenmp
|
||||
CFLAGS = -Ofast -mavx2 -mfma -ffast-math -funroll-loops # -fopenmp
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX)
|
||||
CFLAGS = -Ofast -mavx -ffast-math -funroll-loops # -fopenmp
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),SSE)
|
||||
CFLAGS = -Ofast -msse4.2 -ffast-math -funroll-loops # -fopenmp
|
||||
endif
|
||||
|
||||
#CFLAGS = -O0 -g -std=c99 -fargument-noalias
|
||||
#CFLAGS = -Ofast -march=native -ffast-math -funroll-loops # -fopenmp
|
||||
#CFLAGS = -O3 -march=native -ffast-math -funroll-loops # -fopenmp
|
||||
ASFLAGS = #-masm=intel
|
||||
LFLAGS =
|
||||
DEFINES = -D_GNU_SOURCE -DNO_ZMM_INTRIN
|
||||
INCLUDES = $(LIKWID_INC)
|
||||
LIBS = -lm
|
||||
11
make/include_GROMACS.mk
Normal file
11
make/include_GROMACS.mk
Normal file
@@ -0,0 +1,11 @@
|
||||
GROMACS_PATH=/apps/Gromacs/2018.1-mkl
|
||||
GROMACS_INC ?= -I${GROMACS_PATH}/include
|
||||
GROMACS_DEFINES ?=
|
||||
GROMACS_LIB ?= -L${GROMACS_PATH}/lib64
|
||||
|
||||
ifeq ($(strip $(XTC_OUTPUT)),true)
|
||||
INCLUDES += ${GROMACS_INC}
|
||||
DEFINES += ${GROMACS_DEFINES}
|
||||
LIBS += -lgromacs
|
||||
LFLAGS += ${GROMACS_LIB}
|
||||
endif
|
||||
32
make/include_ICC.mk
Normal file
32
make/include_ICC.mk
Normal file
@@ -0,0 +1,32 @@
|
||||
CC = icc
|
||||
LINKER = $(CC)
|
||||
|
||||
OPENMP = -qopenmp
|
||||
PROFILE = #-profile-functions -g -pg
|
||||
|
||||
ifeq ($(ISA),AVX512)
|
||||
OPTS = -Ofast -xCORE-AVX512 -qopt-zmm-usage=high $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX2)
|
||||
OPTS = -Ofast -xCORE-AVX2 $(PROFILE)
|
||||
#OPTS = -Ofast -xAVX2 $(PROFILE)
|
||||
#OPTS = -Ofast -march=core-avx2 $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX)
|
||||
OPTS = -Ofast -xAVX $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),SSE)
|
||||
OPTS = -Ofast -xSSE4.2 $(PROFILE)
|
||||
endif
|
||||
|
||||
#OPTS = -Ofast -no-vec $(PROFILE)
|
||||
#OPTS = -Ofast -xHost $(PROFILE)
|
||||
CFLAGS = $(PROFILE) -restrict $(OPENMP) $(OPTS)
|
||||
ASFLAGS = #-masm=intel
|
||||
LFLAGS = $(PROFILE) $(OPTS) $(OPENMP)
|
||||
DEFINES = -std=c11 -pedantic-errors -D_GNU_SOURCE
|
||||
INCLUDES =
|
||||
LIBS = -lm
|
||||
33
make/include_ICX.mk
Normal file
33
make/include_ICX.mk
Normal file
@@ -0,0 +1,33 @@
|
||||
CC = icx
|
||||
LINKER = $(CC)
|
||||
|
||||
OPENMP = #-qopenmp
|
||||
PROFILE = #-profile-functions -g -pg
|
||||
|
||||
ifeq ($(ISA),AVX512)
|
||||
OPTS = -Ofast -xCORE-AVX512 -qopt-zmm-usage=high $(PROFILE)
|
||||
#OPTS = -Ofast -march=cascadelake -xCORE-AVX512 -qopt-zmm-usage=high $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX2)
|
||||
OPTS = -Ofast -xCORE-AVX2 $(PROFILE)
|
||||
#OPTS = -Ofast -xHost $(PROFILE)
|
||||
#OPTS = -Ofast -march=core-avx2 $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),AVX)
|
||||
OPTS = -Ofast -xAVX $(PROFILE)
|
||||
endif
|
||||
|
||||
ifeq ($(ISA),SSE)
|
||||
OPTS = -Ofast -xSSE4.2 $(PROFILE)
|
||||
endif
|
||||
|
||||
#OPTS = -Ofast -no-vec $(PROFILE)
|
||||
#OPTS = -Ofast -xHost $(PROFILE)
|
||||
CFLAGS = $(PROFILE) $(OPENMP) $(OPTS)
|
||||
ASFLAGS = #-masm=intel
|
||||
LFLAGS = $(PROFILE) $(OPTS) $(OPENMP)
|
||||
DEFINES = -std=c11 -pedantic-errors -D_GNU_SOURCE -DNO_ZMM_INTRIN
|
||||
INCLUDES =
|
||||
LIBS = -lm
|
||||
29
make/include_ISA.mk
Normal file
29
make/include_ISA.mk
Normal file
@@ -0,0 +1,29 @@
|
||||
ifeq ($(strip $(ISA)), SSE)
|
||||
__ISA_SSE__=true
|
||||
__SIMD_WIDTH_DBL__=2
|
||||
else ifeq ($(strip $(ISA)), AVX)
|
||||
__ISA_AVX__=true
|
||||
__SIMD_WIDTH_DBL__=4
|
||||
else ifeq ($(strip $(ISA)), AVX_FMA)
|
||||
__ISA_AVX__=true
|
||||
__ISA_AVX_FMA__=true
|
||||
__SIMD_WIDTH_DBL__=4
|
||||
else ifeq ($(strip $(ISA)), AVX2)
|
||||
#__SIMD_KERNEL__=true
|
||||
__ISA_AVX2__=true
|
||||
__SIMD_WIDTH_DBL__=4
|
||||
else ifeq ($(strip $(ISA)), AVX512)
|
||||
__ISA_AVX512__=true
|
||||
__SIMD_WIDTH_DBL__=8
|
||||
ifeq ($(strip $(DATA_TYPE)), DP)
|
||||
__SIMD_KERNEL__=true
|
||||
endif
|
||||
endif
|
||||
|
||||
# SIMD width is specified in double-precision, hence it may
|
||||
# need to be adjusted for single-precision
|
||||
ifeq ($(strip $(DATA_TYPE)), SP)
|
||||
VECTOR_WIDTH=$(shell echo $$(( $(__SIMD_WIDTH_DBL__) * 2 )))
|
||||
else
|
||||
VECTOR_WIDTH=$(__SIMD_WIDTH_DBL__)
|
||||
endif
|
||||
10
make/include_LIKWID.mk
Normal file
10
make/include_LIKWID.mk
Normal file
@@ -0,0 +1,10 @@
|
||||
LIKWID_INC ?= -I/usr/local/include
|
||||
LIKWID_DEFINES ?= -DLIKWID_PERFMON
|
||||
LIKWID_LIB ?= -L/usr/local/lib
|
||||
|
||||
ifeq ($(strip $(ENABLE_LIKWID)),true)
|
||||
INCLUDES += ${LIKWID_INC}
|
||||
DEFINES += ${LIKWID_DEFINES}
|
||||
LIBS += -llikwid
|
||||
LFLAGS += ${LIKWID_LIB}
|
||||
endif
|
||||
23
make/include_NVCC.mk
Normal file
23
make/include_NVCC.mk
Normal file
@@ -0,0 +1,23 @@
|
||||
CC = nvcc
|
||||
LINKER = $(CC)
|
||||
|
||||
ANSI_CFLAGS = -ansi
|
||||
ANSI_CFLAGS += -std=c99
|
||||
ANSI_CFLAGS += -pedantic
|
||||
ANSI_CFLAGS += -Wextra
|
||||
|
||||
#
|
||||
# A100 + Native
|
||||
CFLAGS = -O3 -arch=sm_80 -march=native -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
|
||||
# A40 + Native
|
||||
#CFLAGS = -O3 -arch=sm_86 -march=native -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
|
||||
# Cascade Lake
|
||||
#CFLAGS = -O3 -march=cascadelake -ffast-math -funroll-loops --forward-unknown-to-host-compiler # -fopenmp
|
||||
# For GROMACS kernels, we need at least sm_61 due to atomicAdd with doubles
|
||||
# TODO: Check if this is required for full neighbor-lists and just compile kernel for that case if not
|
||||
#CFLAGS = -O3 -g -arch=sm_61 # -fopenmp
|
||||
ASFLAGS = -masm=intel
|
||||
LFLAGS =
|
||||
DEFINES = -D_GNU_SOURCE -DCUDA_TARGET -DNO_ZMM_INTRIN #-DLIKWID_PERFMON
|
||||
INCLUDES = $(LIKWID_INC)
|
||||
LIBS = -lm $(LIKWID_LIB) -lcuda -lcudart #-llikwid
|
||||
17
make/include_ONEAPI.mk
Normal file
17
make/include_ONEAPI.mk
Normal file
@@ -0,0 +1,17 @@
|
||||
CC = icx
|
||||
LINKER = $(CC)
|
||||
|
||||
OPENMP = -qopenmp-simd
|
||||
PROFILE = #-g -pg
|
||||
#OPTS = -Ofast -no-vec
|
||||
#OPTS = -Ofast -xSSE4.2
|
||||
#OPTS = -Ofast -xAVX
|
||||
#OPTS = -Ofast -xCORE-AVX2
|
||||
OPTS = -Ofast -xCORE-AVX512 -qopt-zmm-usage=high
|
||||
#OPTS = -Ofast -xHost
|
||||
CFLAGS = $(PROFILE) $(OPTS) $(OPENMP)
|
||||
ASFLAGS = -masm=intel
|
||||
LFLAGS = $(PROFILE) $(OPTS)
|
||||
DEFINES = -D_GNU_SOURCE -DNOCHUNK
|
||||
INCLUDES =
|
||||
LIBS = -lm
|
||||
Reference in New Issue
Block a user