diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-08 10:03:19 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-08 10:03:19 -0400 |
| commit | 0c75c0cbbf51c71dd70396082beca8cb0c102332 (patch) | |
| tree | bc7b4c170c80f1dc618d100e44c22393ecc86c63 /tools/cuda-ab/perf/perf_workload.cu | |
| parent | 7300003c2c95e037f903e281b9264d69a8496f27 (diff) | |
| download | open-iov-0c75c0cbbf51c71dd70396082beca8cb0c102332.tar.gz open-iov-0c75c0cbbf51c71dd70396082beca8cb0c102332.zip | |
vgsp-rm: PROMOTE_CTX reply acknowledges ctx-buffer initialization
kgrobjPromoteContext_IMPL marks a context buffer initialized only when the
REPLY's promoteEntry[i].bInitialize is clear; echoing the request back kept
the flag set so the RM re-promoted the same contexts indefinitely. Clear
the flag on the echoed reply (per-entry, 5 entries observed per call).
Note: this aligns the reply with the caller contract but did NOT by itself
unblock the observed cudaMalloc stall (client still reaches DEVICE then
spins; zero GR ring content in VRAM) - kept because the old behavior was
wrong and the flag gates kgrctxMarkCtxBufferInitialized.
Diffstat (limited to 'tools/cuda-ab/perf/perf_workload.cu')
0 files changed, 0 insertions, 0 deletions