On Tue, Jun 21, 2016 at 02:52:37PM -0700, Andrew Morton wrote: > On Tue, 21 Jun 2016 02:33:06 +0300 Yury Norov <yury.norov@xxxxxxxxx> wrote: > > > kmalloc_index() uses simple straightforward way to calculate > > bit position of nearest or equal upper power of 2. > > This effectively results in generation of 24 episodes of > > compare-branch instructions in assembler. > > > > There is shorter way to calculate this: fls(size - 1). > > > > The patch removes hard-coded calculation of kmalloc slab and > > uses ilog2() instead that works on top of fls(). ilog2 is used > > with intention that compiler also might optimize constant case > > during compile time if it detects that. > > > > BUG() is moved to the beginning of function. We left it here to > > provide identical behaviour to previous version. It may be removed > > if there's no requirement in it anymore. > > > > While we're at this, fix comment that describes return value. > > kmalloc_index() is always called with a constant-valued `size' (see > __builtin_constant_p() tests) It might change one day. This function is public to any slab user. If you really want to allow call kmalloc_index() for constants only, you'd place __builtin_constant_p() tests inside kmalloc_index(). > so the compiler will evaluate the switch > statement at compile-time. This will be more efficient than calling > fls() at runtime. There will be no fls() for constant at runtime because ilog2() calculates constant values at compile-time as well. From this point of view, this patch removes code duplication, as we already have compile-time log() calculation in kernel, and should re-use it whenever possible.\ Yury. -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>