[dpdk-dev] [PATCH v3] eal: refactor rte_eal_init into sub-functions

[Bruce Richardson]

On Fri, Jan 19, 2024 at 08:58:17AM -0800, Rahul Gupta wrote:
> From: Rahul Gupta <rahulgupt at microsoft.com>
> 
> In continuation to the following email, I am sending this patch.
> (https://inbox.dpdk.org/dev/20231110172523.GA17466@microsoft.com/)
> 
> Initialization requires rte_eal_init + rte_pktmbuf_pool_create which
> can consume a total time of 500-600 ms:
> a) For many devices FLR may take a significant chunk of time
>    (200-250 ms in our use-case), this FLR is triggered during device
>    probe in rte_eal_init().
> b) rte_pktmbuf_pool_create() can consume up to 300-350 ms for
> applications that require huge memory.
> 
> This cost is incurred on each restart (which happens in our use-case
> during binary updates for servicing).
> This patch provides an optimization using pthreads that applications
> can use and which can save 200-230ms.
> 
> In this patch, rte_eal_init() is refactored into two parts-
> a) 1st part is dependent code ie- it’s a perquisite of the FLR and
>    mempool creation. So this code needs to be executed before any
>    pthreads. Its named as rte_eal_init_setup()
> b) 2nd part of code is independent code ie- it can execute in parallel
>    to mempool creation in a pthread. Its named as rte_eal_init_async().
> 
> Existing applications requires to just call-
> rte_eal_init_wait_async_complete() after rte_eal_init() unless they wish
> to leverage the optimization.
> 
This part here is a bit problematic. For something as ubiquitous as
rte_eal_init(), we should not be requiring existing application change.
Instead, can you create a new init function to replace rte_eal_init() in
the async case - and then eal_init can call that and then the
async_complete function, to avoid duplicating code.

The goal should be to have existing apps unaffected, but to allow a path to
faster startup for those that need it. In the original email you linked to
above, we had such a proposal: rte_eal_init_async() &
rte_eal_init_async_done().

Thanks,
/Bruce