| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
| |
Barrier only ensure one work group finish, can't guarantee all work item's atomic ops
have finished before the last atomic_add.
So use atomic_xchg to update first work group's local buffer to other global buffer position.
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Xing, Homer <homer.xing@intel.com>
|
| |
|
|
|
|
|
| |
Because atomic's address used as src and dst, so need to clear this address.
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Zhigang Gong <zhigang.gong@linux.intel.com>
|
| |
|
|
|
|
|
|
| |
The test case include local memory and global memory, atomic operations from
different threads and different work groups.
Signed-off-by: Yang Rong <rong.r.yang@intel.com>
Reviewed-by: Zhigang Gong <zhigang.gong@linux.intel.com>
|
| |
|
