SimpleTPU

A Tensor Processing Unit is designed to accelerate the matrix multiplication, especially for Multilayer perceptron and Convolution Nerual Network.
This implmentaion is mainly following the Google TPU Version 1, which architecture is introduced in https://arxiv.org/ftp/arxiv/papers/1704/1704.04760.pdf.

It may cost a lot of time to implementation TPU using Hardware Description Language (such as VHDL or Verilog HDL), even if I had tried to simplify it. So I try to use the Xilinx HLS ToolKit to complete it.

The plan is divided into three phases.

Phase 1: Completing the main computing module,including
- Lab1:Systolic Array
- Lab2:Relu, Normalization & Pooling
Phase 2: Finish the full design of simpleTPU.
Phase 3: Testing the simpleTPU through some real network, such as MLP and CNN.

Key Features

The key features of Simple TPU including

Int8 mulitply & Int32 accumulators
VLIW based instruction parallel
Vector Architecture based data parallel

Here are some operate which Simple TPU can support.

Operate	Support
Conv3d	in_channels: Resource Constrained out_channels: Resource Constrained kerner_size: Support stride: support padding: Support dilation:Support groups: Architecture Constrained bias :Support
ConvTranspose3d	The same as above
Maxpool2d	kernel_size: Support stride: Support padding: Support
Avgpool2d	The same as above
Relu	Only support Relu as nonlinear function
BatchNorm2d	BatchNorm2d is merge with Conv or Pool when inference
Linear	Resource Constrained
UpscalingNearest2D	Support (calling Avgpool2d multiple times.)
UpscalingBilinear2D	Support (calling Avgpool2d multiple times.)

Performance

The size of mac array in SimpleTPU is 32*32, the clock frequency is 500MHz (timing closure when using Xilinx ultrascale+ FPGA, Speed -2).
$$32\times 32 \times 500 \times 2 = 1 Tops(int8)$$

Installation

env :

Vivado HLS 2018.2

run :

step1: vivado_hls -f run_hls.tcl
step2: lanch vivado HLS and open the project
step3: Run C synthesis, C/RTL cosimulation e.t.c

Synthesis Result:

Simulation Result:

Examlpes

1. MLP

The network structure of mlp is defined as follow.

class MLP(nn.Module):
    def __init__(self):
        super(MLP, self).__init__()
        self.hidden = nn.Linear(784,64)
        self.fc = nn.Linear(64,10)

    def forward(self, x):
        x = x.view(-1,784)
        x = self.hidden(x)
        x = self.fc(x)
        return F.log_softmax(x, dim=1)

Work efficiency of SimpleTPU is about 84%.

LOC	Layers	Nonlinear function	Weights	Batch Size	% of Deployed
10	2 FC	Relu	5M	512	16%

Classfication Result in MNIST.

2. CNN

Because there is no compiler to generate instruction, this plan was suspended. If you want to kown how to calculate convolution using SimpleTPU, lab1 provides a simple example.

Relative Link

https://www.cnblogs.com/sea-wind/p/10993958.html

OmerTariq-KAIST/SimpleTPU