ICODE-ADC/lammps
4
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

make flags for supporting GPU archs consistent, add ampere consumer GPU arch

Browse Source
This commit is contained in:
Axel Kohlmeyer
2021-02-07 15:30:55 -05:00
parent 8134296311
commit 06e484d3e0
5 changed files with 26 additions and 5 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
cmake/Modules/Packages/GPU.cmake
View File

@ -97,9 +97,9 @@ if(GPU_API STREQUAL "CUDA")
if(CUDA_VERSION VERSION_GREATER_EQUAL "10.0")
string(APPEND GPU_CUDA_GENCODE " -gencode arch=compute_75,code=[sm_75,compute_75]")
endif()
# Ampere (GPU Arch 8.0) is supported by CUDA 11 and later
# Ampere (GPU Arch 8.0 and 8.6) is supported by CUDA 11 and later
if(CUDA_VERSION VERSION_GREATER_EQUAL "11.0")
string(APPEND GPU_CUDA_GENCODE " -gencode arch=compute_80,code=[sm_80,compute_80]")
string(APPEND GPU_CUDA_GENCODE " -gencode arch=compute_80,code=[sm_80,compute_80] -gencode arch=compute_86,code=[sm_86,compute_86]")
endif()
if(CUDA_VERSION VERSION_GREATER_EQUAL "12.0")
message(WARNING "Unsupported CUDA version. Use at your own risk.")

3
lib/gpu/Makefile.cuda
View File

@ -38,7 +38,8 @@ NVCC = nvcc
CUDA_ARCH = -gencode arch=compute_50,code=[sm_50,compute_50] -gencode arch=compute_52,code=[sm_52,compute_52] \
-gencode arch=compute_60,code=[sm_60,compute_60] -gencode arch=compute_61,code=[sm_61,compute_61] \
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75]
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75] \
-gencode arch=compute_80,code=[sm_80,compute_80] -gencode arch=compute_86,code=[sm_86,compute_86]
CUDA_INCLUDE = -I$(CUDA_HOME)/include
CUDA_LIB = -L$(CUDA_HOME)/lib64 -L$(CUDA_HOME)/lib64/stubs
CUDA_OPTS = -DUNIX -O3 --use_fast_math $(LMP_INC) -Xcompiler -fPIC

3
lib/gpu/Makefile.cuda_mps
View File

@ -38,7 +38,8 @@ NVCC = nvcc
CUDA_ARCH = -gencode arch=compute_50,code=[sm_50,compute_50] -gencode arch=compute_52,code=[sm_52,compute_52] \
-gencode arch=compute_60,code=[sm_60,compute_60] -gencode arch=compute_61,code=[sm_61,compute_61] \
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75]
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75] \
-gencode arch=compute_80,code=[sm_80,compute_80] -gencode arch=compute_86,code=[sm_86,compute_86]
CUDA_INCLUDE = -I$(CUDA_HOME)/include
CUDA_LIB = -L$(CUDA_HOME)/lib64 -L$(CUDA_HOME)/lib64/stubs
CUDA_OPTS = -DUNIX -O3 --use_fast_math $(LMP_INC) -Xcompiler -fPIC

4
lib/gpu/Makefile.linux
View File

@ -41,6 +41,10 @@ CUDA_ARCH = -arch=sm_50
# Turing hardware
#CUDA_ARCH = -arch=sm_75
# Ampere hardware
#CUDA_ARCH = -arch=sm_80
#CUDA_ARCH = -arch=sm_86
# this setting should match LAMMPS Makefile
# one of LAMMPS_SMALLBIG (default), LAMMPS_BIGBIG and LAMMPS_SMALLSMALL

17
lib/gpu/Makefile.linux_multi
View File

@ -31,9 +31,24 @@ NVCC = nvcc
CUDA_ARCH = -arch=sm_50
#CUDA_ARCH = -arch=sm_52
# Pascal hardware
#CUDA_ARCH = -arch=sm_60
#CUDA_ARCH = -arch=sm_61
# Volta hardware
#CUDA_ARCH = -arch=sm_70
# Turing hardware
#CUDA_ARCH = -arch=sm_75
# Ampere hardware
#CUDA_ARCH = -arch=sm_80
#CUDA_ARCH = -arch=sm_86
CUDA_CODE = -gencode arch=compute_50,code=[sm_50,compute_50] -gencode arch=compute_52,code=[sm_52,compute_52] \
-gencode arch=compute_60,code=[sm_60,compute_60] -gencode arch=compute_61,code=[sm_61,compute_61] \
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75]
-gencode arch=compute_70,code=[sm_70,compute_70] -gencode arch=compute_75,code=[sm_75,compute_75] \
-gencode arch=compute_80,code=[sm_80,compute_80] -gencode arch=compute_86,code=[sm_86,compute_86]
CUDA_ARCH += $(CUDA_CODE)