AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Refactor RO backend data structures (#49)

Browse Source 
- Remove hdp and ipc pointers from BlockHandle, align RO stats with RO contexts

- Add run commands for `rocshmem_g` and `rocshmem_p` API tests in driver.sh

- Allocate rocshmem API return buffers based on number of device contexts.

- Associate status flag address with blocking calls and remove threadId dependency
   - Associated the status flag address with each blocking call request to notify the GPU thread.
   - Removed dependency on threadId for determining the appropriate status flag index.

- Move status flag buffer allocation to backend.

- Initialize allocated memeory to zero

[ROCm/rocshmem commit: df4ad2c04d]
This commit is contained in:
Avinash Kethineedi
2025-03-14 10:49:44 -05:00
committed by GitHub
parent a16492cdf9
commit 7f3879ff31
16 changed files with 389 additions and 357 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocshmem/scripts/functional_tests/driver.sh
+14
View File
@@ -179,6 +179,20 @@ TestRMA() {
ExecTest "teamctxget" 2 1 1 1048576
ExecTest "g" 2 1 1 1048576
ExecTest "g" 2 1 1024 512
ExecTest "g" 2 8 1 1048576
ExecTest "g" 2 16 128 8
ExecTest "g" 2 32 256 512
ExecTest "g" 2 64 1024 8
ExecTest "p" 2 1 1 1048576
ExecTest "p" 2 1 1024 512
ExecTest "p" 2 8 1 1048576
ExecTest "p" 2 16 128 8
ExecTest "p" 2 32 256 512
ExecTest "p" 2 64 1024 8
################################ Non-Blocking ################################
ExecTest "putnbi" 2 1 1 1048576