vmitris / asc-cuda Goto Github PK

Mitris Victor 334CC
Tema 4 ASC
CUDA: 2D Convolution


IMplementation:

	*FOr compiling execute make
	*run  ./run_tests.sh(if out.txt and  grafic.txt are already created, they need to be deleted or because the results are added to the end of the files- being opened with append)
	*After is executed make plot to create the grafic.png ( created plot )
	
Results analyzer:
	Great imprevement in nonshared memory version from the CPU version. The duration for the largest matrix is 48 seconds for CPU and 27 seconds on GPU.
	Se observa ca este o imbunatatire radicala in varianta GPU noshared memory fara de varianta CPU. Pentru cea mai mare matrice 
timpul pe CPU fiind de 48ms in timp ce pe GPU este 27ms.