diff options
| author | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 15:03:33 -0400 |
|---|---|---|
| committer | Arthur Hanson Rasmusson <arthur@vgpu.io> | 2026-10-23 15:03:33 -0400 |
| commit | aa008361479882935798c78ea8ebd791533cca63 (patch) | |
| tree | 7ede5abba2cc0ff18889b27c119188c966de9ed4 /tools/cuda-ab/run/ip4 | |
| parent | 8de9f0e24d9511db9edab879fac1f8f07a3e9fbc (diff) | |
| download | open-iov-aa008361479882935798c78ea8ebd791533cca63.tar.gz open-iov-aa008361479882935798c78ea8ebd791533cca63.zip | |
vgsp-gr r554: the per-VA launch-defer latch needs 64 slots, not 8
r553 soak evidence (victory_reduce, 430 s stall, 0 hangs): the same launch
push VA (0x20043be64) was deferred on every pass for 7+ minutes ("partial
argument set: 7-8 pages unresolved"). A client's UVM pool allocates ~8-10
channels and each command buffer reuses two launch push slots, so more
than 8 distinct push VAs rotate through the guard; with 8 latch slots each
new VA evicted another and the one-deferral-per-VA guarantee thrashed -
the second sighting of any VA found its record gone and deferred again,
so a partial argument set that never resolves held the launch forever.
64 slots cover a full client with margin (1 KB of static state).
Diffstat (limited to 'tools/cuda-ab/run/ip4')
0 files changed, 0 insertions, 0 deletions
