Space-Optimal, Wait-Free Real-Time Synchronization

Size: px
Start display at page:

Download "Space-Optimal, Wait-Free Real-Time Synchronization"

Transcription

1 1 Space-Optmal, Wat-Free Real-Tme Synchronzaton Hyeonjoong Cho, Bnoy Ravndran ECE Dept., Vrgna Tech Blacksburg, VA 24061, USA E. Douglas Jensen The MITRE Corporaton Bedford, MA 01730, USA Abstract We consder wat-free synchronzaton for the sngle-wrter/multple-reader problem n small-memory embedded real-tme systems. We present an analytcal soluton to the problem of determnng the mnmum, optmal space cost requred for ths problem, consderng a-pror knowledge of nterferences the frst such result. We also show that the space costs requred by prevous algorthms can be obtaned by our analytcal soluton, whch subsumes them as specal cases. We also present a watfree protocol that utlzes the mnmum space cost determned by our analytcal soluton. Our evaluaton studes and mplementaton measurements usng the SHaRK RTOS kernel valdate our analytcal results. I. INTRODUCTION Most embedded real-tme systems nvolve mutually exclusve, concurrent access to shared data objects, resultng n contenton for those objects. Resoluton of the contenton drectly affects the system s tmelness, and thus the system s behavor. Mechansms that resolve such contenton can be broadly classfed nto: (1) lock-based e.g., Prorty Inhertance and Celng protocols [1], Stack Resource Polcy [2], DASA [3]; (2) wat-free e.g., protocol [4], Chen s protocol [5], [6], [7]; and (3) lock-free e.g., [8].

2 2 Lock-based protocols have several dsadvantages such as seralzed access to shared objects, resultng n reduced concurrency and thus reduced resource utlzaton [8]. Further, many lockbased protocols typcally ncur addtonal run-tme overhead due to ncreased context swtchng between actvtes blocked on shared objects (.e., blockers ) and actvtes that hold locks of those objects (.e., lock holders ). The ncreased context swtchng occurs when lock-based protocols preempt the currently executng blocker, execute the lock holder untl the holder releases the lock, and then resume the blocker s executon. Another dsadvantage of usng locks s the possblty of deadlocks that can occur when lock holders crash, causng ndefnte starvaton to blockers. Further, many (real-tme) lock-based protocols requre a-pror knowledge of the celngs of the locks [1], [2], whch may be dffcult to obtan n some applcaton contexts. Furthermore, OS data structures (e.g., semaphore control blocks) must be a-pror updated wth that knowledge, resultng n reduced flexblty (e.g., recomplaton to accommodate new actvtes) [8]. These drawbacks have motvated research on wat-free and lock-free object sharng n realtme systems. Wat-free protocols use multple nternal buffers 1 (e.g., a crcular buffer) for wrters and readers [4]. For the sngle-wrter/multple-reader (or SWMR) problem, wat-free protocols typcally use multple buffers for the shared object, where the number of buffers used s proportonal to the maxmum number of tmes the readers can be nterfered by the wrter, when the readers are readng. The maxmum number of nterferences of a reader bounds the number of tmes the wrter can update the object whle the reader s readng. Thus, by usng as many buffers as the worst-case number of nterferences of readers, the readers and the wrter 1 We use the term nternal here to explctly ndcate that a sngle wat-free buffer nternally uses multple buffers for ts atomc operatons. In ths paper, the buffers and the space cost mplctly mean the nternal buffers and the cost of the nternal buffers, respectvely, unless otherwse noted.

3 3 can contnuously read and wrte n dfferent buffers, respectvely, and avod nterference. Lock-free protocols allow readers to concurrently read whle the wrter s wrtng (wthout acqurng locks), but the readers check whether ther readng was nterfered by the wrter. If so, they read agan. Thus, a reader contnuously reads, checks, and retres untl ts read becomes successful. Snce a reader s worst-case number of retres depends upon the worst-case number of tmes the reader s nterfered by the wrter, the addtonal executon-tme overhead ncurred for the retres s bounded by the number of nterferences. Both wat-free and lock-free protocols ncur addtonal costs wth respect to ther lock-based counterparts. Wat-free protocols generally ncur addtonal space costs due to ther multple buffer usage, whch s nfeasble n many small-memory, embedded real-tme systems. Lockfree protocols generally ncur addtonal tme costs due to ther retres, whch s antagonstc to tmelness optmzaton. Pror research have shown how to mtgate these space and tme costs, so that they are feasble for embedded real-tme systems. An excellent survey of ths pror research can be found n [7]. To provde context for our work, we summarze some mportant efforts here: In [4], Kopetz and Resnger present one of the earlest wat-free protocols, where buffer szes n proportonal to worst-case nterferences are used. In [8], Anderson et al. show how to bound the retry loops of lock-free protocols through judcous schedulng. In [5], Chen and Burns present one of the most space-effcent wat-free protocols, where the worst-case preemptons need not be a-pror known. In [9], Sundell and Tsgas descrbe a wat-free protocol for the multple-wrter/multple-reader problem. In [7], Huang et al. mprove the tme and space costs of Chen s protocol. In ths paper, we focus on wat-free synchronzaton for the SWMR problem n small-memory, embedded real-tme systems. We focus on wat-free, as opposed to lock-free, as majorty of the lock-free protocols have hgh computatonal costs [7]. We consder the SWMR problem, as t

4 4 occurs n most embedded real-tme systems [7], and focus on mnmzng ts space costs. We present an analytcal soluton to the problem of determnng the mnmum number of buffers that s requred to ensure the safety and orderlness of wat-free synchronzaton n SWMR. We call ths problem, Wat-Free Buffer sze decson Problem (or WFBP). Note that the optmalty n space that we provde s on the requred number of nternal buffers, and does not nclude the control varables needed for the wat-free protocol s operaton. Ths s because the space cost of nternal buffers domnates that of the control varables, especally when the data sze becomes larger. We prove that our soluton to WFBP subsumes the number of buffers requred by prevous wat-free protocols ncludng Chen s [5] and [4] protocols as specal cases. We analytcally dentfy the condtons under whch our protocol needs less (and equal) number of buffers than other protocols. Further, we present a wat-free protocol that utlzes the mnmum buffer requrement determned by our soluton. To determne the buffer requrements under a broad range of reader/wrter scenaros, we conduct numercal evaluatons. We also mplement our protocol n the SHaRK RTOS [10]. Our evaluatons and mplementaton measurements confrm our soluton to WFBP and valdate our analytcal results. Thus, the paper s contrbutons nclude the analytcal soluton that we present for WFBP and the wat-free protocol that uses the concomtant mnmum number of buffers. Among the class of wat-free protocols that consder a-pror knowledge of nterferences, our optmal space lower bound s the frst such bound that s analytcally establshed. The rest of the paper s organzed as follows: We present our analytcal soluton to WFBP and our wat-free protocol n Secton II. In Secton III, we formally compare our protocol wth Chen s and protocols. We numercally evaluate our protocol n Secton IV, and report our mplementaton experence n Secton V. We conclude the paper n Secton VI.

5 5 II. A SPACE-OPTIMAL WAIT-FREE PROTOCOL A wat-free protocol solves the asynchronous sngle-wrter/multple-reader problem by ensurng that each reader accesses the shared object wthout any nterference from the wrter. To realze the wat-free mechansm, the protocol must hold two propertes: safety and orderlness [5]. The safety property ensures that the shared object does not become corrupted durng readng and wrtng. The orderlness property ensures that all readers always read the latest data that s completely wrtten by the wrter. The basc dea to acheve the two propertes s rooted n the three-slot fully asynchronous mechansm for the sngle-reader/sngle-wrter problem [11]. For ths problem, Chen et al. show that three buffers are requred to keep the latest completely updated buffer for the next readng, whle a wrter and a reader are occupyng buffers respectvely. Ths mechansm allows that a reader can always obtan data from the buffer slot that s last completely updated, whle the wrter s wrtng the new verson of the shared data [5]. The buffers needed for the sngle-wrter/multple-reader problem consst of three types: buffers for readers, a buffer for the latest wrtten data, and a buffer for the next wrte operaton. The buffers for readers must satsfy safety.e., suffcent buffers must be avalable to avod nterference between readng and wrtng. However, ths does not mply that we need as many buffers as there are readers. The two buffers for wrtng are requred to realze orderlness.e., the latest wrtten data must be saved so that a newly actvated reader can access t at any tme. In addton, the latest wrtten data must be kept untl the wrter completely wrtes the next data nto another buffer. We now dscuss how to determne the mnmum number of buffers that are needed for the sngle-wrter/multple-reader problem n the followng subsectons:

6 6 A. Protocol Structure and Task Model Fgure 1 shows a wat-free protocol s common mplementaton. W.2 and R.2 show the code sectons of the wrter and a reader that wrte and read data, respectvely. W.1 s the code secton where the wrter decdes on the buffer for wrtng, and updates a control varable that ndcates the selected buffer. W.3 s the code secton where the wrter ndcates completon of wrtng and the buffer that has the latest data. In R.1, the reader checks for the latest data to read. (a) Wrte (b) Read Fg. 1. Typcal Wat-Free Implementaton The buffer sze requred for the protocol [4] and the mproved protocols n [7] s determned based on the temporal propertes of tasks. These pror works consder the perodc task model, where tasks concurrently share data objects. Aperodc tasks are handled by a perodc server, so the perodc model s not a lmtng assumpton. Assumng that all deadlnes are met (.e., durng under-load stuatons and precludng overloads), the maxmum number of preemptons of the reader by the wrter task n the worst-case can be obtaned. We consder the same task model. B. Number of Buffers n Use We ntroduce some notatons for convenence, most of whch are smlar to those n [5]. We denote the total number of readers as M and the th reader as R. The reader R s j th nstance

7 7 of readng s denoted as R [j]. The wrter s k th wrtng nstance s denoted as W [k]. R [j] (op) stands for a specfc operaton of R [j]. For example, R [j] (READIN G[] = 0) mples the executon of one statement n Chen s algorthm [5]. W [k] (op) also stands for the operaton n W [k]. If R [j] reads what W [k] wrtes, we denote t as w(r [j] ) = W [k]. As prevously mentoned, safety and orderlness can be acheved wth multple buffers for readers, one buffer for the latest wrtten data, and another buffer for the next wrtng. Fg. 2. Number of Buffers n Use Suppose we have 4 readers and 1 wrter, as shown n Fgure 2. At tme t 1, w(r 1 )=W [2], w(r 2 )=W [2], and w(r 3 )=W [1]. Ths mples that two buffers are beng used by the readers. In addton, one buffer s requred to store and save the latest completely wrtten data by W [4], and another s needed for the next wrtng operaton by W [5]. Thus, four buffers are beng used n total, at tme t 1. At tme t 2, w(r 1 )=W [6], w(r 2 )=W [5], w(r 3 )=W [4], and w(r 4 )=W [6]. The latest wrtten data s by W [6], and W [7] s the next operaton. Thus, the total number of buffers used at tme t 2 s four, whch s the mnmum number requred at t 2 for ensurng safety and orderlness propertes. The basc ntuton for determnng the mnmum number of buffers s to construct a worstcase where the requred number of buffers s as large as possble, when the maxmum possble number of nterferences of all readers wth the wrter occurs. We map ths problem to a problem

8 8 called the Dverse Selecton Problem (or DSP) and then solve t. C. Dverse Selecton Problem The DSP denoted as D(R, R( x)) s defned wth the problem range R and the range vector R( x) of all elements n the vector x. R has the lower and upper bounds defned as [l, u]. Each element x n the vector x has the range r =[l, u ]. The soluton to the problem D s represented as a vector x =< x 1,..., x M > where the vector sze n( x) s M. Every x must satsfy ts range constrant r and the problem range constrant R. We defne { x} as a set ncludng all elements of x, but wthout duplcates. Thus, the sze of { x}, n({ x}), s less than or equal to n( x). The objectve of DSP s to determne the maxmum n({ x}) by selectng x, satsfyng all range constrants as dversely as possble. Gven a vector, v =< v 1,..., v,... >, we denote the number of v s havng k value as H( v, k), and the maxmum value among all v elements as T op( v). Gven D(R, R x ), the optmal soluton of D when R = [t 1, t 2 ] s denoted as n max [t 1,t 2 ] ({ x}). For example, f v =< 1, 2, 2, 2, 6 > then, H( v, 2) = 3 and T op( v) = 6. An easy approach to solve DSP s by consderng all possble cases. The number of all possble cases s n(r 1 )... n(r M ), where the R of the problem s gven as < r 1,...r M >. By consderng all cases, we can select a vector x that maxmzes n({ x}). However, such an approach would be computatonally expensve. A more effcent approach to solve DSP can be found by an nductve strategy. Consder a DSP D(R, R( x)), where R = [1, ] and R=< [1, u 1 ],..., [1, u M ] >. If all lower bounds of elements of x are 1, we can defne the upper bound vector u =< u 1,..., u M > nstead of R, for convenence. In the rest of the paper, we call the problem defned by D(R, u) as DSP. Ths s smply because ths assumpton s well-mapped to the problem of decdng the mnmum buffer

9 9 sze for the wat-free protocol. Fg. 3. An Inductve Approach to DSP The soluton to the problem D(R, u) can be represented as n max [1,T op( u)] ({ x}). The dea to decompose the problem s shown n Fgure 3. If the soluton to the problem D([6, 12], u) can be derved from the soluton to the problem D([7, 12], u), we can nductvely determne the fnal soluton to the problem D([1, 12], u). Theorem 2.1 (DSP for the Wat-Free Protocol): In the DSP D(R, u) wth R = [1, N], n max [t+1] ({ x}) = n max ({ x}) + 1, f t+1 k=0 n max ({ x}), otherwse H( u, N k) > nmax ({ x}) where N = T op( u), =[N t, N], and 0 t < N. When t = 0, the n max [0] ({ x}) = 1. Proof: Assume that we have the soluton to the problem D([N t, N], u). When ths problem s extended to D([N (t + 1), N], u), the ranges of several varables x overlap wth the problem range [N (t + 1), N]. The number of newly added varables that we need to consder s H( u, N (t + 1)). When the problem range s extended by 1, the maxmum possble ncrement of n max [t+1] ({ x}) s 1. The ncrement happens only f the number of all x whch have

10 10 ther r overlapped wth [N (t + 1), N] s greater than n max ({ x}). In other words, ths happens when new elements appear n the extended problem scope, or there s an element duplcated wthn [N t, N] at the prevous step. Otherwse, n max [t+1]({ x}) has no change from before. The ncrement means that the value of one element s determned as dversely as possble. The proof s by nducton on t. Bass. We show that the theorem holds when t = 0. When the problem s D([T op( u), T op( u)], u), there must be at least one element x wth the range [1, T op( u)], and the maxmum possble value of n max [0] ({ x}) s 1. Hence, the bass for the nducton holds. Inducton step. Assume that the theorem holds true when R = [N t, N]. We arrve at the optmal soluton of D([N (t + 1), N], u) wth the optmal soluton of D([N t, N], u) as n the base step. Suppose that the derved soluton n max [t+1]({ x}) s not optmal. Then, there must exst another optmal soluton n max [t+1] ({ x} ). Clearly, n max [t+1] ({ x} ) s greater than n max [t+1]({ x}). Now, there are two possble cases: Case 1. If H( u, N (t + 1)) > n max ({ x}), then n max ({ x}) s nmax({ x}) + 1, whch s less than n max [t+1] ({ x} ). Therefore, n max ({ x}) < n max [t+1] ({ x} ) 1. Ths means that there exsts another { x} that has more than n max optmal. [t+1] ({ x}) elements. Ths contradcts the assumpton that n max ({ x}) s Case 2. If H( u, N (t + 1)) = n ({ x}), then n max [t+1]({ x}) s nmax ({ x}), whch s less than n max [t+1] ({ x} ). Snce no element s range becomes newly overlapped and no element has ts duplcate, n max [t+1] ({ x} ) = n max ({ x} ). Ths means that there exsts another n max ({ x} ), whch s greater than n max ({ x}). Ths contradcts the assumpton that n max ({ x}) s optmal. Theorem 2.2 (Soluton Vector for the DSP): In the DSP D(R, u) wth R = [1, N], { x} {N (t + 1)}, f t+1 k=0 H( u, N k) > nmax ({ x}), { x} [t+1] = { x}, otherwse

11 11 where N = T op( u), =[N t, N], and 0 t < N. When t = 0, { x} [0] = {N}. Proof: By Theorem 2.1, { x} can be constructed by addng {N (t + 1)} whenever n max t ({ x}) ncreases by 1. Note that ths { x} s one of the soluton vectors. D. Smlarty to WFBP The DSP has smlarty wth the Wat-Free Buffer sze decson Problem (or WFBP). In ths problem, we are gven M readers and ther maxmum nterferences as < N1 max,..., NM max >. The objectve of WFBP s to determne the worst-case maxmum number of buffers. Fg. 4. A Worst-Case of the WFBP Fgure 4 llustrates how to construct the worst-case where the requred number of buffers are as large as possble wth an example. For convenence, the ndex of the wrter s reversed compared wth Fgure 2. In ths example, R 1 s maxmum nterference s 5, whch s llustrated n a lne. It means that w(r 1 ) may belong to the set {W [1],...,W [6] }. We assume that the worst-case happens at tme t between W [2] and W [1], where W [2] wrtes the latest completely wrtten data, and W [1] s the next wrtng operaton for whch another buffer s needed. For ths reason, we restate WFBP as determnng x =< w(r 1 ),..., w(r M ) > that wll maxmze n({ x} {W [1], W [2] }), where w(r ) {W [1],..., W [N max +1] }. If we abbrevate W [j] as j, the problem s redefned as determnng x =< x 1,..., x M > that wll maxmze n({ x} {1, 2}),

12 12 where x {1,..., N max + 1}. Ths s equvalent to DSP except that n({ x} {1, 2}) s used as the objectve to maxmze, nstead of n({ x}). Therefore, the fnal soluton { x} of a gven WFBP s obtaned wth a sum of the soluton from a mapped DSP and a set {1, 2}. We clam that ths s correct, because the algorthm for DSP that we propose s desgned to fnd { x} whch does not have 1 and 2 as ts elements, f possble. We can guarantee that n ths way, even f the soluton from DSP s summed wth {1} or {2}, t s stll for the worst-case. Corollary 2.3 (Space Optmalty): If a soluton to the WFBP can be obtaned, then t must be the mnmum and space-optmal buffer sze that satsfes the two propertes, safety and orderlness. Proof: The soluton s the number of buffers needed n the worst-case of the gven problem. Even wth one less buffer than the obtaned soluton, we cannot realze all readng and wrtng, and stll satsfy safety and orderlness. Hence, the soluton to the WFBP s the mnmum and space-optmal. E. Algorthm for WFBP We now present an algorthm, Algorthm 1, to solve the WFBP based on the prevous sectons. The algorthm nputs nclude the number of readers M and the maxmum nterference N max []. The sum and the functon doesexst(t) correspond to H( u,...) and H( u, t) n Theorem 2.1. To reduce the tme complexty of doesexst(t), we sort all N max [] before the man loop. doesexst(t) uses a statc varable, and does not search the entre array N max [] each tme. The flag on ndcates whether or not the DSP soluton ncludes. If t does not nclude 1 or 2, the requred buffer sze for the WFBP soluton, n, s ncremented. The tme complexty of ths algorthm s O(MlogM + N max ). We beleve that ths cost s

13 13 Algorthm 1: Algorthm for WFBP 1 nput : # of readers M; max nterference N max [M] 2 output : requred buffer sze n 3 sum=n=0; 4 on 1 =on 2 =false; 5 for = 1 to M do N max []++; 6 sort( N max [1,...,M] ); 7 for t=n max [1] to 1 do 8 sum += doesexst( t, N max [1,...,M] ); 9 f sum>n then 10 n++; 11 f t=2 then on 2 = true; 12 f t=1 then on 1 = true; f on 2=false then n++; f on 1=false then n++; reasonable, as the algorthm s run off-lne for determnng the buffer needs. F. A Wat-Free Implementaton The protocol uses a crcular buffer to realze wat-free synchronzaton. The dea behnd the crcular buffer s that whle a wrter crcularly accesses the buffers, the readers follow the wrter. However, we cannot use the crcular type of buffer because a wrter n our protocol needs to determne a safe buffer, whch can be any of the buffers. The same stuaton arses wth Chen s protocol, where the wrter can access anywhere. Thus, to mplement our protocol, we slghtly modfy Chen s protocol. Our mplementaton scheme s shown n Algorthms 2 and 3. In Algorthms 2 and 3, the GetBuf() functon searches the empty buffer to wrte to the buffers assgned by Algorthm 1. Compared wth the mplementaton n [7], our approach does not need to mplement separate protocols for fast readers and slow readers. Addtonally, we acheve the speed mprovement by reducng the requred buffer sze, whch reduces the number of teratons n GetBuf() s loop, compared wth the orgnal Chen s protocol [5].

14 14 Algorthm 2: Modfed Chen s Protocol for Wrter 1 Data: BUFFER [1,...,NB](NB: # of buffers) ; READING [1,...,n] (n: # of readers) ; LATEST 2 GetBuf() 3 begn 4 bool InUse [1,...,NB]; 5 for =1 to NB do InUse []=false; 6 InUse[LATEST ]=true; 7 for =1 to n do 8 j = READING []; 9 f j 0 then InUse [j]=true; 10 =1; whle InUse [] do ++; 11 return ; 12 end 13 Wrter() 14 begn 15 nteger wdx, ; 16 wdx = GetBuf(); 17 Wrte data nto BUFFER [wdx]; 18 LATEST = wdx; 19 for =1 to n do 20 Compare-and-Swap(READING [],0,wdx); 21 end Algorthm 3: Modfed Chen s Protocol for Reader 1 Data: BUFFER [1,...,NB](NB: # of buffers) ; READING [1,...,n] (n: # of readers) ; LATEST 2 Reader() 3 begn 4 nteger rdx; 5 READING [d]=0; 6 rdx = LATEST; 7 Compare-and-Swap(READING [d],0,rdx); 8 rdx = READING [d]; 9 Read data from BUFFER [rdx]; 10 end III. FORMAL COMPARISON WITH CHEN S AND A. Specal Case Behavor The buffer sze that the protocol [4] requres depends on the maxmum number of nterferences that a reader can suffer from the wrter. It does not depend on the number of readers, because smultaneous readng by the readers accesses the same buffer, rrespectve of the number of readers. On the other hand, the buffer sze that the Chen s protocol [5] requres s

15 15 drectly proportonal to the number of readers, and s ndependent of the number of nterferences. We now show that our protocol subsumes both Chen s protocol and the protocol as specal cases. Lemma 3.1: The buffer sze for Chen s protocol [5] s a specal case of the WFBP soluton gven n Algorthm 1. Proof: Assume that we are gven M readers and no nformaton about nterferences. We can map ths problem to DSP, by settng R as [1, ] and the upper-bounds of x as <,..., >. Accordng to Theorem 2.2, n({ x}) cannot exceed n( x). Thus, the worst-case buffer sze s obtaned as (M + 2), that s n( x)+n({1, 2}). Ths s exactly the same value as that obtaned by Chen s protocol. Lemma 3.2: The buffer sze for protocol [4] s a specal case of the WFBP soluton gven n Algorthm 1. Proof: Assume that we are gven nfnte number of readers wth a knowledge of T op( u) = N max. Ths problem can be modeled as the problem wth R = [1, N max +1] and, u = N max +1 for the worst-case. By Theorem 2.1, H( u, N) =, and whenever t ncreases, n({ x}) ncreases by 1 untl t and n({ x}) reaches to N max and N max +1, respectvely. Thus, the worst-case buffer sze s obtaned as N max + 1,.e., n({1,..., N max + 1} {1, 2}). Ths s exactly the same value as that obtaned by protocol. Theorem 3.3 (Upper Bound of the WFBP soluton): In the WFBP, n max ({ x}) mn(m + 2, N max + 1), where M s the number of readers and N max s the maxmum number of nterferences that a reader can suffer. Proof: Proof follows drectly from Lemmas 3.1 and 3.2. Chen s protocol s attractve because the number of nterferences need not be known a-pror. On the other hand, has the advantage that the requred number of buffers can be further

16 16 reduced f the number of nterferences are much smaller than the number of readers. Addtonally, we note that the number of buffers needed by our algorthm s less than or equal to that of Chen s or protocol. B. Buffer Sze Condtons Accordng to Theorem 3.3, our wat-free protocol always fnds the number of requred buffers whch s less than or equal to that of Chen s protocol or the protocol. We now dentfy the precse condtons under whch the requred buffer sze of our protocol s equal to that of Chen s or. To derve the condtons, we observe two propertes n the WFBP. In the followng theorem, we ntroduce a notaton {{ x}}, whch denotes the set ncludng all possble solutons { x} for the gven DSP. Theorem 3.4 (Chen s Tester): When the number of readers n the wat-free buffer sze decson problem s M and N max > M, {3,..., M + 2} {{ x}}, f and only f n max ({ x}) M + 2. Proof: We prove both necessary and suffcent condtons. Case 1. Assume that when {3,..., M + 2} {{ x}}, n max ({ x}) < M + 2. Snce the sze of the optmal soluton s less than M + 2, the sze of { x} cannot exceed M + 2. Ths contradcts our assumpton that {1, 2} {3,..., M + 2} s a soluton. Case 2. Assume that the set { x} s {x 3,..., x M+2 }, n whch x s are dfferent between each other and algned n ncreasng order. Now, all x must not be 1 or 2, otherwse n max ({ x}) s less than M + 2. Therefore, x 3 should be greater than or equal to 3, and x 4 s greater than x 3. Inductvely, x +1 x + 1, where 3 < M + 2. In other words, snce x x x , the nequalty u x holds. For ths reason, {3,..., M + 2} satsfes the range constrants of all elements.

17 17 By Theorem 3.4, n max ({ x}) < M +2, f {3,..., M +2} / {{ x}}. Ths means that by checkng f {3,..., M + 2} s feasble for the problem, we can determne whether or not t requres M + 2 buffers that Chen s protocol needs. Theorem 3.5 ( Tester): When the number of readers n the wat-free buffer sze decson problem s M and N max M, {2,..., N max + 1} {{ x}}, f and only f n max ({ x}) N max + 1. Proof: We prove both necessary and suffcent condtons. Case 1. Assume that when {2,..., N max + 1} {{ x}}, n max ({ x}) < N max + 1. Snce the sze of the optmal soluton s less than N max + 1, the sze of { x} cannot exceed N max + 1. Ths contradcts our assumpton that {1, 2} {2,..., N max + 1} s a soluton. Case 2. Assume that the set { x} s {x 2,..., x N max +1}, n whch x s are dfferent between each other and algned n ncreasng order. Now, all x must not be 1, otherwse n max ({ x}) s less than N max +1. Therefore, x 2 should be greater than or equal to 2, and x 3 s greater than x 2. Inductvely, x +1 x + 1 where 2 < N max + 1. In other words, snce x x x , the nequalty u x holds. For ths reason, {2,..., N max + 1} satsfes the range constrants of all elements. We can also nvestgate f a gven WFBP needs N max +1 buffers or less by checkng feasblty wth {2,..., N max + 1}. We call {3,..., M + 2} and {2,..., N max + 1} as Chen s tester and tester, respectvely. From Theorems 3.4 and 3.5, we derve a decson procedure that determnes the wat-free protocol wth the lowest buffer sze. Fgure 5 shows ths procedure. To llustrate t, we use the WFBP example n [7], whch s also shown n Table I. By our decson procedure, snce N max > M, Chen s protocol requres smaller number of buffers than. The next step s determnng whether Chen s tester, whch s < 3, 4,..., 9 >

18 18 Fg. 5. Decson Procedure n ths problem, s feasble. It turns out that t s not feasble, as the second element 4 n the tester s out of the range [1, 3] of reader 1. Hence, we expect to fnd smaller number of requred buffers than that of Chen s protocol. TABLE I TASK SET Task Reader 0 2 Reader 1 2 Reader 2 2 Reader 3 3 Reader 4 3 Reader 5 14 Reader 6 49 N max Algorthm 1 determnes that we need 6 buffers for ths problem. We determne a vector { x} = {1, 2, 3, 4, 15, 50} as a worst-case canddate for the WFBP from Theorem 2.2. As mentoned earler, the soluton means that one of the worst-cases occurs when we need buffers for wrters {W [1], W [2], W [3], W [4], W [15], W [50] }. C. Comparson wth Improved Chen s Protocol In [7], Huang et al. suggest a transformaton mechansm to reduce the buffer sze needs of a gven wat-free protocol. The transformaton s appled to many wat-free protocols ncludng Chen s protocol. The transformed Chen s protocol s called Improved Chen s protocol n [7].

19 19 We cannot formally compare our protocol wth Improved Chen s protocol n terms of space cost, because no analytcal foundaton s gven for the transformaton mechansm n [7]. Consequently, a formal comparson s not possble, and only an expermental comparson s possble, where the two protocols can be compared for as many cases as possble. We do ths n Secton IV. Our experments n Secton IV reveal that the buffer sze needs of our protocol and Improved Chen s are the same, for all the cases that we consder. Of course, ths does not mply that Improved Chen s and ours always need the same number of buffers, because t s mpossble that our evaluaton studes n Secton IV cover all the cases. Nevertheless, note that wth Corollary 2.3, we guarantee that the buffer sze needed for wat-free cannot be reduced any further. Addtonal advantage of our protocol s that t s not requred to dvde readers nto fast and slow groups and apply two separate readng operatons as Improved Chen s does. D. Comparson of Tme Complexty Implementaton of and Chen s protocols requre the Compare-And-Swap (CAS) nstructon. The CAS nstructon s used to atomcally modfy control varables of the wat-free protocol by combnng comparson and swap operatons nto a sngle nstructon. The nstructon s avalable n many modern processors and takes constant tme. has no loop wthn both wrte and read operatons. However, Chen s protocol has 3 loops wthn the wrte operaton and no loop wthn the read operaton. Wth n buffers, the tme complexty of Chen s wrtng operaton s O(n). Improved Chen s protocol and our protocol are varatons of Chen s protocol, and hence have smlar tme complextes as that of Chen s wrtng and readng. Accordng to Theorem 3.3, the loop teraton n our protocol s wrte operaton cannot exceed M + 2. Thus, the tme complexty

20 20 TABLE II ASYMPTOTICAL TIME COMPLEXITIES Wat-Free Protocol Read Wrte O(1) O(1) Chen s O(1) O(n) Improved Chen s O(1) O(n) O(1) O(n) of our protocol s O(n), whch s the same as that of Chen s. Snce the asymptotcal speeds are therefore smlar, a speed mprovement can be obtaned (for Chen s, Improved Chen s, and ours) by reducng the buffer sze. Table II summarzes the asymptotcal tme complextes of the protocols. IV. NUMERICAL EVALUATION STUDIES We conduct numercal evaluatons to evaluate the buffer sze needs of our protocol under a broad range of reader/wrter condtons, ncludng ncreasng maxmum nterferences and readers. We also consder, Chen s, and Improved Chen s protocols for comparatve study. We consder Improved Chen s protocol among all protocols n [7], because t s the most spaceeffcent protocol n [7]. We exclude the Double Buffer protocol [7] from our study as t needs nearly two tmes the buffer space than Chen s protocol. (The Double Buffer protocol trades off space for tme.) Thus, our protocol wll clearly outperform the Double Buffer protocol n terms of buffer needs. A. Increasng Interferences We consder a task set wth 1 wrter and multple readers whose maxmum number of nterferences N max s randomly generated wth a normal dstrbuton (wth a fxed standard devaton of 5), and by varyng the average. The protocols are evaluated by ther buffer sze needs the actual amount of needed memory s the number of buffers tmes the message sze n bytes. Each experment s repeated 100 tmes to determne the average buffer szes.

21 21 Number of requred buffers Imp Ave. of maxmum numbers of nterferences Number of requred buffers Imp Ave. of maxmum numbers of nterferences Number of requred buffers Imp Ave. of maxmum numbers of nterferences (a) M = 20 (b) M = 30 (c) M = 40 Fg. 6. Buffer Szes Under Increasng Interferences Wth Normal Dstrbuton for N max Fgure 6 shows the buffer sze needs of each protocol as the average N max s ncreased from 5 to 45, for 20, 30, and 40 readers. From the fgure, we observe that as N max ncreases, the buffer sze needs of ncreases, whereas that of Chen s protocol remans the same (for a gven reader sze), snce ts buffer needs s proportonal only to the number of readers. As the number of readers ncreases from 20 to 40, Chen s protocol needs ncreasng number of buffers. Meanwhle, the number of buffers that our protocol requres never exceeds that of Chen s and s, as Theorem 3.3 holds. Interestngly, the number of buffers that Improved Chen s protocol requres s exactly the same as that of ours. Note that no analyss on the buffer sze needs of Improved Chen s s presented n [7], whereas Theorem 3.3 gves the upper bound on the buffer sze needs of our protocol. We observed exact smlar trends for other fxed standard devatons for N max s dstrbuton, and other dstrbutons for N max. Fgure 7(a) shows the buffer sze needs of each protocol, when N max s generated wth a normal dstrbuton, wth a fxed standard devaton of 10 (nstead of 5), and by varyng the average N max from 5 to 45, for 40 readers. Fgure 7(b) shows the protocols buffer needs under the exact same condtons as those n Fgure 7(a), except that N max s now generated wth an unform dstrbuton.

22 22 Number of requred buffers Imp Ave. of maxmum numbers of nterferences Number of requred buffers Imp Ave. of maxmum numbers of nterferences (a) Normal Dstrbuton (b) Unform Dstrbuton Fg. 7. Buffer Szes Under Dfferent N max Dstrbutons wth 40 Readers From the fgures, we observe that our protocol s buffer needs never exceed that of Chen s and s, and s the same as that of Improved Chen s. B. Heterogenous Readers n Multple Groups From Fgure 6, we also observe that when most readers have small N max, the number of buffers needed by our protocol approaches that of s. Moreover, when most readers have larger N max, the number of buffers needed by our protocol approaches that of Chen s protocol s. Ths motvates us to study the buffer sze needs of our protocol under two groups of readers, one that has small N max s and the other that has large N max s. (A smlar evaluaton s conducted n [7], where readers are classfed as fast and slow. ) We dvde tasks nto the two groups whose averages of the (normal) dstrbuton for N max s are fxed as 5 and 45, respectvely. We then vary the rato of the two groups. For example, 3:1 n the X-axs n Fgure 8(a) means that the readers havng smaller N max are 3 tmes more than the readers havng larger N max. Fgure 8 shows the buffer szes of each protocol as the rato s vared from 3:1 to 1:3, for 20, 30, and 40 readers. We observe that the buffers needed for, Improved Chen s, and our protocol ncrease as the readers wth larger N max ncreases. Ths result s consstent wth that n [7], where Improved Chen s s shown to requre less buffers, as fast readers wth smaller

23 Number of requred buffers Imp Rato of two reader groups Number of requred buffers Imp Rato of two reader groups Number of requred buffers Imp. 3:1 2:1 1:1 1:2 1:3 Rato of two reader groups (a) M = 20 (b) M = 30 (c) M = 40 Fg. 8. Buffer Szes Wth 2 Reader Groups Under Varyng Reader Rato, for 20, 30, and 40 Readers N max ncreases. The results confrm that ours and Improved Chen s requre the mnmum buffer sze when consderng two heterogenous reader groups. We now consder a more complex scenaro wth three reader groups, called fast, slow, and medum, whch are not consdered n [7]. The averages of the (normal) dstrbuton for N max s for the three groups are fxed as 5, 25, and 45, respectvely, and the rato of the three groups are vared from 6:3:1 to 1:3:6. Fgure 9 shows the results Number of requred buffers Imp. 6:3:1 4:2:1 1:1:1 1:2:4 1:3:6 Rato of three reader groups Number of requred buffers Imp. 6:3:1 4:2:1 1:1:1 1:2:4 1:3:6 Rato of three reader groups Number of requred buffers Imp. 6:3:1 4:2:1 1:1:1 1:2:4 1:3:6 Rato of three reader groups (a) M = 20 (b) M = 30 (c) M = 40 Fg. 9. Buffer Szes Wth 3 Reader Groups Under Varyng Reader Rato, for 20, 30, and 40 Readers From the fgure, we observe that as the number of fast readers ncreases, the number of buffers needed decreases. Further, we observe that the buffer sze requred by Improved Chen s s the same as that of ours even when we nclude the medum reader group n our evaluaton.

24 24 V. IMPLEMENTATION EXPERIENCE A wat-free protocol s practcal effectveness s determned by ts space and tme costs. In developng a wat-free protocol, we focus on optmzng space costs, and we establsh the space optmalty of our protocol. Although reducng the protocol s tme costs s not our goal, we now determne the tme costs to establsh our protocol s effectveness. Our wat-free protocol (Algorthms 2 and 3) s a modfcaton of Chen s protocol, augmented wth the buffer sze computed by Algorthm 1. Thus, we expect that our protocol ncurs at most as much tme overhead as that of Chen s. Moreover, the hgher space effcency that our protocol enjoys can lead to hgher tme effcency, because t reduces the search space for determnng the protocol s safe buffer e.g., GetBuf() s loop n Algorthm 2. To evaluate the actual tme costs of our protocol, we mplement our protocol n the SHaRK (Soft Hard Real-Tme Kernel) OS [10], runnng on a 500MHz, Pentum-III processor. Smlar to Secton IV, we also mplement Chen s, Improved Chen s, and protocols for a comparatve study. We also consder lock-based sharng n ths study. Note that all protocols n our study can be adopted for both un-processor and mult-processor systems, although we consder only the performance n the un-processor n ths secton. We consder a task set wth 20 readers and a wrter, and use a message sze of 8 bytes for an nter-process communcaton (or IPC). We measure the average-case executon tme (or ACET) and the worst case executon tme (or WCET) for performng an IPC. The executon tme for an IPC s the tme needed for executng the code segment that accesses the shared object. Wth tradtonal lock-based sharng, ths code segment s the crtcal secton. Note that a wat-free protocol s IPC executon tme ncludes tmes for controllng protocol s varables, accessng the shared object, and potental nterference from other tasks. Thus, WCET tends to be much larger

25 25 than ACET. In Secton IV-B, we vared the rato of two reader groups whose averages of the (normal) dstrbuton for N max s are fxed as 5 and 45, respectvely. We now select two cases from whch the rato of readers havng smaller and larger N max are 4:1 and 1:4, respectvely. These two cases can be represented as 16 fast and 4 slow readers, and 4 fast and 16 slow readers, respectvely, for the purpose of Improved Chen s [7], snce that protocol needs the readers to be classfed as slow and fast. We fx the wrter s perod as 0.2 msec and let the wrter nvoke 6,000,000 tmes durng our experment nterval for computng the ACETs. The perod of the 20 readers ranges from 400 usec to approxmately 10msec. Average executon tme (usec) Lock-based Imp. Average executon tme (usec) Lock-based Imp. (a) 16 Fast and 4 Slow Readers (b) 4 Fast and 16 Slow Readers Fg. 10. ACET of Read/Wrte n SHaRK RTOS Fgure 10 shows the measurements from our mplementaton. We observe that has the smallest ACET, lock-based sharng has the largest ACET, and Chen s, Improved Chen s, and our protocol have almost the same ACET n our mplementaton. has the smallest ACET, because ts mplementaton does not have any loop (and thus less computatonal costs) nsde both the reader and wrter operatons. Lock-based sharng has the largest ACET due to ts blockng tmes. Further, accessng and releasng locks n SHaRK s done through system calls, whch takes longer than wat-free protocols (whch are mplemented wthout system calls).

26 26 Worst case executon tme (usec) Lock-based Imp. Worst case executon tme (usec) Lock-based Imp. (a) 16 Fast and 4 Slow Readers (b) 4 Fast and 16 Slow Readers Fg. 11. WCET of Read/Wrte n SHaRK RTOS In [7], when the number of fast readers are ncreasng, the ACET of Improved Chen s tends to be shorter because the needed buffer sze decreases and, a part of Improved Chen s, performs faster. Ths trend does not appear n our experments. Ths s because the expected speed mprovement s only (approxmately) 0.1 usec. Ths dfference s small enough to be affected by the OS type, code optmzatons, and measurement methodology, among other factors. We observed the smlar results n WCET n Fgure 11. Although reducng the protocol s tme costs s not our goal, we observe that varatons of Chen s ncludng Chen s, Improved Chen s, and ours have much the same ACET and WCET at least n our mplementaton and thus, we beleve that our protocol s tme costs s comparable to that of prevous protocols. We have suggested the decson procedure that determnes the wat-free protocol havng the lowest buffer sze n Secton III-B. Before applyng our protocol, we can determne whch protocol, among Chen s,, and ours, requres the least buffer sze usng the decson procedure descrbed n Fgure 5. We now apply ths decson procedure to the 16 fast/4 slowreader example consdered prevously. Table III shows 16 fast/4 slow readers N max s. At the frst step n the decson procedure, we can easly fnd that N max = 47 > M = 20. It mples that Chen s protocol needs lower

27 27 buffer sze than. Now, the next step s to check f Chen s tester s feasble. Chen s tester s evaluated as {3,...,22} by Theorem 3.4. TABLE III DECISION PROCEDURE ON 16 FAST AND 4 SLOW READERS Task N max + 1 Chen s Tester Feasblty Reader O Reader O Reader O Reader O Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader X Reader O Number of requred buffers Lock-based Imp. Number of requred buffers Lock-based Imp. (a) 16 Fast and 4 Slow Readers (b) 4 Fast and 16 Slow Readers Fg. 12. Buffer Szes Table III ndcates that Chen s tester s not feasble because 18 n the Chen s tester column s not between 1 and 10, for example. Therefore, at the fnal step, we can conclude that our protocol requres less buffers than Chen s. Ths s true as shown n Fgure 12, whch shows the

28 28 number of requred buffer sze for each protocol. VI. CONCLUSIONS In ths paper, we consder the sngle-wrter/multple-reader problem that occurs n embedded real-tme systems. We present an analytcal soluton to the problem of determnng the absolute mnmum buffer requrement of wat-free protocols for ths problem the frst such bound establshed for wat-free protocols that consder a-pror knowledge of nterferences. We also show that the space costs requred by prevous algorthms ncludng Chen s and can also be obtaned by our soluton, whch subsumes them as specal cases. We also present a wat-free protocol that uses the mnmum buffer sze determned by our analytcal soluton. Our evaluaton studes and mplementaton measurements valdate our analytcal results. Some aspects of the work are drectons for further research. Examples nclude extendng the protocol for the multple-wrter/multple-reader problem, and complex concurrent objects such as (non-blockng) stacks and queues. VII. ACKNOWLEDGEMENTS Ths work was sponsored by the US Offce of Naval Research under Grant N and The MITRE Corporaton under Grant A prelmnary verson of ths paper appeared n [12]. REFERENCES [1] L. Sha, R. Rajkumar, and J. P. Lehoczky, Prorty nhertance protocols: An approach to real-tme synchronzaton, IEEE Transactons on Computers, vol. 39, no. 9, pp , [2] T. P. Baker, Stack-based schedulng of real-tme processes, Real-Tme Systems, vol. 3, no. 1, pp , Mar [3] R. K. Clark, Schedulng dependent real-tme actvtes, Ph.D. dssertaton, Carnege Mellon Unversty, [4] H. Kopetz and J. Resnger, The non-blockng wrte protocol nbw: A soluton to a real-tme synchronsaton problem, n IEEE Real-Tme Systems Symposum, 1993, pp

29 29 [5] J. Chen and A. Burns, A fully asynchronous reader/wrter mechansm for multprocessor real-tme systems, Unversty of York, Tech. Rep. YCS-288, May [6] J. H. Anderson, R. Jan, and S. Ramamurthy, Wat-free object-sharng schemes for real-tme unprocessors and multprocessors, n IEEE Real-Tme Systems Symposum, Dec. 1997, pp [7] H. Huang, P. Plla, and K. G. Shn, Improvng wat-free algorthms for nterprocess communcaton n embedded real-tme systems, n USENIX Annual Techncal Conference, 2002, pp [8] J. H. Anderson, S. Ramamurthy, and K. Jeffay, Real-tme computng wth lock-free shared objects, ACM Transactons On Computer Systems, vol. 15, no. 2, pp , [9] H. Sundell and P. Tsgas, Space effcent wat-free buffer sharng n multprocessor real-tme systems based on tmng nformaton, n IEEE Real-Tme Computng Systems and Applcatons, 2000, pp [10] P. Ga, L. Aben, M. Gorg, and G. Buttazzo, A new kernel approach for modular real-tme systems development, n Euromcro Conference on Real-Tme Systems, 2001, pp [11] J. Chen and A. Burns, A three-slot asynchronous reader/wrter mechansm for multprocessor real-tme systems, Unversty of York, Tech. Rep. YCS-186, [12] H. Cho, B. Ravndran, and E. D. Jensen, A space-optmal, wat-free real-tme synchronzaton protocol, n IEEE Euromcro Conference on Real-Tme Systems, July 2005, pp

AADL : about scheduling analysis

AADL : about scheduling analysis AADL : about schedulng analyss Schedulng analyss, what s t? Embedded real-tme crtcal systems have temporal constrants to meet (e.g. deadlne). Many systems are bult wth operatng systems provdng multtaskng

More information

Parallelism for Nested Loops with Non-uniform and Flow Dependences

Parallelism for Nested Loops with Non-uniform and Flow Dependences Parallelsm for Nested Loops wth Non-unform and Flow Dependences Sam-Jn Jeong Dept. of Informaton & Communcaton Engneerng, Cheonan Unversty, 5, Anseo-dong, Cheonan, Chungnam, 330-80, Korea. seong@cheonan.ac.kr

More information

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory

Virtual Memory. Background. No. 10. Virtual Memory: concept. Logical Memory Space (review) Demand Paging(1) Virtual Memory Background EECS. Operatng System Fundamentals No. Vrtual Memory Prof. Hu Jang Department of Electrcal Engneerng and Computer Scence, York Unversty Memory-management methods normally requres the entre process

More information

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz

Compiler Design. Spring Register Allocation. Sample Exercises and Solutions. Prof. Pedro C. Diniz Compler Desgn Sprng 2014 Regster Allocaton Sample Exercses and Solutons Prof. Pedro C. Dnz USC / Informaton Scences Insttute 4676 Admralty Way, Sute 1001 Marna del Rey, Calforna 90292 pedro@s.edu Regster

More information

A Binarization Algorithm specialized on Document Images and Photos

A Binarization Algorithm specialized on Document Images and Photos A Bnarzaton Algorthm specalzed on Document mages and Photos Ergna Kavalleratou Dept. of nformaton and Communcaton Systems Engneerng Unversty of the Aegean kavalleratou@aegean.gr Abstract n ths paper, a

More information

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour

6.854 Advanced Algorithms Petar Maymounkov Problem Set 11 (November 23, 2005) With: Benjamin Rossman, Oren Weimann, and Pouya Kheradpour 6.854 Advanced Algorthms Petar Maymounkov Problem Set 11 (November 23, 2005) Wth: Benjamn Rossman, Oren Wemann, and Pouya Kheradpour Problem 1. We reduce vertex cover to MAX-SAT wth weghts, such that the

More information

Lecture 7 Real Time Task Scheduling. Forrest Brewer

Lecture 7 Real Time Task Scheduling. Forrest Brewer Lecture 7 Real Tme Task Schedulng Forrest Brewer Real Tme ANSI defnes real tme as A Real tme process s a process whch delvers the results of processng n a gven tme span A data may requre processng at a

More information

The Codesign Challenge

The Codesign Challenge ECE 4530 Codesgn Challenge Fall 2007 Hardware/Software Codesgn The Codesgn Challenge Objectves In the codesgn challenge, your task s to accelerate a gven software reference mplementaton as fast as possble.

More information

Concurrent Apriori Data Mining Algorithms

Concurrent Apriori Data Mining Algorithms Concurrent Apror Data Mnng Algorthms Vassl Halatchev Department of Electrcal Engneerng and Computer Scence York Unversty, Toronto October 8, 2015 Outlne Why t s mportant Introducton to Assocaton Rule Mnng

More information

An Optimal Algorithm for Prufer Codes *

An Optimal Algorithm for Prufer Codes * J. Software Engneerng & Applcatons, 2009, 2: 111-115 do:10.4236/jsea.2009.22016 Publshed Onlne July 2009 (www.scrp.org/journal/jsea) An Optmal Algorthm for Prufer Codes * Xaodong Wang 1, 2, Le Wang 3,

More information

Maintaining temporal validity of real-time data on non-continuously executing resources

Maintaining temporal validity of real-time data on non-continuously executing resources Mantanng temporal valdty of real-tme data on non-contnuously executng resources Tan Ba, Hong Lu and Juan Yang Hunan Insttute of Scence and Technology, College of Computer Scence, 44, Yueyang, Chna Wuhan

More information

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields

A mathematical programming approach to the analysis, design and scheduling of offshore oilfields 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 A mathematcal programmng approach to the analyss, desgn and

More information

Array transposition in CUDA shared memory

Array transposition in CUDA shared memory Array transposton n CUDA shared memory Mke Gles February 19, 2014 Abstract Ths short note s nspred by some code wrtten by Jeremy Appleyard for the transposton of data through shared memory. I had some

More information

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance

Tsinghua University at TAC 2009: Summarizing Multi-documents by Information Distance Tsnghua Unversty at TAC 2009: Summarzng Mult-documents by Informaton Dstance Chong Long, Mnle Huang, Xaoyan Zhu State Key Laboratory of Intellgent Technology and Systems, Tsnghua Natonal Laboratory for

More information

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization

Problem Definitions and Evaluation Criteria for Computational Expensive Optimization Problem efntons and Evaluaton Crtera for Computatonal Expensve Optmzaton B. Lu 1, Q. Chen and Q. Zhang 3, J. J. Lang 4, P. N. Suganthan, B. Y. Qu 6 1 epartment of Computng, Glyndwr Unversty, UK Faclty

More information

Real-time Scheduling

Real-time Scheduling Real-tme Schedulng COE718: Embedded System Desgn http://www.ee.ryerson.ca/~courses/coe718/ Dr. Gul N. Khan http://www.ee.ryerson.ca/~gnkhan Electrcal and Computer Engneerng Ryerson Unversty Overvew RTX

More information

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION

CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION 24 CHAPTER 2 PROPOSED IMPROVED PARTICLE SWARM OPTIMIZATION The present chapter proposes an IPSO approach for multprocessor task schedulng problem wth two classfcatons, namely, statc ndependent tasks and

More information

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits

Repeater Insertion for Two-Terminal Nets in Three-Dimensional Integrated Circuits Repeater Inserton for Two-Termnal Nets n Three-Dmensonal Integrated Crcuts Hu Xu, Vasls F. Pavlds, and Govann De Mchel LSI - EPFL, CH-5, Swtzerland, {hu.xu,vasleos.pavlds,govann.demchel}@epfl.ch Abstract.

More information

Real-Time Guarantees. Traffic Characteristics. Flow Control

Real-Time Guarantees. Traffic Characteristics. Flow Control Real-Tme Guarantees Requrements on RT communcaton protocols: delay (response s) small jtter small throughput hgh error detecton at recever (and sender) small error detecton latency no thrashng under peak

More information

CMPS 10 Introduction to Computer Science Lecture Notes

CMPS 10 Introduction to Computer Science Lecture Notes CPS 0 Introducton to Computer Scence Lecture Notes Chapter : Algorthm Desgn How should we present algorthms? Natural languages lke Englsh, Spansh, or French whch are rch n nterpretaton and meanng are not

More information

Support Vector Machines

Support Vector Machines /9/207 MIST.6060 Busness Intellgence and Data Mnng What are Support Vector Machnes? Support Vector Machnes Support Vector Machnes (SVMs) are supervsed learnng technques that analyze data and recognze patterns.

More information

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration

Improvement of Spatial Resolution Using BlockMatching Based Motion Estimation and Frame. Integration Improvement of Spatal Resoluton Usng BlockMatchng Based Moton Estmaton and Frame Integraton Danya Suga and Takayuk Hamamoto Graduate School of Engneerng, Tokyo Unversty of Scence, 6-3-1, Nuku, Katsuska-ku,

More information

GSLM Operations Research II Fall 13/14

GSLM Operations Research II Fall 13/14 GSLM 58 Operatons Research II Fall /4 6. Separable Programmng Consder a general NLP mn f(x) s.t. g j (x) b j j =. m. Defnton 6.. The NLP s a separable program f ts objectve functon and all constrants are

More information

ELEC 377 Operating Systems. Week 6 Class 3

ELEC 377 Operating Systems. Week 6 Class 3 ELEC 377 Operatng Systems Week 6 Class 3 Last Class Memory Management Memory Pagng Pagng Structure ELEC 377 Operatng Systems Today Pagng Szes Vrtual Memory Concept Demand Pagng ELEC 377 Operatng Systems

More information

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique

The Greedy Method. Outline and Reading. Change Money Problem. Greedy Algorithms. Applications of the Greedy Strategy. The Greedy Method Technique //00 :0 AM Outlne and Readng The Greedy Method The Greedy Method Technque (secton.) Fractonal Knapsack Problem (secton..) Task Schedulng (secton..) Mnmum Spannng Trees (secton.) Change Money Problem Greedy

More information

Module Management Tool in Software Development Organizations

Module Management Tool in Software Development Organizations Journal of Computer Scence (5): 8-, 7 ISSN 59-66 7 Scence Publcatons Management Tool n Software Development Organzatons Ahmad A. Al-Rababah and Mohammad A. Al-Rababah Faculty of IT, Al-Ahlyyah Amman Unversty,

More information

Mathematics 256 a course in differential equations for engineering students

Mathematics 256 a course in differential equations for engineering students Mathematcs 56 a course n dfferental equatons for engneerng students Chapter 5. More effcent methods of numercal soluton Euler s method s qute neffcent. Because the error s essentally proportonal to the

More information

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms

Course Introduction. Algorithm 8/31/2017. COSC 320 Advanced Data Structures and Algorithms. COSC 320 Advanced Data Structures and Algorithms Course Introducton Course Topcs Exams, abs, Proects A quc loo at a few algorthms 1 Advanced Data Structures and Algorthms Descrpton: We are gong to dscuss algorthm complexty analyss, algorthm desgn technques

More information

3. CR parameters and Multi-Objective Fitness Function

3. CR parameters and Multi-Objective Fitness Function 3 CR parameters and Mult-objectve Ftness Functon 41 3. CR parameters and Mult-Objectve Ftness Functon 3.1. Introducton Cogntve rados dynamcally confgure the wreless communcaton system, whch takes beneft

More information

Problem Set 3 Solutions

Problem Set 3 Solutions Introducton to Algorthms October 4, 2002 Massachusetts Insttute of Technology 6046J/18410J Professors Erk Demane and Shaf Goldwasser Handout 14 Problem Set 3 Solutons (Exercses were not to be turned n,

More information

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions

Sorting Review. Sorting. Comparison Sorting. CSE 680 Prof. Roger Crawfis. Assumptions Sortng Revew Introducton to Algorthms Qucksort CSE 680 Prof. Roger Crawfs Inserton Sort T(n) = Θ(n 2 ) In-place Merge Sort T(n) = Θ(n lg(n)) Not n-place Selecton Sort (from homework) T(n) = Θ(n 2 ) In-place

More information

Adaptive Resource Allocation Control with On-Line Search for Fair QoS Level

Adaptive Resource Allocation Control with On-Line Search for Fair QoS Level Adaptve Resource Allocaton Control wth On-Lne Search for Far QoS Level Fumko Harada, Toshmtsu Usho, Graduate School of Engneerng Scence Osaka Unversty {harada@hopf, usho@}sysesosaka-uacjp Yukkazu akamoto

More information

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr)

Helsinki University Of Technology, Systems Analysis Laboratory Mat Independent research projects in applied mathematics (3 cr) Helsnk Unversty Of Technology, Systems Analyss Laboratory Mat-2.08 Independent research projects n appled mathematcs (3 cr) "! #$&% Antt Laukkanen 506 R ajlaukka@cc.hut.f 2 Introducton...3 2 Multattrbute

More information

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation

Quality Improvement Algorithm for Tetrahedral Mesh Based on Optimal Delaunay Triangulation Intellgent Informaton Management, 013, 5, 191-195 Publshed Onlne November 013 (http://www.scrp.org/journal/m) http://dx.do.org/10.36/m.013.5601 Qualty Improvement Algorthm for Tetrahedral Mesh Based on

More information

Dynamic Voltage Scaling of Supply and Body Bias Exploiting Software Runtime Distribution

Dynamic Voltage Scaling of Supply and Body Bias Exploiting Software Runtime Distribution Dynamc Voltage Scalng of Supply and Body Bas Explotng Software Runtme Dstrbuton Sungpack Hong EE Department Stanford Unversty Sungjoo Yoo, Byeong Bn, Kyu-Myung Cho, Soo-Kwan Eo Samsung Electroncs Taehwan

More information

Reducing Frame Rate for Object Tracking

Reducing Frame Rate for Object Tracking Reducng Frame Rate for Object Trackng Pavel Korshunov 1 and We Tsang Oo 2 1 Natonal Unversty of Sngapore, Sngapore 11977, pavelkor@comp.nus.edu.sg 2 Natonal Unversty of Sngapore, Sngapore 11977, oowt@comp.nus.edu.sg

More information

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1)

For instance, ; the five basic number-sets are increasingly more n A B & B A A = B (1) Secton 1.2 Subsets and the Boolean operatons on sets If every element of the set A s an element of the set B, we say that A s a subset of B, or that A s contaned n B, or that B contans A, and we wrte A

More information

Multitasking and Real-time Scheduling

Multitasking and Real-time Scheduling Multtaskng and Real-tme Schedulng EE8205: Embedded Computer Systems http://www.ee.ryerson.ca/~courses/ee8205/ Dr. Gul N. Khan http://www.ee.ryerson.ca/~gnkhan Electrcal and Computer Engneerng Ryerson Unversty

More information

Brave New World Pseudocode Reference

Brave New World Pseudocode Reference Brave New World Pseudocode Reference Pseudocode s a way to descrbe how to accomplsh tasks usng basc steps lke those a computer mght perform. In ths week s lab, you'll see how a form of pseudocode can be

More information

Programming in Fortran 90 : 2017/2018

Programming in Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Programmng n Fortran 90 : 2017/2018 Exercse 1 : Evaluaton of functon dependng on nput Wrte a program who evaluate the functon f (x,y) for any two user specfed values

More information

Performance Evaluation of Information Retrieval Systems

Performance Evaluation of Information Retrieval Systems Why System Evaluaton? Performance Evaluaton of Informaton Retreval Systems Many sldes n ths secton are adapted from Prof. Joydeep Ghosh (UT ECE) who n turn adapted them from Prof. Dk Lee (Unv. of Scence

More information

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation

An Iterative Solution Approach to Process Plant Layout using Mixed Integer Optimisation 17 th European Symposum on Computer Aded Process Engneerng ESCAPE17 V. Plesu and P.S. Agach (Edtors) 2007 Elsever B.V. All rghts reserved. 1 An Iteratve Soluton Approach to Process Plant Layout usng Mxed

More information

A New Approach For the Ranking of Fuzzy Sets With Different Heights

A New Approach For the Ranking of Fuzzy Sets With Different Heights New pproach For the ankng of Fuzzy Sets Wth Dfferent Heghts Pushpnder Sngh School of Mathematcs Computer pplcatons Thapar Unversty, Patala-7 00 Inda pushpndersnl@gmalcom STCT ankng of fuzzy sets plays

More information

An efficient iterative source routing algorithm

An efficient iterative source routing algorithm An effcent teratve source routng algorthm Gang Cheng Ye Tan Nrwan Ansar Advanced Networng Lab Department of Electrcal Computer Engneerng New Jersey Insttute of Technology Newar NJ 7 {gc yt Ansar}@ntedu

More information

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT

DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT DESIGNING TRANSMISSION SCHEDULES FOR WIRELESS AD HOC NETWORKS TO MAXIMIZE NETWORK THROUGHPUT Bran J. Wolf, Joseph L. Hammond, and Harlan B. Russell Dept. of Electrcal and Computer Engneerng, Clemson Unversty,

More information

A fair buffer allocation scheme

A fair buffer allocation scheme A far buffer allocaton scheme Juha Henanen and Kalev Klkk Telecom Fnland P.O. Box 228, SF-330 Tampere, Fnland E-mal: juha.henanen@tele.f Abstract An approprate servce for data traffc n ATM networks requres

More information

Meta-heuristics for Multidimensional Knapsack Problems

Meta-heuristics for Multidimensional Knapsack Problems 2012 4th Internatonal Conference on Computer Research and Development IPCSIT vol.39 (2012) (2012) IACSIT Press, Sngapore Meta-heurstcs for Multdmensonal Knapsack Problems Zhbao Man + Computer Scence Department,

More information

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices

Steps for Computing the Dissimilarity, Entropy, Herfindahl-Hirschman and. Accessibility (Gravity with Competition) Indices Steps for Computng the Dssmlarty, Entropy, Herfndahl-Hrschman and Accessblty (Gravty wth Competton) Indces I. Dssmlarty Index Measurement: The followng formula can be used to measure the evenness between

More information

Imperialist Competitive Algorithm with Variable Parameters to Determine the Global Minimum of Functions with Several Arguments

Imperialist Competitive Algorithm with Variable Parameters to Determine the Global Minimum of Functions with Several Arguments Fourth Internatonal Conference Modellng and Development of Intellgent Systems October 8 - November, 05 Lucan Blaga Unversty Sbu - Romana Imperalst Compettve Algorthm wth Varable Parameters to Determne

More information

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009.

Assignment # 2. Farrukh Jabeen Algorithms 510 Assignment #2 Due Date: June 15, 2009. Farrukh Jabeen Algorthms 51 Assgnment #2 Due Date: June 15, 29. Assgnment # 2 Chapter 3 Dscrete Fourer Transforms Implement the FFT for the DFT. Descrbed n sectons 3.1 and 3.2. Delverables: 1. Concse descrpton

More information

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following.

Complex Numbers. Now we also saw that if a and b were both positive then ab = a b. For a second let s forget that restriction and do the following. Complex Numbers The last topc n ths secton s not really related to most of what we ve done n ths chapter, although t s somewhat related to the radcals secton as we wll see. We also won t need the materal

More information

Mixed-Criticality Scheduling on Multiprocessors using Task Grouping

Mixed-Criticality Scheduling on Multiprocessors using Task Grouping Mxed-Crtcalty Schedulng on Multprocessors usng Task Groupng Jankang Ren Lnh Th Xuan Phan School of Software Technology, Dalan Unversty of Technology, Chna Computer and Informaton Scence Department, Unversty

More information

S1 Note. Basis functions.

S1 Note. Basis functions. S1 Note. Bass functons. Contents Types of bass functons...1 The Fourer bass...2 B-splne bass...3 Power and type I error rates wth dfferent numbers of bass functons...4 Table S1. Smulaton results of type

More information

Intra-Parametric Analysis of a Fuzzy MOLP

Intra-Parametric Analysis of a Fuzzy MOLP Intra-Parametrc Analyss of a Fuzzy MOLP a MIAO-LING WANG a Department of Industral Engneerng and Management a Mnghsn Insttute of Technology and Hsnchu Tawan, ROC b HSIAO-FAN WANG b Insttute of Industral

More information

A comparison of MPCP and MSRP when sharing resources in the Janus multiple-processor on a chip platform

A comparison of MPCP and MSRP when sharing resources in the Janus multiple-processor on a chip platform A comparson of MPCP and MSRP when sharng resources n the Janus multple-processor on a chp platform Paolo Ga, Marco D Natale, Guseppe Lpar, Scuola Superore Sant Anna, Psa, Italy {pj,marco,lpar}@sssup.t

More information

Efficient Distributed File System (EDFS)

Efficient Distributed File System (EDFS) Effcent Dstrbuted Fle System (EDFS) (Sem-Centralzed) Debessay(Debsh) Fesehaye, Rahul Malk & Klara Naherstedt Unversty of Illnos-Urbana Champagn Contents Problem Statement, Related Work, EDFS Desgn Rate

More information

An Efficient Genetic Algorithm with Fuzzy c-means Clustering for Traveling Salesman Problem

An Efficient Genetic Algorithm with Fuzzy c-means Clustering for Traveling Salesman Problem An Effcent Genetc Algorthm wth Fuzzy c-means Clusterng for Travelng Salesman Problem Jong-Won Yoon and Sung-Bae Cho Dept. of Computer Scence Yonse Unversty Seoul, Korea jwyoon@sclab.yonse.ac.r, sbcho@cs.yonse.ac.r

More information

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments

Comparison of Heuristics for Scheduling Independent Tasks on Heterogeneous Distributed Environments Comparson of Heurstcs for Schedulng Independent Tasks on Heterogeneous Dstrbuted Envronments Hesam Izakan¹, Ath Abraham², Senor Member, IEEE, Václav Snášel³ ¹ Islamc Azad Unversty, Ramsar Branch, Ramsar,

More information

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS

A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Proceedngs of the Wnter Smulaton Conference M E Kuhl, N M Steger, F B Armstrong, and J A Jones, eds A MOVING MESH APPROACH FOR SIMULATION BUDGET ALLOCATION ON CONTINUOUS DOMAINS Mark W Brantley Chun-Hung

More information

y and the total sum of

y and the total sum of Lnear regresson Testng for non-lnearty In analytcal chemstry, lnear regresson s commonly used n the constructon of calbraton functons requred for analytcal technques such as gas chromatography, atomc absorpton

More information

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices

An Application of the Dulmage-Mendelsohn Decomposition to Sparse Null Space Bases of Full Row Rank Matrices Internatonal Mathematcal Forum, Vol 7, 2012, no 52, 2549-2554 An Applcaton of the Dulmage-Mendelsohn Decomposton to Sparse Null Space Bases of Full Row Rank Matrces Mostafa Khorramzadeh Department of Mathematcal

More information

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task

Term Weighting Classification System Using the Chi-square Statistic for the Classification Subtask at NTCIR-6 Patent Retrieval Task Proceedngs of NTCIR-6 Workshop Meetng, May 15-18, 2007, Tokyo, Japan Term Weghtng Classfcaton System Usng the Ch-square Statstc for the Classfcaton Subtask at NTCIR-6 Patent Retreval Task Kotaro Hashmoto

More information

Wishing you all a Total Quality New Year!

Wishing you all a Total Quality New Year! Total Qualty Management and Sx Sgma Post Graduate Program 214-15 Sesson 4 Vnay Kumar Kalakband Assstant Professor Operatons & Systems Area 1 Wshng you all a Total Qualty New Year! Hope you acheve Sx sgma

More information

CSE 326: Data Structures Quicksort Comparison Sorting Bound

CSE 326: Data Structures Quicksort Comparison Sorting Bound CSE 326: Data Structures Qucksort Comparson Sortng Bound Bran Curless Sprng 2008 Announcements (5/14/08) Homework due at begnnng of class on Frday. Secton tomorrow: Graded homeworks returned More dscusson

More information

Verification by testing

Verification by testing Real-Tme Systems Specfcaton Implementaton System models Executon-tme analyss Verfcaton Verfcaton by testng Dad? How do they know how much weght a brdge can handle? They drve bgger and bgger trucks over

More information

5 The Primal-Dual Method

5 The Primal-Dual Method 5 The Prmal-Dual Method Orgnally desgned as a method for solvng lnear programs, where t reduces weghted optmzaton problems to smpler combnatoral ones, the prmal-dual method (PDM) has receved much attenton

More information

Fast Computation of Shortest Path for Visiting Segments in the Plane

Fast Computation of Shortest Path for Visiting Segments in the Plane Send Orders for Reprnts to reprnts@benthamscence.ae 4 The Open Cybernetcs & Systemcs Journal, 04, 8, 4-9 Open Access Fast Computaton of Shortest Path for Vstng Segments n the Plane Ljuan Wang,, Bo Jang

More information

Parallel Branch and Bound Algorithm - A comparison between serial, OpenMP and MPI implementations

Parallel Branch and Bound Algorithm - A comparison between serial, OpenMP and MPI implementations Journal of Physcs: Conference Seres Parallel Branch and Bound Algorthm - A comparson between seral, OpenMP and MPI mplementatons To cte ths artcle: Luco Barreto and Mchael Bauer 2010 J. Phys.: Conf. Ser.

More information

Learning the Kernel Parameters in Kernel Minimum Distance Classifier

Learning the Kernel Parameters in Kernel Minimum Distance Classifier Learnng the Kernel Parameters n Kernel Mnmum Dstance Classfer Daoqang Zhang 1,, Songcan Chen and Zh-Hua Zhou 1* 1 Natonal Laboratory for Novel Software Technology Nanjng Unversty, Nanjng 193, Chna Department

More information

An Entropy-Based Approach to Integrated Information Needs Assessment

An Entropy-Based Approach to Integrated Information Needs Assessment Dstrbuton Statement A: Approved for publc release; dstrbuton s unlmted. An Entropy-Based Approach to ntegrated nformaton Needs Assessment June 8, 2004 Wllam J. Farrell Lockheed Martn Advanced Technology

More information

Parallel matrix-vector multiplication

Parallel matrix-vector multiplication Appendx A Parallel matrx-vector multplcaton The reduced transton matrx of the three-dmensonal cage model for gel electrophoress, descrbed n secton 3.2, becomes excessvely large for polymer lengths more

More information

arxiv: v3 [cs.ds] 7 Feb 2017

arxiv: v3 [cs.ds] 7 Feb 2017 : A Two-stage Sketch for Data Streams Tong Yang 1, Lngtong Lu 2, Ybo Yan 1, Muhammad Shahzad 3, Yulong Shen 2 Xaomng L 1, Bn Cu 1, Gaogang Xe 4 1 Pekng Unversty, Chna. 2 Xdan Unversty, Chna. 3 North Carolna

More information

USING GRAPHING SKILLS

USING GRAPHING SKILLS Name: BOLOGY: Date: _ Class: USNG GRAPHNG SKLLS NTRODUCTON: Recorded data can be plotted on a graph. A graph s a pctoral representaton of nformaton recorded n a data table. t s used to show a relatonshp

More information

User Authentication Based On Behavioral Mouse Dynamics Biometrics

User Authentication Based On Behavioral Mouse Dynamics Biometrics User Authentcaton Based On Behavoral Mouse Dynamcs Bometrcs Chee-Hyung Yoon Danel Donghyun Km Department of Computer Scence Department of Computer Scence Stanford Unversty Stanford Unversty Stanford, CA

More information

CSE 326: Data Structures Quicksort Comparison Sorting Bound

CSE 326: Data Structures Quicksort Comparison Sorting Bound CSE 326: Data Structures Qucksort Comparson Sortng Bound Steve Setz Wnter 2009 Qucksort Qucksort uses a dvde and conquer strategy, but does not requre the O(N) extra space that MergeSort does. Here s the

More information

An Efficient Garbage Collection for Flash Memory-Based Virtual Memory Systems

An Efficient Garbage Collection for Flash Memory-Based Virtual Memory Systems S. J and D. Shn: An Effcent Garbage Collecton for Flash Memory-Based Vrtual Memory Systems 2355 An Effcent Garbage Collecton for Flash Memory-Based Vrtual Memory Systems Seunggu J and Dongkun Shn, Member,

More information

A Facet Generation Procedure. for solving 0/1 integer programs

A Facet Generation Procedure. for solving 0/1 integer programs A Facet Generaton Procedure for solvng 0/ nteger programs by Gyana R. Parja IBM Corporaton, Poughkeepse, NY 260 Radu Gaddov Emery Worldwde Arlnes, Vandala, Oho 45377 and Wlbert E. Wlhelm Teas A&M Unversty,

More information

An Investigation into Server Parameter Selection for Hierarchical Fixed Priority Pre-emptive Systems

An Investigation into Server Parameter Selection for Hierarchical Fixed Priority Pre-emptive Systems An Investgaton nto Server Parameter Selecton for Herarchcal Fxed Prorty Pre-emptve Systems R.I. Davs and A. Burns Real-Tme Systems Research Group, Department of omputer Scence, Unversty of York, YO10 5DD,

More information

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements

2x x l. Module 3: Element Properties Lecture 4: Lagrange and Serendipity Elements Module 3: Element Propertes Lecture : Lagrange and Serendpty Elements 5 In last lecture note, the nterpolaton functons are derved on the bass of assumed polynomal from Pascal s trangle for the fled varable.

More information

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS

EVALUATION OF THE PERFORMANCES OF ARTIFICIAL BEE COLONY AND INVASIVE WEED OPTIMIZATION ALGORITHMS ON THE MODIFIED BENCHMARK FUNCTIONS Academc Research Internatonal ISS-L: 3-9553, ISS: 3-9944 Vol., o. 3, May 0 EVALUATIO OF THE PERFORMACES OF ARTIFICIAL BEE COLOY AD IVASIVE WEED OPTIMIZATIO ALGORITHMS O THE MODIFIED BECHMARK FUCTIOS Dlay

More information

Technical Report. i-game: An Implicit GTS Allocation Mechanism in IEEE for Time- Sensitive Wireless Sensor Networks

Technical Report. i-game: An Implicit GTS Allocation Mechanism in IEEE for Time- Sensitive Wireless Sensor Networks www.hurray.sep.pp.pt Techncal Report -GAME: An Implct GTS Allocaton Mechansm n IEEE 802.15.4 for Tme- Senstve Wreless Sensor etworks Ans Koubaa Máro Alves Eduardo Tovar TR-060706 Verson: 1.0 Date: Jul

More information

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide

Lobachevsky State University of Nizhni Novgorod. Polyhedron. Quick Start Guide Lobachevsky State Unversty of Nzhn Novgorod Polyhedron Quck Start Gude Nzhn Novgorod 2016 Contents Specfcaton of Polyhedron software... 3 Theoretcal background... 4 1. Interface of Polyhedron... 6 1.1.

More information

Agenda & Reading. Simple If. Decision-Making Statements. COMPSCI 280 S1C Applications Programming. Programming Fundamentals

Agenda & Reading. Simple If. Decision-Making Statements. COMPSCI 280 S1C Applications Programming. Programming Fundamentals Agenda & Readng COMPSCI 8 SC Applcatons Programmng Programmng Fundamentals Control Flow Agenda: Decsonmakng statements: Smple If, Ifelse, nested felse, Select Case s Whle, DoWhle/Untl, For, For Each, Nested

More information

A Background Subtraction for a Vision-based User Interface *

A Background Subtraction for a Vision-based User Interface * A Background Subtracton for a Vson-based User Interface * Dongpyo Hong and Woontack Woo KJIST U-VR Lab. {dhon wwoo}@kjst.ac.kr Abstract In ths paper, we propose a robust and effcent background subtracton

More information

Solving two-person zero-sum game by Matlab

Solving two-person zero-sum game by Matlab Appled Mechancs and Materals Onlne: 2011-02-02 ISSN: 1662-7482, Vols. 50-51, pp 262-265 do:10.4028/www.scentfc.net/amm.50-51.262 2011 Trans Tech Publcatons, Swtzerland Solvng two-person zero-sum game by

More information

TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment

TripS: Automated Multi-tiered Data Placement in a Geo-distributed Cloud Environment TrpS: Automated Mult-tered Data Placement n a Geo-dstrbuted Cloud Envronment Kwangsung Oh, Abhshek Chandra, and Jon Wessman Department of Computer Scence and Engneerng Unversty of Mnnesota Twn Ctes Mnneapols,

More information

Achieving class-based QoS for transactional workloads

Achieving class-based QoS for transactional workloads Achevng class-based QoS for transactonal workloads Banca Schroeder Mor Harchol-Balter Carnege Mellon Unversty Department of Computer Scence Pttsburgh, PA USA @cs.cmu.edu Arun Iyengar Erch

More information

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search

Sequential search. Building Java Programs Chapter 13. Sequential search. Sequential search Sequental search Buldng Java Programs Chapter 13 Searchng and Sortng sequental search: Locates a target value n an array/lst by examnng each element from start to fnsh. How many elements wll t need to

More information

Feature Reduction and Selection

Feature Reduction and Selection Feature Reducton and Selecton Dr. Shuang LIANG School of Software Engneerng TongJ Unversty Fall, 2012 Today s Topcs Introducton Problems of Dmensonalty Feature Reducton Statstc methods Prncpal Components

More information

A Generic and Compositional Framework for Multicore Response Time Analysis

A Generic and Compositional Framework for Multicore Response Time Analysis A Generc and Compostonal Framework for Multcore Response Tme Analyss Sebastan Altmeyer Unversty of Luxembourg Unversty of Amsterdam Clare Maza Grenoble INP Vermag Robert I. Davs Unversty of York INRIA,

More information

Simulation Based Analysis of FAST TCP using OMNET++

Simulation Based Analysis of FAST TCP using OMNET++ Smulaton Based Analyss of FAST TCP usng OMNET++ Umar ul Hassan 04030038@lums.edu.pk Md Term Report CS678 Topcs n Internet Research Sprng, 2006 Introducton Internet traffc s doublng roughly every 3 months

More information

Scheduling. In general, a scheduling scheme provides two features: An algorithm for ordering the use of system resources (in particular the CPUs)

Scheduling. In general, a scheduling scheme provides two features: An algorithm for ordering the use of system resources (in particular the CPUs) Schedulng Goal To understand the role that schedulng and schedulablty analyss plays n predctng that real-tme applcatons meet ther deadlnes Topcs Smple process model The cyclc executve approach Process-based

More information

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations*

Configuration Management in Multi-Context Reconfigurable Systems for Simultaneous Performance and Power Optimizations* Confguraton Management n Mult-Context Reconfgurable Systems for Smultaneous Performance and Power Optmzatons* Rafael Maestre, Mlagros Fernandez Departamento de Arqutectura de Computadores y Automátca Unversdad

More information

Avoiding congestion through dynamic load control

Avoiding congestion through dynamic load control Avodng congeston through dynamc load control Vasl Hnatyshn, Adarshpal S. Seth Department of Computer and Informaton Scences, Unversty of Delaware, Newark, DE 976 ABSTRACT The current best effort approach

More information

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters

Proper Choice of Data Used for the Estimation of Datum Transformation Parameters Proper Choce of Data Used for the Estmaton of Datum Transformaton Parameters Hakan S. KUTOGLU, Turkey Key words: Coordnate systems; transformaton; estmaton, relablty. SUMMARY Advances n technologes and

More information

Efficient Broadcast Disks Program Construction in Asymmetric Communication Environments

Efficient Broadcast Disks Program Construction in Asymmetric Communication Environments Effcent Broadcast Dsks Program Constructon n Asymmetrc Communcaton Envronments Eleftheros Takas, Stefanos Ougaroglou, Petros copoltds Department of Informatcs, Arstotle Unversty of Thessalonk Box 888,

More information

Support Vector Machines

Support Vector Machines Support Vector Machnes Decson surface s a hyperplane (lne n 2D) n feature space (smlar to the Perceptron) Arguably, the most mportant recent dscovery n machne learnng In a nutshell: map the data to a predetermned

More information

A SYSTOLIC APPROACH TO LOOP PARTITIONING AND MAPPING INTO FIXED SIZE DISTRIBUTED MEMORY ARCHITECTURES

A SYSTOLIC APPROACH TO LOOP PARTITIONING AND MAPPING INTO FIXED SIZE DISTRIBUTED MEMORY ARCHITECTURES A SYSOLIC APPROACH O LOOP PARIIONING AND MAPPING INO FIXED SIZE DISRIBUED MEMORY ARCHIECURES Ioanns Drosts, Nektaros Kozrs, George Papakonstantnou and Panayots sanakas Natonal echncal Unversty of Athens

More information

SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR

SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR SENSITIVITY ANALYSIS IN LINEAR PROGRAMMING USING A CALCULATOR Judth Aronow Rchard Jarvnen Independent Consultant Dept of Math/Stat 559 Frost Wnona State Unversty Beaumont, TX 7776 Wnona, MN 55987 aronowju@hal.lamar.edu

More information

BIN XIA et al: AN IMPROVED K-MEANS ALGORITHM BASED ON CLOUD PLATFORM FOR DATA MINING

BIN XIA et al: AN IMPROVED K-MEANS ALGORITHM BASED ON CLOUD PLATFORM FOR DATA MINING An Improved K-means Algorthm based on Cloud Platform for Data Mnng Bn Xa *, Yan Lu 2. School of nformaton and management scence, Henan Agrcultural Unversty, Zhengzhou, Henan 450002, P.R. Chna 2. College

More information