LeFlowを調べてみました
3 likes1,371 views
The document discusses leflow, a tool that enables flexible FPGA high-level synthesis of TensorFlow deep neural networks, detailing its optimization process using LLVM IR for better FPGA compatibility. It highlights the current limitations of leflow while suggesting potential areas for improvement, such as specific FPGA-targeted kernels in the XLA backend and enhanced memory partitioning algorithms for machine learning. Additionally, it mentions future enhancements like fixed-point representation support and improved debugging infrastructure for developers.
![最適化
Algorithm 1
%0 = bitcast i8** %params to [2 x float]**
%arg0 = load [2 x float] %0, align 8
=> @arg0 を global に変更し、ゼロで初期化
Algorithm 2
%arg0 = global [2 x float] zeroinitializer, align 8](https://image.slidesharecdn.com/leflow-180722031411/85/LeFlow-6-320.jpg)
![MNIST
with tf.devide(“device:XLA_CPU:0”):
y = tf.nn.softmax(tf.add(tf.matmul(input, weights)[0], bias))](https://image.slidesharecdn.com/leflow-180722031411/85/LeFlow-7-320.jpg)
LeFlowを調べてみました
- 1. LeFlow Enabling Flexible FPGA High-Level Synthesis of Tensorflow Deep Neural Networks @Vengineer 2018/07/22 (LeFlow追記) TensorFlow XLAのロゴ https://github.com/tensorflow/tensorflow/tree/master/tensorflow/compiler/xla
- 2. ブログ (2007年～) : Vengineerの戯言 http://blogs.yahoo.co.jp/verification_engineer SlideShare : https://www.slideshare.net/ssuser479fa3 Twitter (2009年～) : ＠Vengineer
- 3. Enabling Flexible FPGA High-Level Synthesis of Tensorflow Deep Neural Networks 論文 : https://arxiv.org/abs/1807.05317 Submitted on 14 Jul 2018 github : https://github.com/danielholanda/LeFlow 最適化 XLAの最適化無しのLLVM IRを使って、LegUpで最適化を行う。 LeFlow
- 4. LeFlowのフォロー ・TensorFlowのコード が Verilog HDL に！ ・TensorFlow XLA側の修正無し (ただし、ソースコードへのパッチは必要) (2ファイル) ・LegUpのそのまま使える ・Intel (Altera)へのサンプルコード有 (Xilinxでもできる模様)
- 5. 生成されるHDLの構成
- 6. 最適化 Algorithm 1 %0 = bitcast i8** %params to [2 x float]** %arg0 = load [2 x float] %0, align 8 => @arg0 を global に変更し、ゼロで初期化 Algorithm 2 %arg0 = global [2 x float] zeroinitializer, align 8
- 7. MNIST with tf.devide(“device:XLA_CPU:0”): y = tf.nn.softmax(tf.add(tf.matmul(input, weights)[0], bias))
- 8. LeFlowの処理内容 src/LeFlow # Create folder and generate Makefile # Clean folder to erase previously generated files # Generate IR from tensorflow # Remove unused files and name things properly # Convert to old LLVM syntax (Tensorflow and LegUp use different versions of LLVM) # Unrolling loops # Restructure function signature, change variables scope and reorganizes code # Reqrites unsupported operations # Partitioning the memories # Convert human-readable .ll file to bitcode (.bc file) # Start LegUp compilation # Instrumenting verilog testbench
- 10. LeFlowの処理内容 src/LeFlow # Create folder and generate Makefile # Clean folder to erase previously generated files # Generate IR from tensorflow # Remove unused files and name things properly # Convert to old LLVM syntax (Tensorflow and LegUp use different versions of LLVM) # Unrolling loops # Restructure function signature, change variables scope and reorganizes code # Reqrites unsupported operations # Partitioning the memories <= ここがポイント(メモリを分割) # Convert human-readable .ll file to bitcode (.bc file) # Start LegUp compilation # Instrumenting verilog testbench
- 12. Current Limitations and Opportunities 1). LeFlow currently uses kernels that were implemented in XLA and were originally meant to be used by CPUs. Although compiler optimizations and scheduling are able to retrieve a substantial amount of parallelism from those implementations, LeFlow would heavily benefit from an XLA back-end with kernels targeting FPGAs ; 2). The high dimensionality of inputs/weights and the amount of parallel accesses that are typical in machine learning applications is a challenge for modern automatic memory partitioning algorithms . LeFlow would specially benefit from a machine learning specific automatic memory partitioning algorithm. 3). One of the key possibilities that make deep learning networks efficient in FPGAs is t he opportunity to use a customizable fixed-point bit width . Adding fixed-point support to LeFlow will be an important step in the development of this toolkit. Additionally, techniques to automatically profile the application and choose the appropriate representation could be easily explored in software with Tensorflow and deployed in hardware. 4). Although it is straightforward to use Tensorflow to debug the functionality of an implementation, it is currently difficult for software developers to debug the generated hardware in terms of the original Python code. A performance debugging infrastructure suitable for software developers is another interesting venue for research .
- 13. ブログ (2007年～) : Vengineerの戯言 http://blogs.yahoo.co.jp/verification_engineer SlideShare : https://www.slideshare.net/ssuser479fa3 Twitter (2009年～) : ＠Vengineer ありがとうございました