• Aucun résultat trouvé

On the construction of the asymmetric Chudnovsky multiplication algorithm in finite fields without derivated evaluation

N/A
N/A
Protected

Academic year: 2021

Partager "On the construction of the asymmetric Chudnovsky multiplication algorithm in finite fields without derivated evaluation"

Copied!
6
0
0

Texte intégral

(1)

HAL Id: hal-01705865

https://hal.archives-ouvertes.fr/hal-01705865

Submitted on 13 Jun 2018

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Distributed under a Creative Commons Attribution - NonCommercial - NoDerivatives| 4.0

International License

On the construction of the asymmetric Chudnovsky

multiplication algorithm in finite fields without

derivated evaluation

Alexis Bonnecaze, Stéphane Ballet, Nicolas Baudru, Mila Tukumuli

To cite this version:

Alexis Bonnecaze, Stéphane Ballet, Nicolas Baudru, Mila Tukumuli. On the construction of the

asym-metric Chudnovsky multiplication algorithm in finite fields without derivated evaluation. Comptes

Rendus. Mathématique, Centre Mersenne (2020-..) ; Elsevier Masson (2002-2019), 2017, 355 (7),

pp.729 - 733. �10.1016/j.crma.2017.06.002�. �hal-01705865�

(2)

Contents lists available atScienceDirect

C. R.

Acad.

Sci.

Paris,

Ser. I

www.sciencedirect.com

Number theory/Computer science

On

the

construction

of

the

asymmetric

Chudnovsky

multiplication

algorithm

in

finite

fields

without

derivated

evaluation

Construction

effective

de

l’algorithme

asymétrique

de

multiplication

de

Chudnovsky

dans

les

corps

finis

Stéphane Ballet

a

,

Nicolas Baudru

b

,

Alexis Bonnecaze

a

,

Mila Tukumuli

a aAixMarseilleUniv,CNRS,CentraleMarseille,I2M,Marseille,France

bAixMarseilleUniv,CNRS,CentraleMarseille,LIF,Marseille,France

a

r

t

i

c

l

e

i

n

f

o

a

b

s

t

r

a

c

t

Articlehistory:

Received21November2016 Acceptedafterrevision2June2017 Availableonline16June2017 PresentedbytheEditorialBoard

TheChudnovskyalgorithmforthemultiplicationinextensionsoffinitefieldsprovidesa bilinearcomplexityuniformlylinearwithrespecttothedegreeoftheextension.Recently, Randriambololona hasgeneralized the method,allowingasymmetry intheinterpolation procedureand leadingtonewupperboundsonthebilinearcomplexity.Inthisnote,we describethe construction ofthisasymmetricmethodwithoutderived evaluation. To do this,wetranslatethisgeneralizationintothelanguageofalgebraicfunctionfieldsandwe giveastrategyofconstructionandimplementation.

©2017Académiedessciences.PublishedbyElsevierMassonSAS.Thisisanopenaccess articleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

r

é

s

u

m

é

L’algorithme de multiplication dans les corps finis de Chudnovsky a une complexité

bilinéaireuniformémentlinéaireenledegrédel’extension.Randriambololonaarécemment généralisé cette méthode en introduisant l’asymétrie dans la procédure d’interpolation et en obtenant ainsi de nouvelles bornes surla complexité bilinéaire. Dans cette note, nous décrivons la construction de cette méthode asymétrique sans évaluation dérivée. Pourcefaire,noustraduisonscettegénéralisationdanslelangagedescorpsdefonctions algébriques,etnousdonnonsunestratégiedeconstructionetd’implantation.

©2017Académiedessciences.PublishedbyElsevierMassonSAS.Thisisanopenaccess articleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

E-mailaddresses:stephane.BALLET@univ-amu.fr(S. Ballet),nicolas.BAUDRU@univ-amu.fr(N. Baudru),alexis.BONNECAZE@univ-amu.fr(A. Bonnecaze),

tukumulimila@gmail.com(M. Tukumuli).

http://dx.doi.org/10.1016/j.crma.2017.06.002

1631-073X/©2017Académiedessciences.PublishedbyElsevierMassonSAS.ThisisanopenaccessarticleundertheCCBY-NC-NDlicense (http://creativecommons.org/licenses/by-nc-nd/4.0/).

(3)

730 S. Ballet et al. / C. R. Acad. Sci. Paris, Ser. I 355 (2017) 729–733

1. Introduction

Letq beaprimepower,

F

q thefinitefieldwithq elementsand

F

qn thedegreen extensionof

F

q.Amongallalgorithms

ofmultiplicationsin

F

qn,thosebasedontheChudnovsky–Chudnovsky[6]methodareknowntoprovidethelowestbilinear

complexity.Thismethodisbasedoninterpolationonalgebraiccurvesdefinedoverafinitefieldandprovidesabilinear com-plexity,whichislinearinn.Theoriginalalgorithmusesonlypointsofdegree1,withmultiplicity1.BalletandRolland[4,5]

andArnaud[1]improvedthealgorithm,introducinginterpolationatpointsofhigherdegreeorhighermultiplicity.The sym-metryoftheoriginalconstructioninvolves2-torsionpointsthatrepresentanobstacletotheimprovementofupperbilinear complexity bounds. Toeliminate thisdifficulty, Randriambololona[8]allowed asymmetry inthe interpolation procedure, andthenPieltantandRandriambololona[7]derivednewbounds,uniforminq,ofthebilinearcomplexity.Unlikesymmetric constructions,noeffectiveimplementationofthisasymmetricconstructionhasbeendoneyet.When g

=

1,itisknown[3]

thatanasymmetricalgorithmcanalwaysbesymmetrized.However,forgreatervaluesofg,itmaynotbethecase.Thus,it isofinteresttoknowan effectiveconstructionofthisasymmetricalgorithm.Sofar,noeffectiveimplementationhasbeen proposedforsuchanalgorithm.

1.1. Multiplicationalgorithmandtensorrank

Themultiplicationoftwoelementsof

F

qn isan

F

q-bilinearapplicationfrom

F

qn

×F

qn onto

F

qn.Thenitcanbeconsidered

as an

F

q-linearapplication fromthe tensorproduct

F

qn

Fq

F

qn onto

F

qn. Consequently, itcan also be considered asan

element Tmof

F

qn

F q

F

qn



Fq

F

qn where



denotesthedual.When Tmiswritten

Tm

=

r



i=1

xi

yi

ci

,

(1)

wherether elements xi aswell asthe r elements yi areinthedual

F

qn of

F

qn whilether elementsci arein

F

qn,the

followingholdsforanyx

,

y

∈ F

qn:x

·

y

=



ri=1x

i

(

x

)

yi

(

y

)

ci.Thedecomposition(1)isnotunique.

Definition1.1.Everyexpressionx

·

y

=



ri=1xi

(

x

)

yi

(

y

)

cidefinesabilinearmultiplicationalgorithm

U

ofbilinearcomplexity

μ

(U)

=

r.Suchanalgorithmissaidsymmetricifxi

=

yiforalli.

Definition1.2. Theminimalnumberofsummands ina decompositionofthetensor Tm ofthemultiplicationiscalledthe

bilinearcomplexity(resp.symmetricbilinearcomplexity)ofthemultiplicationandisdenotedby

μ

q

(

n

)

(resp.

μ

symq

(

n

)

):

μ

q

(

n

)

=

min

U

μ

(

U

)

where

U

isrunningoverallbilinearmultiplicationalgorithms(resp.allbilinearsymmetricmultiplicationalgorithms)in

F

qn

over

F

q.

1.2. Organisationofthenote

InSection2,wegiveanexplicittranslationofthegeneralizationoftheChudnovskyalgorithmgivenbyRandriambololona

[8,Theorem3.5].TheninSection3,bydefininganewdesignofthisalgorithm,wegiveastrategyofconstructionand imple-mentation.Inparticular,thankstoasuitablerepresentationoftheRiemann–Rochspaces,wepresentthefirstconstructionof asymmetriceffectivealgorithmsofmultiplicationinfinitefields.Thesealgorithmsaretailoredtohardwareimplementation andthey allow computationstobe parallelized whilemaintaininga lownumberof bilinearmultiplications. InSection 4, wegiveananalysisofthenotasymptoticalcomplexityofthisalgorithm.

2. MultiplicationalgorithmsoftypeChudnovsky:generalizationofRandriambololona

In this section, we present a generalization of Chudnovsky-type algorithms, introduced in [8, Theorem 3.5] by Ran-driambololona, which is possibly asymmetric. Since our aim is to describe explicitly the effective construction of this asymmetric algorithm, we transformthe representation ofthis algorithm,initially made inthe abstract geometrical lan-guage,inthemoreexplicitlanguageofalgebraicfunctionfields.

Let F

/

F

q be an algebraic functionfield overthefinitefield

F

q ofgenus g

(

F

)

.Wedenote by N1

(

F

/

F

q

)

thenumberof

places ofdegreeone ofF over

F

q.IfD isadivisor,

L(

D

)

denotes theRiemann–Rochspaceassociatedwith D.Wedenote

by

O

Q the valuationring oftheplace Q andby FQ its residueclassfield

O

Q

/

Q ,whichisisomorphic to

F

qdeg Q,where

deg Q isthedegreeoftheplace Q .

Intheframeworkofalgebraicfunction fields,theresult[8,Theorem3.5]ofRandriambololonacanbestatedasin The-orem 2.1.Notethat wedo nottake intoaccountderived evaluations,sincewe arenot interestedinasymptoticresults.It means thatwedescribe thisasymmetric algorithmwiththedivisor G

=

P1

+ · · · +

PN wherethe Pi arepairwisedistinct

(4)

Let usdefine the following Hadamardproduct in

F

ql1

× F

ql2

× · · · × F

qlN, where the li’s denote positive integers, by

(

u1

,

. . . ,

uN

)

 (

v1

,

. . . ,

vN

)

= (

u1v1

,

. . . ,

uNvN

)

.

Theorem2.1.LetF

/

F

qbeanalgebraicfunctionfieldofgenusg over

F

q.Supposethatthereexistsaplace Q ofdegreen.Let

P =

{

P1

,

. . . ,

PN

}

beasetofN placesofarbitrarydegreenotcontainingtheplaceQ .Supposethatthere existtwoeffectivedivisorsD1

,

D2

ofF

/

F

qsuchthat:

(i) theplaceQ andtheplacesof

P

arenotinthesupportofthedivisorsD1andD2;

(ii) thenaturalevaluationmapsEifori

=

1

,

2 definedasfollowsaresurjective

Ei

:



L

(

Di

)

−→ F

qn



FQ

f

−→

f

(

Q

)

(iii) thenaturalevaluationmapdefinedasfollowsisinjective

T

:



L

(

D1

+

D2

)

−→ F

qdeg P1

× F

qdeg P2

× · · · × F

qdeg P N f

−→

(

f

(

P1

),

f

(

P2

), . . . ,

f

(

PN

))

Thenforanytwoelementsx

,

y in

F

qn,wehave:

xy

=

EQ

T|Im T1



T

E11

(

x

)



T

E21

(

y

)



,

whereEQ denotesthecanonicalprojectionfromthevaluationring

O

Q oftheplace Q initsresidueclassfieldFQ,

thestandard

compositionmap,T|Im T1 therestrictionoftheinversemapofT ontheimageofT ,Ei1theinversemapoftherestrictionofthemapEion

thequotientgroup

L(

Di

)/

ker Eiand



theHadamardproductin

F

qdeg P1

× F

qdeg P2

× · · · × F

qdeg P N;and

μ

q

(

n

)



N

i=1

μ

q

(

deg Pi

)

.

3. Effectivealgorithm

3.1. Methodandstrategyofimplementation

Theconstructionofthealgorithmisbasedonthechoiceoftheplace Q ,theeffectivedivisorsD1 andD2,thebasesof

spaces

L(

D1

)

,

L(

D2

)

and

L(

D1

+

D2

)

andthebasisoftheresidueclassfield FQ.

Inpractice, followingtheideas of [2],divisors D1 and D2 arechosen asplaces ofdegree n

+

g

1.Furthermore,we

requiresomeadditionalproperties,whicharedescribedbelow.

3.2. FindinggoodplacesD1,D2,andQ

Inordertoobtainthegoodplaces,weproceedasfollows:

– wedraw atrandom anirreducible polynomial

Q(

x

)

ofdegreen in

F

q

[

X

]

andcheck thatthispolynomialis primitive

andtotallydecomposedinthealgebraicfunctionfield F

/

F

q;

– wechooseaplace Q ofdegreen abovethepolynomial

Q(

x

)

;

– wechooseaplace Q ofdegreen amongtheplacesofF

/

F

q lyingabovethepolynomial

Q(

x

)

;

– wedrawatrandomaplace D1 ofdegreen

+

g

1 andcheckthatD1

Q isanon-specialdivisorofdegreeg

1,i.e.

dim

L(

D1

Q

)

=

0;

– wedrawatrandomaplace D2 ofdegreen

+

g

1 andcheckthatD2

Q isanon-specialdivisorofdegreeg

1,i.e.

dim

(

D2

Q

)

=

0.

3.3. Choosinggoodbasesofthespaces TheresiduefieldFQ.

Wechoosethecanonicalbasis

B

Q generatedbyaroot

α

ofthepolynomial

Q(

x

)

,namely

B

Q

= (

1

,

α

,

α

2

,

...,

α

n−1

)

.From

nowon,weidentify

F

qn to FQ,astheresidueclassfield FQ oftheplace Q isisomorphictothefinitefield

F

qn. TheRiemann–Rochspaces

L(

D1

)

and

L(

D2

)

.

Wechooseasbasisof

L(

Di

)

thereciprocalimage

B

Di ofthebasis

B

Q

= (φ

1

,

. . . ,

φ

n

)

of FQ bytheevaluationmap Ei,

namely

B

Di

= (

E

1

i

i

),

. . . ,

Ei1

n

))

.

Letusdenote

B

Di

= (

fi,1

,

...,

fi,n

)

with fi,1

=

1 fori

=

1

,

2. TheRiemann–Rochspace

L(

D1

+

D2

)

.

Notethatsince D1 andD2areeffectivedivisors,wehave

L(

D1

)

L(

D1

+

D2

)

and

L(

D2

)

L(

D1

+

D2

)

.

(5)

732 S. Ballet et al. / C. R. Acad. Sci. Paris, Ser. I 355 (2017) 729–733

Proposition3.1.LetD1,D2andQ beplaceshavingthepropertiesdescribedin(3.2).Considerthemap



:

L(

D1

+

D2

)

FQ such

that

(

f

)

=

f

(

Q

)

forf

L(

D1

+

D2

)

.Thereexistsavectorspace

M

ker



ofdimensiong suchthat

L

(

D1

+

D2

)

=

L

(

D1

)

L

r

(

D2

)

M

,

where

L

r

(

D2

)

issuchthat

L(

D2

)

= F

q

L

r

(

D2

)

and

denotesthedirectsum.Inparticular,ifg

=

0,then

M

=

K er



isequal

to

{

0

}

.

We chooseasbasisof

L(

D1

+

D2

)

thebasis

B

D1+D2 definedby

B

D1+D2

= (

f1

,

. . . ,

fn

,

fn+1

,

. . . ,

f2n+g−1

)

where

B

D1

=

(

f1

,

. . . ,

fn

)

isthebasisof

L(

D1

)

,

(

fn+1

,

. . . ,

f2n−1

)

isabasisof

L

r

(

D2

)

such that fn+j

=

f2,j+1

B

D2 with

B

D1 and

B

D2

definedinSection3.3and

B

M

= (

f2n

,

. . . ,

f2n+g−1

)

isabasisof

M

.

3.4. Productoftwoelementsin

F

qn

Letx

= (

x1

,

. . . ,

xn

)

andy

= (

y1

,

. . . ,

yn

)

betwoelementsof

F

qn givenbytheircomponentsover

F

qrelativetothechosen

basis

B

Q.Accordingtothepreviousnotation,wecanconsiderthatx andy areidentifiedasrespectively fx

=



n

i=1xif1,i

L(

D1

)

and fy

=



ni=1yif2,i

L(

D2

)

.

Theproduct fxfy ofthetwoelements fx and fyistheirproductinthevaluationring

O

Q.Thisproductliesin

L(

D1

+

D2

)

,since D1 andD2 areeffectivedivisors.Weconsiderthat x and y arerespectivelytheelements fx and fy embedded

intheRieman–Roch space

L(

D1

+

D2

)

,viarespectivelytheembeddings Ii

:

L(

Di

)

−→

L(

D1

+

D2

)

,definedby I1

(

fx

)

and

I2

(

fy

)

asfollows.If, fx and fyhaverespectivelycoordinates fxi and fyi in

B

D1+D2,wherei

∈ {

1

,

. . . ,

2n

+

g

1

}

,wehave:

I1

(

fx

)

= (

fx1

:=

x1

,

. . . ,

fxn

:=

xn

,

0

,

. . . ,

0

)

andI2

(

fy

)

= (

fx1

:=

y1

,

0

,

. . . ,

0

,

fyn+1

:=

y2

,

. . . ,

fy2n−1

:=

yn

,

0

,

. . .

0

)

.Now itis

clearthatknowingx (resp. y)or fx (resp. fy)bytheircoordinatesisthesamething.

Theorem3.2.LetPMsbetheprojectionof

L(

D1

+

D2

)

onto

M

s

=

L(

D1

)

L

r

(

D2

)

,andlet



bethemapdefinedasin

Proposi-tion 3.1.Then,foranyelementsx

,

y

∈ F

qn,theproductofx byy issuchthat

xy

=  ◦

PMs



T|Im T1



T

I1

E−11

(

x

)



T

I2

E−21

(

y

)



,

where

denotesthestandardcompositionmap,T|Im T1 therestrictionoftheinversemapofT ontheimageofT ,and



theHadamard productasinTheorem 2.1.

Wecannowpresentthesetupalgorithm(Algorithm 1),whichisdoneonlyonce.

Algorithm1Setupalgorithm.

INPUT: F/Fq, Q,D1,D2,P1,. . . ,PN.

OUTPUT: T and T−1.

(i) TherepresentationofthefinitefieldFq=<a>,wherea isafixedprimitiveelement.

(ii) ThefunctionfieldF/Fq,theplaceQ ,thedivisorsD1andD2andthepointsP1,. . . ,PNaresuchthatConditions(ii)and(iii)inTheorem 2.1are

satisfied.Inaddition,werequirethat1≤iNdeg Pi=2n+g−1.

(iii) RepresentFqn inthecanonicalbasisBQ= {1,α,α2,...,αn−1},whereFqn=<α>withαaprimitiveelementasinSection3.3.

(iv) Constructabasis(f1,. . . ,fn,fn+1,. . . ,f2n+g−1)ofL(D1+D2),where(f1,. . . ,fn)isthebasisofL(D1),(f1,fn+1,. . . ,f2n−1)thebasisofL(D2)

and(f2n,. . . ,f2n+g−1)thebasisofM,definedinSection3.3.

(v) ComputethematricesT andT−1.

(vi) Computethematrix.

Themultiplicationalgorithm(Algorithm 2)ispresentedhereafter.

4. Complexityanalysis

Intermsofnumberofmultiplicationsin

F

q,thecomplexity ofthismultiplicationalgorithmisasfollows:calculationof

z andt needs2

(

2n2

+

ng

n

)

multiplications,calculationofu needs

(

2n

+

2g

2

+

r

)

sup1≤ir

μq(i)

i bilinearmultiplications

andcalculation of2n

1 firstcomponents of w needs

(

2n

+

g

1

)(

2n

1

)

multiplications (remark that,inAlgorithm 2, we justhaveto computethe 2n

1 first componentsof w). Thecalculation ofxy needs n

+

g multiplications.The total complexityintermsofmultiplicationsisboundedby8n2

+

n

(

4g

5

)

+ (

2n

+

2g

2

+

r

)

sup

1≤i≤r

μq(i)

i .

The generalconstructionoftheset-upalgorithminvolvessome randomchoiceofdivisorshavingprescribedproperties overanexponentiallylargesetofdivisors.Togetapolynomiallyconstructiblealgorithmwithlinearcomplexity,oneneeds toconstructexplicitly(i.e.polynomially)pointsofcorrespondingdegreesn oncurvesofarbitrarygenuswithmanyrational points. Unfortunately,sofaritisunknown howtoproducesuch points(cf.[9,Section 4,Remark5] and[8,Remark 6.6]). Hence,theasymptoticcomplexityofsuchaconstructionisanopenproblem.

(6)

Algorithm2Multiplicationalgorithm. INPUT: x= (x1,. . . ,xn)and y= (y1,. . . ,yn). OUTPUT: xy. (i) Compute ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ z1,d1 . . . zn,dn zn+1,dn+1 . . . zN,dN ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ = ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ z1 . . . zn zn+1 . . . z2n+g−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ =T ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ x1 . . . xn 0 . . . 0 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ and ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ t1,d1 . . . tn,dn tn+1,dn+1 . . . tN,dN ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ = ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ t1 . . . tn tn+1 . . . t2n+g−1 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ =T ⎛ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎜ ⎝ y1 0 y2 . . . yn 0 ⎞ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎠ ,

whereNi=1di=2n+g−1,(zi,j,ti,j)∈ (Fqd j)2,(zi,ti)∈ (Fq)2and0 standsforthenullvector.

(ii) ComputetheHadamardproductu= (u1,d1,. . . ,uN,dN)= (u1,. . . ,u2n+g−1),whereui,di=zi,diti,di,inFqd1× Fqd2× · · · × FqdN asinTheorem 2.1.

(iii) Computew= (w1,. . . ,w2n+g−1)=T−1(u).

(iv) Extractw= (w1,. . . ,w2n−1)(instep(iii),justthe2n−1 firstcomponentshavetobecomputed).

(v) Returnxy= (w).

References

[1]N.Arnaud,Évaluationdérivées,multiplicationdanslescorpsfinisetcodescorrecteurs,PhDThesis,UniversitédelaMéditerranée,Institutde mathéma-tiquesdeLuminy,France,2006.

[2]S.Ballet,CurveswithmanypointsandmultiplicationcomplexityinanyextensionofFq,FiniteFieldsAppl.5 (4)(1999)364–377.

[3]S.Ballet,A.Bonnecaze,M.Tukumuli,OntheconstructionofellipticChudnovsky-typealgorithmsformultiplicationinlargeextensionsoffinitefields,J. AlgebraAppl.15 (1)(2016)1650005.

[4]S.Ballet,R.Rolland,Multiplicationalgorithminafinitefieldandtensorrankofthemultiplication,J.Algebra272 (1)(2004)173–185.

[5]S.Ballet,R.Rolland,Onthebilinearcomplexityofthemultiplicationinfinitefields,in:ProceedingsoftheConferenceArithmetic,GeometryandCoding Theory(AGCT2003),in:Semin.Congr.,vol. 11,SociétémathématiquedeFrance,2005,pp. 179–188.

[6]D.V.Chudnovsky,G.V.Chudnovsky,Algebraiccomplexitiesandalgebraiccurvesoverfinitefields,J.Complex.4(1988)285–316.

[7]J.Pieltant,H.Randriambololona,Newuniformandasymptoticupperboundsonthetensorrankofmultiplicationinextensionsoffinitefields,Math. Comput.84(2015)2023–2045.

[8]H.Randriambololona,BilinearcomplexityofalgebrasandtheChudnovsky–Chudnovskyinterpolationmethod,J.Complex.28(2012)489–517.

[9]I.Shparlinski,M.Tsfasman,S.Vladut,Curveswithmanypointsandmultiplicationinfinitefields,in:H.Stichtenoth,M.A.Tsfasman(Eds.),CodingTheory andAlgebraicGeometry,ProceedingsofAGCT-3Conference,Luminy,France,17–21June1991,in:Lect.NotesMath.,vol. 1518,Springer-Verlag,Berlin, 1992,pp. 145–169.

Références

Documents relatifs

Together with Debry [1] and with Sgobba and Tronto [5] the author proved that there are finitely many divisibility parameters such that the above Kummer degree can be expressed

use generalizations of Hensel’s lifting to compute Puiseux series [45] or for polynomial factorization in L[[X]][Y ] [44, 28], but no complexity analysis is given.. In [32],

In particular, using the definition field descent on the field with 2 or 3 elements of a Garcia- Stichtenoth tower of algebraic function fields which is asymptotically optimal in

Moreover they combined their notion of multiplication friendly codes with two newly introduced primitives for function fields over finite fields [11], namely the torsion limit

In Section 5 we analyze the complexity of the blockwise product for two polynomials supported by monomials of degree at most d − 1: if A contains Q in its center, then we show

The practical running time in these applications is dominated by the multiplication of matrices with large integer entries, and it is vital to have a highly efficient implementation

L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des

trade secrets and intellectual property rights are but a partial defence against disclosure and may be relied on only to the extent strictly necessary to protect