HIP/ROCm Kernel Implementations

Contents

Chapter 3

• BlurImage: Implements image blurring using HIP.
• greyScaleConverter: Converts images to grayscale using HIP.
• matmul: Matrix multiplication using HIP.

Chapter 4

• gpu_features.cpp: Queries and displays GPU features.

Chapter 5

• gpu_features.cpp: Queries and displays GPU features.
• matmul_tiled_gen.cpp: General matrix multiplication using tiling.
• matmul_tiled_sq.cpp: Square matrix multiplication using tiling.

Chapter 7

• conv_1.cpp: 2D convolution using HIP.

Chapter 8

• stencil_kernel.hpp: Header file for stencil kernels.
• kernel_shared.cpp: Shared memory stencil kernel.
• kernel_tc.cpp: Thread coarsening stencil kernel.

Chapter 9

• kernel.hpp: Header file for histogram kernels.
• kernel_basic.cpp: Basic histogram kernel.
• kernel_priv_shared.cpp: Histogram kernel with privatization using shared memory.
• kernel_private.cpp: Histogram kernel with privatization using RAM.
• kernel_tc_cont.cpp: Histogram kernel with coarsening using contiguous partitioning.
• kernel_tc_interleaved.cpp: Histogram kernel with coarsening using interleaved partitioning.
• kernel_aggregation.cpp: Placeholder for histogram kernel with aggregation.
• main.cpp: Main file to test histogram kernels.

Chapter 10

• kernel.hpp: Header file for sum reduction kernels.
• kernel_arbitary_len.cpp: Sum reduction kernel for arbitrary length arrays.
• kernel_convergent.cpp: Sum reduction kernel with less control divergence.
• kernel_shared_mem.cpp: Sum reduction kernel using shared memory.
• kernel_simple.cpp: Simple sum reduction kernel.
• main.cpp: Main file to test sum reduction kernels.

Library

• PngHandlib: Library for handling PNG images.

Prerequisites

• HIP/ROCm installed on your system.
• A compatible AMD GPU. (In the book, CUDA is used instead of HIP, so check the book if you have an NVIDIA GPU.)