More Improvements on Linux Kernel

We have seen several performance improvements introduced on every merge window just after a new Linux Kernel is being released to the public by Linus Torvalds. This trends will always an interesting (and sometimes controversial) topics whenever new ideas came out since performance itself in most cases cannot be achieved without taking some risk and probably could break others as well. In short, gaining performance requires some sacrifices before you get the optimum results.

A new patch is being developed at this moment and it's predicted to be pushed to the main repository when Linux Kernel 3.1 gets released and a new merge window for Linux Kernel 3.2 is open for public. Phoronix already posted about this news. Follow the discussion on LKML

Another Substantial Improvement

Linus has just written a quick note about his pull on Nick Piggin's work on RCU lookup that he considered scary, but interesting and impressive one. Scary since it's some very core code, but he also likes it because it significantly improve the performance on his machine (he only conduct a small test with find).

I'm looking forward to see major performance improvements on Linux Kernel 2.6.38 which should be out around April

Another Linux Improvement Patches Attempt

There's always room for improvement. I think that quotes also apply to Linux Kernel development. Every new kernel release always tries to bring new features, but behind all those new features, there's always an attempt to improve performance.

Several weeks ago, Phoronix wrote an interesting article about 200 lines of code which could improve desktop performance by automatically creating task groups per TTY. While this patch set has done a great job, another attempt has been made to further improve it.

This month, Andrea has updated his patch set against 2.6.37-rc1 and based on benchmarking done by Phoronix using Phoronix Test Suite, it does give an improvement compared to other kernels used on the test process. So, what's the idea behind this patch set?

Taken from Phoronix's article:

The Transparent Hugepage Support in the Linux kernel works by reducing the number of TLB (Translation Lookaside Buffer) entries that such applications need and at the same time increasing the cap that a TLB cache can provide.

For more information about THS, please have a look on article by LWN and also documentation in the Kernel.

I think this will be released in 2.6.37 or probably in 2.6.38 if it's too late for 2.6.37. I'm looking forward to see more improvements in the future

Linux Desktop Will be More Responsive

Linux kernel developers has tried so hard to improve desktop responsiveness by adding more patches in the current development version of 2.6.36. Phoronix has written an article about it. Even so, it will not be fully deployed in 2.6.36 and most of the patches will be pushed to the mainline kernel when merge window for 2.6.37 has been opened, which means that we will see those improvements next year, as one release will take about three months and it's considered too late to push this in 2.6.36 as Linus is very strict lately (which is good in some circumstances) about merging new features.

Performance Improvement on 2.6.36

So far, i'm quite impressed with 2.6.35 Linux Kernel, since it's more stable and IMHO, it's much more faster rather than 2.6.34. But, looking fore 2.6.36 merge window information on LWN makes me more curious about what will be implemented in 2.6.36 which can improve performance on system overall.

I really wished they fixed some performance regression found on 2.6.35