Re: [PATCH] compressdev: implement API

[Ahmed Mansour <ahmed.mansour@nxp.com>]

On 3/5/2018 9:32 AM, Verma, Shally wrote:
>
>> -----Original Message-----
>> From: Ahmed Mansour [mailto:ahmed.mansour@nxp.com]
>> Sent: 03 March 2018 01:19
>> To: Trahe, Fiona <fiona.trahe@intel.com>; Verma, Shally <Shally.Verma@cavium.com>; dev@dpdk.org
>> Cc: De Lara Guarch, Pablo <pablo.de.lara.guarch@intel.com>; Athreya, Narayana Prasad <NarayanaPrasad.Athreya@cavium.com>;
>> Gupta, Ashish <Ashish.Gupta@cavium.com>; Sahu, Sunila <Sunila.Sahu@cavium.com>; Challa, Mahipal
>> <Mahipal.Challa@cavium.com>; Jain, Deepak K <deepak.k.jain@intel.com>; Hemant Agrawal <hemant.agrawal@nxp.com>; Roy
>> Pledge <roy.pledge@nxp.com>; Youri Querry <youri.querry_1@nxp.com>
>> Subject: Re: [dpdk-dev] [PATCH] compressdev: implement API
>>
>> On 3/2/2018 4:53 AM, Trahe, Fiona wrote:
>>>> On 3/1/2018 9:41 AM, Trahe, Fiona wrote:
>>>>> Hi Shally
>>>>>
>>>>> //snip//
>>>>>> [Shally] This looks better to me. So it mean app would always call xform_init() for stateless and attach
>>>> an
>>>>>> updated priv_xform to ops (depending upon if there's shareable or not). So it does not need to have
>>>>>> NULL pointer on priv_xform. right?
>>>>>>
>>>>> [Fiona] yes. The PMD must return a valid priv_xform pointer.
>>>> [Ahmed] What I understood is that the xform_init will be called once
>>>> initially. if the @flag returned is NONE_SHAREABLE then the application
>>>> must not attach two inflight ops to the same @priv_xform? Otherwise the
>>>> application can attach many ops in flight to the @priv_xform?
>>> [Fiona Yes. App calls the xform_init() once on a device where it plans to send stateless ops.
>>> If PMD returns shareable, then it doesn't need to call again and can attach this to every stateless op going to that device.
>>> If PMD returns SINGLE_OP then it must call xform_init() before every other
>>> stateless op it wants to have inflight simultaneously. This does not mean it must be called before every op,
>>> but probably will set up a batch of priv_xforms  - it can reuse each priv_xform once the op finishes with it.
>> [Ahmed] @Shally Can this complexity of managing the NONE_SHAREABLE mode
>> be pushed into the PMD? A flexible stockpile can be kept and maintained
>> by the PMD and it can be increased or decreased based on
>> low-water/high-water thresholds
> [Shally] It is doable to manage within PMD but need to do hands on to evaluate effectiveness. So far, we have never exercised this way and left it to application to attach different session (or stream) to op for maximum performance gain. So, I would say, may it be ok to have flag feature in first place and deprecate later, if it not required?! Or just have API without any flag option and add a feature flag to indicate PMD support for SHAREABLE/NON-SHAREABLE xform_priv handle?!
[Ahmed] Either way looks ok to me. I see your point about performance.
If this is in the PMD it will have to constantly guess how much memory
the user needs and accommodate dynamically. The user can implement a
similar scheme or if the application is simple they can pre-allocate and
reduce CPU allocation de-allocation overhead.