This book investigates several problems facing compilers for supercomputers. One problem is building a comprehensive data dependence graph; others are associated with reductions and recurrent relations, including conditional recurrences and mixed control-data dependence cycles. The book proposes data dependence testing methods and, by labelling the arcs with direction vectors that show the flow of data within the loop structure of the programme, this labelled data dependence graph is used in several high-level compiler loop transformations. The author explains how to apply these optimizations for different super-computer architectures. The book also investigates some problems associated with reductions and recurrence relations, including conditional recurrences and mixed control-data dependence cycles. A new formulation for the wavefront methods of executing a loop is also given. Two other problems studied are the management of compiler temporary arrays and vectorizing WHILE loops.
"synopsis" may belong to another edition of this title.
Seller: Roland Antiquariat UG haftungsbeschränkt, Weinheim, Germany
Softcover. 165 p. Good condition. The reading pages are clean and unmarked. Slight signs of storage and use. Retired library copy with corresponding labelling. Spine slightly rubbed. Otherwise a good copy. 9780273088011 Sprache: Englisch Gewicht in Gramm: 1000. Seller Inventory # 305237
Seller: BUCHSERVICE / ANTIQUARIAT Lars Lutzer, Wahlstedt, Germany
Condition: gut. 1989. Optimizing Supercompilers for Supercomputers In englischer Sprache. pages. Seller Inventory # BN497898