Skymizer connects ONNX to all deep learning accelerator ASICs

Skymizer, a compiler company founded in 2013, will launch the open source compiler “ONNC” (Open Neural Network Compiler) to ONNX backed by its unique compiler technologies.

Hundreds of AI chips are releasing in the near future, the latest figures indicate 34 IC and IP vendors will provide various AI chips and deep learning accelerator (DLA) ASICs in 2018. These all reflect the urgent need for an open compiler to support different AI chips.

Skymizer foresaw the trend and developed the compiler ONNC. Based on ONNX, ONNC is an efficient way to connect all current AI chips, especially DLA ASICs, with ONNX. Skymizer will open source ONNC before the end of July 2018.

Open Neural Network Exchange Format (ONNX) is a standard for representing deep learning models that enables models to be transferred between frameworks. Skymizer introduces ONNC that supports ONNX format and mainstream AI frameworks such as Caffe and Tensorflow. ONNC’s dominant advantage to current AI frameworks is that it provides direct support to DLA ASIC chips by ability to describe variants of performance cost models of hardware and by general optimization passes. DLA ASIC vendors can reuse these optimization passes by describing its special performance cost model in ONNC. ONNX and ONNC together help DLA ASIC vendors support various AI frameworks within a short time, improves DLA’s performance and shortens developing schedule.

“AI innovations need the open ecosystem, ONNX, who guarantees interoperability among frameworks,” said Luba Tang, CEO of Skymizer. “ONNC aims to connect all deep learning accelerators to ONNX, by general approach, in short time.”

Skymizer will release ONNC, an open source neural network compiler before the end of July 2018.

Introduction to evaluate tool “SkySquire”

SkySquire is a versatile tool, or say a Swiss army knife, to help you revise the default behavior of softwares. For example, you can enable ‘-O3’ as a default option of your gcc/clang. With SkySquire, you don’t need to handle with all details of Makefiles and building system any more.

Continue reading Introduction to evaluate tool “SkySquire”

How to evaluate Skymizer Intelligent Compiler for your projects

According to our study, the bigger performance gap between O0 to O3 is, the higher performance Skymizer Intelligent Compiler can achieve eventually.

Skymizer Intelligent Compiler combines machine learning, high-performance computing and compiler technologies to find the best way to optimize your projects. Its high level concept is a sort of iterative compiler which drives performance guide optimization (PGO) and linker optimization (on linker level, not LTO). Our machine learning system studies the whole compilation process of your project, from compiler to linker, and recommends you the best optimization approach.

Continue reading How to evaluate Skymizer Intelligent Compiler for your projects

SkySquire:a very easy way to eable LTO/PGO

SkySquire is a versatile tool, or say a Swiss army knife, to help you revise the default behavior of softwares. For example, you can enable PGO (profile-guide optimization) or LTO (link-time optimization) on a very easy way. With Evaluate, you don’t need to handle with all details of Makefiles and building system any more. All that you need to do is to set -flto and -fprofile-generate/use as a default mark of gcc by using SkySquire.

More Information about SkySquire
Introduction to Evaluate
How to evaluate Skymizer Intelligent Compiler for your projects