100_Days_Of_GPU

Code Repository for the 100 Days of GPU Challenge

Day 1

Coding: Vector Addition Kernel for a Data of 1024 Elements

Reading: Read Chapter 1 of PMPP Book

Day 2

Coding:-

Encrypting and Decrypting a Character Array of lower-case letters.

Used the CUDA Events API to record the time and compute the bandwidth utilization.

Reading:

Read Chapter 2 of PMPP Book

Read about Performance Metrics in CUDA

Day 3

Coding:-

Coded the Matrix Multiplication Kernel Basic

Reading:-

Read about Matmul from this link

Solved the Exercises of Chapter 2 of PMPP book.

Day 4

Coding:-

Coded the Color to Grayscale Image Converter Kernel

Reading:-

Read Chapter 3 (Half) of the PMPP book.

Day 5

Coding: -

Coded the Image Blur Kernel

Reading:-

Read about Image Blur Kernel from PMPP book.

Day 6

Coding: -

Coded the Matrix Multiplication Kernel that calculates the output with per thread activity: -

Row-Wise

Col-Wise

Reading: -

Solved the Exercise of Chapter 3 in PMPP book.

Day 7

Coding: -

Coded the Matrix-Vector Multiplication Kernel

Reading: -

Started Reading Chapter 4 of the PMPP Book.

Day 8

Coding: -

Coded the Reduction Kernel using Atomic Operation

Reading: -

Read the Chapter 4 of the PMPP Book.

Day 9

Coding: -

Coded the Reduction Kernel that uses interleaved addressing

Reading: -

Reading about reduction kernels from this link

Day 10

Coding: -

Coded the Reduction Kernel that uses sequential addressing to reduce thread divergence.

Reading: -

Reading about reduction kernels from this link

Day 11

Coding: -

Coded the Reduction Kernel that uses sequenced addressing to reduce bank conflicts.

Reading: -

Reading about reduction kernels from this link

Day 12

Coding: -

Coded the Reduction Kernel that uses sequenced addressing to reduce bank conflicts and increases per thread activity by adding an element while loading

Reading: -

Reading about reduction kernels from this link

Day 13

Coding: -

Coded the Reduction Kernel that uses sequenced addressing to reduce bank conflicts and increases per thread activity by adding an element while loading. Also added a warp reduce optimization to optimize the execution for the last warp.

Reading: -

Reading about reduction kernels from this link

Day 14

Coding: -

Coded the CSR Representation in CPU but using the CSR to print the Degree of each Node using a CUDA Kernel.

Reading: -

Reading about CSR Format from this paper

Day 15

Coding: -

Added Edge Weights to the CSR Kernel

Added Timing for GPU/CPU Comparison

Used the Synthetic Benchmark Suite to Generate a Graph of 250k edges and 15k vertices and did performance measurements

Reading: -

Reading about CSR Format from this paper

Synthetic Benchmark Suite

Day 16

Coding: -

Wrote a GPU Kernel to Print Per-Vertex Edge Weight Mean

Used the Synthetic Benchmark Suite to Generate a Graph of 250k edges and 15k vertices and did performance measurements: -

Command: ./csr 15229 245952 .\bio-CE-CX.edges

Reading: -

Reading about CSR Format from this paper

Synthetic Benchmark Suite

Day 17

Coding: -

Wrote a GPU Kernel to Implement BFS

Reading: -

Reading about BFS algorithm in this paper

Day 18

Coding: -

Wrote a GPU Kernel to Implement Matrix Multiplication using Tiling

Day 19

Coding: -

Wrote a GPU Kernel to Invert Colors of an image

Day 20

Coding: -

Wrote a GPU Kernel to Compute the ReLU function for a set of floating point numbers.

Day 21

Coding: -

Wrote a GPU Kernel to Compute the Leaky ReLU function for a set of floating point numbers.

Performance Analysis vs CPU version

Day 22

Coding: -

Wrote a GPU Kernel to Compute the Matrix Transpose by loading through rows and storing by columns.

Day 23

Coding: -

Wrote a GPU Kernel to Compute the Matrix Transpose by loading through columns and storing by rows.