Re: [PATCH v1] mm/contpte: Optimize loop to reduce redundant operations

[Ryan Roberts <ryan.roberts@arm.com>]

On 14/04/2025 09:51, Dev Jain wrote:
> 
> 
> On 14/04/25 1:36 pm, Ryan Roberts wrote:
>> On 12/04/2025 06:05, Lance Yang wrote:
>>> On Sat, Apr 12, 2025 at 1:30 AM Dev Jain <dev.jain@arm.com> wrote:
>>>>
>>>> +others
>>>>
>>>> On 11/04/25 2:55 am, Barry Song wrote:
>>>>> On Mon, Apr 7, 2025 at 9:23 PM Xavier <xavier_qy@163.com> wrote:
>>>>>>
>>>>>> This commit optimizes the contpte_ptep_get function by adding early
>>>>>>    termination logic. It checks if the dirty and young bits of orig_pte
>>>>>>    are already set and skips redundant bit-setting operations during
>>>>>>    the loop. This reduces unnecessary iterations and improves performance.
>>>>>>
>>>>>> Signed-off-by: Xavier <xavier_qy@163.com>
>>>>>> ---
>>>>>>    arch/arm64/mm/contpte.c | 13 +++++++++++--
>>>>>>    1 file changed, 11 insertions(+), 2 deletions(-)
>>>>>>
>>>>>> diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
>>>>>> index bcac4f55f9c1..ca15d8f52d14 100644
>>>>>> --- a/arch/arm64/mm/contpte.c
>>>>>> +++ b/arch/arm64/mm/contpte.c
>>>>>> @@ -163,17 +163,26 @@ pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
>>>>>>
>>>>>>           pte_t pte;
>>>>>>           int i;
>>>>>> +       bool dirty = false;
>>>>>> +       bool young = false;
>>>>>>
>>>>>>           ptep = contpte_align_down(ptep);
>>>>>>
>>>>>>           for (i = 0; i < CONT_PTES; i++, ptep++) {
>>>>>>                   pte = __ptep_get(ptep);
>>>>>>
>>>>>> -               if (pte_dirty(pte))
>>>>>> +               if (!dirty && pte_dirty(pte)) {
>>>>>> +                       dirty = true;
>>>>>>                           orig_pte = pte_mkdirty(orig_pte);
>>>>>> +               }
>>>>>>
>>>>>> -               if (pte_young(pte))
>>>>>> +               if (!young && pte_young(pte)) {
>>>>>> +                       young = true;
>>>>>>                           orig_pte = pte_mkyoung(orig_pte);
>>>>>> +               }
>>>>>> +
>>>>>> +               if (dirty && young)
>>>>>> +                       break;
>>>>>
>>>>> This kind of optimization is always tricky. Dev previously tried a similar
>>>>> approach to reduce the loop count, but it ended up causing performance
>>>>> degradation:
>>>>> https://lore.kernel.org/linux-mm/20240913091902.1160520-1-dev.jain@arm.com/
>>>>>
>>>>> So we may need actual data to validate this idea.
>>>>
>>>> The original v2 patch does not work, I changed it to the following:
>>>>
>>>> diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
>>>> index bcac4f55f9c1..db0ad38601db 100644
>>>> --- a/arch/arm64/mm/contpte.c
>>>> +++ b/arch/arm64/mm/contpte.c
>>>> @@ -152,6 +152,16 @@ void __contpte_try_unfold(struct mm_struct *mm,
>>>> unsigned long addr,
>>>>    }
>>>>    EXPORT_SYMBOL_GPL(__contpte_try_unfold);
>>>>
>>>> +#define CHECK_CONTPTE_FLAG(start, ptep, orig_pte, flag) \
>>>> +       int _start; \
>>>> +       pte_t *_ptep = ptep; \
>>>> +       for (_start = start; _start < CONT_PTES; _start++, ptep++) { \
>>>> +               if (pte_##flag(__ptep_get(_ptep))) { \
>>>> +                       orig_pte = pte_mk##flag(orig_pte); \
>>>> +                       break; \
>>>> +               } \
>>>> +       }
>>>> +
>>>>    pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
>>>>    {
>>>>           /*
>>>> @@ -169,11 +179,17 @@ pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
>>>>           for (i = 0; i < CONT_PTES; i++, ptep++) {
>>>>                   pte = __ptep_get(ptep);
>>>>
>>>> -               if (pte_dirty(pte))
>>>> +               if (pte_dirty(pte)) {
>>>>                           orig_pte = pte_mkdirty(orig_pte);
>>>> +                       CHECK_CONTPTE_FLAG(i, ptep, orig_pte, young);
>>>> +                       break;
>>>> +               }
>>>>
>>>> -               if (pte_young(pte))
>>>> +               if (pte_young(pte)) {
>>>>                           orig_pte = pte_mkyoung(orig_pte);
>>>> +                       CHECK_CONTPTE_FLAG(i, ptep, orig_pte, dirty);
>>>> +                       break;
>>>> +               }
>>>>           }
>>>>
>>>>           return orig_pte;
>>>>
>>>> Some rudimentary testing with micromm reveals that this may be
>>>> *slightly* faster. I cannot say for sure yet.
>>>
>>> Yep, this change works as expected, IIUC.
>>>
>>> However, I'm still wondering if the added complexity is worth it for
>>> such a slight/negligible performance gain. That said, if we have
>>> solid numbers/data to back it up, all doubts would disappear ;)
>>
>> I agree with Barry; we need clear performance improvement numbers to consider
>> this type of optimization. I doubt there will be measurable improvement for 4K
>> and 64K base pages (because the the number of PTEs in a contpte block are 16 and
>> 32 respectively). But 16K base pages may benefit given there are 128 PTEs in a
>> contpte block for that case.
>>
>> Also FWIW, I'm struggling to understand CHECK_CONTPTE_FLAG(); Perhaps something
>> like this would suffice?
> 
> The idea of CHECK_CONTPTE_FLAG() is to check only for the other flag until it's
> found; the diff below will still keep checking for both until both are found.

Oh I see. Although you're effectively just saving a tbnz instruction, which will
be in the noise.

Anyway, come back with numbers and we can talk :)

> 
>>
>> ----8<----
>> diff --git a/arch/arm64/mm/contpte.c b/arch/arm64/mm/contpte.c
>> index 55107d27d3f8..7787b116b339 100644
>> --- a/arch/arm64/mm/contpte.c
>> +++ b/arch/arm64/mm/contpte.c
>> @@ -169,11 +169,17 @@ pte_t contpte_ptep_get(pte_t *ptep, pte_t orig_pte)
>>          for (i = 0; i < CONT_PTES; i++, ptep++) {
>>                  pte = __ptep_get(ptep);
>>
>> -               if (pte_dirty(pte))
>> +               if (pte_dirty(pte)) {
>>                          orig_pte = pte_mkdirty(orig_pte);
>> +                       if (pte_young(orig_pte))
>> +                               break;
>> +               }
>>
>> -               if (pte_young(pte))
>> +               if (pte_young(pte)) {
>>                          orig_pte = pte_mkyoung(orig_pte);
>> +                       if (pte_dirty(orig_pte))
>> +                               break;
>> +               }
>>          }
>>
>>          return orig_pte;
>> ----8<----
>>
>> Thanks,
>> Ryan
>>
>>
>>>
>>> Thanks,
>>> Lance
>>>
>>>>
>>>>>
>>>>>>           }
>>>>>>
>>>>>>           return orig_pte;
>>>>>> -- 
>>>>>> 2.34.1
>>>>>>
>>>>>
>>>>> Thanks
>>>>> Barry
>>>>>
>>>>
>>
>