Skip to content

Latest commit

 

History

History
 
 

math-libraries-intro

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
 
 
 
 
 
 
 
 

README.md

GEMM Examples

============================

Matrix Multiplication performed using OpenBLAS and cuBLAS.

Getting Started

============================

Packages Used

  • CUDA Toolkit 11.3
  • OpenBLAS 0.2.19

Hardware Specifications

CPU: Intel(R) Xeon(R) CPU E5-2698 v3 @ 2.30GHz

GPU: Tesla V100-PCIE 32GB

Set Environment Variables

export OPENBLAS_NUM_THREADS=32

Set GPU Clocks

  • Copy the code block below into the .bashrc file
scs () {

        module load cuda/11.2.1

        DATE=$(date +"%m%d%y-%H%M%S")
        G_NAME=$(nvidia-smi -i 0 --query-gpu=gpu_name --format=csv,nounits,noheader | sed 's/ /-/g')
        G_CLK=$(nvidia-smi -i 0 --query-gpu=clocks.max.sm --format=csv,nounits,noheader)
        M_CLK=$(nvidia-smi -i 0 --query-gpu=clocks.max.memory --format=csv,nounits,noheader)
        P_LIMIT=$(nvidia-smi -i 0 --query-gpu=power.max_limit --format=csv,nounits,noheader)
        DRIVER=$(nvidia-smi -i 0 --query-gpu=driver_version --format=csv,nounits,noheader)

        sudo nvidia-smi
        sudo nvidia-smi -pm ENABLED
        sudo nvidia-smi --auto-boost-default=0
        sudo nvidia-smi -ac ${M_CLK},${G_CLK}
        sudo nvidia-smi -lgc ${G_CLK},${G_CLK}
        sudo nvidia-smi -pl ${P_LIMIT}
        sudo nvidia-smi -q -d POWER,CLOCK
}
  • Update with source ~/.bashrc
  • Run scs to execute

Running Examples

============================

Build

git clone the repo

cd code-samples/posts/math-libraries-intro

make

OpenBLAS

Run

./openblas-example

Sample Output

=============================


  This example computes real matrix C=alpha*A*B+beta*C using
 OpenBLAS dgemm, where A, B, and  C are matrices and
 alpha and beta are double precision scalars

 Initializing data for matrix multiplication C=A*B for matrix
 A(4092x4092) and matrix B(4092x4092)

 Allocating memory for matrices aligned on 64-byte boundary for better
 performance

 Time Elapsed: 414.35 ms

 Example completed.

cuBLAS

Run

./cublas-example

Sample Output

=============================


 This example computes real matrix C=alpha*A*B+beta*C using
 cuBLAS dgemm, where A, B, and  C are matrices and
 alpha and beta are double precision scalars

 Initializing data for matrix multiplication C=A*B for matrix
 A(4092x4092) and matrix B(4092x4092)

 Allocating memory for matrices aligned on 64-byte boundary for better
 performance

 Computing matrix product using cuBLAS dgemm function

 Computations completed.

 Time Elapsed: 19.80 ms

 Deallocating memory

 Example completed.