Skip to content

Commit

Permalink
Merge pull request merzlab#348 from ohearnk/build-updates-intel-oneapi
Browse files Browse the repository at this point in the history 
Add build support for Intel OneAPI / LLVM compilers
  • Loading branch information
@agoetz
agoetz authored Mar 21, 2024
2 parents ef4d027 + b316ae3 commit a15d07c
Show file tree
Hide file tree
Showing 9 changed files with 104 additions and 21 deletions.
20 changes: 14 additions & 6 deletions cmake/AmberCompilerConfig.cmake
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2,28 +2,32 @@

# Handle the COMPILER option
# --------------------------------------------------------------------
set(ALL_COMPILER_VALUES GNU INTEL PGI CRAY CLANG MSVC AUTO MANUAL)
set(ALL_COMPILER_VALUES GNU INTEL INTELLLVM ONEAPI PGI CRAY CLANG MSVC AUTO MANUAL)

# help message displayed when COMPILER is unset or invalid
set(COMPILER_HELP "
Supported ${PROJECT_NAME} compilers:
-----------------------------------------------------------------------
COMPILER value | C executable | C++ executable | Fortran executable | tested versions
--------------------------------------------------------------------------------------
GNU | gcc | g++ | gfortran | 4.8.5+
INTEL | icc | icpc | ifort | 19
PGI | pgcc | pgc++ | pgf90 |
GNU | gcc | g++ | gfortran | 6.0 +
INTEL | icc | icpc | ifort | 19 - 22
INTELLLVM | icx | icpx | ifx | 2024
ONEAPI | icx | icpx | ifx | 2024
PGI | pgcc | pgc++ | pgf90 | 14.9, 15.4, 16.5
CLANG | clang | clang++ | gfortran |
CRAY | cc | CC | ftn |
CRAY | cc | CC | ftn | 8.4.6*
--------------------------------------------------------------------------------------
AUTO | <uses the default CMake-chosen compilers>
MANUAL | <uses the CC, CXX, and FC environment variables, or the
CMAKE_<LANGUAGE>_COMPILER CMake variables>
--------------------------------------------------------------------------------------
Note that INTEL is Intel classic, INTELLLVM is Intel oneAPI, and ONEAPI is INTELLLVM.
")


#create actual option
set(COMPILER "" CACHE STRING "Compiler to build Amber with. Valid values: GNU, INTEL, PGI, CRAY, CLANG, MSVC, AUTO, MANUAL. If 'auto', autodetect the host compiler, or use the CC,CXX,and FC variables if they are set.
set(COMPILER "" CACHE STRING "Compiler to build Amber with. Valid values: GNU, INTEL, INTELLLVM, ONEAPI, PGI, CRAY, CLANG, MSVC, AUTO, MANUAL. If 'auto', autodetect the host compiler, or use the CC,CXX,and FC variables if they are set.
This option can ONLY be set the first time CMake is run. If you want to change it, delete the build directory and start over.")

set(COMPILER_VALID FALSE)
Expand Down Expand Up @@ -139,6 +143,10 @@ if(FIRST_RUN)
set_compiler(C icc)
set_compiler(CXX icpc)
set_compiler(Fortran ifort)
elseif(${COMPILER} STREQUAL INTELLLVM OR ${COMPILER} STREQUAL ONEAPI)
set_compiler(C icx)
set_compiler(CXX icpx)
set_compiler(Fortran ifx)
elseif(${COMPILER} STREQUAL PGI)
set_compiler(C pgcc)
set_compiler(CXX pgc++)
Expand Down
73 changes: 73 additions & 0 deletions cmake/CompilerFlags.cmake
View file
Original file line number Diff line number Diff line change
Expand Up @@ -333,6 +333,79 @@ if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "Intel")
set(OPT_CXXFLAGS -O3)
endif()

#IntelLLVM aka Intel oneAPI
#---------------------------------------------------------------------------------------------------------------------------------------------------------------------

if("${CMAKE_C_COMPILER_ID}" STREQUAL "IntelLLVM")
set(CMAKE_C_FLAGS_DEBUG "-g -debug all")
set(OPT_CFLAGS -ip -O3)

# How flags get set for optimization depend on whether we have a MIC processor,
# the version of Intel compiler we have, and whether we are cross-compiling
# for multiple versions of SSE support. The following coordinates all of this.
# This was done assuming that MIC and SSE are mutually exclusive and that we want
# SSE instructions included only when optimize = yes. Note that use of an
# SSE_TYPES specification needs to be given in place of xHost not in addition to.
# This observed behavior is not what is reported by the Intel man pages. BPK

if(SSE)
# BPK removed section that modified O1 or O2 to be O3 if optimize was set to yes.
# We already begin with the O3 setting so it wasn't needed.
# For both coptflags and foptflags, use the appropriate settings
# for the sse flags (compiler version dependent).
if(${CMAKE_C_COMPILER_VERSION} VERSION_GREATER 11 OR ${CMAKE_C_COMPILER_VERSION} VERSION_EQUAL 11)
if(NOT "${SSE_TYPES}" STREQUAL "")
list(APPEND OPT_CFLAGS "-ax${SSE_TYPES}")
else()
list(APPEND OPT_CFLAGS -xHost)
endif()
else()
list(APPEND OPT_CFLAGS -axSTPW)
endif()

endif()
endif()

if("${CMAKE_Fortran_COMPILER_ID}" STREQUAL "IntelLLVM")

if(WIN32)
add_flags(Fortran /D_CRT_SECURE_NO_WARNINGS)
set(OPT_FFLAGS "/Ox")
set(CMAKE_Fortran_FLAGS_DEBUG "/Zi")
else()
set(CMAKE_Fortran_FLAGS_DEBUG "-g -debug all")
set(OPT_FFLAGS -ip -O3)

if(SSE)
if("${CMAKE_Fortran_COMPILER_VERSION}" VERSION_GREATER 11 OR ${CMAKE_Fortran_COMPILER_VERSION} VERSION_EQUAL 11)
if(NOT "${SSE_TYPES}" STREQUAL "")
list(APPEND OPT_FFLAGS "-ax${SSE_TYPES}")
else()
list(APPEND OPT_FFLAGS -xHost)
endif()
else()
list(APPEND OPT_FFLAGS -axSTPW)
endif()
endif()

# warning flags
add_flags(Fortran "-warn all" "-warn nounused")

option(IFORT_CHECK_INTERFACES "If enabled and Intel Fortran is in use, then ifort will check that types passed to functions are the correct ones, and produce warnings or errors for mismatches." FALSE)

if(NOT IFORT_CHECK_INTERFACES)
# disable errors from type mismatches
add_flags(Fortran -warn nointerfaces)
endif()

endif()
endif()

if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "IntelLLVM")
set(CMAKE_CXX_FLAGS_DEBUG "-g -debug all")
set(OPT_CXXFLAGS -O3)
endif()

# PGI
#---------------------------------------------------------------------------------------------------------------------------------------------------------------------
if("${CMAKE_C_COMPILER_ID}" STREQUAL "PGI")
Expand Down
2 changes: 1 addition & 1 deletion src/calMP2.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -485,7 +485,7 @@ subroutine MPI_calmp2
else
do i=1,mpisize-1
! receive integrals from slave nodes
call MPI_RECV(temp4d,nbasis*ivir*iocc*nsteplength,mpi_double_precision,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
call MPI_RECV(temp4d,nbasis*ivir*iocc*nsteplength,mpi_double_precision,i,i,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
! and sum them into operator
do j1=1,nbasis
do k1=1,ivir
Expand Down
4 changes: 2 additions & 2 deletions src/dnc/dnc_hfoperator.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -296,7 +296,7 @@ end subroutine hfoperatordc
subroutine mpi_hfoperatordc(oneElecO)
use allmod
use quick_gaussian_class_module
use quick_cutoff_module, only: cshell_density_cutoff
use quick_cutoff_module, only: cshell_density_cutoff
use quick_cshell_eri_module, only: getCshellEriEnergy
use mpi
implicit double precision(a-h,o-z)
Expand Down Expand Up @@ -427,7 +427,7 @@ subroutine mpi_hfoperatordc(oneElecO)
else
do i=1,mpisize-1
! receive opertors from slave nodes
call MPI_RECV(temp2d,nbasis*nbasis,mpi_double_precision,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
call MPI_RECV(temp2d,nbasis*nbasis,mpi_double_precision,i,i,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
! and sum them into operator
do ii=1,nbasis
do jj=1,nbasis
Expand Down
6 changes: 3 additions & 3 deletions src/dnc/dnc_scf.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -313,11 +313,11 @@ subroutine electdiisdc(jscf,PRMS)
do i=1,mpi_dc_fragn(ittt)
itt=mpi_dc_frag(ittt,i)
call MPI_RECV(codcsub(1:NNmax,1:NNmax,itt),NNmax*NNmax, &
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
call MPI_RECV(codcsubtran(1:NNmax,1:NNmax,itt),NNmax*NNmax, &
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
call MPI_RECV(evaldcsub(itt,1:NNmax),NNmax,mpi_double_precision, &
ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
enddo
enddo
endif
Expand Down
2 changes: 1 addition & 1 deletion src/modules/quick_mpi_module.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -34,7 +34,7 @@ module quick_mpi_module
logical :: master = .true. ! flag to show if the node is master node
logical :: bMPI = .true. ! flag to show if MPI is turn on
logical :: libMPIMode = .false. ! if mpi is initialized somewhere other than quick
integer, allocatable :: MPI_STATUS(:)
integer, allocatable :: QUICK_MPI_STATUS(:)
integer, parameter :: MIN_1E_MPI_BASIS=6
integer, allocatable :: mgpu_ids(:)
integer :: mgpu_id
Expand Down
2 changes: 1 addition & 1 deletion src/modules/quick_timer_module.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -373,7 +373,7 @@ subroutine timer_output(io)
! MPI_timer_cumer=timer_cumer
! max_timer_cumer=timer_cumer
! do i=1,mpisize-1
! call MPI_RECV(tmp_timer_cumer,1,mpi_timer_cumer_type,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
! call MPI_RECV(tmp_timer_cumer,1,mpi_timer_cumer_type,i,i,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
! MPI_timer_cumer%TTotal=MPI_timer_cumer%TTotal+tmp_timer_cumer%T2e+tmp_timer_cumer%TMP2+ &
! tmp_timer_cumer%T1e+ tmp_timer_cumer%TDiag+tmp_timer_cumer%TGrad
! MPI_timer_cumer%TTotal=MPI_timer_cumer%TTotal+tmp_timer_cumer%TDiag
Expand Down
10 changes: 6 additions & 4 deletions src/mpi_setup.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,7 @@ subroutine initialize_quick_mpi()
call MPI_GET_PROCESSOR_NAME(pname,namelen,mpierror)
call MPI_BARRIER(MPI_COMM_WORLD,mpierror)

if(.not. allocated(MPI_STATUS)) allocate(MPI_STATUS(MPI_STATUS_SIZE))
if(.not. allocated(QUICK_MPI_STATUS)) allocate(QUICK_MPI_STATUS(MPI_STATUS_SIZE))

if (mpirank.eq.0) then
master=.true.
Expand Down Expand Up @@ -375,7 +375,7 @@ subroutine mgpu_setup()
mgpu_ids(1)=mgpu_id

do i=1,mpisize-1
call MPI_RECV(mgpu_ids(i+1),1,mpi_integer,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
call MPI_RECV(mgpu_ids(i+1),1,mpi_integer,i,i,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
enddo

endif
Expand Down Expand Up @@ -588,8 +588,10 @@ subroutine get_mpi_ssw
else

do i=1,mpisize-1
call MPI_RECV(quick_xcg_tmp%tmp_sswt,quick_dft_grid%init_ngpts,mpi_double_precision,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
call MPI_RECV(quick_xcg_tmp%tmp_weight,quick_dft_grid%init_ngpts,mpi_double_precision,i,i,MPI_COMM_WORLD,MPI_STATUS,IERROR)
call MPI_RECV(quick_xcg_tmp%tmp_sswt,quick_dft_grid%init_ngpts,mpi_double_precision,i,i,MPI_COMM_WORLD,&
QUICK_MPI_STATUS,IERROR)
call MPI_RECV(quick_xcg_tmp%tmp_weight,quick_dft_grid%init_ngpts,mpi_double_precision,i,i,MPI_COMM_WORLD,&
QUICK_MPI_STATUS,IERROR)

do j=1,quick_dft_grid%init_ngpts
quick_xcg_tmp%sswt(j)=quick_xcg_tmp%sswt(j)+quick_xcg_tmp%tmp_sswt(j)
Expand Down
6 changes: 3 additions & 3 deletions src/obsolete/electdii.f90
View file
Original file line number Diff line number Diff line change
Expand Up @@ -832,11 +832,11 @@ subroutine electdiisdc(jscf,PRMS)
do i=1,mpi_dc_fragn(ittt)
itt=mpi_dc_frag(ittt,i)
call MPI_RECV(codcsub(1:NNmax,1:NNmax,itt),NNmax*NNmax, &
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
call MPI_RECV(codcsubtran(1:NNmax,1:NNmax,itt),NNmax*NNmax, &
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
mpi_double_precision,ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
call MPI_RECV(evaldcsub(itt,1:NNmax),NNmax,mpi_double_precision, &
ittt,itt,MPI_COMM_WORLD,MPI_STATUS,IERROR)
ittt,itt,MPI_COMM_WORLD,QUICK_MPI_STATUS,IERROR)
enddo
enddo
endif
Expand Down

0 comments on commit a15d07c

Please sign in to comment.