- Replaces logic relying on clock64() instead to
use an atomic counter that the last wg in grid
and multi-grid groups wait on while the the
non-last work groups increment the counter
to "done"
- Implements a new logic to verify the multi-grid
sync()
Change-Id: I7780d8124e5f144b124e5d191d0f412483a3b565
This change fixes build warning when compiled with -Wcast-function-type flag
Signed-off-by: Sebastian Luzynski <Sebastian.Luzynski@amd.com>
Change-Id: If60abf5c6855879a07375c166d7cb9ff6cd6fe5d
- make blockdim to become 0 and hence need validation.
- run cooperative kernel with param lesser # of threads
than actual # of elements and hence need correction in
validation part.
Change-Id: Ie48a038fe6f8cf25d7f6460703648a360f511f35
The test checks the pattern of thing "A" should be always faster than
thing "B". Which might not be the case for several reasons and will result
in unreliable, randomly failing test.
Change-Id: I6fa9e2c110f9ef48ef63a720e7a64c54e7f2a72f
- Remove unused variables
- Validate return value of HIP APIs
- Disable failed tests in PSDB
- Use HIP_CHECK_ERROR instead of ASSERT_EQUAL
Change-Id: I9dd0792ecd8c79760b6c948388a953ac8c97b5c2
* SWDEV-385161 - Deprecate usage of env vars in HIP samples and tests env vars such as HIP_PATH, ROCM_PATH, HIP_COMPILER, HIP_RUNTIME, HSA_PATH etc are removed.
Change-Id: I280e20b291cfb14675846aae72d97fa5b3ad567d
- Migrate basic Cooperative Groups tests and integrate to catch
- Refactor basic Cooperative Groups tests
- Rename tiled partition related files and fix minor bug
- Add LaunchCooperativeKernal and LaunchCooperativeKernelMultiDevice tests
- Refactor hipCGThreadBlockTileType to use common function
- Fix updated file not added during merge
- Add coalesced_group type tests
- Add coalesced_group shuffle_up and shuffle_down tests
- Add coalesced_group shuffle tests - test fails
- Implement common code for cooperative group tests
- Fixed compilation errror in cooperative_groups_common.hh
- Implement busy wait device function
- Reimplement tests for grid_group APIs
- Add tests for grid_group member and non-member APIs
- Refactor existing test for grid_group sync testing
- Add thread and block dimensions generators
- Add check of grid and block dimensions
- Modify doxygen comments
- Move cpu_grid.h and supporting functions to catch/include
- Use warp_size from properties in grid/block dims generators
- Fix condition for warp size 32 on AMD
- Fix cpu_grid.h for warp function tests
- Add missing include into cpu_grid.h
- Code cleanup
- Fix doxygen comments
- Add missing include in utils header
Marko Arandjelovic
Sourabh Betigeri
Vladana Stojiljkovic
Jatin Chaudhary
Rahul Manocha
Alex Xie
Sebastian Luzynski
Jaydeep Patel
Branislav Brzak
Mirza Halilcevic
Nives Vukovic
Rakesh Roy
Maneesh Gupta
Rupam Chetia
ROCm CI Service Account
nives-vukovic
Paulius Velesko