

Study with the several resources on Docsity
Earn points by helping other students or get them with a premium plan
Prepare for your exams
Study with the several resources on Docsity
Earn points to download
Earn points by helping other students or get them with a premium plan
A concise overview of advanced computer architecture concepts, presenting definitions for terms such as basic block, branch bias, cache pollution, content addressable memory, and dynamic branch prediction. It also covers topics like fetch bundles, lanes in superscalar processors, linked data structures, miss caches, pointer chasing, prefetching, program order, saturating counters, shift registers, static branch prediction, structural hazards, superscalar architecture, and victim caches. This resource is useful for students and professionals seeking a quick reference guide to these essential concepts in computer architecture.
Typology: Study notes
1 / 2
This page cannot be seen from the preview
Don't miss anything!


basic block - correct answera sequence of instruction with exactly one entrance and one exit. If execution reaches the first instruction in a basic block then it is guaranteed that all of the instructions within basic block will be executed branch bias - correct answera conditional branch instruction that almost always exhibits the same taken/not-taken status each time it executes cache pollution - correct answerwhen a useless prefetch block replaces a useful cache block (demand or prefetches correctly) content addressable memory - correct answera generic digital logic memory a full- associative RAM. All entries are searched in parallel for the input value. Must be kept small capacity. Used in several different places in modern architectures Demand access - correct answera memory access caused by actual program instructions. Could be an actual instruction fetch. Or an access using an effective address calculated by load and/or store instructions dynamic branch prediction - correct answereach a CTI is fetched, predict the outcome (taken vs. not-taken) based on past history fetch bundle - correct answergroup of instructions fetched at the same time lane - correct answerthe path an instruction could possible take through a superscalar processor. For a 2-wide superscalar there are two lanes. For 3-wide superscalar, there are three lanes linked data structures - correct answerthe generic term for any software data structure that retains not only data value(s). But also reference(s) to other items of data that are logically adjacent. The reference is an address, which is the link to the other items. miss cache - correct answercopies information into miss cache when ever it misses pointer chasing - correct answerin order to access the 2nd node you must first access the 1st node in order to access the 3rd node you must first access the 2nd node and on and on prefetch - correct answercheck the memory hierarchy for cache block data before the program actually needs the data. Bring block data into higher level cache early. Used not just for instruction fetches. But any memory access
prefetch access - correct answera memory access caused by a prefetch A guess the address used for a prefetch could be useless for the actual program program order - correct answerthe order of instructions as they appear in the original assembly code saturating counter - correct answera counter that does not wrap around when adding to the highest possible value, or when subtracting from the lowest possible value shift register - correct answera register that is written 1 bit at a time but read all but read all bits wide at a time when a write occurs all bit values move over by one position the lowest most value is ousted and the new bit value is written to the upper bit position static branch prediction - correct answereach time a CTI is fetched assume the same outcome (taken vs. not-taken) structural hazard - correct answerwhen multiple instructions need to use the same circuit in the same cycle superscalar - correct answera processor that supports multiple instructions within a given pipeline stage at the same tim victim cache - correct answerupdate with information from the L1D cache