You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We discussed about that problem with Lourival and decided to try use pool(9).
Firstly, I dived into current kmem_intr_alloc/free implementation that already using pool internally. Analyzing of code shows that current implementation of lua_alloc function that using kmem_intr_alloc/free internally (that use pool internally :)) seems efficient because LUAL_BUFFERSIZE less that maximum allocatable chunk of memory in kmem allocator pool (16KiB).
I have some ideas and I would like to discuss about them.
kmem allocator implementation seems very efficient and lua_alloc consequently too. But we have performance problem. It seems that something wrong with my hands. :) But all aims that problem not in I/O operations...
At that time I creating minimal test code and want to simulate lua_alloc calls as it was. After that compare results with userland similar tests.
May be you have some ideas or hypothesis? I spend a lot of time on experiments and measurements. Currently, I have enough time to work on our project. ;)
The text was updated successfully, but these errors were encountered:
Dear colleagues,
Issue (in fork repository) with performance tests and results.
We discussed about that problem with Lourival and decided to try use pool(9).
Firstly, I dived into current kmem_intr_alloc/free implementation that already using pool internally. Analyzing of code shows that current implementation of lua_alloc function that using
kmem_intr_alloc/free
internally (that use pool internally :)) seems efficient becauseLUAL_BUFFERSIZE
less that maximum allocatable chunk of memory in kmem allocator pool (16KiB).I have some ideas and I would like to discuss about them.
lua_alloc
consequently too. But we have performance problem. It seems that something wrong with my hands. :) But all aims that problem not in I/O operations...lua_alloc
calls as it was. After that compare results with userland similar tests.The text was updated successfully, but these errors were encountered: