Titlebook: Automatic SIMD Vectorization of SSA-based Control Flow Graphs; Ralf Karrenberg Book 2015 Springer Fachmedien Wiesbaden 2015 Data-parallel

潔凈

聚集

Overview,In this chapter, we present an overview of the basic . algorithm, the challenges faced when vectorizing real-world code, and the most important performance implications of the transformation. In Chapter 5, we give a detailed description of the analyses involved, followed by the actual presentation of the WFV algorithm in Chapter 6.

平庸的人或物

Whole-Function Vectorization,In this chapter, we present the main transformation phases of the Whole- Function Vectorization algorithm: Mask Generation, Select Generation, Partial CFG Linearization, and Instruction Vectorization.

AGGER

Dynamic Code Variants,During or after vectorization of a function, additional optimizations that exploit . of the function can be applied. Such an optimization duplicates an existing code path, improves the new code by assuming certain properties, and guards the new path by a runtime check that ensures that the assumed properties hold. We call such a guarded path a ..

Factual

an approach that allows a compiler to automatically create code that exploits data-parallelism using SIMD instructions. Data-parallel applications such as particle simulations, stock option price estimation or video decoding require the same computations to be performed on huge amounts of data. Wit

chapel

–scent

Evaluation, system aimed at highly efficient code without loss of portability and flexibility [Karrenberg et al. 2010]. Third, in ., a compiler that allows the user to specify which optimizations should be run on what parts of the code. In addition, we quantify the impact of the key techniques that exploit the results of the analyses described in Chapter 5.

笨拙的我

TRACE

洞察力