Abstract

International audience; The xMAS micro-architecture modeling language has been introduced by Intel to facilitate the formal representation and analysis of on-chip interconnect fabrics. In this paper, we introduce xMAStime, a new domain-specific language inspired by xMAS. xMAStime allows the modeling of full micro-architectures comprising certain classes of CPU pipelines, caches, and RAM. Given an in-order pipeline model in xMAStime, we automatically generate both a Cycle-Accurate, Bit-Accurate (CABA) hardware simulator and a timed instruction set simulator where time is accounted with safe upper bounds, as in the pipeline analysis step of Worst-Case Execution Time (WCET) analysis. The approach relies on the theory of endochronous systems, which allows us to ensure functional equivalence and timing consistency between the two generated simulators, using a delay-insensitivity argument. xMAStime is implemented over Lucid Synchrone-a dataflow synchronous language featuring a higher order type system and type inference, which facilitate the definition of our DSL. We use the new DSL to model and synthesize simulation code for a full-fledged MIPS32-based architecture.

Abstract

International audience; Increasing data movement costs motivate the integration of polyhedral loop optimizers in the standard flow (-O3) of production compilers. While polyhedral optimizers have been shown to be effective when applied as source-to-source transformation, the single static assignment form used in modern compiler mid-ends makes such optimizers less effective. Scalar dependencies (dependencies carried over a single memory location) are the main obstacle preventing effective optimization. We present DeLICM, a set of transformations which, backed by a polyhedral value analysis, eliminate problematic scalar dependences by 1) relocating scalar memory references to unused array locations and by 2) forwarding computations that otherwise cause scalar dependences. Our experiments show that DeLICM effectively eliminates dependencies introduced by compiler-internal canonicalization passes, human programmers, optimizing code generators, or inlining – without the need for any additional memory allocation. As a result, polyhedral loop optimizations can be better integrated into compiler pass pipelines which is essential for metaprogramming optimization.