id
stringlengths
7
12
sentence1
stringlengths
5
1.44k
sentence2
stringlengths
6
2.06k
label
stringclasses
4 values
domain
stringclasses
5 values
train_126700
Careful use of slot rendering can eliminate a number of runtime switches.
the OpenCL compute kernels and OpenGL shaders can more effectively utilize the cache on the GPU.
reasoning
Software and its engineering
train_126701
In order to determine the class of a previously unclassi.ed method signature, we merge the rule sets from all classes into one ordered ruleset, optimized for precision.
we perform the following steps: 1.
reasoning
Software and its engineering
train_126702
It is desirable to use such veri.cations as early as possible to prevent the application of other expensive, possibly manual, activities.
an integration of V&V activities within the build script is necessary.
reasoning
Software and its engineering
train_126703
As a result, these elements turn into partial (Pt) or variable elements (V).
the analysed categories (F, Pt and V) provide insights to the developers of which elements should be restructured as potential common or variable elements in the family s source code (Q2).
reasoning
Software and its engineering
train_126704
Views As stated above, views provide an abstraction of the man­aged actors state for the managers.
the views must be observed by the managers.
reasoning
Software and its engineering
train_126705
Then s -~a s' if and only ifs ~N-~ s'.
n the operational semantics presented in Section 4.3 is compositional.
reasoning
Software and its engineering
train_126706
The great advantage of mathematics is that the rules are simpler than those of natural language, and the vocabulary is much smaller.
when presented with something unfamiliar it is possible to work out a solution for yourself, by logical deduction and invention rather than by consulting books or experts.
reasoning
Software and its engineering
train_126707
It can be decided at compile time that the second inner loop, which covers three rows ofA and one rowofB,is the largest loop that can residein the uncom­pressed region.
handle checks are hoisted to the beginning of the second inner loop, as shown in (b).
reasoning
Software and its engineering
train_126708
In the deterministic case we can assume that object instances departing from similar states process events in a similar form leaving a similar final state.
if we can determine the next object state, why not do it in each peer or client?
reasoning
Software and its engineering
train_126709
However, it is easier to study the fundamentals of labeled control-.ow points and harmless advice in a setting with the simplest possible pcds.
we will start our investigation in a setting where pcds are simply sets of labels {l1,...,lk}and advice is written as {{l1,...,lk}.x .e}.
reasoning
Software and its engineering
train_126710
In essence, Smalltalk and Java share the common object-oriented execution model where objects interoperate by sending messages.
lan­guage interoperability can best be established at this level, Figure 1: Smalltalk interacts with Java through the languages FFI interfaces with C. buf := JavaWorld.java.io.BufferedReader new_Reader: (JavaWorld.java.io.FileReader new_String: sometext.txt ).
reasoning
Software and its engineering
train_126711
In his classic paper on the essence and accidents of software engineering, Brooks stated that the hardest single part of building a sofware system is deciding precisely what to build...
the most important function that the software builder performs for the client is the iterative extraction and re.nement of the Permission to make digital or hard copies of part or all of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page.
reasoning
Software and its engineering
train_126712
We have used a similar approach to generate programs that solve physics problems [26] and to combine and specialize generic soft­ware components to form a variety of application programs [30].
we believe that the approach described here generalizes to program generation from components in a variety of areas.
reasoning
Software and its engineering
train_126713
Hence, the fastest agents collect at least M ·|F| uncollected values during the .rst cvmin events (rather than during the .rst 2 · cvmin events).
after X := 2 · cvmin ·r(n-|F|-1) l- cvmin events (denoted M·|F|by a segment e'), all the values are collected.
reasoning
Software and its engineering
train_126714
Your skill should tell you if it is ugly.
when you refactor, bring the redundancies together using an appropriate object/functional design.
reasoning
Software and its engineering
train_126715
Feature model con.guration gets even more complicated if we consider that this activity can be performed not only by a single stakeholder, but also by multiple stakeholders ge­ographically distributed [15, 12, 8].
this calls for approaches that support this con.guration process.
reasoning
Software and its engineering
train_126716
Thirdly, the sub-domains of a product line architecture may use, change and manage their own technologies, and evolve separately but in a controlled way.
heterogeneous technologies can be used in a product line, and components may be covered by COTS components, ordered from OCMs or be allocated to subcontractors.
reasoning
Software and its engineering
train_126717
Similarly, due to the re.exive M-edge (v, M, v), there al­ways exists a G1-edge (v, G1, w) for each (v, G2(4), w), where G2(4) = a (M?a)*M?.
it is similar to category (3).
reasoning
Software and its engineering
train_126718
)(Ol[t - 1] - µl[t - 1]) , (12) By repeated application of (12) we get Ol[t] - µl[t] = (1 - .
)t (Ol[0] - µl[0]) (13) for a given E> 0, if Ol[0] - µl[0] t> log1/(1-.)
reasoning
Software and its engineering
train_126719
Ac­cording to our .rst design principle, thehypervisor should be as small as possible, because it belongs to the trusted computing base of every other component in the system.
we implemented policies and all functionality that is neither security-nor performance-critical outside the hypervisor.
reasoning
Software and its engineering
train_126720
Optimizations, by contrast, have no clear ordering and are prone to phase ordering problems.
they need to be combined for max­imum effectiveness.
reasoning
Software and its engineering
train_126721
4.1 Challenges Because of the many variants of platform-and context­speci.c adaptations that may be necessary in arbitrary ap­plications there is an exponential number of possible com­binations.
implementing adaptations for speci.c combinations is prohibitive.
reasoning
Software and its engineering
train_126722
x)) The idea behind these de.nitions is to implement each state monad of the type ST (a,t) as a function that takes a witness and the region a as arguments and returns a witness, the region a, and a value of the type t. It is easy to see that if a state monad program is well-typed by the monadic type system, then it is also well-typed with our type system using the above de.nitions for the monadic primitives.
our approach is more expressive than monads.
reasoning
Software and its engineering
train_126723
Resource Discovery Framework Self-Chord Routing Information and LRU Policy Metadata Snapshots in Grid Communities Type of Grid Peer-to-Peer Hierarchical Hierarchical Ant travels on a graph Yes, moves along a logical ring Yes, moves on a hierarchical tree Yes, moves on a hierarchical tree Ant memory No Yes Yes Ant start state Yes Yes Yes Ant termination state Yes Yes Yes Ant transition rules Yes Yes Yes Ant probabilistic decision rule No No No Use of pheromones No No Yes, by way of Self-led Critical Friends Pheromone decay No No No Mobility Yes Yes Yes Autonomy Yes Yes Yes Communication Limited, with nodes only Limited, with nodes and LRU Limited, with nodes and metadata snapshots Adaptation Yes Yes Yes Persistence No No No Local interaction Limited, with nodes only Limited, with nodes and LRU Limited, with nodes and metadata snapshots in this method suggests that the concept of stigmergy is not adhered to.
indirect communications between ants are weak and there is no exploitation of global knowledge.
reasoning
Software and its engineering
train_126724
In contrast, Velodrome (our prototype checker for Java) detects 133 non-atomic methods (and misses the remaining 21 by not generalizing from the observed traces), but produces zero false alarms.
in the termi­nology of information retrieval, Velodrome provides slightly less recall but vastly increased precision a trade-off that we believe is quite appropriate for all but the most safety-critical applications.
reasoning
Software and its engineering
train_126725
The latter can be more powerful and is necessary for higher-order test generation to always subsume sound concretization, as will be shown in the next section.
we adopt this second option in what follows.
reasoning
Software and its engineering
train_126726
However, we have made it su.­ciently clear in the introduction that our approach is quite di.erent from existing ones.
we deliberately do not compare our work with the existing literature.
reasoning
Software and its engineering
train_126727
It needs to be noted that we may not always select the minimum RPM level for a given array access because there may not be suf.cient number of iterations in the loop nest where this array reference appears2 .
at the end of this .rst step of our approach, we determine the preferable disk speed for each array reference.
reasoning
Software and its engineering
train_126728
Ei New host states MNH : Extensions do interact with the host language by including host language constructs in the extension­introduced constructs.
productions in PEi may include host language nonterminals and terminals on the right hand side.
reasoning
Software and its engineering
train_126729
We have additionally modi.ed the SU to incorporate task malleability.
the task option value is given as Cij vij = .iNi (8) Ci where .i is the priority, Ci is the maximum number of pro­cessors requested, and Cij is the number of processors re­quested in option j.
reasoning
Software and its engineering
train_126730
For the literal part, f(x) = in the example, we had to choose a language that a humanwould easily type, and LATEX looked like the only reason­able choice: MathML was ruled out by its verbosity.
although we de.ned few quotation kinds for documents, anadded complexity came from more levels of possible nesting,as a consequence of three input kinds that can contribute totexts in documents: literal strings, playing a role akin to the character data (CDATA) of XML, to write natural-language text; literal LATEX code fragments, which appear as escapesfrom literal strings to denote a mathematical formula; symbolic code fragments, which appear as escapes fromLATEX code to denote a symbolic calculation whose resulthas to be displayed.
reasoning
Software and its engineering
train_126731
The entities described here are exhaustive and conform to the SIP philosophy.
the use of SIP makes this architecture more useful and interoperable with any other SIP clients or servers.
reasoning
Software and its engineering
train_126732
Invisible watermarking (or simply watermarking) has become a well-accepted technique for protecting digital im­ages and other forms of multimedia content, and the search for better watermarking methods remains an extremely ac­tive area of research [4, 5, 9, 14, 18, 22].
as satellite imagery and GIS have found ever increasing commercial ap­plicability, watermarking is invariably being considered for use in these applications.
reasoning
Software and its engineering
train_126733
We denote this portion of cost as ..
the total cost for redistribution is: Cost=(a+f.W)+.
reasoning
Software and its engineering
train_126734
Baker s original collector copies objects atomically when it evacuates them from fromspace to tospace.
the duration of pauses caused by copy­ing depends on the size of objects and is not bounded by a small constant.
reasoning
Software and its engineering
train_126735
A void result (Ø) is assigned to the op­posite port automatically.
the then and the else output ports receive a nested array with the same struc­ture and size, as de.ned by the input nesting levels, ports depths and iteration strategies used, but complementary in the sense that if a value is available in one of the ouput, the corresponding item is empty in the other output and vice versa.
reasoning
Software and its engineering
train_126736
Considering that both the ATM and the Card Protocol are deterministic it is easy to see that the composite PLTS can never exhibit the trace If we compute the composite model and minimise with respect to the strong semantic equivalence (defined in the previous section), the resulting PLTS will no longer have the following pairs of undefined behaviour (compare with Table 2): {(insertCard, 2), (insertCard, 9), (insertCard, 10), (ejectCard, 3), (ejectCard, 4), (ejectCard, 5), (ejectCard, 6), (ejectCard, 7), (requestTakeCard, 10), (takeCard, 9)}.
we now have a composite behaviour model that has fewer gaps requiring stakeholder intervention.
reasoning
Software and its engineering
train_126737
5 shows an example of a client that uses a Counter object c. During the initialisation phase of the object c the PointsTo(c.data, 1, 0) predicate is obtained from which the permission part, Perm(c.data, 1, 0), is transferred into the lock.
the client obtains the Hist predicate for data (line 4).
reasoning
Software and its engineering
train_126738
Our test sequence data structure saves the results of the backward tracking search which generated the path from the hot spot to the start node, including all the edges that had already been searched.
it is possible to reverse the search and return to a point where more than one choice of edge is available.
reasoning
Software and its engineering
train_126739
In general, we want to minimize the number of sanitizers we infer.
if there are several solutions to the set of constraints B1, we want to favor solutions that infer fewer sanitizers, while satisfy­ing B1, with higher probability.
reasoning
Software and its engineering
train_126740
A fault is de.ned as a change to an individual executable .le in response to a fault MR.
if a given fault MR causes N di.erent .les to be changed, then we count that as N distinct faults.
reasoning
Software and its engineering
train_126741
Finally, we do not require users to be aware of the par­titioning scheme or the transaction execution model when writing transaction code.
users should be able to write a single atomic block, and the system would make sure the appropriate code branches will execute at the corresponding partitions.
reasoning
Software and its engineering
train_126742
Changes in the execution envi­ronment could modify the program behavior in ways that our technique would not be able to identify.
we assume that the execution environment is the same across versions.
reasoning
Software and its engineering
train_126743
the application and various controllers for the connected hardware, need to be changed in order to facilitate the new self-checkout.
the new compo­nent RFIDController must be connected to the event bus giving rise to new ports and changes to the event types.
reasoning
Software and its engineering
train_126744
Even though this strategy is slower than our counter-based implemen­tation, it is still at the same level with abc.
we expect all applications that are making use of cflow pointcuts to bene.t from our approach, even if they are using a large number of pointcuts.
reasoning
Software and its engineering
train_126745
This, however, is an issue that can be directly addressed in an opinion poll.
metrics can be refined and misleading interpretation can be avoided.
reasoning
Software and its engineering
train_126746
However, users might want a more global metric for yield.
figure 8 plots the network-wide total recovered energy for the experiments of Figure 7.
reasoning
Software and its engineering
train_126747
In real-time systems, the penalty for missing a task dead- line is enormous.
it is critical in real-time sys- tems to have I/O devices powered up and running at the correct times to guarantee that all application tasks meet their deadlines.
reasoning
Software and its engineering
train_126748
The reduction in the number of .ows per peer results in an increase in the per-.ow download rate of 41.5%.
while the total number of .ows in the system has shrunk by 20.7%, the BitTorrent traf.c has had a net increase of 12.1% between 2009 and 2010.
reasoning
Software and its engineering
train_126749
As mentioned in Section 2.1, for well-structured uses of locks and unlocks, such a reordering is safe.
this fence represents un­necessary overhead, possibly doubling the cost of lock acquisition, without bene.t to reasonably written code.
reasoning
Software and its engineering
train_126750
When developers change a target program, they also change the unit test cases associated to the program so that all the unit test cases succeed.
test-based pointcuts automatically re.ect the changes to the target program as long as the associated tests are maintained.
reasoning
Software and its engineering
train_126751
If process r later on fails to capture v.lock, because the lock is still owned by process p, then r obtains the pointer ptr to Wp.
r can call Wp.Wait() in order to wait until the lock has been released.
reasoning
Software and its engineering
train_126752
)Typically, JML speci.ca­tions will either not exist at all, or only exist for a small fraction of the methods in an application and its libraries.
eSC/Java tries to be useful even with partial infor­mation.
reasoning
Software and its engineering
train_126753
However, OCEAN restricts the Finite Number of Parameters Infinite Parameters Space C1 Arbitrary Performance Prediction Model Undecidable Problem C2 Arbitrary Polynomial Performance Prediction Model Undecidable Problem C3 Fixed Polynomial Performance Model Finite Parameters Space Decidable Problem Ex: ATLAS [19] C4 Finite Parameters Space Decidable Problem Ex : OCEAN project [2] Figure 2: Classes of Best-Parameters Problems exploration to .nite set of parameter intervals.
the number of parameter combinations becomes .nite, allowing a triv­ial exhaustive search of the best parameter values: each optimized program resulting from a particular value of the optimization pa­rameters is generated and evaluated.
reasoning
Software and its engineering
train_126754
This is achieved by processing context-related and other messages one at a time, picking them up from the agent mailbox.
it is not possible that context-related messages interfere with the execution activated by a standard message.
reasoning
Software and its engineering
train_126755
Re­gardless of additional available information, the developer will always be interested in the source code itself.
in this paper we focus on static information available from the source code, in the form of programmer-de.ned structure and lexical identi.er names.
reasoning
Software and its engineering
train_126756
This is why the human-in-the-loop approach was implemented, along with the other more automated parts.
the user can pick a specific aircraft that he/she wants to pilot.
reasoning
Software and its engineering
train_126757
Given Lemma 2, the net element translation rules can be applied to all N in MAS.
pROG is a complete translation of the multi-agent net MAS.
reasoning
Software and its engineering
train_126758
Most common tools used by developers, including compilers, debuggers, and pro- fliers do not support the sort of fine-grained storage systems provided by a VSC system.
the system must provide some mechanism for externalizing sets of fine-grained arti- facts as standard source files for manipulation by conven- tional tools.
reasoning
Software and its engineering
train_126759
Moreover, at the time the signal turns green, a GLOSA­enabled vehicle will be cruising through the intersection with an initial non-zero speed, as opposed to a regular vehicle that would have to start from a complete halt.
gLOSA may im­prove an individual vehicle s travel time.
reasoning
Software and its engineering
train_126760
Second, by using a deep embedding, the construction of the monadic computation is separated from its interpretation.
multiple interpretations can be given to a computa­tion [2, 24].
reasoning
Software and its engineering
train_126761
Error codes are represented as integers but conceptu­ally they are atomic values on which arithmetic operations are meaningless.
if op is an arithmetic or bitwise operation, then we can safely assume that e1 and e2 do not contain errors.
reasoning
Software and its engineering
train_126762
The checker had an enormous number of false positives, most due to a single limitation of our analysis: a small number of functions only freed one argument based on the value of another argument, but our analysis decided that these functions always freed their ar­gument.
rather than having an error rate of one error per few hundred callsites, these functions had rates closer to .fty errors per hundred callsites.
reasoning
Software and its engineering
train_126763
The MTF is a per-HVM .ag that is global to the VM, and unlike like the TF bit in the EFLAGS register, the MTF is not changed at thread context switches.
in an HVM guest, a single-step begun in one thread or context might continue into another thread or context.
reasoning
Software and its engineering
train_126764
A quanti.ed boolean expression has exactly one quanti.er, while a quanti.ed reachability or an unreachability expression can have up to two quanti.ers (one in the source and one in the destination expression).
we use two expansion functions: one for each quanti.er.
reasoning
Software and its engineering
train_126765
[3] have found that aspect-instance look-up took more than 26% of the total execution time in selected use cases.
it is important to research such policies and optimization opportunities.
reasoning
Software and its engineering
train_126766
However, since a fault can be trig­gered in many ways, failing traces due to the same fault can be quite di.erent.
previous methods actu­ally group together traces exhibiting similar behaviors, like similar branch coverage, rather than traces due to the same fault.
reasoning
Software and its engineering
train_126767
Thumb-1 Instructions Overlap ARM Instructions (a) (b) (c) (d) Figure 8: Four kinds of illegal instructions instruction.
these bytes cannot be disassembled as an instruction.
reasoning
Software and its engineering
train_126768
From (6.1) none of the processes in Gi(D)= Gi(E) -{a} is in cov(a, EII).
from part (a) of the induction hypothesis all of the processes in Gi(D)= Gi(E) -{a} are in the same state at the end of EII as at the end of DII, so if a process other than a is poised to access r at the end of EII it is also poised to access r at the end of DII .
reasoning
Software and its engineering
train_126769
Since we can only read ele­ments of type Object from this list, we can say that this list be­haves almost exactly the same as a list of type List.
it will be a good idea to check whether type con.icts caused by the wildcard ?
reasoning
Software and its engineering
train_126770
approaches as selective mutation.
when detecting a software fault, developers need to Recently, a fault localization approach based on mutation identify the faulty program parts in the application s source code analysis has been proposed [22].
reasoning
Software and its engineering
train_126771
Our notion of respon­siveness is weaker than synchronous languages like Esterel [4] (which require input events to be handled synchronously during every clock tick, and are hence too strong to be implemented in asynchronous software), but stronger than purely asynchronous languages like Rhapsody [12] (where asynchronous events can be queued arbitrarily long before being handled).
our notion of responsiveness lies in an interesting design point between syn­chrony and asynchrony.
reasoning
Software and its engineering
train_126772
The access to RESTful Web services uses standard HTTP protocol (GET, POST, PUT, and DELETE).
existing Web technologies can be leveraged to facilitate RESTful Web service development, security, and performance.
reasoning
Software and its engineering
train_126773
ADAPTIVE CONTROL OF px AND pm Biological evolution shows that px and pm should be adapted and should depend on the evolution state[5].
in order to enhance Copyright is held by the author/owner(s).
reasoning
Software and its engineering
train_126774
A Without any assumptions on the mobility pattern of the nodes, it is impossible to guarantee the existence of JIT paths.
we are motivated to assume that every exe­cution satis.es the following property, which was originally introduced in [2] and is also related to the dynamic diame­ter of [11].
reasoning
Software and its engineering
train_126775
If a .eld fd of an ob ject o is guarded by thisThread, it is accessed by only one thread and hence cannot be involved in a data race.
no lock is required to access fd.
reasoning
Software and its engineering
train_126776
Furthermore, as target models can be platform-independent, the kinds of target models used in our approach are rather de.ned by autonomic managers than by the underlying infrastructure.
autonomic managers preferably use target mod­els than a complex source model to monitor or adapt a sys­tem.
reasoning
Software and its engineering
train_126777
This was also the case when the cluster size is larger than three.
we set the minimum cluster size for our approach to be applied to three in our experiments.
reasoning
Software and its engineering
train_126778
Thus, any method that is matched by the reference re.ective block s pattern is matched by the declaration re.ective block s pattern, regardless of what Y is.
there is always a method n in LogMe.
reasoning
Software and its engineering
train_126779
RELATEDWORK Real-timegarbage collection research datesbacktothe 1970s where collectorsforLISPandMLhavebeendeveloped.
avast number of papers on real-timegarbage collection havebeen pub­lished.
reasoning
Software and its engineering
train_126780
In the GEMM example, (s4) s shape is a pre.x of (s1).
they can be fused into a larger unit.
reasoning
Software and its engineering
train_126781
For example, a high-level task may require two cores, but may or may not care which cores they are.
we in­clude in Poli-C a way of specifying these sorts of requests, which we call policies.
reasoning
Software and its engineering
train_126782
This rule is e.ective in moving down the search space toward lower energy regions, like water from a stream running into an empty reservoir, while .lling local depressions along the way.
these tactical procedures of the CONFLEX search are referred to as the Reservoir Fil ling Algorithm.
reasoning
Software and its engineering
train_126783
In both approaches, we sample a constant number of data points (i.e., tasks) and perform the analysis treating the sample set as representative of the entire set of points.
we avoid having an unmanageable linear algorithmic complexity with respect to the number of points; both ap­proaches scale with a complexity of the log of the number of tasks.
reasoning
Software and its engineering
train_126784
Results and Analysis The central features of the four techniques studied in our experiment are 1) whether the four techniques can be effective in reducing the size of the test suite; and 2) whether the four techniques can perform their tasks in acceptable time.
our comparison concentrated on the following things: 3.1 Representative Sets The sizes of the representative sets are depicted in Fig.
reasoning
Software and its engineering
train_126785
Instead of focusing on the setup of a project, developers should be able to rapidly prototype media rich applications.
the time to get a project up and running should be kept short, i.e., through the use of easily installable and useable templates.
reasoning
Software and its engineering
train_126786
The reason for Mitosis not being as good as our technique is that it only uses one version of each iteration.
if more than one hot path exists, which is true for most benchmarks as indicated in Fig.
reasoning
Software and its engineering
train_126787
Unfortunately, comput­ing multiple versions of information for a method causes the method to be analyzed many times.
it may signif­icantly increase the client s cost.
reasoning
Software and its engineering
train_126788
2An element of a partition is called block.
in this paper, ablock refers to a set of PDG nodes.
reasoning
Software and its engineering
train_126789
However, for computing the relevant slice for this input, the predicate at line 5 will be identi.ed as a potential dependence since if it had evaluated to false outcome, the value of variable a output at line 14 could be affected.
the relevant slice will contain the statements at lines 5, 3 and 1 in addition to the statement at line 4 that is in the full dynamic slice.
reasoning
Software and its engineering
train_126790
This implies that p can write a message to q s local memory andspinonits localmemoryuntil q responds (and vice versa).
after the establishment of the link, p and q can e.ciently execute a reliable two-way communication protocol.
reasoning
Software and its engineering
train_126791
The forthcoming BPMN version 2.0 promises to bridge the gap between the notation and the real execution of a business process, therefore rendering WSBPEL obsolete.
wSBPEL will not be regarded in the further discussion.
reasoning
Software and its engineering
train_126792
The services modelled describing their e.ects make user­environment interaction more natural and simpler because they allow the user to be isolated from the operational cha­racteristics and the service implementation, as well as from the functional features and what it does.
the user can focus exclusively on the e.ects that the service produces in the environment.
reasoning
Software and its engineering
train_126793
Interoperability of sharing knowledge base using MLM depends on healthcare standard knowledge representation and on standard terminologies usage.
our system covers interoperability by using SNOMED CT standard terminologies and knowledge representation in standard MLM.
reasoning
Software and its engineering
train_126794
However, there is no indication in the literature that constructivism has been used in the VE au­thoring context before.
this project includes a plan to test the tool with novice designers, in order to pro­vide empirical evidence of the success of constructivism for novice VE designers.
reasoning
Software and its engineering
train_126795
Thus, I(i ' ,j) was already captured at Tj ' , and we get that fCi ' ,j (P,S) = 0.
si.i ' ,j will increase fC (P,S), since fC (P,Si.i ' ,j )=fC i,j (P,S)+ fC = (P,S) - fCi ' ,j (P,S) fC (P,S)+1 If the .rst case does not apply, fCi ' ,j (P,S) = 1 and Si.i ' ,j will preserve fC (P,S), since fC (P,Si.i ' ,j )= fC (P,S)+ fC fC fC i,j (P,S) - fCi ' ,j (P,S)= (P,S)+1 - 1= (P,S).
reasoning
Software and its engineering
train_126796
The semantics of failure recovery are controlled by the adapter so that they may vary from user to user.
recovery is discussed below in section 6.
reasoning
Software and its engineering
train_126797
While this option needs to be available for faculty, we quickly found out that it is unrealistic to expect students who are in the process of learning about data structures to be able to modify a separate viewer class in order to see an instance of their own data structure.
we turned our attention to building a mechanism that could determine if an instance was a linked list or binary tree based on a set of heuristics, and then automatically generate an appropriate view.
reasoning
Software and its engineering
train_126798
The procedure just described is implemented by relatively simple combinatorial logic that does not extend the critical path in the execute stage.
steele-style address generation in hardware is as fast as standard address generation.
reasoning
Software and its engineering
train_126799
During replay, FuzzSim outputs a timestamp whenever it .nds a new bug.
we can easily plot and compare di.erent scheduling algorithms by comparing the number of bugs produced under the same time budget.
reasoning
Software and its engineering