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

enable cuda unit tests #354

Merged
merged 135 commits into from
Oct 14, 2022
Merged

enable cuda unit tests #354

merged 135 commits into from
Oct 14, 2022

Conversation

evaleev
Copy link
Member

@evaleev evaleev commented Jul 13, 2022

No description provided.

evaleev and others added 30 commits May 2, 2022 15:56
@evaleev
blaspp seems to always define empty BLA_VENDOR in CACHE so only look …
2988db4 
…at its value, not its presence
@evaleev
update unit tests to replace DistArray::range() with DistArray::tiles…
0afdbfe 
…_range() per 8cf3327
@evaleev
ta_test: UM expr tests need looser FP64 comparison tolerance
760bfa3
@evaleev
introduced check_serial{,-tiledarray} targets
4be43ff
@evaleev
gitlab ci: run serial unit tests with CUDA + qualify all targets with…
862c014 
… -tiledarray where possible
@evaleev
ccd example avoids using deprecated DistArray::range()
e263f99
@evaleev
print nvcc version info and output of nvidia-smi
e0dfb45
@evaleev
vector<T>::data() may not return null ptr if empty
c172a98
@evaleev
bump up UT test timeout to accomodate CUDA builds
9c355bd
@evaleev
added UT um_expressions_suite/scal_add_block
3c14fa4
@asadchev
Distributed einsum
0366fa7
@evaleev
einsum: fixed hadamard reduction (e.g. ij,ij->i) for sparse inputs
5a9631f
@asadchev
Einsum: move explicit subworld fence outside the loop
e1f2d41
@evaleev
Merge branch 'anisimov/feature/cutt-to-librett' into evaleev/fix/cuda…
b534969 
…-unit-tests
@evaleev
[cmake] TA_ENABLE_RANGEV3 -> TA_RANGEV3
c320b7a
@evaleev
[cmake] find/build TTG by setting TA_TTG=ON ... can customize TA_TRAC…
d9b2cb0 
…KED_TTG_{TAG,URL}
@evaleev
const-sanitized pmap + TensorImpl (hence DistArray) holds shape via s…
3065cfb 
…hared_ptr<const Shape>

- shared_ptr<Pmap> -> shared_pmap<const Pmap>
- introduced {TensorImpl,DistArray}::shape_shared
- introduced {TensorImpl,DistArray}::pmap_shared to mirror shape_shared ... unfortunately pmap and shape return types do not mirror
@evaleev
fixed DistArray::future typedef
c7454a8
@evaleev
Tensor ctor taking existing (shared_ptr) buffer is public, useful for…
fbfbc49 
… move data in from existing data structures
@evaleev
introduced Range1, an overdue iterable replacement of std::pair
988d679
@evaleev
replaced std::pair with (compatible) Range1
3b5267d
@evaleev
bump BTAS tag + only make boost targets IMPORTED_GLOBAL if not alread…
db3347b 
…y so

pulls in ValeevGroup/BTAS@ab866a7
@evaleev
If have TTG, discover it before looking for MADWorld ourselves (to ma…
d91104c 
…ke sure that it's told to use PaRSEC) + initialize/finalize it from within TA::{initialize,finalize}()
@evaleev
Initial draft of TTG-based TRTRI, not yet being tested in linalg.cpp
df4bbf3
@evaleev
TTG-based cholesky and cholesky_inv are implemented and function, but…
4bda70b 
… do not pass correctness check and hang in fence ... for now hardwired all tests in linalg to 4x4 matrices with 1 tile
@evaleev
tiled TTG-based col-major cholesky works
385af92
@evaleev
generalize TTG cholesky code for row-major capable TTG flows
068568e
@evaleev
updated to work with new TTG code using LD-including MatrixTile
c7667fe
@evaleev
introduced Range1::{lo,up}bound
eda3dd2
@evaleev
minor cleanup
b869011
@evaleev evaleev force-pushed the evaleev/fix/cuda-unit-tests branch 2 times, most recently from 26d199d to 5868494 Compare October 1, 2022 14:59
@evaleev evaleev force-pushed the evaleev/fix/cuda-unit-tests branch 8 times, most recently from 886f00f to 9e2e561 Compare October 13, 2022 18:35
evaleev added 12 commits October 13, 2022 14:52
@evaleev
use single CUDA stream for unit tests for now
c536b1b
@evaleev
[ci] attempt to resolve https://gitlab.com/ValeevGroup/tiledarray/-/j…
15f54ae 
…obs/3112494437#L1284
@evaleev
fixed ta_reduce_cuda for asymmetric matrices, resolves #366
8f2a363
@evaleev
try using valeevgroup/ubuntu:cuda image
32fa553
@evaleev
no need to install cuda when using valeevgroup/ubuntu:cuda image
0c07229
@evaleev
[ci] invoke nvidia-smi to dump driver + CUDA info to the log
1d73c64
@evaleev
Revert "[ci] invoke nvidia-smi to dump driver + CUDA info to the log"
08df869 
This reverts commit 1fc764c.
@evaleev
Revert "no need to install cuda when using valeevgroup/ubuntu:cuda im…
4e0ce09 
…age"

This reverts commit 15e1d5b.
@evaleev
Revert "try using valeevgroup/ubuntu:cuda image"
dd12732 
This reverts commit 2223c8e.
@evaleev
[ci] define RUNNER_TAGS so that only CUDA jobs end up on our local ru…
7a7bc76 
…nner
@evaleev
cleanup
caba336
@evaleev
hush warnings re use of Eigen's Tensor header when CUDA enabled
77c95ff
@evaleev evaleev force-pushed the evaleev/fix/cuda-unit-tests branch 4 times, most recently from d0f5c92 to ffc56de Compare October 14, 2022 09:13
@evaleev evaleev force-pushed the evaleev/fix/cuda-unit-tests branch from ffc56de to 6bd84aa Compare October 14, 2022 10:12
@evaleev evaleev merged commit 8d50fe5 into anisimov/feature/cutt-to-librett Oct 14, 2022
@evaleev evaleev deleted the evaleev/fix/cuda-unit-tests branch October 14, 2022 11:57
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
2 participants
@evaleev @asadchev