19
I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim, Yunheung Paek and Doosan Cho** Seoul National University, Korea *UNIST, Korea **Sunchon National University, Korea

I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

Embed Size (px)

Citation preview

Page 1: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable

FabricsJonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park,

Yongjoo Kim, Yunheung Paek and Doosan Cho**Seoul National University, Korea

*UNIST, Korea**Sunchon National University, Korea

Page 2: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

2Udo KebschullUniversity of Heidelberg

Outline• CGRA & Augmentation

• Overall Design Flow• Our Approach (I2CRF)

• Problem definition(Inexact graph matching)• Mapping with A* search

• Experiment• Conclusion

Page 3: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

3Udo KebschullUniversity of Heidelberg

Reconfigurable Architecture• Reconfiguration is emerging

• increasing needs for flexible and high speed computing fabrics

• CGRAs (Coarse-Grained Reconfigurable Architectures)

• operation level granularity• high performance• S/W development is easy

MorphoSys ADRES

Page 4: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

4Udo KebschullUniversity of Heidelberg

Augmentation• General CGRA - Mapping

• CGRA Arch. + Applications Configurations

• Application specific CGRAs - Synthesis • Applications New Arch. + Configurations

• Augmentation• Base CGRA + Applications New Arch.+Configurations

• Customizable Features• The number of PEs• The set of PE operation• Heterogeneity or Homogeneity• Memory subsystem architectures• Interconnection network

Interconnect Exploration for Energy Versus Performance Tradeoffs for Coarse Grained Reconfigurable Architectures, TVLSI 200914% (130nm) 30%(45nm)

Energy consumption

Page 5: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

5Udo KebschullUniversity of Heidelberg

Overall design flow - I2CRF

Kernel

Evaluation

Application-Specific Reconfigurable Architecture

Arch ExtensionMapping (A* Search for Minimum-Cost Edit

Path)

+(Accum.)

I2CRF (Incremental Interconnect Customization for Reconfigurable Fabrics )

Base CGRA

Interconnections

Not Satisfied

Vertex Clustering

Page 6: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

6Udo KebschullUniversity of Heidelberg

I2CRF• Incremental architecture change by adding

interconnections to the base architecture• Strengths

• Regularity is maintained through the base architecture

• But provides specialization for the target applications

• Fast specialization and no limitation for design space

• The architecture change occurs while kernel is mapped.

Page 7: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

7Udo KebschullUniversity of Heidelberg

The difference Compared with general mapping

PE 1

PE 2

PE 3

PE 4

PE 5

PE 6

1

6

3

4 5

2

1 2

6

43

• Existing application mapping for CGRA • Find a graph X C that is isomorphic to K

• Augmentation and Mapping• Find the a graph Y that is isomorphic to K and a subset of C`

which is most similar to C

Kernel graph, K Base CGRA graph, C

5

×

1 2

6

43

5

General Mapping

Augmentation and Mapping

Page 8: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

8Udo KebschullUniversity of Heidelberg

Problem Definition - Inexact Graph Matching Problem

• How to find C which is most similar to C0 : Inexact graph matching• Similarity between two graph can be measured by calculating the

cost of graph edit path• Edit path is the set of edit operations that transform G1 into

another G2• Edit operations

– Node(or edge) substitution : NS, ES ( identical or non-identical )– Node(or edge) insertion : NI, EI– Node(or edge) deletion : ND, ED– All the other edit operations are induced by Node substitution.

1 2

3 5

6 7

4

a b c

d e f

g h i

NS1 e2 a3 h4 d5 b6 g7 f

e1

a2

h3

b5

g6

d4f7

Identical ESNon-identical ES & NI

ED EI<G1><G2>

Page 9: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

9Udo KebschullUniversity of Heidelberg

Graph Edit Cost Model• Ce - The cost of Edge deletion

• Interconnection insertion cost• Cv - The cost of Node insertion

• Routing PE insertion cost

• Routing PE can replace interconnection insertion in case there are extra PEs

• Do not need augmentation– can reduce the amount of architecture extension

• Cv is much cheaper than Ce

Page 10: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

10Udo KebschullUniversity of Heidelberg

A* Search for Min Cost Edit Path• Inexact graph matching problem is NP-complete

How to search the mapping space for the min cost path : A* Search algorithm• Root : Kernel graph• Leaf : Sub-CGRA graph• s : current mapping state• g(s) : The sum of the costs(Ce, Cv) of the graph

edit operations from root to current state s• h(s) : The estimated cost from current state s to

a leaf state• Assessment of the partial mapping s

• g(s) + h(s)

Page 11: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

11Udo KebschullUniversity of Heidelberg

Vertex Scattering • Make clusters of vertex and assign each cluster to

row• Strengths of Vertex scattering

• Search space reduction• Considering shared resource constraints

2 1

34

PE 1

PE 2

PE 4

PE 5

PE 3

PE 6

2 1

345

5

2

1

3

4 5

KernelClustering & Row assignment Final

mapping

Row 1

Row 2

Page 12: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

12Udo KebschullUniversity of Heidelberg

h(s) & Vertex Scattering • Heuristic function, h(s) …

• guides the fast search of mapping space• needs cost estimation methods

• Detecting difficult-to-map edges• After vertex scattering• Forks, Over-length edges cannot be mapped to a mesh

without routing PE or a custom interconnection links• H(s) # of forks & over-length edges (=Nr )

• Unroutable difficult-to-map edge (c1) has more cost than routable (c2)

1

2

3

5

6

4 7

Page 13: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

13Udo KebschullUniversity of Heidelberg

ExamplePE 1

PE 2

PE 4

PE 5

PE 3

PE 6

c1 = cv = 1c2 = ce = 3

1 4

32

1 4

32

s=0{ }

s=1{(11)}

s=2{(12)}

s=4{(42)}

s=5{(43),($2)}

s=8{(24)}

s=7{(25)}

s=3{(13)}

g( s ) + h( s ) = 0 + 1

s=6{(26)}

s=9{(33),($5}

0+1 0+1 0+1

0+1 1+1

0+10+10+1

4+0 s=10{(35),($4)}

1+0

1 4

3 23

Page 14: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

14Udo KebschullUniversity of Heidelberg

Experimental Setup• We test I2CRF on a CGRA called

RSPA• mesh base interconnection• Each row has 2 shared

multipliersEach row can perform 2 loads and 1 store

• PE can be used for routing• Benchmarks from

• Livermore loops, MultiMedia and DSPStone

• Comparison to Mesh, 1-hop, Diagonal, and Mixed

Page 15: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

15Udo KebschullUniversity of Heidelberg

Performance Improvement

• IPC of 16 is equivalent to 100% utilization• PE utilization and the IPC are increased by more than 70% on

average compared to Mesh or by 41% on average compared to Mixed

compress

LowPass

waveletca

lc1ca

lc2ca

lc3Sobel

SORbdist dct

dist2

hydroICCG

inner_product

n_complex_

up...

prewitt SADsta

teG.M

.0

2

4

6

8

10

12

14

16

18Mesh Diagonal 1-Hop Mixed I2CRF

IPC

(P

E U

tiliz

atio

n, 1

6=10

0%)

Page 16: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

16Udo KebschullUniversity of Heidelberg

Customization Overhead

• Through our interconnection increment, …• # of new interconnection links is very small• Very marginal increase in the overall Mux complexity

Page 17: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

17Udo KebschullUniversity of Heidelberg

Optimization Time

• Find competitive custom interconnection architecture with configuration in reasonable time.

1

10

100

1000

10000

Tim

e (s

ec)

Page 18: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

18Udo KebschullUniversity of Heidelberg

Conclusion• We presented an interconnection customization

method for CGRAs • Our method exploits the similarity between the

interconnection customization problem and inexact graph

• Non-homogeneous extensions to a base interconnection architecture may present some challenges and possibly penalty in back-end VLSI design matching

• We plan to find out the extent of the difficulty due to the non-homogeneity as well as find novel ways to mitigate any impact if necessary

Page 19: I2CRF: Incremental Interconnect Customization for Embedded Reconfigurable Fabrics Jonghee W. Yoon, Jongeun Lee*, Jaewan Jung, Sanghyun Park, Yongjoo Kim,

19Udo KebschullUniversity of Heidelberg

Thank you for your attention!