Assignment 3 Solution

Submit responses to all tasks which don’t specify a le name to Canvas in a le called assign-ment3.txt, docx, pdf, rtf, odt (choose one of the formats). All source les should be submitted in the HW03 subdirectory on the master branch of your CAE git repo with no subdirectories.

All commands or code must work on Euler with only the cuda module loaded unless speci ed otherwise. Commands and/or code may behave di erently on your computer, so be sure to test on Euler before you submit.

Please submit clean code. Consider using a formatter like clang-format.

    • Before you begin, copy the provided les from HW03 of the ME759-2020 repo. Do not change any of the provided les because we will write clean copies over them when grading.

        1. Write a C++ program using CUDA in a le called task1.cu which launches a GPU kernel with 1 block and 4 threads. Inside the kernel, each thread should use std::printf to write out Hello World! I am thread x. (followed by a newline), where x is the index of the thread that prints the message. (Follow your kernel call with a call to cudaDeviceSynchronize() so that the host waits for the kernel to nish printing before returning from main.)

Compile: nvcc task1.cu -Xcompiler -O3 -Xcompiler -Wall -Xptxas -O3 -o task1

Run: ./task1

Expected output (lines could be out of order): Hello World! I am thread 0.

Hello World! I am thread 1. Hello World! I am thread 2. Hello World! I am thread 3.






































1

    2. Write a C++ program using CUDA in a  le called task2.cu which does the following:

From the host, allocates an array of 16 ints on the device called dA.

Launches a kernel with 2 blocks, each block having 8 threads. Each thread computes the sum of its thread index and its block index and writes the result into dA.

Copies back the data stored in the device array dA into a host array called hA.

Prints (from the host) the 16 values stored in the host array separated by a single space each.

Compile: nvcc task2.cu -Xcompiler -O3 -Xcompiler -Wall -Xptxas -O3 -o task2

Run: ./task2

Expected output (followed by newline): 0 1 2 3 4 5 6 7 1 2 3 4 5 6 7 8


























































2

    3. a) Implement in a le called vadd.cu, the vadd kernel function as declared and described in vadd.cuh. This function should take in two arrays, a and b, and add them together, storing the result in b. Each thread should do at most one of the additions, adding the right element from a with the corresponding element from b.

        b) Write a  le task3.cu which does the following.

Creates two arrays of length n where n is read from the rst command line argument with whatever values you like.

Calls your vadd kernel with a 1D execution con guration that uses 512 threads per block.

Prints the amount of time taken to execute the kernel in seconds using CUDA events1.

Prints the rst element of the resulting array. Prints the last element of the resulting array.

Compile: nvcc task3.cu vadd.cu -Xcompiler -O3 -Xcompiler -Wall -Xptxas -O3 -o task3

Run (where n is a positive integer): ./task3 n Example expected output (followed by newline):

0.2

35.3

1.3

        c) On an Euler compute node, run task3 for each value n = 210; 211; ; 230 and generate a plot task3.pdf which plots the time taken by your vadd as a function of n. Overlay another plot which shows the scaling results when using 1024 threads per block.









































    • Recall the document timing.md.


3