memset vs SecureZeroMemory

[Adam Roderick J]

Hi All, i have a DLL, there is some code as mentioned below:-

char szName[MAX_PATH];
CString csName = "3424234"; // size can be judged only on run time
wcstombs( szName, csName.operator LPCTSTR(), _tcslen( csName.operator LPCTSTR()));

Now i want to initialize szName array to 0, to aviod the unexpected behaviour. Now my question is whether i use memset or SecureZeroMemory? Which is better, and why? Any help will be appreciated.

Величие не Бога может быть недооценена.

modified on Tuesday, February 9, 2010 12:39 AM

[KingsGambit]

char szName[MAX_PATH] = {0};

[_Superman_]

SecureZeroMemory may be a little slow on the x86 platform as it initializes each element of the array in a loop. But for x64 it uses an intrinsic, which will be very fast. You could also use the C-Runtime memcpy_s function. To initialize the array to zero at the time of creation, the reply by Rejeesh definitely works in Visual Studio.

«_Superman_» I love work. It gives me something to do between weekends.
Microsoft MVP (Visual C++)

[Adam Roderick J]

Thanks for the reply. So you mean in 64 bit processor, SecureZeroMemory will not make much change in time consumption.

Величие не Бога может быть недооценена.

[Adam Roderick J]

But i am looking for details regarding SecureZeroMemory and memset or memcpy, any way thanks for your help :)

Величие не Бога может быть недооценена.

[KingsGambit]

I think _superman_ has already sent you a better answer which I think explains the differences. I usually initialize all arrays to zero as I posted before.

[_Superman_]

That's correct. Intrinsics are similar to writing inline assembly instructions. x64 does not support inline assembly and so you have to use intrinsics. Read more about intrinsics in the following links - Intrinsic function[^] Compiler Intrinsics[^]

«_Superman_» I love work. It gives me something to do between weekends.
Microsoft MVP (Visual C++)

[Rajesh R Subramanian]

One thing others didn't point out: memset is a CRT call, and SecureZeroMemory is an API call. That's it. I don't think one would be "faster" over the other (practically) to initialize an array of size MAX_PATH. If you had considered ZeroMemory, there are chances where the compiler can completely optimize out (remove) the call to this function when it detects that the initialised buffer isn't being used anywhere. However, SecureZeroMemory would ensure that the buffer be initialised, even if it isn't being used at all (the optimiser won't rule this call out).

“Follow your bliss.” – Joseph Campbell

[Lost User]

Adam Roderick J 09 wrote:

Now i want to initialize szName array to 0, to aviod the unexpected behaviour.

1. You should be using the wcstombs_s() function. 2. You should use that function as documented here[^] to calculate the required size of your destination buffer, and allocating that appropriately. In your example above, if the length of your string exceeds MAX_PATH then you have an immediate memory leak. 3. Clearing the buffer to zeroes before copying the string provides no benefit.

MVP 2010 - are they mad?

[Chris Losinger]

(i think you've got your x86 and x64 reversed here)

image processing toolkits | batch image processing