uops.info

x86 Operand Microarchitecture Port (circuit theory)
DOI: 10.1145/3297858.3304062 Publication Date: 2019-04-04T18:38:43Z
ABSTRACT
Modern microarchitectures are some of the world's most complex man-made systems. As a consequence, it is increasingly difficult to predict, explain, let alone optimize performance software running on such microarchitectures. basis for predictions and optimizations, we would need faithful models their behavior, which are, unfortunately, seldom available. In this paper, present design implementation tool construct latency, throughput, port usage x86 instructions. To end, first discuss common notions instruction throughput usage, introduce more precise definition latency that, in contrast previous definitions, considers dependencies between different pairs input output operands. We then develop novel algorithms infer based automatically-generated microbenchmarks that accurate than existing work. facilitate rapid construction optimizing compilers tools prediction, our provided machine-readable format. provide experimental results processors all generations Intel's Core architecture, i.e., from Nehalem Coffee Lake, various cases where differs considerably prior
SUPPLEMENTAL MATERIAL
Coming soon ....
REFERENCES (33)
CITATIONS (53)
EXTERNAL LINKS
PlumX Metrics
RECOMMENDATIONS
FAIR ASSESSMENT
Coming soon ....
JUPYTER LAB
Coming soon ....