AI/rocm-systems
2
0
Tükrözés 0
Kód Hibajegyek Egyesítési kérések Actions Packages Projektek Kiadások Wiki Tevékenység

Update Barrier_All and Sync_All APIs (#72)

Forráskód böngészése 
* Fix deadlock in `rocshmem_ctx_wg_barrier_all` API in IPC conduit by adding per-context pSync buffers and context IDs
  - Added separate pSync buffers for each device context
  - Resolved deadlock when invoking barrier API (`rocshmem_ctx_wg_barrier_all`) concurrently from multiple contexts

* Update barrier_all functional tests for multi-context support

* Add thread, wavefront, and workgroup-level barrier_all APIs in IPC and RO conduits
  - Implemented barrier_all APIs at thread, wavefront, and workgroup granularity
  - Added support in both IPC and RO conduits
  - Updated functional tests to cover all `barrier_all` APIs

* Add thread, wavefront, and workgroup-level sync_all APIs in IPC and RO conduits
  - Implemented sync_all APIs for thread, wavefront, and workgroup scopes
  - Added support into both IPC and RO conduits
  - Added functional tests to cover all `sync_all` APIs

[ROCm/rocshmem commit: c652f58cef]
This commit is contained in:
Avinash Kethineedi
2025-04-02 11:58:55 -05:00
committed by GitHub
szülő 0cde5f53dc
commit 426bbf525b
22 fájl változott, egészen pontosan 508 új sor hozzáadva és 53 régi sor törölve
Download Patch File Download Diff File Expand all files Collapse all files
projects/rocshmem/src/backend_bc.cpp
+4
Fájl megtekintése
@@ -130,6 +130,8 @@ void Backend::dump_stats() {
printf("Quiets %llu\n", device_stats.getStat(NUM_QUIET));
printf("ToAll %llu\n", device_stats.getStat(NUM_TO_ALL));
printf("BarrierAll %llu\n", device_stats.getStat(NUM_BARRIER_ALL));
printf("WAVE_BarrierAll %llu\n", device_stats.getStat(NUM_BARRIER_ALL_WAVE));
printf("WG_BarrierAll %llu\n", device_stats.getStat(NUM_BARRIER_ALL_WG));
printf("Wait Until %llu\n", device_stats.getStat(NUM_WAIT_UNTIL));
printf("Wait Until Any %llu\n", device_stats.getStat(NUM_WAIT_UNTIL_ANY));
printf("Wait Until All %llu\n", device_stats.getStat(NUM_WAIT_UNTIL_ALL));
@@ -153,6 +155,8 @@ void Backend::dump_stats() {
printf("Tests %llu\n", device_stats.getStat(NUM_TEST));
printf("SHMEM_PTR %llu\n", device_stats.getStat(NUM_SHMEM_PTR));
printf("SyncAll %llu\n", device_stats.getStat(NUM_SYNC_ALL));
printf("WAVE_SyncAll %llu\n", device_stats.getStat(NUM_SYNC_ALL_WAVE));
printf("WG_SyncAll %llu\n", device_stats.getStat(NUM_SYNC_ALL_WG));
const auto& host_stats{globalHostStats};
printf("HOST STATS\n");