GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 | Boston University - Edubirdie

University:
Boston University
Course:
MET CS 201 | Introduction to Programming
Academic year:

2024
Views:

27

Pages:

17
Author:

Cali Bowman

Shared Memory - Making use of it ‣ Idea: We could load only once to shared memory, and operate there __global__ void update (float *u, float *u_prev, int N, float dx, float dt, float c) { // Each thread will load one element int i = threadIdx.x; int I = threadIdx.x + BLOCKSIZE * blockIdx.x; __shared__ float u_shared[BLOCKSIZE]; if (I > =N){return;} u_shared[i] = u[I]; __syncthreads(); if (I > 0) { u[I] = u_shared[i] ‐ c*dt/dx*(u_shared[i] ‐ u_shared[i‐1]); } } Shared Memory - Making use of it ‣ Idea: We could load only once to shared memory, and operate there __global__ void update (float *u, float *u_prev, int N, float dx, float dt, float c) { // Each thread will load one element int i = threadIdx.x; int I = threadIdx.x + BLOCKSIZE * blockIdx.x; __shared__ float u_shared[BLOCKSIZE]; Allocate shared array if (I > =N){return;} u_shared[i] = u[I]; __syncthreads(); if (I > 0) { u[I] = u_shared[i] ‐ c*dt/dx*(u_shared[i] ‐ u_shared[i‐1]); } } Shared Memory - Making use of it ‣ Idea: We could load only once to shared memory, and operate there __global__ void update (float *u, float *u_prev, int N, float dx, float dt, float c) { // Each thread will load one element int i = threadIdx.x; int I = threadIdx.x + BLOCKSIZE * blockIdx.x; __shared__ float u_shared[BLOCKSIZE]; Allocate shared array if (I > =N){return;} u_shared[i] = u[I]; Load to shared mem __syncthreads(); if (I > 0) { u[I] = u_shared[i] ‐ c*dt/dx*(u_shared[i] ‐ u_shared[i‐1]); } } Shared Memory - Making use of it ‣ Idea: We could load only once to shared memory, and operate there __global__ void update (float *u, float *u_prev, int N, float dx, float dt, float c) { // Each thread will load one element int i = threadIdx.x; int I = threadIdx.x + BLOCKSIZE * blockIdx.x; __shared__ float u_shared[BLOCKSIZE];

Related Documents

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 1

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 2

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 3

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 4

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 5

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 6

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 7

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 8

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 9

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 10

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 11

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 12

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 13

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 14

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 15

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 16

GPU Computing With CUDA - Efficient Shared Memory Use Pt 3 - Page 17

of 17

Recommended Documents

Introduction to MPI - Mixing Blocking

Introduction to MPI - Mixing Blocking

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Tuning MATLAB For Better Perfomance - Code Optimization

Tuning MATLAB For Better Perfomance - Code Optimization

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Tuning MATLAB For Better Perfomance SOR Update Function

Tuning MATLAB For Better Perfomance SOR Update Function

Boston University MET CS 201 | Introduction to Programming

Lecture Note

MPI Beginning Lecture - Introduction Pt 1

MPI Beginning Lecture - Introduction Pt 1

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Lecture About Artificial Intelligence

Lecture About Artificial Intelligence

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Security Camp Cloud Security

Security Camp Cloud Security

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Tuning MATLAB For Better Perfomance

Tuning MATLAB For Better Perfomance

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Introduction and Computational Thinking Pt 1

Introduction and Computational Thinking Pt 1

Boston University MET CS 201 | Introduction to Programming

Lecture Note

LaTeX Command Summary Vocabulary

LaTeX Command Summary Vocabulary

Boston University MET CS 201 | Introduction to Programming

Cheat Sheet

Introduction to MPI Lecture Note

Introduction to MPI Lecture Note

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Introduction To C++ Part 1.3

Introduction To C++ Part 1.3

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Introduction to Databases Lecture Note

Introduction to Databases Lecture Note

Boston University MET CS 201 | Introduction to Programming

Lecture Note

New Documents from this Course

WEB Design

Boston University MET CS 201 | Introduction to Programming

Research

To Change the PMG Problem Conﬁguration

To Change the PMG Problem Conﬁguration

Boston University MET CS 201 | Introduction to Programming

Research

Introduction to Matlab Parallel Computing Tooblox Pt 3

Introduction to Matlab Parallel Computing Tooblox Pt 3

Boston University MET CS 201 | Introduction to Programming

Case

A Scalable Framework for Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters with Weight and Workload Balancing

A Scalable Framework for Acceleration of CNN Training on Deeply-Pipelined FPGA Clusters with Weight and Workload Balancing

Boston University MET CS 201 | Introduction to Programming

Research

Using Crash Hoare Logic for Certifying the FSCQ File System Ch 4

Using Crash Hoare Logic for Certifying the FSCQ File System Ch 4

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Using Crash Hoare Logic for Certifying the FSCQ File System Ch 3

Using Crash Hoare Logic for Certifying the FSCQ File System Ch 3

Boston University MET CS 201 | Introduction to Programming

Lecture Note

Introduction to MATLAB Lecture 4 Pt 2

Introduction to MATLAB Lecture 4 Pt 2

Boston University MET CS 201 | Introduction to Programming

Lab

Processing of 2D RGB and Depth Data into 3D Space

Processing of 2D RGB and Depth Data into 3D Space

Boston University MET CS 201 | Introduction to Programming

Research

Introduction to MATLAB Lecture 4 Pt 3

Introduction to MATLAB Lecture 4 Pt 3

Boston University MET CS 201 | Introduction to Programming

Lab