0641a4a29e
make MPI_Init and MPI_Finalize independent of the test fixtures
...
[ROCm/rocshmem commit: 0f7dc70894
2024-10-01 18:33:36 +00:00
c097da70c4
Poll the signal from one thread instead of all
...
[ROCm/rocshmem commit: 0659f8d93c
2024-09-27 15:17:57 -05:00
25d7d7fccd
Change notifier max thread block value to account for MI300 CPX
...
[ROCm/rocshmem commit: db221b022a
2024-09-27 11:17:53 -05:00
24a527dcda
Reset config options to original values
...
[ROCm/rocshmem commit: 56b2ed699b
2024-09-27 11:17:11 -05:00
325ce3cba7
Bugfixes for the ipc unit tests
...
[ROCm/rocshmem commit: f85c46ec0a
2024-09-26 13:40:05 -05:00
bed676f89d
fix assembly switch/case instruction
...
move the case statement out of the architecture specific section.
[ROCm/rocshmem commit: c133ea18a5
2024-09-20 20:25:40 +00:00
fe3ecde6f6
Vectorize WQe segments writes
...
Signed-off-by: Muhammad Awad <MuhammadAbdelghaffar.Awad@amd.com >
[ROCm/rocshmem commit: 3162d49b56
2024-09-17 20:34:18 -05:00
56c1626df1
Update fine-grained simple tests
...
[ROCm/rocshmem commit: 46fdb1851c
2024-09-10 09:35:41 -07:00
e64264d233
Add missing header file
...
[ROCm/rocshmem commit: 86a2f34539
2024-09-10 09:35:02 -07:00
bf79c21ea8
Conservatively use SEQ_CST atomics in IPC conduit
...
[ROCm/rocshmem commit: 7411c45591
2024-09-10 09:34:45 -07:00
10d351b6a1
Intermediate commit for rebase
...
[ROCm/rocshmem commit: 2806e1be79
2024-09-10 07:10:22 -07:00
74c4a248cc
Add an extra assertion check for nullptr
...
[ROCm/rocshmem commit: 678564ba3c
2024-09-10 07:10:22 -07:00
688826937f
Minor updates to Nofifier sync method
...
[ROCm/rocshmem commit: 45c29e7734
2024-09-10 07:10:21 -07:00
aed0da61d0
Add sync method to notifier class
...
[ROCm/rocshmem commit: 359d6be797
2024-09-10 07:10:21 -07:00
9b0e4dc05d
Change notifier fixture to prep for other fixtures
...
[ROCm/rocshmem commit: 1289d50be5
2024-09-10 07:10:21 -07:00
68384da019
Update Notifier fixture to Block
...
[ROCm/rocshmem commit: 5b42cff96c
2024-09-10 07:10:21 -07:00
13ec689cdf
Updates to Notifier
...
[ROCm/rocshmem commit: 51c33b2a66
2024-09-10 07:10:21 -07:00
4896235ada
Change read/write to load/store in Nofitier API
...
[ROCm/rocshmem commit: 039ea82777
2024-09-10 07:10:21 -07:00
ef4a14e947
Fix problems with Notifier
...
[ROCm/rocshmem commit: 0c53a075f2
2024-09-10 07:10:21 -07:00
cec6ba91a0
Partial notifier
...
[ROCm/rocshmem commit: c4b7e0d91b
2024-09-10 07:10:21 -07:00
4610659bf1
Add simple fine test
...
[ROCm/rocshmem commit: da93542c40
2024-09-10 07:10:21 -07:00
94122cd208
Merge pull request #27 from ROCm/ipc_bringup
...
Ipc bringup
[ROCm/rocshmem commit: e9fb01ab6b
2024-09-10 09:06:51 -05:00
72c048d80a
Merge pull request #25 from mawad-amd/fix-bad-auto
...
Fix bad auto
[ROCm/rocshmem commit: 375d145913
2024-09-09 16:30:17 -05:00
c88645e402
Merge branch 'ipc_bringup' into ipc_atomics
...
[ROCm/rocshmem commit: e58077e3cf
2024-09-09 14:22:55 -05:00
1ddc82fc28
remove pSync from internal_bcast functions
...
remove the pSync arguments from the internal_broadcast functions,
they are not used anyway.
[ROCm/rocshmem commit: dfcacdc4a3
2024-09-09 12:06:30 -07:00
9532e084fc
Add tilled version of puts and gets at wavefront level to the functional test suite
...
* Implemented tiled version of put*_wave and get*_wave functions
* Maintain single kernel that supports both tiled and untiled versions
* Disable IPC in the default RO build script
[ROCm/rocshmem commit: b6d31ac7ef
2024-09-07 16:06:36 -07:00
3d26792831
Add tilled version of puts and gets at the workgroup level to the functional test suite
...
[ROCm/rocshmem commit: d226922733
2024-09-07 15:58:14 -07:00
2a87c646f5
Fix bad auto
...
Signed-off-by: Muhammad Awad <MuhammadAbdelghaffar.Awad@amd.com >
[ROCm/rocshmem commit: b811b2b683
2024-09-05 18:39:02 -05:00
31c43bd58d
add functional tests for puts and gets at wavefront level
...
* These functional tests are simple puts and gets, where every wave will get/put the same amount of data
* Enabled workgroup level puts and gets tests
[ROCm/rocshmem commit: ff954237dd
2024-09-05 14:52:48 -07:00
e381ea5478
remove local_pe calculation from puts, gets and atomics functions
...
* All the PEs are assumed to be accessible using IPC backend
[ROCm/rocshmem commit: 7bbf34d334
2024-09-05 11:52:00 -07:00
46176eeeff
Comment out ping pong test
...
* ping pong test fails sporadically
* issues with roc_shmem_wait_until
[ROCm/rocshmem commit: 9c9ef4ffd3
2024-08-28 12:40:51 -07:00
d71286ace4
Update IPC object
...
* Update the state of the IPC object in the RO context class with the instance created in the RO backend
[ROCm/rocshmem commit: 862421e0ca
2024-08-28 12:31:01 -07:00
caf0a7dfad
utils.hpp: silence warnings
...
silence some warnings when compiling with debug flags.
[ROCm/rocshmem commit: 1ffe62bb2e
2024-08-28 09:42:14 -07:00
6f93dd9585
ipc/context_ipc_device.cpp: set barrier_sync
...
set the barrier_sync variable on the context during
object creation
[ROCm/rocshmem commit: aae6295460
2024-08-28 09:41:05 -07:00
14a8bf726a
Add atomics
...
* Add atomic_add, atomic_set, atomic_cas, atomic_fetch_add and atomic_fetch_cas to IPC backend
[ROCm/rocshmem commit: e1e1ac6df6
2024-08-28 08:30:46 -07:00
1335d98f1f
Update IPC object
...
* Update the IPC object in the context class with the instance created in the IPC backend
[ROCm/rocshmem commit: 45a8cb3354
2024-08-28 08:14:38 -07:00
9a380a65ff
first cut on collectives and sync
...
code is based on the GPUIB implementations of the routines, which seem
however generic enough to work also for the IPC conduit.
Some code is in for broadcast, fcollect, and alltoall.
[ROCm/rocshmem commit: 0de3b5e6fc
2024-08-27 15:03:38 -07:00
f0127c9e21
remove device wait_until functions
...
adding the device versions of the wait_until* and test functions in the
ipc folder leads to linking errors of the functional tests. Remove them
and use for now the upper level versions of the functions, similarly to
the RO conduit. Might have to revisit this later again.
[ROCm/rocshmem commit: e2e30b5339
2024-08-27 15:03:32 -07:00
1432a6be71
Add buffers required for collectives
...
[ROCm/rocshmem commit: a9571ec002
2024-08-22 09:28:09 -07:00
6f8d2df73a
Add IPC teams
...
[ROCm/rocshmem commit: a59bdd4f6b
2024-08-22 09:15:44 -07:00
6228db57e4
Add gets and puts functionality to IPC context
...
[ROCm/rocshmem commit: c8b0f2378e
2024-08-15 13:17:44 -07:00
fe866aaea6
Update context_ipc_gtest.cpp to use IPCbackend
...
[ROCm/rocshmem commit: c9dbcf80c2
2024-08-15 11:54:56 -07:00
d06394fa2c
remove ipc_policy.{hpp & cpp} and context_ipc.{hpp & cpp}
...
* move ipc_policy.{hpp & cpp} to src
* rename context_ipc.{hpp & cpp} to context_ipc_device.{hpp & cpp}
[ROCm/rocshmem commit: b68867ee17
2024-08-15 08:52:06 -07:00
60fb77de73
Add IPC backend
...
* add backend_ipc.{cpp & hpp}
* rename context_ipc.{cpp & hpp} to context_ipc_device.{cpp & hpp}
* add host interface to IPC backend
* add context_ipc_host.{cpp & hpp} to support host interface
* add USE_RO compile flag to enable support for single backend interface at a time
* add ipc_single script to build rocSHMEM with IPC backend
[ROCm/rocshmem commit: 49779863c2
2024-08-14 22:59:02 -07:00
a13e2599d9
Code refactor
...
move ipc_policy.hpp and ipc_policy.cpp files to src, since they are used by all the conduits.
[ROCm/rocshmem commit: 24375a949e
2024-08-14 20:44:35 -07:00
bda68c964e
Add ipc unit_tests
...
[ROCm/rocshmem commit: 58c5a98b5d
2024-08-07 12:18:12 -07:00
a542e0a332
Fix memcpy_wave bug
...
[ROCm/rocshmem commit: c865fcd4bb
2024-08-07 11:37:45 -07:00
962f4bedf8
Add assembly for gfx94x
...
[ROCm/rocshmem commit: bb6e48dded
2024-07-31 14:14:24 -05:00
7e446cfc81
unit_tests: add ipc_context tests
...
add the initial outline of an ipc_context unit test. The current test
only invokes the ipc_context constructor, more tests will be added later
as the class is being populated.
Also, at the moment the unit test takes an ROBackend as an argument for
the constructor, not sure whether this will be the final solution.
[ROCm/rocshmem commit: b1bc4a497a
2024-07-30 15:22:03 -07:00
21d8301c83
Move inline asm into separate file
...
[ROCm/rocshmem commit: 862ef5713f
2024-07-30 14:53:19 -05:00
avinashkethineedi
Brandon Potter
Edgar Gabriel
Muhammad Awad
Brandon Potter