Objective(s)

• Learn how to prepare, launch and analyze the result of a kernel call in 1D
• Learn how to prepare, launch and analyze the result of a kernel call in 2D

Duration

You should complete this part within 30 minutes.

Exercise 1: Launch 1D kernels

Instructions

In this exercise your will have to prepare a device buffer and launch some kernels to change its content. Then, you will copy the device buffer to the host for inspection.

Code template

Read and edit the file 1-1d_kernel_calls.cu from the student resources archive student_resources.tar.gz.

You have to write the code for the kernel in the main() function. You are expected to insert your code in the sections demarcated with //@@. You should not have to make any change in other code sections. In particular, you do not have to write kernels yet: we already coded them and they are available in the kernels.o object file. The documentation / API of the kernels in provided in the kernel.h header.

Validation

Running make test-1-1d_kernel_calls should build and run your program. The integrated test should succeed.

Relevant documentation

• CUDA C programming guide, section 2.1 Kernels
• CUDA C programming guide, section 2.2 Thread Hierarchy
• Sample code from the CUDA C programming guide
• cudaMalloc() API documentation
• cudaMemcpy() API documentation
• cudaFree() API documentation

Exercise 2: Launch 2D kernels

Instructions

In this exercise your will have to prepare a device buffer and launch some kernels to change its content. Then, you will copy the device buffer to the host for inspection.

Code template

Read and edit the file 2-2d_kernel_calls.cu from the student resources archive student_resources.tar.gz.

You have to write the code for the kernel in the main() function. You are expected to insert your code in the sections demarcated with //@@. You should not have to make any change in other code sections. In particular, you do not have to write kernels yet: we already coded them and they are available in the kernels.o object file. The documentation / API of the kernels in provided in the kernel.h header.

Validation

Running make test-2-2d_kernel_calls should build and run your program. The integrated test should succeed.

Relevant documentation

• CUDA C programming guide, section 2.1 Kernels
• CUDA C programming guide, section 2.2 Thread Hierarchy
• Sample code from the CUDA C programming guide
• cudaMallocPitch() API documentation
• cudaMemcpy2D() API documentation
• cudaFree() API documentation