Automatic generation of peephole superoptimizers
Automatic generation of peephole superoptimizers
复制标题
DOI：
10.1145/1168918.1168906
复制
发表时间：
2006-11-01
影响因子：
--
通讯作者：
Aiken, Alex
中科院分区：
文献类型：
--
作者：
Bansal, Sorav;Aiken, Alex
窥视优化器通常是使用人类编写的模式匹配规则来构建的，这种方法需要专业知识和时间，而且在利用所有优化机会方面缺乏系统性。我们探索了使用蛮力超优化全自动构造窥视孔优化器。虽然我们的自动系统发现的优化可能没有人类编写的相应优化那么通用，但我们的方法有可能自动学习数千到数百万个优化的数据库，而不是目前的窥视孔优化器中发现的数百个优化。我们的实验表明，我们的优化器能够利用现有编译器所没有的性能机会；特别是，我们在一些计算密集型内核上的加速比传统的优化编译器快了1.7到10倍。
Peephole optimizers are typically constructed using human-written pattern matching rules, an approach that requires expertise and time, as well as being less than systematic at exploiting all opportunities for optimization. We explore fully automatic construction of peephole optimizers using brute force superoptimization. While the optimizations discovered by our automatic system may be less general than human-written counterparts, our approach has the potential to automatically learn a database of thousands to millions of optimizations, in contrast to the hundreds found in current peephole optimizers. We show experimentally that our optimizer is able to exploit performance opportunities not found by existing compilers; in particular, we show speedups from 1.7 to a factor of 10 on some compute intensive kernels over a conventional optimizing compiler.