Re: [dpdk-dev] [PATCH] test/ring: code rework to reduce compilation time

[Thomas Monjalon <thomas@monjalon.net>]

30/04/2020 16:43, Ananyev, Konstantin:
> 
> Hi Honnappa,
>  
> > Hi Konstantin,
> > 	I like the way the tests are organized and it looks good.
> > 
> > I am just wondering about the way it is being tested here. The intent to write the test cases the way they are currently is to mimic how the
> > APIs would be used mostly. IMO, the APIs would be used with a constant value for element size so that the compiler will throw away the
> > unwanted code (in the functions where the actual copy is being done).
> > 
> > With your method here, it looks to me like all the branches in the copy functions are kept and the branch decisions are done at run time.
> > Is  my understanding correct?
> 
> You mean branching on esize[] values?
> Actually from what I've seen that happens for both cases:
> before and after the patch (gcc 7.3 -O3).
> 
> Main intention in my changes was to avoid using test_ring_enqueue/test_ring_dequeue,
> as it seems too many branches here and it takes compiler a lot of effort to resolve all
> of them at compile time.
> So I replaced it with  array of function pointers (test_enqdeq_impl[]) and iterating over it.
> That way compiler knows straightway which function to use.

In case we choose this solution, please make a v2 including such explanations.