CODEWITHSUNDEEP
c++cudagpu-shared-memory
Isaac Kleinman
Isaac Kleinman

Reputation: 4452

CUDA: Shared Memory Assignment

Suppose I define a shared variable in a cuda kernel as follows:

__shared__ int var;

Now, let's say at some point in my kernel I'd like to assign some value, say 100 to var. Saying

var = 100;

results in all threads in the block executing this assignment.

How can I have the assignment take place only once? Is the following my only option?

if( threadIdx.x == 0)
    var = 100;

Upvotes: 1

Views: 443

Answers (1)

talonmies
talonmies

Reputation: 72352

Your only option is actually this:

if( threadIdx.x == 0)
    var = 100;

__syncthreads();

If you omit a synchronisation barrier, there is no guarantee that all threads in the block will read the value of var after the assignment statement is executed.

Upvotes: 10

Related Questions