abstract efficient support for complex numbers in java.pdf
还剩
9页未读,
继续阅读
下载文档到电脑,马上远离加班熬夜!
亲,喜欢就下载吧,价低环保!
内容要点:
References l-11 PI 13:1 r41 [I [61 r71 PI 191 [lOI [llI [W J. C. Adams, W. S. Brainerd, J. T. Martin, B. T. Smith, and J. L. Wagener. Fortran 90 Handbook: Complete ANSI/IS0 Reference. McGraw-Hill, 199class libraries and just waiting for optimization. Second, the amount of work necessary to semantically expand one of these classes is small - on the order of the amount of time necessary to implementPerformance of MATMUL on R&6000 590 (Mflops) Performance of LU on fW6000 590 (Mflops) 1 136.4 0.9 s e 0.8 2 0.7 $0.6 B 0.5 .g 0.4 z 0.3 0.2 0.1 plain c++ complex array Fortran Code version 1 0.9 : z 0Performance of MICROSTRIP on FE/6000 590 (Mflops) ( 1 0.9 SO.8 s 20.7 iO.6 -5 0.5’ .go.4 $0.3 0.2; 0.1 I ii l-l 1 .o plain c++ complex array Fortran Code version Figure 7: Summary of results for MICROfoo I object foo I times value aA I A plus a value value 4 I I b b C (4 (b) Figure 6: The AST for a complex arithmetic expression. be created, even if each use of the object being assigned re- quires The first optimization provided by semantic expansion is the elimination of method invocation overhead. The com- piler replaces method invocations by code that directly im- plements the semantics of tIn comparison, the memory utilization grows rapidly in AC mode, exhausting the total memory allocated for the appli- cation in a few iterations, thus forcing garbage collection to occur. Thus, the exeComplex[l[l b; b[i][j].assign(a[i+llIjl.plus(a[i-1 ComplexArray2D a; ComplexArrayZD b; ?omplex[l [I a; b.set(i,j,a.get(i+i,j~.plus~a.get(i-l,j~~.plus(a.get~i,j+l)).plus(a.get(i,j-l)).times(O.25~~;
发表评论
暂无评论,赶快抢占沙发吧。