Branch Prediction using Advanced Neural Methods

1
Branch Prediction using Advanced Neural Methods

• Sunghoon Kim
• CS252 Project

2
Introduction

• Dynanmic Branch Prediction
• No doubt about its importance in Speculation
  performance
• Given history of branch behaviors, predict branch
  behaviors at the next step
• Common solutions gshare, bimode,hybrid
• Replace saturating counters with neural methods?

3
Neural methods

• Capable of classification (predicting into which
  set of classes a particular instance will fall)
• Learns correlations between inputs and output,
  and generalized learning to other inputs
• Potential to solve problems of most two-levels
  predictors

4
Simulation Models - Gshare
GHT
Predict
2-bit saturation counter
BrachPCgtgt2
Update counters
PHT

• 20-bit Global history shift register
• Per-address history table with 2-bit saturation
  counter

5
Simulation Models - Perceptron
PHT
GHT
BrachPCgtgt2
OR
Predict
BrachPCgtgt2
Training weights bias

• 14-bit Global history shift register
• Per-address history table with 8-bit weights and
  bias
• Indexed by Gshare or BranchPC alone

6
Simulation Models - Backpropagation
GHT
BrachPCgtgt2
OR
Predict
Training weights bias
BrachPCgtgt2
PHT

• 10-bit GHR
• Sigmoid transfer function
• Floating point computation
• Floating point weights and biases
• 20 neurons one hidden layer

7
Simulation Models Radial Basis Networks
GHT
PHT
BrachPCgtgt2
OR
Predict
Training weights bias
BrachPCgtgt2

• Transfer function for radial basis neuron
  exp(-n2)
• Distance function between an input vector and a
  weight vector

8
Simulation Models Elman Networks
GHT
PHT
BrachPCgtgt2
OR
Predict
Training weights bias
BrachPCgtgt2

• Feedback from the hidden layer outputs to the
  first layer

9
Simulation Models Learning Vector Quantization
Networks
GHT
PHT
BrachPCgtgt2
OR
Predict
Training weights bias
BrachPCgtgt2

• Distance function as Radial but without biases
• Competitive function gives one only to an winning
  input (biggest value) and zeroe to the other

10
Simulation Environment

• SimpleScalar Tool
• Some of SPEC2000 benchmarks
• Execute 100,000,000 instructions and dump
  conditional branch histories
• 5000 branch instructions are used for training
• Make all budgets for PHTs the same
• Floating point is 4 byte

11
Results
12
Hardware constraints

• Predictors must predict within a (few) cycle
• Gshare easy to achieve
• Perceptron Integer adders, possible
  alternative, more accurate if more layers
• Other advanced neural net Hard to implement,
  Floating point functional units,

13
Future Works

• Replace floating point weights and biases with
  scaled integer ones?
• Replace floating point function with
  approximately equivalent integer function, using
  Taylors series?
• Without budget consideration, what will be the
  best performance of advanced neural network
  methods?
• Look at codes carefully if there are mistakes

14
Conclusions

• There is not much of benefit using advanced
  neural networks on the same budget as Gshare,
  sometimes worse
• Elman Networks method is the best
• Hard to implement in hardware unless floating
  point computation are easy to do
• NN can be alternative predictors if well designed