exec: Introduce AddressSpaceDispatch.mru_section

Under heavy workloads the lookup will likely end up with the same
MemoryRegionSection from last time. Using a pointer to cache the result,
like ram_list.mru_block, significantly reduces cost of
address_space_translate.

During address space topology update, as->dispatch will be reallocated
so the pointer is invalidated automatically.

Perf reports a visible drop on the cpu usage, because phys_page_find is
not called.  Before:

   2.35%  qemu-system-x86_64       [.] phys_page_find
   0.97%  qemu-system-x86_64       [.] address_space_translate_internal
   0.95%  qemu-system-x86_64       [.] address_space_translate
   0.55%  qemu-system-x86_64       [.] address_space_lookup_region

After:

   0.97%  qemu-system-x86_64       [.] address_space_translate_internal
   0.97%  qemu-system-x86_64       [.] address_space_lookup_region
   0.84%  qemu-system-x86_64       [.] address_space_translate

Signed-off-by: Fam Zheng <famz@redhat.com>
Message-Id: <1456813104-25902-8-git-send-email-famz@redhat.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
This commit is contained in:
Fam Zheng 2016-03-01 14:18:24 +08:00 committed by Paolo Bonzini
parent 29cb533d8c
commit 729633c2bc

16
exec.c
View File

@ -135,6 +135,7 @@ typedef struct PhysPageMap {
struct AddressSpaceDispatch {
struct rcu_head rcu;
MemoryRegionSection *mru_section;
/* This is a multi-level map on the physical address space.
* The bottom level has pointers to MemoryRegionSections.
*/
@ -351,14 +352,25 @@ static MemoryRegionSection *address_space_lookup_region(AddressSpaceDispatch *d,
hwaddr addr,
bool resolve_subpage)
{
MemoryRegionSection *section;
MemoryRegionSection *section = atomic_read(&d->mru_section);
subpage_t *subpage;
bool update;
section = phys_page_find(d->phys_map, addr, d->map.nodes, d->map.sections);
if (section && section != &d->map.sections[PHYS_SECTION_UNASSIGNED] &&
section_covers_addr(section, addr)) {
update = false;
} else {
section = phys_page_find(d->phys_map, addr, d->map.nodes,
d->map.sections);
update = true;
}
if (resolve_subpage && section->mr->subpage) {
subpage = container_of(section->mr, subpage_t, iomem);
section = &d->map.sections[subpage->sub_section[SUBPAGE_IDX(addr)]];
}
if (update) {
atomic_set(&d->mru_section, section);
}
return section;
}