Use of Previously Acquired Positioning of Optimizations for Phase Ordering Exploration

Abstract

This paper presents a new approach to efficiently search for suitable compiler pass sequences, a challenge known as phase ordering. Our approach relies on information about the relative positions of compiler passes in compiler pass sequences previously generated for a set of functions when compiling for a specific processor. We enhanced two iterative compiler pass exploration schemes, one relying on simple sequential compiler pass insertion and other implementing an auto-tuned simulated annealing process, with a data structure that holds information about the relative positions of compiler sequences; in order to reduce the set of compiler passes considered for insertion in a given position of a given candidate compiler pass sequence to include only the passes that have a higher probability of performing well on that relative position in the compiler sequence, speeding up the exploration time as a result. We tested our approach with two different compilers and two different targets; the ReflectC and the LLVM compilers, targeting a MicroBlaze processor and a LEON3 processor, respectively. The experimental results show that we can considerably reduce the number of algorithm iterations by a factor of up to more than an order of magnitude when targeting the MicroBlaze or the LEON3, while finding compiler sequences that result in binaries that when executed on the target processor/simulator are able to outperform (i.e. use less CPU cycles) all the standard optimization levels (i.e., we compare against the most performing optimization level flag on each kernel, e.g. -O1, -O2 or -O3 in the case of LLVM) by a geometric mean performance improvement of 1.23x and 1.20x when targeting the MicroBlaze processor, and 1.94x and 2.65x when targetting the LEON3 processor; for each of the two exploration algorithms and two kernel sets considered.

A Survey on Compiler Autotuning using Machine Learning

“…Clustering / Downsampling [19,29,111,122,142,170,171,185,186,212,248] Speedup Prediction [18, 19, 21, 28, 50, 67, 77, 78, 96, 99, 100, 136, 152, 155, 199-202, 235, 236, 259] Compiler Sequence Prediction [3, 6, 14-16, 21, 28, 29, 38, 41, 42, 50-55, 61, 67, 69, 70, 79, 83, 91, 98, 99, 101, 105-108, 116, 122, 123, 132, 142, 146-149, 151, 160, 162, 163, 166, 169-171, 178, 183-185, 196, 198, 201, 202, 206, 208, 209, 212, 213, 217, 221, 235-237, 240, 248, 250, 253, 259] Tournament / Intermediate [18,151,199,201,202,251] Feature Prediction [66,73,77,157,202,253] try to nd ways to decrease the enormous size of the optimization space by orders of magnitudes. This technique, known as downsampling, has been used in many recent works.…”

“…Iterative compilation is the most commonly used exploration technique for the compiler optimization eld. Many recent works found this technique interesting and successful either (i) alone [42,143], (ii) combined with machine learning techniques [3,21,51], or (iii) combined with other search and meta-heuristics techniques such as random exploration [42] and DSE techniques [170,185].…”

Supporting the Scale-Up of High Performance Application to Pre-Exascale Systems: The ANTAREX Approach

“…Compared to other approaches that usually focus on code injection (e.g., [29]), LARA provides access to other types of actions, e.g., code refactoring, compiler optimizations, and inclusion of additional information, all of which can guide compilers to generate more efficient implementations. Additional types of actions may be defined in the language specification and associated weaver, such as software/hardware partitioning [30] or compiler optimization sequences [31].…”

The ANTAREX tool flow for monitoring and autotuning energy efficient HPC systems

“…The LARA language and framework have been developed during this project and we support not only code insertions but also specific weaver actions that deal with code transformations, refactorings, software/hardware partitioning [8], compiler optimization sequences [9], among other concerns.…”