Gem5-GPU - PowerPoint Presentation

Slide1

Gem5-GPU Installation

CS5100 Advanced

Computer ArchitectureSlide2

Introduction of

Gem5-GPU

It

merges

2 popular simulators: gem5 and

gpgpu

-sim

Simulates

CPUs, GPUs, and the interactions between

them

Models

a flexible memory system with support for heterogeneous processors and

coherence

Supports

full-system simulation through GPU driver emulationSlide3

Set up gem5-gpu environment

Platform:

Ubuntu 14.04LTS

64bit

R

ecommended

RAM size:

4G

Recommended Hard Disk size: 80G

Download path: https

://mega.co.nz/#!

wVdgFJQL!EPwc4Q9tTmrJtiBrWbtSSvv74ah8ljIB0iX1EvyMAPE

gcc

4.7

& g++

4.7 installation

sudo

add-apt-repository

ppa:ubuntu-toolchain-r

/test

sudo

apt-get

update

sudo

apt-get install

gcc-4.7

sudo

apt-get install g++-

4.7

sudo

update-alternatives --install /usr/bin/gcc gcc /

usr/bin/gcc-4.7

3

0

sudo

update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-

4.7

3

0

sudo

update-alternatives --

config

gcc

sudo

update-alternatives --

config

g

++Slide4

Set up gem5-gpu environment

Install essential packages

sudo apt-get install zlib1g-dev bison mercurial scons libprotobuf-dev python-dev swig protobuf-compiler

Install CUDA toolkit 3.2

wget http://developer.download.nvidia.com/compute/cuda/3_2_prod/toolkit/cudatoolkit_3.2.16_linux_64_ubuntu10.04.run

chmod +x ./cudatoolkit_3.2.16_linux_64_ubuntu10.04.run

sudo ./cudatoolkit_3.2.16_linux_64_ubuntu10.04.run

export CUDAHOME=/usr/local/cuda

export PATH=$PATH:/usr/local/cuda/bin

export LD_LIBRARY_PATH=/usr/local/cuda/lib64Slide5

Set up gem5-gpu environment

edit .

hgrc

vim .

hgrc

type the following in the document

[extensions]

mq

=

[

ui

]

username = <your username> ←fill in your own usernameSlide6

Set up gem5-gpu environment

Clone gem5 and gem5-patches

mkdir

gem5-gpu

cd gem5-gpu

hg

qclone

http://repo.gem5.org/gem5 -p http://gem5-gpu.cs.wisc.edu/repo/gem5-patches

cd gem5/

hg update -r 11060

hg

qpush

-a

cd ..Slide7

Set up gem5-gpu environment

Clone GPGPU-Sim and GPGPU-Sim patches

hg

qclone

http://gem5-gpu.cs.wisc.edu/repo/gpgpu-sim -p http://gem5-gpu.cs.wisc.edu/repo/gpgpu-sim-patches

cd

gpgpu

-sim

hg

qpush

-a

cd

..

Clone gem5-gpu glue code

hg clone http://

gem5-gpu.cs.wisc.edu/repo/gem5-gpu

Build

cd gem5

scons

build/X86_VI_hammer_GPU/gem5.opt --default=X86 EXTRAS=../gem5-gpu/

src

:../

gpgpu

-sim/ PROTOCOL=

VI_hammer

GPGPU_SIM=TrueSlide8

Set up gem5-gpu environment

Obtain CUDA SDK

wget

http://developer.download.nvidia.com/compute/cuda/3_2_prod/sdk/gpucomputingsdk_3.2.16_linux.run

chmod

+x ./gpucomputingsdk_3.2.16_linux.run

sudo

./gpucomputingsdk_3.2.16_linux.run

export NVIDIA_CUDA_SDK_LOCATION=/home/<username>/

NVIDIA_GPU_Computing_SDK

/C

Obtain Benchmarks

~/gem5-gpu $

hg clone https://

gem5-gpu.cs.wisc.edu/repo/benchmarks/

Do

this

under

your

root

directory

(gem5-gpu/

in

this

example)Slide9

Set up gem5-gpu environment

Compile

~/gem5-gpu$ cd

benchmarcks

/

libcuda

/

m

ake

If

you

encounter

make

error,

please

export

CUDA

path

in

page

4

again

cd /home/<username>/

NVIDIA_GPU_Computing_SDK

/C/common

makeSlide10

Example of Compiling benchmark

U

se

gcc

4.4 & g++ 4.4 to compile benchmarks

sudo

apt-get install gcc-4.4

sudo

apt-get install g++-4.4

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.4 40

sudo update-alternatives --install /usr/bin/g++ g++ /usr/bin/g++-4.4 40

sudo

update-alternatives --

config

gcc

sudo

update-alternatives --

config

g

++

Example

of Compiling a Benchmark

~/

gem5-gpu/benchmarks

$ cd

rodinia

/

backprop

make

gem5-fusionSlide11

Example of running benchmark

Example of Running a Benchmark

cd gem5/

build/X86_VI_hammer_GPU/gem5.opt ../gem5-gpu/

configs

/

se_fusion.py

-c ../benchmarks/

rodinia

/

backprop

/gem5_fusion_backprop -o "16“

Or type the commend under if you have

problem

/home/pads/gem5-gpu/gem5/build/X86_VI_hammer_GPU/gem5.opt /home/pads/gem5-gpu/gem5-gpu/

configs

/

se_fusion.py

-c /home/pads/gem5-gpu/benchmarks/

rodinia

/

backprop

/gem5_fusion_backprop -o "

16”

Run

other

benchmarks

16

is

the

parameter

that

backprop

benchmark

needed,

different

benchmark

will

need

different

type

of

input

Find

the command argument at README or run file in the benchmark

folderSlide12

Example of running benchmark with

input

data

Download

input data

https://drive.google.com/file/d/0B-xa6WK_zXMHaXc3SWtDNkd3bUE/view?usp=sharing

Compile

image

compression

benchmark

dwt

~/gem5-gpu/benchmarks$ cd

rodinia

/dwt

make

gem5-fusion

Run

dwt

build/X86_VI_hammer_GPU/gem5.opt

../gem5-gpu/configs/se_fusion.py -c ../

benchmarks/

rodinia

/dwt/gem5_fusion_dwt

-o

“

--

dimension 192x192

<

your input data folder>

/dwt2d/192.bmp

<your

input

data folder

>

/data/dwt2d/out”

Parameters

in

command

correspond

to

“--dimension

<

size

>

<

Input_path

]>

<

Output_path

>

”Slide13

Possible error

When

you

run

benchmark

with

input

data

you

may

encounter

following

error

Please

go to gem5/build/X86_VI_hammer_GPU/mem/packet_queue.cc and modify limit of transmitList.size() in line 117, then rebuild gem5 Slide14

Reference

Getting

Started

https://

gem5-gpu.cs.wisc.edu/wiki/start

Build Benchmarks

https://

gem5-gpu.cs.wisc.edu/wiki/benchmarks

Simulator

version

information

http

://

goo.gl/S1ts6