997eb69b5a
modified team based to_all -> reduce
2024-11-06 09:46:43 -06:00
b2b0d559cb
Merge branch 'ROCm:develop' into active_set_APIs
2024-11-05 23:02:44 +00:00
fe767d9abf
remove cooperative groups
2024-10-30 20:10:21 +00:00
5975b8c621
Update broadcast function to use stride calculations instead of log_stride
2024-10-29 19:10:05 +00:00
e1ff06913c
Remove device-side active-set-based broadcast API interface from rocSHMEM
2024-10-29 19:04:49 +00:00
abec29bd6a
Update all_reduce algorithm to use internal put/get functions for updating pWrk and pSync arrays
...
* Change log_stride calcualtions to stride calculations
* Update all_reduce example code to use team based interface
2024-10-28 22:10:18 +00:00
c22048112e
Remove the device-side active-set-based reduction API interface from rocSHMEM
2024-10-28 21:35:14 +00:00
794b888d69
Merge pull request #43 from ROCm/LWPRHMEM-75-API-differences-bug-fix
...
Lwprhmem 75 api differences bug fix
2024-10-28 15:45:15 -04:00
98afb41263
API bug fix in IB conduit
2024-10-24 11:52:03 -05:00
e210020e9b
API change bug fix
2024-10-24 11:52:03 -05:00
11df5427a6
add ascii art for ring allredude
2024-10-24 15:08:32 +00:00
a4b4281f50
fix odd-case allreduce scenarios
...
if the number of elements to be used in the allreduce operation is not
exact multiple of the work-array buffer size and number of pe's, we need
to adjust the algorithm to:
- initially perform a ring_allreduce on n_segments * chunk_size (which
is the integer division of the number of elements and the work-buffer
size, i.e. will not cover the entire buffer)
- perform another ring_allreduce where chunk_size is reduced to match
the remaining elements
- if the remaining elements from the previous step cannot evenly be
divded by the number of pe's, we need to perform a direct_allreduce on
the outstanding number of elements.
2024-10-24 15:08:32 +00:00
87db7f7d38
fix barrier synchronization on gfx90a
2024-10-24 15:08:28 +00:00
1fbb89bc73
ipc: add ring_allreduce algorithms
...
add the ring allreduce algorithm to the ipc conduit in order to be able
to execute slightly largers reductions.
2024-10-24 15:07:17 +00:00
ba21cb7b85
ipc/to_all: add direct allreduce algorithm
...
add a simple version of an allreduce algorithm as a starting point.
2024-10-24 15:07:14 +00:00
8a16968cf2
Merge pull request #41 from avinashkethineedi/collective_routine_buffers
...
Fine grained memory buffers for work/sync arrays
2024-10-23 23:33:48 -05:00
d5ea5868e3
Fix quiet and fence of default context
...
* Update tinfo of default context
2024-10-22 16:18:05 +00:00
6685d0ab60
Add fine grained memory buffers for work/sync arrays
...
* Add interanl put_mem/get_mem{_wave, _wg} functions to read/write to work/sync arrays
* Add condition check to ensure all MPI processes are on the same compute node for IPC conduit
2024-10-21 15:28:39 +00:00
8b3854b252
updated atomic_fetch() parameters
2024-10-11 13:34:28 -07:00
722a5f0731
updated *_wait* APIs to use int rather than roc_shmem_cmps
2024-10-11 13:34:28 -07:00
bcf3fdff10
*_wait* routines changed parameter from ptr to ivars to match OpenSHMEM
2024-10-11 13:34:28 -07:00
e419a8b963
Merge pull request #29 from ROCm/improve-ib-latency
...
Vectorize WQE segments writes
2024-10-11 11:55:48 -05:00
509277c034
fixed notifier bug
2024-10-10 06:45:43 -07:00
b1134e8633
added notifier->sync() when we are not using cooperative groups
...
updated scope bug
2024-10-09 13:11:28 -07:00
63667a3167
Added Cooperative Groups configure option and header
2024-10-09 13:11:12 -07:00
1baa071edf
Fix initialization order bug
2024-10-09 13:11:12 -07:00
e2f6a65284
fixed barrier issue on MI250X
2024-10-08 13:18:04 -07:00
92fb1abaf2
Add team information to the context
...
* Update roc_shmem_ctx_fence API to use team-relative PE numbering
* Update backend to populate team_opaque member of ROC_SHMEM_CTX_DEFAULT (used to store information about the team wrt TEAM_WORLD)
2024-10-04 17:56:15 +00:00
979aed105a
Add fence and quiet functionality
...
* Perform atomic stores to enforce memory ordering
2024-10-03 06:28:12 +00:00
787cf0ff3f
Merge pull request #31 from BKP/ipc_bringup_fine_unit_09-26-24
...
Add IPC Simple Buffer Fine-grained Unit Tests
2024-10-01 15:12:30 -05:00
24b928a007
Poll the signal from one thread instead of all
2024-10-01 15:01:37 -05:00
f85c46ec0a
Bugfixes for the ipc unit tests
2024-09-26 13:40:05 -05:00
c133ea18a5
fix assembly switch/case instruction
...
move the case statement out of the architecture specific section.
2024-09-20 20:25:40 +00:00
3162d49b56
Vectorize WQe segments writes
...
Signed-off-by: Muhammad Awad <MuhammadAbdelghaffar.Awad@amd.com >
2024-09-17 20:34:18 -05:00
86a2f34539
Add missing header file
2024-09-10 09:35:02 -07:00
7411c45591
Conservatively use SEQ_CST atomics in IPC conduit
2024-09-10 09:34:45 -07:00
2806e1be79
Intermediate commit for rebase
2024-09-10 07:10:22 -07:00
45c29e7734
Minor updates to Nofifier sync method
2024-09-10 07:10:21 -07:00
359d6be797
Add sync method to notifier class
2024-09-10 07:10:21 -07:00
5b42cff96c
Update Notifier fixture to Block
2024-09-10 07:10:21 -07:00
51c33b2a66
Updates to Notifier
2024-09-10 07:10:21 -07:00
039ea82777
Change read/write to load/store in Nofitier API
2024-09-10 07:10:21 -07:00
0c53a075f2
Fix problems with Notifier
2024-09-10 07:10:21 -07:00
c4b7e0d91b
Partial notifier
2024-09-10 07:10:21 -07:00
e9fb01ab6b
Merge pull request #27 from ROCm/ipc_bringup
...
Ipc bringup
2024-09-10 09:06:51 -05:00
375d145913
Merge pull request #25 from mawad-amd/fix-bad-auto
...
Fix bad auto
2024-09-09 16:30:17 -05:00
e58077e3cf
Merge branch 'ipc_bringup' into ipc_atomics
2024-09-09 14:22:55 -05:00
dfcacdc4a3
remove pSync from internal_bcast functions
...
remove the pSync arguments from the internal_broadcast functions,
they are not used anyway.
2024-09-09 12:06:30 -07:00
b811b2b683
Fix bad auto
...
Signed-off-by: Muhammad Awad <MuhammadAbdelghaffar.Awad@amd.com >
2024-09-05 18:39:02 -05:00
7bbf34d334
remove local_pe calculation from puts, gets and atomics functions
...
* All the PEs are assumed to be accessible using IPC backend
2024-09-05 11:52:00 -07:00
Yiltan Hassan Temucin
avinashkethineedi
Yiltan
Yiltan Temucin
Edgar Gabriel
Brandon Potter
Brandon Potter
Muhammad Awad