• 検索結果がありません。

Research on Combined Approaches of Graph Theory and Mathematical Programming for High-Level and Physical Synthesis of VLSI

N/A
N/A
Protected

Academic year: 2022

シェア "Research on Combined Approaches of Graph Theory and Mathematical Programming for High-Level and Physical Synthesis of VLSI"

Copied!
4
0
0

読み込み中.... (全文を見る)

全文

(1)

R e s e a r c h o n C o m b i n e d

A p p r o a c h e s o f G r a p h T h e o r y a n d M a t h e m a t i c a l P r o g r a m m i n g f o r

H i g h - L e v e l a n d P h y s i c a l S y n t h e s i s o f V L S I

Cong HAO

2017 4

(2)

2

A l o n g w it h e x p l o s i v e V L S I ap p li c a t i o ns i n m o d e r n t e c h n ol o g i es , V L S I d e s i g n us i n g s of t w a r e t o o ls be c om e s of t h e u t m os t i m p or t a nc e, w h i c h is r e f er r ed t o as e l ec t r o n ic d e s i g n a u t o m a t io n ( ED A ) . ED A d e s i gn f low, a l s o c a l l ed I C d es i g n f l ow, is a p r oc es s w h i c h c o ns i s t s o f s e v e r a l a u t om a t e d s t ep s t o ac c om p l i s h t h e d es i g n o f a n I C . Si n c e E D A t o o ls p l a y a n im p o r t an t r o l e i n d e v e l op i n g I C s w i t h hi g h p er f o r m a n c e, l ow c os t a nd f a s t t i m e- t o - m a r k et , t h e d e v e l op m en t o f h i g h ef f ic i e nc y a l g or i t h m s us ed i n ED A d e s i g n f l ow f or v ar i o us op t i m i z at i o n o b j ec t i v es a r e at t r a c t i n g m o r e a nd m o r e r e s ea r c h i n t er es t s . T h e g en e r al ED A d e s i g n f l ow c o nt a i n s s e v e r a l r ep r es e nt a t i v e s t e p s i nc l u d i n g h i g h l e ve l s y n t h e s i s ( H LS ) , l o g i c s y n t h es is , a n d p h y s ic a l s y n t h es i s , et c . , t o m e e t v ar i o us d e s i g n r e q u i r e m e n t s s u c h a s p o w e r, c h i p a r e a, c l oc k f r e q uen c y a nd s i g n a l d e l a y.

Si n c e m id - 1 9 7 0s , t h e i n ve s t i g a t io n s o f E D A d es i g n m e t ho d ol o g i es h a v e b e e n c a r r i ed o n f or d e c ad es , a n d s i g n i f i c a nc e ac h i e v em e nt s h a v e be e n m a d e f or a u t o m at ed d es i g ns i ns t e a d o f m a n u al d es i g ns . A lo n g w i t h t h e d e v e l op m e n t o f c o m m er c ial ED A t o o ls , E D A c om p a n i e s ar e a ls o g r ow i n g f a s t a n d c on t i n uo us l y. B es id es t h e ac h i e v em e nt s , t h e r e ar e s t i l l u ns o l v ed p r o b le m s , e it h er b ec a us e of t h e c o m p le x i t y of t h e p r o b l em s o r of t he r a p id l y g r o w i n g p r o bl e m s i z e. F or e x am p le i n H LS , t h e s ol u t i on q u a li t y i s s t il l l ow e r t h a n m a n u al d es i g ns , w h i c h ha s a l ar g e r oo m f or i m p r o ve m e n t ; f o r i nt e r c o n n ec t io n op t im i z at i o n, s om e im p or t a nt p r o bl e m s s u c h as t h e p o r t as s i g nm e nt p r o b l em , a r e i g n or ed a nd n ot b ei n g f u l l y i n v es t i g at e d . T h er e f o r e, a l g or it hm s w it h h i g h op t im a l it y a r e s t i l l e x p e c t e d , w h i c h e q u a l or e ve n e xc e l m a n u a l d e s i g n s . I n a d d i t io n , i n p h ys i c al s y nt h es i s , t o g et h er w i t h t h e g r ow i n g d es i g n a n d c i r c u i t s i z e, s o m e p r o bl em s b e c o m e e x t r em el y h u g e, w hi c h m a k e s i t d if f i c u lt t o b e s o l v ed u s i n g e x is t i n g al g o r it h m s . T h er e f o r e, a l g or i t hm s w i t h hi g h ef f ic i e nc y ar e e x p ec t ed t o h a nd l e l a r g e s c a le d p r o b l em s in p h ys i c al s y n t he s i s .

M o t i v at ed b y t h e e x is t i n g is s u es , i n t h is r es e ar c h, t w o t op i c s a r e t o b e s t ud i ed . O n e i s t h e l ow p ow er H L S t ec h n o lo g y, i n c l u d in g s c h ed u l i n g, r e s o u r c e al l o c a t i o n, bi n d i n g, a n d i n t e r c o n n ec t i on op t i m i z a t i o n. A n ot h e r is t h e p h ys i c a l s y nt h es is f o r 3D I C t ec h n ol o g y, w he r e t h e TS V i n s e r t i o n f o r 3D - I C is s t ud i ed . T h e y ar e s o l v ed b y c o m m o n t e c h n ol o g i es , i n c l ud i n g g r ap h t h e or y, m at h em a t i c a l p r o g r am m i n g a nd it e r a t i v e m et h o d s , w i t h t h e p r i nc i p le of im p r o v i n g t h e al g o r it h m op t i m al i t y a nd ef f ic i e nc y. T hi s t h es i s is o r g a n i ze d a s t h e f ol l ow s .

C h ap t er 1 , [ I n t r o d u c t i o n ] , f i r s t g i v es a br i ef i n t r o d u c t io n t o ED A d e s i g n f l ow. T he n s o m e e x i s t i n g p r o b le m s i n c u r r e n t ED A d e s i g n f l ow a r e

(3)

3

ad d r e s s ed . M o t i v a t e d b y t he e x is t i n g p r ob l em s , t h e p r in c ip l es of t h i s r e s e a r c h ar e g i v e n , w h ic h i s t o i m p r o v e a l g or i t hm o p t i m a l it y f o r s m al l a nd m e d i u m s i z ed p r o b le m s , a nd t o im p r o v e a l g or i t h m e f f ic i e nc y f or la r g e s i z ed p r o bl e m s . T h e n, t h e t hr e e t o p i c s of t h is r es e a r c h ar e i n t r o d u c e d, f o l lo w e d b y t he i r c o m m o n t ec h n o lo g i es. B es id es , n e c e s s ar y p r e li m in a r i e s ar e b r ie f l y g i v e n, i nc l u d i n g t h e b ac k g r o u n d k n ow l ed ge of H LS , 3D I C a n d T S V i ns e r t i o n p r o b l em . F in a l l y, t h e o r ga ni z a t io n o f t hi s t h es is i s g i v e n.

C h ap t er 2, [ A U n i f i ed S c h ed u l in g A p p r o ac h w it h M u l t i p l e Vd d or / a nd Vt h i n H L S] , d i s c us s es t h e d y n am ic a nd l e ak a g e p ow er m i n im i z at i o n p r o b le m us i n g m u l t i p le Vd d an d Vt h t ec h n o lo g y i n op er a t i o n s c h ed u l i n g, a s w e l l a s t h e c om b i n ed s c h ed u l i n g an d b i n d i n g m e t h od . A lt h o u g h t h is i s a m e d i u m s c a le d p r o b le m , t h e op t im a li t y of e x is t i n g a l g o r i t hm s ar e s t i l l n ot h i g h e n ou g h , s a y e q u al t o o r ev e n e x c e l m a n u a l d e s i g ns .

T h er e f o r e, i n t h is c h ap t e r, a hi g h l y o p t im i z e d u n if i e d s c h ed u l i n g ap p r o a c h w h i c h is ap p l ic a b le t o v a r i o us op t i m i z at i o n p r o bl e m s i s p r o p o s e d , i nc l u d i n g : ( 1 ) d y n am i c p ow e r a nd r es o ur c e us a g e c o - op t i m i za t i o n; ( 2) l ea k a ge p ow e r op t i m i za t i on ; a n d ( 3 ) d y n a m i c p ow er a nd l e ak a g e p ow e r c o - op t i m i za t i o n. To d ea l w it h d if f e r e n t o b j ec t i v es w i t h h i g h f l e x i b i li t y, t h r e e p r o b l em s a r e d i v i d e d i nt o t w o c om m o n s u b - p r o bl e m s, i nc l u d i n g d e la y as s i g n m en t a nd r es o ur c e d e ns it y v ar i a n c e m i n i m i z a t i on . T he n a v er t e x p o t e nt i a l b a s e d m o b i li t y a ll o c at i o n m od e l is p r op os ed t o s ol v e t w o s u b - p r o b l em s s im u lt a n e o us l y. O n t h e m o b i lit y g r a p h , t h e ne t w o r k s i m p l e x m e t ho d is ap p li e d , w h ic h op t im i z e s b ot h d y n am ic p ow er a n d r es o ur c e us a g e b y ad j u s t i n g t h e v e r t e x p o t e n t i a ls . T h e m o b i li t y g r a p h is it e r a t i v el y s o l v ed b y lo c a l s e ar c h u n t il t he a l g or i t h m m ee t s s t op c r i t e r i a. T h e c o m bi n e d s c h e d u l i n g an d bi n d i n g f o r p ow er m i n im i z at i o n i s a l s o i n v es t i g a t e t o f u r t h er r e d u c e o v er a l l p o w er c o ns um p t i o n .

E x p er im e n t a l r es u lt s s h ow t h a t , f o r d y nam i c p ow er a nd r e s o u r c e c o - op t i m i za t i o n, t h e p r op os e d s c he d u l i n g a p p r o a c h p r od u c es op t im u m s o l ut i o ns f o r a l l 6 b e nc h m a r k s w it h 1 5 g r o u p s of d a t a ; f o r l ea k a ge p ow er op t im i z at i o n it a ls o g r ea t l y e x c e ls t he l a t es t e x i s t i n g w o r k , b y 2 0 % l ea k a ge p ow e r r e d u c t i o n a nd 52 t i m e s s p e e d up .

C h ap t er 3, [ I nt e r c o n n e c t i o n A l lo c at ion be t w e e n F u nc t i o n a l U ni t s a nd R e g is t er s i n H L S] , d is c us s e s t h e i nt e r c o n n ec t i o n op t im i z at i o n t e c h n iq u es i n H LS , w hi c h h as n ot b e e n f u l l y i n ves t i g a t ed . A lt h o u g h it is a m e d i u m s c a le d p r o bl e m , i t s op t im a l it y h as a l a r ge i m p a c t o n c h i p p ow e r, ar e a a n d s i g n a l d e l a y. T h e r e f or e h i g h o p t im a l it y a l go r i t h m s ar e e x p e c t e d .

(4)

4

I n t h is c h a p t e r, t h e p o r t as s i g nm e nt p r o b le m f o r i n t e r c o n n ec t i on c o m p l e x it y r ed u c t i o n is d i s c us s ed . Fir s t , t h e p o r t a s s i g nm e n t p r o bl em i s f o r m u l at e d o n a c o ns t r a i nt g r ap h , a nd a p r ac t ic a l m e t h o d i s p r o p os ed t o f i n d a v a li d a nd i n it i a l s o l ut i o n. F o r s o l ut i o n op t im i z at i on , a n el e m e n t ar y s p a n ni n g t r ee t r a n s f or m a t i on b as ed l oc al s e a r c h al g o r it hm i s p r o p o s e d . To im p r o v e t h e ef f ic i e nc y o f o p t i m i z a t i o n, a m a t r i x f or m u l at i o n i s p r op os e d , a nd t h e n e t w or k s i m p l e x m e t h od i s a d op t e d w i t h p i v ot i n g o p er a t i o ns t o p e r f o r m o p t im i z a t i on . T h e t w o- s t ep s u c c e s s i v e p i v o t in g i s a ls o d i s c u s s e d t o f u r t h e r im p r o v e a l g o ri t hm e f f i c i e n c y a nd op t i m a li t y.

E x p er im e n t a l r es u lt s s h ow t h at o n t he r a nd o m l y g e ne r at e d t e s t c a s es , t h e m a t r i x b as ed al g o r it h m s h ow s h i g hes t s o l u t i o n o p t im a l it y a n d is f i v e t im es f as t e r t h an t h e el em e n t a r y t r a ns f o r m a t io n m et h od . O n t he r e a l h i g h - l e v e l s y nt h e s i s b e nc h m a r k s , t h e m a t r i x b a s e d m e t ho d r ed u c ed 1 4% i n t e r c o n n e c t i o ns , w h i l e t h e p r e v i o us a l g o r i t hm r e d u c e d 8% .

C h ap t er 4 , [ A M u l t i - L e ve l A l g o r it hm f o r 3D - I C T S V As s i g n m e n t i n P h ys i c a l S y nt h es i s ] , d is c us s e s t he p r o b le m of T hr o u g h S i li c on Vi a ( T S V ) i ns er t i on o n 3D - I Cs i n p h y s ic a l s y nt h e s i s . T h is is a h u g e s c a l e p r o bl em w h i c h r e q u ir es h i g h ef f ic i e nc y a l go r i t h m s w h il e t h e s o l ut i o n o p t im a l it y is n ot s ac r if i c e d .

T h r o u g h- s i li c o n v i a ( TS V ) as s i g nm e nt p r o bl em is o n e of t h e k e y d e s i g n c h a ll e n g es of 3- D I C w h i c h i s c r u c ia l t o t h e w ir e le n g t h a nd s i g n a l d e l a y. I n t h is c h a p t er t h e 3- D I C T S V as s i g n m en t i s f or m ul a t ed as a n I n t e g e r M i n im um C os t M ul t i C o m m od it y ( IM C M C ) p r o b le m , a nd a m u l t i - le v e l a l g or i t hm i s p r op os e d . I t c o a r s ens t he I M C M C ne t w o r k l e ve l b y l e v e l, a p p l i es a r o u g h f l ow a s s i g n m en t o n e ac h l e v e l o f c o a r s e n ed g r a p h, a nd g e ne r a t e s o n l y p r o m is i n g ed g es t o r ed u c e t he I M C M C n et w o r k s i z e. To im p r o v e t h e T S V a s s i g n m e nt q ua l i t y, a m i x e d s i n g l e a n d m u l t i- c o m m od i t y f l ow m e t h od is p r o p o s e d . B e s i d es , a n e x t e n d e d l a y er b y l a y er a l g or i t hm i s p r op o s e d t o op t im i z e t he TS V as s i g n m e n t .

E x p er im e n t a l r e s u lt s s h ow t h a t t h e p r op os e d m u lt i- l e v e l p r o p os a l ac h i e v es 3 7 t i m e s s p ee d up c om p ar ed t o t h e p r e v i o us w or k , a n d m e a nw hi l e r e d uc e s t h e t o t a l w i r e l e n gt h by 7. 0% . T h e e x t e nd ed la y e r b y la y e r op t im i z at i o n f u r t h er r e d uc es 0. 6% t o t a l w ir e l e n g t h .

C h ap t er 5 , [ C o n c l us i o ns ] , c o n c l ud es t hi s t h es is a n d d is c us s es s om e op e n is s ue s of t h is r es e ar c h, f o ll ow e d b y f u t ur e w o r k s .

参照

関連したドキュメント

Bellman., Fuzzy dynalnic programming and its extensions, TIMS. /Studies

The goal of our method is to simultaneously determine the operational scheduling i.e., when and which version of FUs each operation is assigned to and allocations i.e., the

Simultaneous Allocation and Binding Considering Multiplexers in High-Level Synthesis Yuko Hara-Azumi,†1,†2,†3,†4 Hiroyuki Tomiyama†3 and Hiroaki Takada †4 This paper

 Transform this convex cost flow to a minimum cost flow problem , and then solve it using the cycle canceling algorithm 3.1 Convex cost network construction Given a scheduling

The high-level parallel programning language Nano-2[3] was designed for shared memory multiprocessors and supports several features to write reliable parallel programs..

Given a scheduled flow graph with a constant number of execution traces and the corresponding conflict graph IC, the problem of finding a maximum compatible set of operations

[9], our work is based on Pluto [7], because of the following three advantages of Pluto: 1 simultaneous optimization of outer parallelism which leads to speedups

Then our scheduling/binding problem is, for given a CDFG GN, E, a clock period constraint T clk , a control step constraint S max , a set of functional units, huddle