English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Pipelining between data loading and computation is a critical tensor programoptimization for GPUs. In order to unleash the high performance of latest GPUs,we must perform a synergetic optimization of multi-stage pipelining across themulti-level buffer hierarchy of GPU. Existing frameworks rely on hand-writtenlibraries such as cuBLAS to perform pipelining optimization, which isinextensible to new operators and un-composable with prior tensor compileroptimizations. This paper presents ALCOP, the first framework that iscompiler-native and fully supports multi-stage multi-level pipelining. ALCOPovercomes three critical obstacles in generating code for pipelining: detectionof pipelining-applicable buffers, program transformation for multi-levelmulti-stage pipelining, and efficient schedule parameter search byincorporating static analysis. Experiments show that ALCOP can generateprograms with 1.23x speedup on average (up to 1.73x) over vanilla TVM. Onend-to-end models, ALCOP can improve upon TVM by up to 1.18x, and XLA by up to1.64x. Besides, our performance model significantly improves the efficiency ofthe schedule tuning process and can find schedules with 99% of the performancegiven by exhaustive search while costing 40x fewer trials.

ALCOP: Automatic Load-Compute Pipelining in Deep Learning Compiler for AI-GPUs