Owispatent A technique for compiling a plan to decrease the possibility of cache thrashing is supplied Title: Method and equipment for decreasing cache thrashing Abstract: A method for compiling a plan to reduce the chance of cache thrashing is provided. The method includes identifying a loop in a program, figuring out each vector memory reference in the loop, and determining dependencies between the vector memory reference billiga louis vuitton vaska sverige in the loop. A technique, comprising: identifying a loop in a plan identifying every vector memory reference in the loop figuring out dependencies between vector memory references in the loop, such as determining unidirectional and round dependencies distributing the vector memory references into a plurality of depth loops configured to allocate the vector memory references into a plurality of temporary arrays, sized and located, so that none of polo ralph lauren sverige the vector memory references are cache synonyms, whereby the vector memory references that have circular dependencies therebetween are integrated in a typical detail loop, and whereby the depth loops are requested in accordance to the unidirectional dependencies in between the memory references examining an execution profile of the program following stated distributing and primarily based on the execution profile, determining whether to repeat said figuring out a loop, stated figuring out each vector2. A method, as established forth in claim 1, additional comprising allocating a plurality of temporary storage locations within a cache and determining the size of each temporary storage area primarily based on the dimension of the cache and the quantity of temporary storage3. A technique, as set forth in claim one, further comprising at least 1 section loop including the plurality of detail loops. 4. A technique, as established forth in claim 1, whereby distributing the vector memory references into a plurality of depth loops additional includes distributing the vector memory references into a plurality of depth loops nike air force 1 low that every include at least one vector memory reference that could advantage from cache management. 5. A technique, as set forth in declare 1, additional comprising inserting cache management instructions into at minimum one of said depth loops to manage motion of information related with the vector memory reference in between a cache and main memory. 6. A method, as established billiga nike skor online forth in declare one, further comprising inserting prefetch Directions into at least one of said detail loops to manage movement of data associated with the vector memory reference in between a cache and main memory. 7. A method, as set forth in declare 1, additional comprising performing loop unrolling on at minimum 1 of stated detail loops to control movement of information related with the vector memory reference between a cache and primary memory. eight. A method, as set forth in declare 1, further comprising inserting at minimum 1 of a prefetch instruction and a cache management instruction into at minimum one of stated detail loops to control movement of information related with the vector memory polo outlet reference in between a cache and

文章標籤
創作者介紹

new nfl jerseys, nfl jerseys for dogs, nfl jerseys on sale

duiwo00 發表在 痞客邦 PIXNET 留言(0) 人氣()