| Message ID | 20210422054323.150993-10-aneesh.kumar@linux.ibm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | Speedup mremap on ppc64 | expand |
On Thu, 22 Apr 2021 11:13:23 +0530 "Aneesh Kumar K.V" <aneesh.kumar@linux.ibm.com> wrote: > mremap HAVE_MOVE_PMD/PUD optimization time comparison for 1GB region: > 1GB mremap - Source PTE-aligned, Destination PTE-aligned > mremap time: 1127034ns > 1GB mremap - Source PMD-aligned, Destination PMD-aligned > mremap time: 508817ns > 1GB mremap - Source PUD-aligned, Destination PUD-aligned > mremap time: 23046ns Well that's nice. How significant is this in practice? How common is it for applications to successfully align the region? Do real-world applications actually benefit from this? Are there userspace libraries (malloc() etc) which should be reworked to fully exploit this?
diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype index 3ce907523b1e..2e666e569fdf 100644 --- a/arch/powerpc/platforms/Kconfig.cputype +++ b/arch/powerpc/platforms/Kconfig.cputype @@ -97,6 +97,8 @@ config PPC_BOOK3S_64 select PPC_HAVE_PMU_SUPPORT select SYS_SUPPORTS_HUGETLBFS select HAVE_ARCH_TRANSPARENT_HUGEPAGE + select HAVE_MOVE_PMD + select HAVE_MOVE_PUD select ARCH_ENABLE_THP_MIGRATION if TRANSPARENT_HUGEPAGE select ARCH_SUPPORTS_NUMA_BALANCING select IRQ_WORK
mremap HAVE_MOVE_PMD/PUD optimization time comparison for 1GB region: 1GB mremap - Source PTE-aligned, Destination PTE-aligned mremap time: 1127034ns 1GB mremap - Source PMD-aligned, Destination PMD-aligned mremap time: 508817ns 1GB mremap - Source PUD-aligned, Destination PUD-aligned mremap time: 23046ns Signed-off-by: Aneesh Kumar K.V <aneesh.kumar@linux.ibm.com> --- arch/powerpc/platforms/Kconfig.cputype | 2 ++ 1 file changed, 2 insertions(+)