O. Temam, W. Jalby Characterizing the Behavior of Sparse Algorithms on Caches Home/Search Document Not in Database Summary Related Articles Check |
This paper is cited in the following contexts:
The Cache Visualisation Tool - van der Deijl, Kanbier (Correct)
....are shown. 3.3 Controlling the CVT There are two rows of buttons to control the CVT. The top row is to control the CVT in terms letting the CVT simulate a certain kind of cache in the way the user wants, like listening to a CD the way a user wants ) A[1:10,1:10] A[1,1] has logical number 0 A[1,2] has logical number 1 : A[2,1] has logical number 10 A[2,2] has logical number 11 : A[10,10] has logical number 99 When we only want to see statistic of the first row of this matrix, we specify: Array Name : A First Element: 0 Last Element : 9 When we only want to see statistics of the first ....
....CVT There are two rows of buttons to control the CVT. The top row is to control the CVT in terms letting the CVT simulate a certain kind of cache in the way the user wants, like listening to a CD the way a user wants ) A[1:10,1:10] A[1,1] has logical number 0 A[1,2] has logical number 1 : A[2,1] has logical number 10 A[2,2] has logical number 11 : A[10,10] has logical number 99 When we only want to see statistic of the first row of this matrix, we specify: Array Name : A First Element: 0 Last Element : 9 When we only want to see statistics of the first column of this matrix, we ll need ....
[Article contains additional citation context not shown here]
O. Temam, W. Jalby Characterizing the Behavior of Sparse Algorithms on Caches
Speculative Prefetching - Jegou, Temam (1993) (22 citations) Self-citation (Temam) (Correct)
....of accesses in these codes. Prefetching benefits such codes because there is some random regularity in irregular codes. This is true for CPR, but even more for SpMV since it has been shown that sparse computations exhibit some regularity and locality due to the structure of sparse matrices (cf [13]) Hit ratios The main effect of Speculative Prefetching is to increase the global hit ratio (i.e hit ratio of cache plus prefetch cache) The cache and global hit ratio are indicated on figure 15. For most regular codes, the prefetch cache absorbs a major part of cache misses. Except for CPR, ....
O. Temam, W. Jalby: Characterizing the Behavior of Sparse Algorithms on Caches, Proc. of Supercomputing '92.
Characterizing the Behavior of Sparse Algorithms on Caches - Temam, Jalby (1992) (27 citations) Self-citation (Temam) (Correct)
....of sparse matrices. In section 2 of this paper, spatial and temporal locality of SpMxV is qualitatively evaluated and potential problems are identified. Then, in section 3, a meaningful share of the paper is devoted to modeling the non regular references which appear in SpMxV are presented (see [12] for details of the model) because a necessary prerequisite to evaluating and optimizing a primitive is a good understanding of its behavior. Besides, the purpose of this section is to show that it is possible to predict the behavior and performance of sparse codes, and to actually quantify their ....
....matrices (matrix 1138 BUS of the Harwell Boeing suite) DO I=1,N DO J=1,N Y(I) Y(I) A(I,J) X(J) ENDDO ENDDO DO I=1,N REG = Y(I) DO J=D(I) D(I 1) 1 REG = REG Matrix(J) X(Index(J) ENDDO Y(I) REG ENDDO Figure 4: Original loop nest; storage by row. Figure 5: Problem parameters (cf. [12]) 2.1 Data locality A first step to understanding the behavior of SpMxV on cache is to study the locality of the data used by this primitive. Since there are N nz references to arrays X, Index and Matrix, and 2N references to arrays Y and D, the total number of references is 3 N nz 4 N (cf. ....
[Article contains additional citation context not shown here]
O. Temam, W. Jalby: Characterizing the behavior of sparse algorithms on caches, INRIA Report No. 1666, April 1992.
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC