Re: [PATCH v12 2/2] mm: page_alloc: reduce unnecessary binary search in memblock_next_valid_pfn
On 2019/7/23 16:33, Mike Rapoport wrote: > On Tue, Jul 23, 2019 at 01:51:13PM +0800, Hanjun Guo wrote: >> From: Jia He >> >> After skipping some invalid pfns in memmap_init_zone(), there is still >> some room for improvement. >> >> E.g. if pfn and pfn+1 are in the same memblock region, we can simply pfn++ >> instead of doing the binary search in memblock_next_valid_pfn. >> >> Furthermore, if the pfn is in a gap of two memory region, skip to next >> region directly to speedup the binary search. > How much speed up do you see with this improvements relatively to simple > binary search in memblock_next_valid_pfn()? The major speedup on my platform is the previous patch in this patch set, not this one, I think it's related to sparse memory mode for different platforms. Thanks Hanjun >
Re: [PATCH v12 2/2] mm: page_alloc: reduce unnecessary binary search in memblock_next_valid_pfn
On Tue, Jul 23, 2019 at 01:51:13PM +0800, Hanjun Guo wrote:
> From: Jia He
>
> After skipping some invalid pfns in memmap_init_zone(), there is still
> some room for improvement.
>
> E.g. if pfn and pfn+1 are in the same memblock region, we can simply pfn++
> instead of doing the binary search in memblock_next_valid_pfn.
>
> Furthermore, if the pfn is in a gap of two memory region, skip to next
> region directly to speedup the binary search.
How much speed up do you see with this improvements relatively to simple
binary search in memblock_next_valid_pfn()?
> Signed-off-by: Jia He
> Signed-off-by: Hanjun Guo
> ---
> mm/memblock.c | 37 +++--
> 1 file changed, 31 insertions(+), 6 deletions(-)
>
> diff --git a/mm/memblock.c b/mm/memblock.c
> index d57ba51bb9cd..95d5916716a0 100644
> --- a/mm/memblock.c
> +++ b/mm/memblock.c
> @@ -1256,28 +1256,53 @@ int __init_memblock memblock_set_node(phys_addr_t
> base, phys_addr_t size,
> unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn)
> {
> struct memblock_type *type = &memblock.memory;
> + struct memblock_region *regions = type->regions;
> unsigned int right = type->cnt;
> unsigned int mid, left = 0;
> + unsigned long start_pfn, end_pfn, next_start_pfn;
> phys_addr_t addr = PFN_PHYS(++pfn);
> + static int early_region_idx __initdata_memblock = -1;
>
> + /* fast path, return pfn+1 if next pfn is in the same region */
> + if (early_region_idx != -1) {
> + start_pfn = PFN_DOWN(regions[early_region_idx].base);
> + end_pfn = PFN_DOWN(regions[early_region_idx].base +
> + regions[early_region_idx].size);
> +
> + if (pfn >= start_pfn && pfn < end_pfn)
> + return pfn;
> +
> + /* try slow path */
> + if (++early_region_idx == type->cnt)
> + goto slow_path;
> +
> + next_start_pfn = PFN_DOWN(regions[early_region_idx].base);
> +
> + if (pfn >= end_pfn && pfn <= next_start_pfn)
> + return next_start_pfn;
> + }
> +
> +slow_path:
> + /* slow path, do the binary searching */
> do {
> mid = (right + left) / 2;
>
> - if (addr < type->regions[mid].base)
> + if (addr < regions[mid].base)
> right = mid;
> - else if (addr >= (type->regions[mid].base +
> - type->regions[mid].size))
> + else if (addr >= (regions[mid].base + regions[mid].size))
> left = mid + 1;
> else {
> - /* addr is within the region, so pfn is valid */
> + early_region_idx = mid;
> return pfn;
> }
> } while (left < right);
>
> if (right == type->cnt)
> return -1UL;
> - else
> - return PHYS_PFN(type->regions[right].base);
> +
> + early_region_idx = right;
> +
> + return PHYS_PFN(regions[early_region_idx].base);
> }
> EXPORT_SYMBOL(memblock_next_valid_pfn);
> #endif /* CONFIG_HAVE_MEMBLOCK_PFN_VALID */
> --
> 2.19.1
>
--
Sincerely yours,
Mike.
[PATCH v12 2/2] mm: page_alloc: reduce unnecessary binary search in memblock_next_valid_pfn
From: Jia He
After skipping some invalid pfns in memmap_init_zone(), there is still
some room for improvement.
E.g. if pfn and pfn+1 are in the same memblock region, we can simply pfn++
instead of doing the binary search in memblock_next_valid_pfn.
Furthermore, if the pfn is in a gap of two memory region, skip to next
region directly to speedup the binary search.
Signed-off-by: Jia He
Signed-off-by: Hanjun Guo
---
mm/memblock.c | 37 +++--
1 file changed, 31 insertions(+), 6 deletions(-)
diff --git a/mm/memblock.c b/mm/memblock.c
index d57ba51bb9cd..95d5916716a0 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -1256,28 +1256,53 @@ int __init_memblock memblock_set_node(phys_addr_t base,
phys_addr_t size,
unsigned long __init_memblock memblock_next_valid_pfn(unsigned long pfn)
{
struct memblock_type *type = &memblock.memory;
+ struct memblock_region *regions = type->regions;
unsigned int right = type->cnt;
unsigned int mid, left = 0;
+ unsigned long start_pfn, end_pfn, next_start_pfn;
phys_addr_t addr = PFN_PHYS(++pfn);
+ static int early_region_idx __initdata_memblock = -1;
+ /* fast path, return pfn+1 if next pfn is in the same region */
+ if (early_region_idx != -1) {
+ start_pfn = PFN_DOWN(regions[early_region_idx].base);
+ end_pfn = PFN_DOWN(regions[early_region_idx].base +
+ regions[early_region_idx].size);
+
+ if (pfn >= start_pfn && pfn < end_pfn)
+ return pfn;
+
+ /* try slow path */
+ if (++early_region_idx == type->cnt)
+ goto slow_path;
+
+ next_start_pfn = PFN_DOWN(regions[early_region_idx].base);
+
+ if (pfn >= end_pfn && pfn <= next_start_pfn)
+ return next_start_pfn;
+ }
+
+slow_path:
+ /* slow path, do the binary searching */
do {
mid = (right + left) / 2;
- if (addr < type->regions[mid].base)
+ if (addr < regions[mid].base)
right = mid;
- else if (addr >= (type->regions[mid].base +
- type->regions[mid].size))
+ else if (addr >= (regions[mid].base + regions[mid].size))
left = mid + 1;
else {
- /* addr is within the region, so pfn is valid */
+ early_region_idx = mid;
return pfn;
}
} while (left < right);
if (right == type->cnt)
return -1UL;
- else
- return PHYS_PFN(type->regions[right].base);
+
+ early_region_idx = right;
+
+ return PHYS_PFN(regions[early_region_idx].base);
}
EXPORT_SYMBOL(memblock_next_valid_pfn);
#endif /* CONFIG_HAVE_MEMBLOCK_PFN_VALID */
--
2.19.1
