Cuda Emulator

Content

CSE 591 591:: GPU Pro Progr grammi ammi ng Setting Se tting Up GPU GPU Programming Environment

• Before Start: CPU Emulator • Installation • Compilation Options • Browsing SDK examples

Ziyi Zi yi Zh Zhen eng g

Computer Science Department Stony Brook University

Content

Start: Emul ator • Before Start: • Installation • Compilation options

• Browsing SDK examples

Before Start Start : Emulation

For those who want to use CUDA but do not have CUDAenabled GPU CPU Emulator

•

nvcc …. -deviceem -deviceemu u -D_DEVICEE -D_DEVICEEMU MU …

debugging to help code code development • Aiming at debugging Parallel Nsight ( requiring a CUDA-enabled CUDA-enabled GPU too) • Replaced by Parallel

Nvidia starts to remove remove CPU emulator emulator support on CUDA CUDA 3.0, March 2010 • Latest CUDA version is CUDA 3.2 September 2010 • Need to install CUDA 2.3, June 2009 (toolkit and SDK)  Older NVCC, older APIs  Can use emulation version of CUFFT, CUBLAS  No CUSPARSE, CURNG

CUDA for CPUs

CUDA for X86 platform

CUDA C++ compiler in research

Commercial CUDA C compiler

MCUDA

• Under development by Portland Group (PGI) • No GPU required • Will be demonstrated at the SC10 Supercomputing

• • • •

Developed by Wen-mei Hwu’s group Aiming at comparing GPUs and optimized CPUs performance CUDA code  optimized C++ code for multi-core CPUs Linux based

conference in November 13-15, 2010.

• We are ahead of time

Download http://impact.crhc.illinois.edu/mcuda.php

papers http://www.ifp.illinois.edu/~minhdo/publications/parallelvideo.pdf

Not required in the course. Use it only when you want to fairly compare the performance between CPU and GPU

Content

• Before Start: Emulator • Installation • Compilation Options • Browsing SDK examples

Witho ut CUDA-Enabled GPU

Step

1. http://developer.nvidia.com/object/cuda_2_3_downloads.html

2. Install CUDA Toolkit 2.3 3. Install CUDA SDK code examples 2.3

With CUDA-Enable GPU

STEP

1. 2. 3. 4.

http://developer.nvidia.com/object/cuda_3_2_toolkit_rc.htm Download appropriate GPU driver Install CUDA Toolkit 3.2 Install GPU Computing SDK code examples 3.2

Av ail abl e Resources

• • • • •

NVCC Visual Studio syntax highlighting CUDA BLAS (CUBLAS) and FFT (CUFFT) libraries CUDA Visual Profiler CUDA-GDB for Linux

Not in CUDA 2.3 but included in later version

• • • • •

OPENCL DirectCompute CUDA Fortran compiler CUDA LAPACK library CUDA CUSPARSE and CURNG libraries

ATI/AMD Card + OPENCL

ATI/AMD Card + CUDA

STEP

Convert CUDA code into OPENCL code

1. http://developer.amd.com/gpu/atistreamsdk/pages/default.

then build OPENCL code and executed on ATI/AMD card

aspx

2. Download ATI Stream SDK 2.2

Additional STEP

1. http://developer.amd.com/gpu/atistreamsdk/pages/default.aspx 2. Download ATI Stream SDK 2.2 3. http://www.multiscalelab.org/swan 4. Download Swan (27 May 2010)

Content


CUDA Programming Environment Windows, Linux, Mac OS

Associated Environment Variables Automatic set by Toolkit CUDA_BIN_PATH CUDA_INC_PATH CUDA_LIB_PATH

Microsoft Visual Studio 2008

Need MS Visual Studio?

• Go this website: https://www.dreamspark.com to get Visual Studio 2008 online.

• Must validate via student ID.

Integrate NVCC with Visual Studio

MS Visual Studio 2008

• Either using CUDA build rules (installed by CUDA SDK) GUI 

interface

Generate compilation commands (options, parameters)

• Or write custom build rules • Served as an IDE (integrated development environment) • Served as an C/C++ compiler and linker for the host program.



Command line interface



Directly Writing compilation commands



Such as :

"C:\CUDA\bin\nvcc.exe" -ccbin "$(VCInstallDir)bin" -c -DWIN32 -D_CONSOLE D_MBCS -Xcompiler /EHsc,/W3,/nologo,/Wp64,/O2,/Zi,/MT -I"C:\CUDA\include" -I./ -I../../common/inc -o $(ConfigurationName)\$(InputName).obj $(InputFileName)

CUDA Build Rules 2.3

CUDA Build Rules 2.3

1. Right click a projection 2. Choose “Custom Build Rules …” 3. Choose a CUDA rule 2.3 if available in your system which will be available after you installing the CUDA SDK 2.3 4. Right click a .cu file 5. Choose “Property” 6. Click CUDA rule 2.3

Setting Building Option by Command

CUDA Proj ect

1. Click General 2. For “Tool” : choose “Custom Build Tool”

• Create one from scratch?

3. Then Choose “Custom Build Step” 4. Enter your building command

• Modify existing projects in SDK • CUDA visual studio wizard http://sourceforge.net/projects/cudavswizard/ Third party, independent updates, no document support

CPU Emulatio n Mode fo r CUDA 2.3

For projections in CUDA SDK 2.3 In visual studio configuration: Chose EmuRelease or EmuDebug Instead of Release or debug

Content


For your own projections with CUDA 2.3

1. Add a building configuration 2. change build rules settings, (or simply adding -deviceemu -D_DEVICEEMU into complication command line)

Bandwidth Test

Memory transfer on

Device Query

Graphics Hardware Capability

CPU  GPU GPU  GPU

On a 8600m GT card

GPU  CPU

Capability 1.1

On a 8600m GT card

# Multi-Processor

8

# cores

32

Block limit per dimension

512 x 512 x 64

CPU  GPU

1236 MB/s

Maximum # thread per block

512

GPU  GPU

11836 MB/s

Grid limit per dimension

65535 x 65535 x 1

GPU  CPU

380 MB/s

Matrix Multiplication

Template

8600m GT v.s. Core2 Duo 2.4 GHz

8600m GT v.s. Core2 Duo 2.4 GHz

GPU :

0.62 ms

GPU :

179 ms

CPU in emulation mode :

around 850ms

CPU in emulation mode :

66 ms

1000x faster

3 times slower? Multiply 32 numbers by another 32 numbers

Matrix A: 80x48

unsigned int num_threads = 32;

Matrix B: 48x128

dim3 grid( 1, 1, 1);

Matrix C: 80x128

Computational intensive 

GPU is better than CPU

dim3 threads( num_threads, 1, 1);

32 multiplications Not computational intensive at all CPU is better than GPU

Cuda Emulator

Recommend Documents