id	author	title	date	pages	extension	mime	words	sentence	flesch	summary	cache	txt
ap-826	Šimeček, I.	Performance Aspects of Sparse Matrix-Vector Multiplication	2006	6	.pdf	application/pdf	2473	130	63	Loop unrolling-and-jam applied to the example code (Uf � 2) s1 � 0.0; s2 � 0.0; for i � 1 to n step 2 do s1 � s1�A[i]; s2 � s2�A[i+1]; s � s1�s2; 1.2.3 Software pipelining The initial instruction(s) of the first iteration is/are moved into the prologue phase, and the final instruction(s) of the last iteration is/are moved into the epilogue phase. This makes the serial code longer, so that the instructions can be better scheduled, the internal pipeline can be better utilized, and the number of condition tests drops from n to � �n Uf .	cache/ap-826.pdf	txt/ap-826.txt