|
0baaddbe98
|
Pipeline mode commit on Aurora. 5+ TF/s on 16^3x32 per tile at 384
nodes.
More concurrency/fine grained scheduling is possible.
|
2025-02-04 19:27:26 +00:00 |
|
|
8cf809e231
|
Best results on Aurora so far
|
2025-01-31 16:14:45 +00:00 |
|
|
94019a922e
|
Significantly better performance on Aurora without using pipeline mode
|
2025-01-30 16:36:46 +00:00 |
|
Peter Boyle
|
8f70cfeda9
|
Clean up
|
2024-10-18 13:56:53 -04:00 |
|
|
c5c67b706e
|
cl::sycl -> SYCL
|
2024-10-10 22:04:12 +00:00 |
|
|
066544281f
|
Deprecate UVM
|
2024-09-17 13:34:27 +00:00 |
|
Peter Boyle
|
c4b9f71357
|
CPU compile ordering is important
|
2024-05-21 02:22:32 +01:00 |
|
Peter Boyle
|
9a1ad6a5eb
|
Merge branch 'develop' of https://github.com/paboyle/Grid into develop
|
2024-05-17 11:33:46 -04:00 |
|
Peter Boyle
|
057f86c1de
|
2 queues works ok in performance
|
2024-05-07 18:42:50 +00:00 |
|
Peter Boyle
|
cc04dc42dc
|
Merge branch 'develop' into feature/scidac-wp1
|
2024-03-06 14:55:21 -05:00 |
|
|
783a66b348
|
Deterministic reduction please
|
2024-03-06 00:01:37 +00:00 |
|
Peter Boyle
|
3f1636637d
|
Merge pull request #453 from dbollweg/feature/sliceSum_gpu
Feature/slice sum gpu
|
2024-02-28 14:04:43 -05:00 |
|
Christoph Lehner
|
9f89486df5
|
remove unnecessary code path
|
2024-02-28 19:56:23 +01:00 |
|
Christoph Lehner
|
22b43b86cb
|
Make GPT test suite work with SYCL
|
2024-02-28 12:57:17 +01:00 |
|
Peter Boyle
|
9f40467e24
|
Warning squash
|
2024-02-27 11:40:36 -05:00 |
|
Dennis Bollweg
|
6cd2d8fcd5
|
Replace cuda/hip memcpy with Grid functions
|
2024-02-26 09:55:07 -05:00 |
|
Peter Boyle
|
5e5b471bb2
|
Put/Get and DEviceToDevice
|
2024-02-21 14:47:06 -05:00 |
|
|
303b83cdb8
|
Scaling benchmarks, verbosity and MPICH aware in acceleratorInit()
For some reason Dirichlet benchmark fails on several nodes; need to
debug this.
|
2024-02-13 19:48:03 +00:00 |
|
Peter Boyle
|
6f51b49ef8
|
Use stderr
|
2024-01-22 17:41:09 -05:00 |
|
Peter Boyle
|
e07cb2b9de
|
Accelerator memory
|
2024-01-17 16:24:31 -05:00 |
|
Peter Boyle
|
d34b207eab
|
Avoid HIP warnings
|
2023-10-24 10:57:04 -04:00 |
|
Peter Boyle
|
33097681b9
|
FTHMC compiled and merged to develop
|
2023-10-14 00:42:55 +03:00 |
|
Peter Boyle
|
9626a2c7c0
|
Asynch handling
|
2023-10-13 18:21:56 +03:00 |
|
Peter Boyle
|
c564611ba7
|
Annoying hack that is useful to preserve for profiling
|
2023-09-29 17:11:12 -04:00 |
|
Peter Boyle
|
b4f2ca81ff
|
Copy queue and compute queue same as better concurrency
|
2023-04-11 12:18:21 -07:00 |
|
Peter Boyle
|
da503fef0e
|
Name change on barrier routine
|
2023-04-11 12:14:04 -07:00 |
|
Peter Boyle
|
4a382fad3f
|
Use distinct SYCL queue for copies
|
2023-04-04 07:41:41 -07:00 |
|
Peter Boyle
|
af64c1c6b6
|
Had managed to drop the accelerator_barrier() in the Wilson Compressor gather
|
2023-03-30 17:34:44 -04:00 |
|
Peter Boyle
|
866f48391a
|
Temporary fix for develop incorrect results
|
2023-03-30 17:10:13 -04:00 |
|
Peter Boyle
|
496d04cd85
|
Weaken the Fence
|
2023-03-29 18:58:51 -04:00 |
|
Peter Boyle
|
b5b759df73
|
Merge branch 'develop' into feature/dirichlet
|
2023-03-21 16:05:46 -04:00 |
|
Peter Boyle
|
861e5d7f4c
|
SYCL version update. Why do they keep making incompatible changes
|
2023-03-14 12:10:02 -07:00 |
|
Peter Boyle
|
03508448f8
|
Remove verbose
|
2022-10-04 11:12:15 -07:00 |
|
Peter Boyle
|
1177b8f661
|
Merge branch 'develop' into feature/dirichlet
|
2022-08-31 19:05:57 -04:00 |
|
Peter Boyle
|
95b640cb6b
|
10TF/s on 32^3 x 64 on single node
|
2022-08-04 15:43:52 -04:00 |
|
Peter Boyle
|
2cb5bedc15
|
Copy stream HIP improvements
|
2022-08-04 15:24:03 -04:00 |
|
Peter Boyle
|
188d2c7a4d
|
PVC default, ignore ATS
|
2022-08-02 08:38:53 -07:00 |
|
Peter Boyle
|
84110166e4
|
Fix the fence
|
2022-08-02 08:00:43 -07:00 |
|
Peter Boyle
|
d32b923b6c
|
Fencing on a stream in SYCL is needed. Didn't know that ... gulp
|
2022-08-02 07:58:04 -07:00 |
|
Peter Boyle
|
5f8892bf03
|
Mistake pointed out by Camilo
|
2022-07-19 09:31:51 -07:00 |
|
Peter Boyle
|
f14e7e51e7
|
Grid accelerator
|
2022-07-12 10:56:22 -07:00 |
|
Peter Boyle
|
3544965f54
|
Stream doesn't work
|
2022-07-07 17:49:20 +01:00 |
|
Peter Boyle
|
bd99fd608c
|
Introduce a non-default stream for compute operatoins
|
2022-07-01 09:42:53 -04:00 |
|
Peter Boyle
|
136d843ce7
|
Crusher updates
|
2022-05-25 12:36:09 -04:00 |
|
Peter Boyle
|
5012adfebf
|
Merge branch 'develop' into feature/dirichlet
|
2022-04-05 16:26:19 -04:00 |
|
Peter Boyle
|
92a83a9eb3
|
Performance improve for Tesseract
|
2022-03-16 17:14:36 +00:00 |
|
Peter Boyle
|
5340e50427
|
HMC running with new formulation
|
2022-03-01 17:10:25 -05:00 |
|
Peter Boyle
|
e16fc5b2e4
|
Threaded intranode comms transfer - ideally between NUMA domains
|
2022-03-01 11:17:24 -05:00 |
|
Julio Maia
|
86f4e17928
|
Changing thread block order and adding launch_bounds
|
2022-02-07 11:29:37 -06:00 |
|
Peter Boyle
|
7f7d06d963
|
Merge branch 'develop' of https://github.com/paboyle/Grid into develop
|
2021-12-07 09:06:42 -08:00 |
|