GPU Operators use work vectors #1673
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
jeremylt
force-pushed
the
jeremy/use-work-vecs
branch
5 times, most recently
from
f9d01df to
9bb9298
Compare
jeremylt
commented
Sep 26, 2024
| @@ -862,7 +862,7 @@ int CeedVectorPointwiseMult(CeedVector w, CeedVector x, CeedVector y) { | |||
| CeedCall(CeedVectorGetLength(w, &length_w)); | |||
| CeedCall(CeedVectorGetLength(x, &length_x)); | |||
| CeedCall(CeedVectorGetLength(y, &length_y)); | |||
| CeedCheck(length_w == length_x && length_w == length_y, ceed, CEED_ERROR_UNSUPPORTED, | |||
| CeedCheck(length_x >= length_x && length_y >= length_w, ceed, CEED_ERROR_UNSUPPORTED, | |||
There was a problem hiding this comment.
Needed to allow usage of longer cached vectors than strictly needed
jeremylt
commented
Sep 26, 2024
| @@ -331,11 +331,6 @@ static int CeedBasisApplyAtPointsCheckDims(CeedBasis basis, CeedInt num_elem, co | |||
| if (x_ref != CEED_VECTOR_NONE) CeedCall(CeedVectorGetLength(x_ref, &x_length)); | |||
| if (u != CEED_VECTOR_NONE) CeedCall(CeedVectorGetLength(u, &u_length)); | |||
|
|
|||
| // Check compatibility of topological and geometrical dimensions | |||
There was a problem hiding this comment.
Also needed to allow use of cache variables that are longer than strictly required
|
Blocks #1646 |
jeremylt
force-pushed
the
jeremy/use-work-vecs
branch
2 times, most recently
from
5ba7b9b to
1945a8d
Compare
|
Branch passes Ratel CI too $ make prove -j CEED_BACKENDS=/gpu/cuda/shared
-----------------------------------------
| ____ __ __ |
| / __ \ ____ _ / /_ ___ / / |
| / /_/ / / __ `/ / __/ / _ \ / / |
| / _, _/ / /_/ / / /_ / __/ / / |
| /_/ |_| \__,_/ \__/ \___/ /_/ |
-----------------------------------------
-----------------------------------------
Dependencies:
CEED_DIR = /home/jeremy/Dev/libCEED
PETSC_DIR = /home/jeremy/Dev/petsc
PETSC_ARCH = arch-cuda-mpich
Optional Dependencies:
ENZYME_LIB = (not found)
-----------------------------------------
Running unit tests
- Testing with libCEED backends: /gpu/cuda/shared
- Testing on 1 processes
prove -j 16 --exec 'python3 tests/junit.py --petsc-arch arch-cuda-mpich --ceed-backends /gpu/cuda/shared --mode tap --nproc 1 --pool-size 1' t000-init t001-view t002-view t003-ts-monitor t004-ts-checkpoint t010-eigensolver t050-mpm t100-static-elasticity t101-static-elasticity t102-static-elasticity t103-static-elasticity t110-static-elasticity t111-static-elasticity t120-static-elasticity t121-static-elasticity t122-static-elasticity t123-static-elasticity t211-quasistatic-elasticity t221-quasistatic-elasticity t222-quasistatic-elasticity ex01-static ex02-quasistatic ex03-dynamic
t050-mpm ..................... ok
t010-eigensolver ............. ok
t000-init .................... ok
t001-view .................... ok
t002-view .................... ok
t003-ts-monitor .............. ok
t101-static-elasticity ....... ok
t122-static-elasticity ....... ok
t123-static-elasticity ....... ok
t100-static-elasticity ....... ok
t111-static-elasticity ....... ok
t211-quasistatic-elasticity .. ok
t103-static-elasticity ....... ok
t222-quasistatic-elasticity .. ok
t110-static-elasticity ....... ok
t120-static-elasticity ....... ok
t102-static-elasticity ....... ok
t221-quasistatic-elasticity .. ok
t121-static-elasticity ....... ok
ex03-dynamic ................. ok
t004-ts-checkpoint ........... ok
ex01-static .................. ok
ex02-quasistatic ............. ok
All tests successful.
Files=23, Tests=197, 482 wallclock secs ( 0.12 usr 0.04 sys + 1145.66 cusr 120.38 csys = 1266.20 CPU)
Result: PASS |
|
Note to me for follow-up: Make an issue about adding this code cleanup and simplification to the CPU side of the house |
jeremylt
force-pushed
the
jeremy/use-work-vecs
branch
from
f1d6ed5 to
96093a6
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This takes lessons from the Gen backends and processes the fields in the best order to prevent duplicate work while allowing for reuse of buffers, specifically the new cached vector buffer.