An Overview of Kernel Lock Improvements [pdf]
events.linuxfoundation.org
An Overview of Kernel Lock Improvements [pdf]
1–10 of 12 posts
Re: An Overview of Kernel Lock Improvements [pdf]
#2This is great; I'm still reading through, but, I had no idea that the performance drop was that steep!
Re: An Overview of Kernel Lock Improvements [pdf]
#3great
Re: An Overview of Kernel Lock Improvements [pdf]
#4This is great; I'm still reading through, but, I had no idea that the performance drop was that steep!
Do you mean the drop they later used to analyze cache line coherency when adding cores? They improved this if I understood the latter results correctly.
Re: An Overview of Kernel Lock Improvements [pdf]
#5Note: for those interested, MCS locks and qspinlocks are discussed in this fine article: http://lwn.net/Articles/590243/
Re: An Overview of Kernel Lock Improvements [pdf]
#6This is the key: each locker spins on local memory rather than the lock word at page 63.
Re: An Overview of Kernel Lock Improvements [pdf]
#7It would be interesting to see the same benchmarks for more normal amounts of cores. Most system does not have 240 cores, after all..
Re: An Overview of Kernel Lock Improvements [pdf]
#8It would be interesting to see the same benchmarks for more normal amounts of cores. Most system does not have 240 cores, after all..
The latest Intel hi-end desktop cpu has 8 cores. With hyper-threading; which makes 16 threads.
I think in the next 10 years we will reach a number that is close.
Re: An Overview of Kernel Lock Improvements [pdf]
#9Can anyone explain the use of 'round-robin' to describe mulit-node scenarios and 'fill-first' for single node scenarios. I initially assumed they were describing thread schedulers, but that doesn't make clear sense in these tests. Thanks in advance.
Re: An Overview of Kernel Lock Improvements [pdf]
#10It would be interesting to see the same benchmarks for more normal amounts of cores. Most system does not have 240 cores, after all..
The latest Intel hi-end desktop cpu has 8 cores. With hyper-threading; which makes 16 threads. I think in the next 10 years we will reach a number that is close.
If you count Xeon Phi, Intel are already up to 61.
http://ark.intel.com/products/family/71840/Intel-Xeon-Phi-Co...