Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Reimplementation of local band-to-tridiagonal #938

Merged
merged 9 commits into from
Jul 21, 2023
Merged

Reimplementation of local band-to-tridiagonal #938

Show file tree
Hide file tree
Changes from 8 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
eaca145
First working version. Note: sweeps have still an artificial dependen…
rasolca Jul 13, 2023
448d2c2
removed constrain
rasolca Jul 14, 2023
bc202e2
partial refactoring: snake_case functions
rasolca Jul 17, 2023
2651c6d
Keep pipeline of workers (implementation is more similar to coming di…
rasolca Jul 18, 2023
372e46f
fix doc and a couple of snake_case
rasolca Jul 19, 2023
a105e6a
format
rasolca Jul 19, 2023
1c93733
fix variable name
rasolca Jul 19, 2023
277100e
Apply suggestions from code review
rasolca Jul 20, 2023
188f562
fix typo and format
rasolca Jul 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
10 changes: 5 additions & 5 deletions include/dlaf/eigensolver/band_to_tridiag.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,8 +71,8 @@ namespace eigensolver {
/// @pre band_size is a divisor of mat_a.blockSize().cols(), and band_size >= 2
/// @pre mat_a is not distributed.
template <Backend B, Device D, class T>
TridiagResult<T, Device::CPU> bandToTridiag(blas::Uplo uplo, SizeType band_size,
Matrix<const T, D>& mat_a) {
TridiagResult<T, Device::CPU> band_to_tridiag(blas::Uplo uplo, SizeType band_size,
Matrix<const T, D>& mat_a) {
DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
Expand Down Expand Up @@ -142,16 +142,16 @@ TridiagResult<T, Device::CPU> bandToTridiag(blas::Uplo uplo, SizeType band_size,
/// @pre band_size is a divisor of mat_a.blockSize().cols() and band_size >= 2,
/// @pre mat_a is distributed according to grid.
template <Backend backend, Device device, class T>
TridiagResult<T, Device::CPU> bandToTridiag(comm::CommunicatorGrid grid, blas::Uplo uplo,
SizeType band_size, Matrix<const T, device>& mat_a) {
TridiagResult<T, Device::CPU> band_to_tridiag(comm::CommunicatorGrid grid, blas::Uplo uplo,
SizeType band_size, Matrix<const T, device>& mat_a) {
DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
DLAF_ASSERT(band_size >= 2, band_size);

// If the grid contains only one rank force local implementation.
if (grid.size() == comm::Size2D(1, 1))
return bandToTridiag<backend, device, T>(uplo, band_size, mat_a);
return band_to_tridiag<backend, device, T>(uplo, band_size, mat_a);

switch (uplo) {
case blas::Uplo::Lower:
Expand Down
Loading
Loading