Skip to content

DLA-Future#1

Merged
RMeli merged 130 commits intomaster-tmpfrom
dlaf
Oct 26, 2023
Merged

DLA-Future#1
RMeli merged 130 commits intomaster-tmpfrom
dlaf

Conversation

@RMeli
Copy link
Owner

@RMeli RMeli commented Oct 26, 2023

PR to master-tmp in order to squash all commits, for an easier git rebase (required by CP2K).

The current dlaf branch is saved as dlaf-backup.

Dr. Mathieu Taillefumier and others added 30 commits January 20, 2023 15:45
Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>
Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>
* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up
@RMeli RMeli self-assigned this Oct 26, 2023
@RMeli RMeli merged commit a76d27d into master-tmp Oct 26, 2023
@RMeli RMeli deleted the dlaf branch October 26, 2023 11:12
RMeli added a commit that referenced this pull request Oct 27, 2023
* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup
RMeli added a commit that referenced this pull request Oct 30, 2023
* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup
RMeli added a commit that referenced this pull request Oct 30, 2023
* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup
RMeli added a commit that referenced this pull request Nov 6, 2023
* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup
RMeli added a commit that referenced this pull request Nov 15, 2023
* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup
RMeli added a commit that referenced this pull request Nov 22, 2023
* DLA-Future (#1)

* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* fix cmake conflicts

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* make pretty

* Revert "MPI: Specify imported routines and constants with MPI_F08"

This reverts commit aecd330.

* Revert "Revert "MPI: Specify imported routines and constants with MPI_F08""

This reverts commit f224465.

* fix MPI

* initialize HIP/CUDA before MPI

* initialize HIP/CUDA before MPI

* better mpi init

---------

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>
RMeli added a commit that referenced this pull request Dec 1, 2023
* DLA-Future (#1)

* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* fix cmake conflicts

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* make pretty

* Revert "MPI: Specify imported routines and constants with MPI_F08"

This reverts commit aecd330.

* Revert "Revert "MPI: Specify imported routines and constants with MPI_F08""

This reverts commit f224465.

* fix MPI

* initialize HIP/CUDA before MPI

* initialize HIP/CUDA before MPI

* better mpi init

---------

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>
RMeli added a commit that referenced this pull request Dec 11, 2023
* DLA-Future (#1)

* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* fix cmake conflicts

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* make pretty

* Revert "MPI: Specify imported routines and constants with MPI_F08"

This reverts commit aecd330.

* Revert "Revert "MPI: Specify imported routines and constants with MPI_F08""

This reverts commit f224465.

* fix MPI

* initialize HIP/CUDA before MPI

* initialize HIP/CUDA before MPI

* better mpi init

---------

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>
RMeli added a commit that referenced this pull request Dec 11, 2023
* DLA-Future (#1)

* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* fix cmake conflicts

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* make pretty

* Revert "MPI: Specify imported routines and constants with MPI_F08"

This reverts commit aecd330.

* Revert "Revert "MPI: Specify imported routines and constants with MPI_F08""

This reverts commit f224465.

* fix MPI

* initialize HIP/CUDA before MPI

* initialize HIP/CUDA before MPI

* better mpi init

---------

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>
RMeli added a commit that referenced this pull request Dec 12, 2023
* DLA-Future (#1)

* More cmake fixes

Signed-off-by: Dr. Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* Add dla-future support for cholesky

Signed-off-by: Mathieu Taillefumier <mathieu.taillefumier@free.fr>

* More work on cmake

* New Keyword: MIN_PAIR_LIST_RADIUS can be used to force non-zero blocks (cp2k#2511)

in KS and Overlap matrix.

* Regtesting: Fix DeprecationWarning

* Docker: Do not flag slow tests in minimal build

* PAO: Add equivariant parametrization

* PAO: Use correct distribution in pao_calc_AB_equi

* PAO: Add missing mp_sum to pao_calc_AB_equi

* Regtesting: Fix issue with Python 3.6

* Consider target CPU in OpenBLAS toolchain build

This should partially fix cp2k#2517

* Update track imag density (#6)

* PAO: Fix unused parameter and CMake build

* Keep compatibility with bash v3

* Add info for sourceable arch files and links to HowTos

* Refactor allocations of imaginary parts for RTP (cp2k#2531)

* unify allocation of imaginary parts of matrix_h_im and matrix_ks_im with their real counterparts

* refactor allocations of imaginary parts for RTP

* clean-up

* More work on cmake

* FIST: Add NequIP equivariant neural network potentials

* FIST: Fix CMake build and conventions for NequIP

* Remove unused parameters

* Remove unused label from beta_gamma_psi.F

* Toolchain: Add -Werror=unused-parameter and -Werror=unused-label

* Docker: Fix typo in cmake variable and add check for warnings

* Regtesting: Introduce tests/UNIT_TESTS file

* Let nequip_unittest fail when libtorch is missing

* Remove some more unused parameters

* Regtesting: Add Valgrind option

* Fix various simple Valgrind issues

* update keyword descriptions for RTP

* fix strings

* Fix typo in arch/Linux-intel-x86_64.psmp 

Shared part is not really tested afaict, but for the sake of completeness let's fix this typo.

* Add more target mappings for gcc to OpenBLAS

* Fix funky libxsmm library order

* fix libxc

* cleanup dlaf cholesky

* dirty water: dirty version working for H2O-32

* dlaf

* make preffiy

* Revert "make preffiy"

This reverts commit d27b375.

* revert fftw3 change

* add some dlaf output

* remove copies using new eigensolver api

* port @msimberg improvements

* clang formatting

* add improvements suggested by @msimberg

* rename cholesky_dlaf to dlaf

* remove single-threades scope from cp2k, it is now in dlaf

* general cleanup

* clang format

* further cleanup

* remove unused single precision eigensolver

* remove MKL-specific single threaded blas/lapack scope

* use new dlaf C API

* fix dlaf calls

* improve cuSOLVER-stype integration

* make cp2k compilable without dlaf

* print pika binding

* refactoring and cleanup

* remove local grid creation/free

* zero eigenvalues buffer

* cleanup

* add missing licence from c file

* refactoring

* cholesky wrapper

* cholesky wrapper

* add scalapack fallback and falback parameter dlaf_neigvec_min

* make pretty

* add unit test for dlaf

* remove spurious file

* fix usage of CP2K as a library

* remove initialization and pika threads

* add dlaf timing without upper to full

* actually fix mpi

* remove string utilities

* prettify

* Revert "prettify"

This reverts commit cd1e89a.

* cleanup:

* remove redistribution

* manual cleanup

* prettify cmake

* make pretty exluding dlaf.F

* add some details to install and __DLAF so that make pretty does not complain

* prettify install

* more consistent cabort

* remove intrinsic

* Revert "remove intrinsic"

This reverts commit 2458e00.

* format doc

* cleanup

* revert some changes to MPI init

* cleanup

* prettify

* remove redistribution

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit f7fd23c.

* Revert "Revert "pretty-test""

This reverts commit b9a9879.

* Revert "pretty-test"

This reverts commit d582f3f.

* pretty-test

* pretty-test

* pretty-test

* pretty-test

* Revert "pretty-test"

This reverts commit b88b373.

* pretty-test

* remove snippet

* add timers -- workaround for cp2k#3071

* prettify

* doc cleanup

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* fix cmake conflicts

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* Update src/fm/cp_dlaf_utils.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* remove ugly workaround

* cmake config

* line wrap install

* Update src/input_cp2k_global.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* gitignore

* add guards

* remove pika:print-bind in favour of environment variable

* __dlaf

* make pretty

* Update src/fm/cp_fm_dlaf_api.F

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>

* actually fix cmake

* make pretty

* Revert "MPI: Specify imported routines and constants with MPI_F08"

This reverts commit aecd330.

* Revert "Revert "MPI: Specify imported routines and constants with MPI_F08""

This reverts commit f224465.

* fix MPI

* initialize HIP/CUDA before MPI

* initialize HIP/CUDA before MPI

* better mpi init

---------

Co-authored-by: Mikael Simberg <mikael.simberg@iki.fi>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

8 participants