SODA: Stencil with Optimized Dataflow Architecture

Publication

• Yuze Chi, Jason Cong. Exploiting Computation Reuse for Stencil Accelerators. In DAC, 2020. [PDF] [Code]
• Yuze Chi, Jason Cong, Peng Wei, Peipei Zhou. SODA: Stencil with Optimized Dataflow Architecture. In ICCAD, 2018. (Best Paper Candidate) [PDF] [Slides] [Code]

SODA DSL Example

# comments start with hashtag(#)

kernel: blur      # the kernel name, will be used as the kernel name in HLS
burst width: 512  # I/O width in bits, for Xilinx platform 512 works the best
unroll factor: 16 # how many pixels are generated per cycle

# specify the dram bank, type, name, and dimension of the input tile
# the last dimension is not needed and a placeholder '*' must be given
# dram bank is optional
# multiple inputs can be specified but 1 and only 1 must specify the dimensions
input dram 0 uint16: input(2000, *)

# specify an intermediate stage of computation, may appear 0 or more times
local uint16: tmp(0, 0) = (input(-1, 0) + input(0, 0) + input(1, 0)) / 3

# specify the output
# dram bank is optional
output dram 1 uint16: output(0, 0) = (tmp(0, -1) + tmp(0, 0) + tmp(0, 1)) / 3

# how many times the whole computation is repeated (only works if input matches output)
iterate: 2

Getting Started

Prerequisites

• Python 3.6+ and corresponding pip

sudo apt install software-properties-common python3-pip
sudo add-apt-repository ppa:deadsnakes/ppa
sudo apt update
sudo apt install python3.6

sudo apt install python3 python3-pip

sudo yum install python3 python3-pip

Install SODA

Install from PyPI

python3 -m pip install --user --upgrade sodac

• Replace python3 with a more specific Python version higher than or equal to python3.6, if necessary.
• Make sure ~/.local/bin is in your PATH, or replace sodac with python3 -m soda.sodac below.

Generate Vivado HLS kernel code

sodac tests/src/blur.soda --xocl-kernel blur_kernel.cpp

Generate Intel OpenCL kernel code

sodac tests/src/blur.soda --iocl-kernel blur_kernel.cl

Generate Xilinx Object file with AXI MMAP Master Interface

Requries vivado_hls.

sodac tests/src/blur.soda --xocl-hw-xo blur_kernel.hw.xo

Generate Xilinx Object file with AXI Stream Interface

Requries vivado_hls.

sodac tests/src/blur.soda --xocl-hw-xo blur_kernel.hw.xo --xocl-interface axis

Apply Computation Reuse

sodac tests/src/blur.soda --computation-reuse --xocl-kernel blur_kernel.cpp

Code Snippets

Configuration

• 5-point 2D Jacobi: t0(0, 0) = (t1(0, 1) + t1(1, 0) + t1(0, 0) + t1(0, -1) + t1(-1, 0)) * 0.2f
• tile size is (2000, *)

Each function in the below code snippets is synthesized into an RTL module. Their arguments are all hls::stream FIFOs; Without unrolling, a simple line-buffer pipeline is generated, producing 1 pixel per cycle. With unrolling, a SODA microarchitecture pipeline is generated, procuding 2 pixeles per cycle.

Without Unrolling

#pragma HLS dataflow
Module1Func(
  /*output*/ &from_t1_offset_0_to_t1_offset_1999,
  /*output*/ &from_t1_offset_0_to_t0_pe_0,
  /* input*/ &from_super_source_to_t1_offset_0);
Module2Func(
  /*output*/ &from_t1_offset_1999_to_t1_offset_2000,
  /*output*/ &from_t1_offset_1999_to_t0_pe_0,
  /* input*/ &from_t1_offset_0_to_t1_offset_1999);
Module3Func(
  /*output*/ &from_t1_offset_2000_to_t1_offset_2001,
  /*output*/ &from_t1_offset_2000_to_t0_pe_0,
  /* input*/ &from_t1_offset_1999_to_t1_offset_2000);
Module3Func(
  /*output*/ &from_t1_offset_2001_to_t1_offset_4000,
  /*output*/ &from_t1_offset_2001_to_t0_pe_0,
  /* input*/ &from_t1_offset_2000_to_t1_offset_2001);
Module4Func(
  /*output*/ &from_t1_offset_4000_to_t0_pe_0,
  /* input*/ &from_t1_offset_2001_to_t1_offset_4000);
Module5Func(
  /*output*/ &from_t0_pe_0_to_super_sink,
  /* input*/ &from_t1_offset_0_to_t0_pe_0,
  /* input*/ &from_t1_offset_1999_to_t0_pe_0,
  /* input*/ &from_t1_offset_2000_to_t0_pe_0,
  /* input*/ &from_t1_offset_4000_to_t0_pe_0,
  /* input*/ &from_t1_offset_2001_to_t0_pe_0);

In the above code snippet, Module1Func to Module4Func are forwarding modules; they constitute the line buffer. The line buffer size is approximately two lines of pixels, i.e. 4000 pixels. Module5Func is a computing module; it implements the computation kernel. The whole design is fully pipelined; however, with only 1 computing module, it can only produce 1 pixel per cycle.

Unroll 2 Times

#pragma HLS dataflow
Module1Func(
  /*output*/ &from_t1_offset_1_to_t1_offset_1999,
  /*output*/ &from_t1_offset_1_to_t0_pe_0,
  /* input*/ &from_super_source_to_t1_offset_1);
Module1Func(
  /*output*/ &from_t1_offset_0_to_t1_offset_2000,
  /*output*/ &from_t1_offset_0_to_t0_pe_1,
  /* input*/ &from_super_source_to_t1_offset_0);
Module2Func(
  /*output*/ &from_t1_offset_1999_to_t1_offset_2001,
  /*output*/ &from_t1_offset_1999_to_t0_pe_1,
  /* input*/ &from_t1_offset_1_to_t1_offset_1999);
Module3Func(
  /*output*/ &from_t1_offset_2000_to_t1_offset_2002,
  /*output*/ &from_t1_offset_2000_to_t0_pe_1,
  /*output*/ &from_t1_offset_2000_to_t0_pe_0,
  /* input*/ &from_t1_offset_0_to_t1_offset_2000);
Module4Func(
  /*output*/ &from_t1_offset_2001_to_t1_offset_4001,
  /*output*/ &from_t1_offset_2001_to_t0_pe_1,
  /*output*/ &from_t1_offset_2001_to_t0_pe_0,
  /* input*/ &from_t1_offset_1999_to_t1_offset_2001);
Module5Func(
  /*output*/ &from_t1_offset_2002_to_t1_offset_4000,
  /*output*/ &from_t1_offset_2002_to_t0_pe_0,
  /* input*/ &from_t1_offset_2000_to_t1_offset_2002);
Module6Func(
  /*output*/ &from_t1_offset_4001_to_t0_pe_0,
  /* input*/ &from_t1_offset_2001_to_t1_offset_4001);
Module7Func(
  /*output*/ &from_t0_pe_0_to_super_sink,
  /* input*/ &from_t1_offset_1_to_t0_pe_0,
  /* input*/ &from_t1_offset_2000_to_t0_pe_0,
  /* input*/ &from_t1_offset_2001_to_t0_pe_0,
  /* input*/ &from_t1_offset_4001_to_t0_pe_0,
  /* input*/ &from_t1_offset_2002_to_t0_pe_0);
Module8Func(
  /*output*/ &from_t1_offset_4000_to_t0_pe_1,
  /* input*/ &from_t1_offset_2002_to_t1_offset_4000);
Module7Func(
  /*output*/ &from_t0_pe_1_to_super_sink,
  /* input*/ &from_t1_offset_0_to_t0_pe_1,
  /* input*/ &from_t1_offset_1999_to_t0_pe_1,
  /* input*/ &from_t1_offset_2000_to_t0_pe_1,
  /* input*/ &from_t1_offset_4000_to_t0_pe_1,
  /* input*/ &from_t1_offset_2001_to_t0_pe_1);

In the above code snippet, Module1Func to Module6Func and Module8Func are forwarding modules; they constitute the line buffers of the SODA microarchitecture. Although unrolled, the line buffer size is still approximately two lines of pixels, i.e. 4000 pixels. Module7Func is a computing module; it is instanciated twice. The whole design is fully pipelined and can produce 2 pixel per cycle. In general, the unroll factor can be set to any number that satisfies the throughput requirement.

Best Practices

• For non-iterative stencil, unroll factor shall be determined by the DRAM bandwidth, i.e. saturate the external bandwidth, since the resource is usually not the bottleneck
• For iterative stencil, to use more PEs in a single iteration or to implement more iterations is yet to be explored
• Note that 2.0 will be a double number. To generate float, use 2.0f. This may help reduce DSP usage
• SODA is tiling-based and the size of the tile is specified in the input keyword. The last dimension is omitted because it is not needed in the reuse buffer generation

Projects Using SODA

• Licheng Guo, Jason Lau, Yuze Chi, Jie Wang, Cody Hao Yu, Zhe Chen, Zhiru Zhang, Jason Cong . Analysis and Optimization of the Implicit Broadcasts in FPGA HLS to Improve Maximum Frequency. In DAC, 2020. [PDF]
• Jiajie Li, Yuze Chi, Jason Cong. HeteroHalide: From Image Processing DSL to Efficient FPGA Acceleration. In FPGA, 2020. [PDF] [Slides]
• Young-kyu Choi, Yuze Chi, Jie Wang, Jason Cong. FLASH: Fast, ParalleL, and Accurate Simulator for HLS. In TCAD, 2020. [PDF]
• Yi-Hsiang Lai, Yuze Chi, Yuwei Hu, Jie Wang, Cody Hao Yu, Yuan Zhou, Jason Cong, Zhiru Zhang. HeteroCL: A Multi-Paradigm Programming Infrastructure for Software-Defined Reconfigurable Computing. In FPGA, 2019. (Best Paper Candidate) [PDF] [Slides] [Code]
• Yuze Chi, Young-kyu Choi, Jason Cong, Jie Wang. Rapid Cycle-Accurate Simulator for High-Level Synthesis. In FPGA, 2019. [PDF] [Slides]