Re: [Patch v3 0/3] iommu: reduce spinlock contention on fast path

[John Garry]

On 06/12/2019 21:38, Cong Wang wrote:
> This patchset contains three small optimizations for the global spinlock
> contention in IOVA cache. Our memcache perf test shows this reduced its
> p999 latency down by 45% on AMD when IOMMU is enabled.
>
> Cong Wang (3):
>    iommu: avoid unnecessary magazine allocations
>    iommu: optimize iova_magazine_free_pfns()
>    iommu: avoid taking iova_rbtree_lock twice
> ---
>   drivers/iommu/iova.c | 75 ++++++++++++++++++++++++++------------------
>   1 file changed, 45 insertions(+), 30 deletions(-)

I retested, and got a ~1.1% gain in throughput for my storage test - 
results here if interested https://pastebin.com/dSQxYpN8

Thanks,
John