wip
This commit is contained in:
@@ -201,10 +201,12 @@ __global__ void compute(Chunk *data, int l, int r) {
|
||||
// printf("Compressing : %d\n", l);
|
||||
}
|
||||
else {
|
||||
// Launch child kernels without synchronization (host will handle sync)
|
||||
compute<<<n/2,16>>>(data, l, l+n/2);
|
||||
cudaDeviceSynchronize();
|
||||
compute<<<n/2,16>>>(data, l+n/2, r);
|
||||
cudaDeviceSynchronize();
|
||||
|
||||
// Wait for all threads in this block to finish
|
||||
__syncthreads();
|
||||
|
||||
data[l].flags |= PARENT;
|
||||
|
||||
|
||||
Reference in New Issue
Block a user