[dpdk-dev] Regarding rte_memzone_reserve with len =0

[dpdk-dev] Regarding rte_memzone_reserve with len =0

[Gonzalez Monroy, Sergio]

Hi,

I was wondering about the use case of rte_memzone_reserve_xxxx APIs with 
len=0.

 From the docs (http://dpdk.org/doc/api/rte__memzone_8h.html):
len          The size of the memory to be reserved. If it is 0, the 
biggest contiguous zone will be reserved.

What are the use cases?
When would you want a memzone of undetermined size?

Any thoughts appreciated.

Sergio

Re: [dpdk-dev] Regarding rte_memzone_reserve with len =0

[Olivier MATZ]

Hi Sergio,

On 05/06/2015 06:10 PM, Gonzalez Monroy, Sergio wrote:
> Hi,
>
> I was wondering about the use case of rte_memzone_reserve_xxxx APIs with
> len=0.
>
>  From the docs (http://dpdk.org/doc/api/rte__memzone_8h.html):
> len          The size of the memory to be reserved. If it is 0, the
> biggest contiguous zone will be reserved.
>
> What are the use cases?
> When would you want a memzone of undetermined size?
>
> Any thoughts appreciated.

As the application does not have access to the lengths of memory
segments, probably the initial idea is when an application wants
to allocate more memory that the biggest segment.

Example, the application wants to allocate 1G (even fragmented):

- the easy case is when it can be done in one call to
   rte_memzone_reserve(1G)

- else, the application can iterate like in this sample:

      remain = 1 * 1024 * 1024 * 1024;
      while (remain > 0) {
          mz = rte_memzone_reserve(remain);
          if (mz != NULL)
             return 0;
          mz = rte_memzone_reserve(remain);
          if (mz == NULL)
             return -1;
          remain -= mz->len;
      }

Regards,
Olivier

Re: [dpdk-dev] Regarding rte_memzone_reserve with len =0

[Gonzalez Monroy, Sergio]

On 12/05/2015 15:48, Olivier MATZ wrote:
> Hi Sergio,
>
> On 05/06/2015 06:10 PM, Gonzalez Monroy, Sergio wrote:
>> Hi,
>>
>> I was wondering about the use case of rte_memzone_reserve_xxxx APIs with
>> len=0.
>>
>>  From the docs (http://dpdk.org/doc/api/rte__memzone_8h.html):
>> len          The size of the memory to be reserved. If it is 0, the
>> biggest contiguous zone will be reserved.
>>
>> What are the use cases?
>> When would you want a memzone of undetermined size?
>>
>> Any thoughts appreciated.
>
> As the application does not have access to the lengths of memory
> segments, probably the initial idea is when an application wants
> to allocate more memory that the biggest segment.
>
> Example, the application wants to allocate 1G (even fragmented):
>
> - the easy case is when it can be done in one call to
>   rte_memzone_reserve(1G)
>
> - else, the application can iterate like in this sample:
>
>      remain = 1 * 1024 * 1024 * 1024;
>      while (remain > 0) {
>          mz = rte_memzone_reserve(remain);
>          if (mz != NULL)
>             return 0;
>          mz = rte_memzone_reserve(remain);
You meant rte_memzone_reserve(0) here, right?
> if (mz == NULL)
>             return -1;
>          remain -= mz->len;
>      }
>
Thanks Olivier, that makes sense.

Sergio
> Regards,
> Olivier
>

Re: [dpdk-dev] Regarding rte_memzone_reserve with len =0

[Olivier MATZ]

On 05/13/2015 12:14 PM, Gonzalez Monroy, Sergio wrote:
> On 12/05/2015 15:48, Olivier MATZ wrote:
>> Hi Sergio,
>>
>> On 05/06/2015 06:10 PM, Gonzalez Monroy, Sergio wrote:
>>> Hi,
>>>
>>> I was wondering about the use case of rte_memzone_reserve_xxxx APIs with
>>> len=0.
>>>
>>>  From the docs (http://dpdk.org/doc/api/rte__memzone_8h.html):
>>> len          The size of the memory to be reserved. If it is 0, the
>>> biggest contiguous zone will be reserved.
>>>
>>> What are the use cases?
>>> When would you want a memzone of undetermined size?
>>>
>>> Any thoughts appreciated.
>>
>> As the application does not have access to the lengths of memory
>> segments, probably the initial idea is when an application wants
>> to allocate more memory that the biggest segment.
>>
>> Example, the application wants to allocate 1G (even fragmented):
>>
>> - the easy case is when it can be done in one call to
>>   rte_memzone_reserve(1G)
>>
>> - else, the application can iterate like in this sample:
>>
>>      remain = 1 * 1024 * 1024 * 1024;
>>      while (remain > 0) {
>>          mz = rte_memzone_reserve(remain);
>>          if (mz != NULL)
>>             return 0;
>>          mz = rte_memzone_reserve(remain);
> You meant rte_memzone_reserve(0) here, right?

yes, sorry for the typo