Interprocedural Type Specialization of JavaScript Programs Without Type Analysis

TL;DR

This paper introduces an interprocedural extension of lazy basic block versioning for JavaScript JIT compilers, effectively eliminating redundant type checks without costly static analysis, leading to significant performance improvements.

Contribution

It extends lazy basic block versioning to propagate type information across function calls, outperforming static analysis in eliminating type checks and improving execution speed.

Findings

94.3% of type tag tests eliminated on average

Up to 56% speedup achieved

Outperforms Truffle/JS in execution and compilation time

Abstract

Dynamically typed programming languages such as Python and JavaScript defer type checking to run time. VM implementations can improve performance by eliminating redundant dynamic type checks. However, type inference analyses are often costly and involve tradeoffs between compilation time and resulting precision. This has lead to the creation of increasingly complex multi-tiered VM architectures. Lazy basic block versioning is a simple JIT compilation technique which effectively removes redundant type checks from critical code paths. This novel approach lazily generates type-specialized versions of basic blocks on-the-fly while propagating context-dependent type information. This approach does not require the use of costly program analyses, is not restricted by the precision limitations of traditional type analyses. This paper extends lazy basic block versioning to propagate type information interprocedurally, across function call boundaries. Our implementation in a JavaScript JIT compiler shows that across 26 benchmarks, interprocedural basic block versioning eliminates more type tag tests on average than what is achievable with static type analysis without resorting to code transformations. On average, 94.3% of type tag tests are eliminated, yielding speedups of up to 56%. We also show that our implementation is able to outperform Truffle/JS on several benchmarks, both in terms of execution time and compilation time.