prov/efa: remove unneeded memset

[a-szegel]

The memset assigns the same length as the variable we are setting.... no chance anything bad will happen.

[wzamazon]

The reason memset can be removed is because there is no gap in struct efa_prv_reverse_av_key

[shijin-aws]

The reason memset can be removed is because there is no gap in struct efa_prv_reverse_av_key

@wzamazon what gap do you mean?

[shijin-aws]

Will it be safer to struct efa_cur_reverse_av_key cur_key = {0}; at the beginning?

Ignore this, it doesn't matter since it only has ahn and qpn and we already assign them explicitly

[wzamazon]

I meant the struct struct efa_cur_reverse_av_key consist of two int16_t so there is no gap in the struct. If it has one int32, and one int16 then there will be a gap. because compiler will do padding.

libfabric/prov/efa/src/efa_av.h

Line 66 in 52b24eb

struct efa_cur_reverse_av_key {

[shijin-aws]

I thought the memset isn't needed because efa_cur_reverse_av_key only has ahn and qpn and we already assign them explicitly

[darrylabbate]

The padding would only be relevant if the object is ever cast or punned to some other struct. Otherwise, you don't (and should rarely) need to be acutely aware of the member padding/alignment.

It should be noted that compilers are pretty good at detecting superfluous memset(&foo, 0, ...), so I wouldn't be surprised if this were optimized out anyway.