The objective of this project is the parallelization of the Cholesky's algorithm using the OpenMP and Pthread tools.
RA: 147512 Name: Nathália Menini Cardoso dos Santos Email: nathmenini@gmail.com
RA: 192744 Name: Miguel Antonio Rodriguez Santander Email: m.rodriguezs1990@gmail.com
A Makefile has been created with the following commands:
make: This command creates a folder calledbinariesin which compiles the three codes created for the benchmark.make clean: This commnad delete all the content of thebinariesfolder.make run: This command runs the benchmark with the input files in theinputfolder, which will be run with different number of threads (16, 24 and 32).
Comand:
makeThe files will be compiled are:
cholesky_serial.c: original version of Cholesky's descomposition.cholesky_serial_for_parallel.c: serial version, which is parallelizable (slower than the original).cholesky_openmp.c: parallelization using theOpenMPtools.cholesky_pthreads.c: parallelization usingPthreadstool, in which the number of calculations is divided for each thread.cholesky_pthreads_prodcon.c: implementation of producer-consumer architecture withPthreads.
The inputs of the codes will be by standard input, which will consist of two lines, the first will be the square matrix size n, and the second will be n*n numbers representing the matrix.
The number of threads will be give by argument input.
Example:
The execution of serial code:
./binaries/cholesky_serial.bin < ./input/matrix_5000x5000.inThe execution of parallel code:
./binaries/cholesky_pthreads.bin 32 < ./input/matrix_5000x5000.inCommand:
make cleanThis command delete all compiled codes.
Comand:
make runThis commnad execute the file run.sh, which is configured for execute the different implementations with each input files and with different number of threads. For the benchmark the number of threads will be 16, 24 and 32.
The speedup is calculated as: parallel time / serial time
./input/matrix_5000x5000.in
--------------------------
Number of threads: 16
OpenMP SPEEDUP: 2.0587971
Pthread SPEEDUP: 1.8232368
Pthread Producer consumer SPEEDUP: 0.0179181
Number of threads: 24
OpenMP SPEEDUP: 1.6803953
Pthread SPEEDUP: 1.4147363
Pthread Producer consumer SPEEDUP: 0.0114220
Number of threads: 32
OpenMP SPEEDUP: 1.4583506
Pthread SPEEDUP: 1.0219771
Pthread Producer consumer SPEEDUP: 0.0114220
./input/matrix_6000x6000.in
--------------------------
Number of threads: 16
OpenMP SPEEDUP: 2.0587971
Pthread SPEEDUP: 1.8232368
Pthread Producer consumer SPEEDUP: 0.0179181
Number of threads: 24
OpenMP SPEEDUP: 1.6803953
Pthread SPEEDUP: 1.4147363
Pthread Producer consumer SPEEDUP: 0.0114220
Number of threads: 32
OpenMP SPEEDUP: 1.4583506
Pthread SPEEDUP: 1.0219771
Pthread Producer consumer SPEEDUP: 0.0114220
./input/matrix_7000x7000.in
--------------------------
Number of threads: 16
OpenMP SPEEDUP: 2.0587971
Pthread SPEEDUP: 1.8232368
Pthread Producer consumer SPEEDUP: 0.0179181
Number of threads: 24
OpenMP SPEEDUP: 1.6803953
Pthread SPEEDUP: 1.4147363
Pthread Producer consumer SPEEDUP: 0.0114220
Number of threads: 32
OpenMP SPEEDUP: 1.4583506
Pthread SPEEDUP: 1.0219771
Pthread Producer consumer SPEEDUP: 0.0114220