Initializing structs with memset

[andrey]

Why is it a standard in nordic-supplied code to initialize structs with memset like so? An example would be:

ble_gatts_char_md_t char_md;
memset(&char_md, 0, sizeof(char_md));

Instead of a cleaner way:

ble_gatts_char_md_t char_md = {0};

I understand that using memset also clears padding bits, but don't understand the reason for doing so.

[RK]

Well there's style, and space.

ble_gatts_char_md_t char_md = {};

forces the compiler to emit code to explicitly set each element of the structure to zero when it's created, for large structures that's a fair amount of instructions. memset, which is usually linked in anyway, is just a function call, less code space.

[Charles Manning]

I don't think that is quite true.

The {0} version will create a binary version of the structure and the = will cause a copy.

xx = {0};

is eqivalent to

y = {0};memcpy(&x,&y, sizeof(y));

;

But the result is the same: The code is larger (and slower) since making a large structure needs to make the huge structure then also do a copy. A copy requires reading then writing each word in the structure. That is slower than just setting it to zero.

[Sensors]

Just did a little experiment with this. There was no change in code size between the two approaches using Keil uVision (armcc).