Improved assignment description

Bernhard Kerbl · Bernhard Kerbl · commit 53b4539a0593 · 2021-11-24T21:13:58.000+01:00
diff --git a/17_CooperativeGroups/src/main.cu b/17_CooperativeGroups/src/main.cu
@@ -159,17 +159,20 @@ int main()
 /*
 Exercises:
 1) Write a kernel where each thread first computes its ID in a register.
-Within each group of 4 consecutive threads, threads should then share their 
+Within each group of 4 consecutive threads, threads should then share their
 ID with all others, using shuffling. Write this kernel once with, once without
 cooperative groups, and confirm correctness via output.
-2) Launch a COOPERATIVE KERNEL and use grid-wide synchronization to make sure 
+2) Launch a COOPERATIVE KERNEL and use grid-wide synchronization to make sure
 all threads in the entire grid are at the same point in the program. Can you
-think of any use cases for this?
-3) Write a simple program with the following tasks A, B, C, each with N threads. 
-In A, each thread t should compute and store t*t in its output A_out[t]. In B, 
-each thread t should compute A_out[N - t - 1] - t and store it in its output 
-B_out[t]. In C, each thread t should compute B_out[N - t - 1] + 4 and store it 
-in its output C_out[t]. Implement this once using one kernel for each task A, 
+think of any use cases for this? Your device will need to support the attribute
+cudaDevAttrCooperativeLaunch for this, check if it has it before starting.
+3) Write a simple program with the following tasks A, B, C, each with N threads.
+In A, each thread t should compute and store t*t in its output A_out[t]. In B,
+each thread t should compute A_out[N - t - 1] - t and store it in its output
+B_out[t]. In C, each thread t should compute B_out[N - t - 1] + 4 and store it
+in its output C_out[t]. Implement this once using one kernel for each task A,
 and once with a single kernel that uses grid synchronization between tasks.
 In the single kernel, do you need additional threadfences and/or volatiles?
+Again, in order to do grid sync, your device will need to support the
+cudaDevAttrCooperativeLaunch attribute, check if it has it before starting.
 */
