Matrix Multiply

TAU has support for the OpenACC directives available in PGI 12.3 and greater. Configure TAU:

./configure -c++=pgCC -cc=pgcc -fortran=pgi -cuda=<path to CUDA 4.1 or greater>
make install

export TAU_MAKEFILE=<path to TAU>/x86_64/lib/Makefile.tau-cupti-pgi
export TAU_OPTIONS='-optVerbose -optShared'

Compile

tau_f90.sh -ta=nvidia,time -Minfo matmult.f90 -o matmult

Run:

tau_exec -T serial,cupti -cupti ./matmult
module

Use TAU analysis tool to view performance data:

pprof
paraprof

Openacc