Adaptive Algorithms using Bounded Memory are Inherently Non-Uniform
Adaptive Algorithms using Bounded Memory are Inherently Non-Uniform
Burkhard Englert
California State University Long Beach,Dept.of Comp.Engr.&Comp.Science
Long Beach,CA90840
Email:benglert@https://www.wendangku.net/doc/cd1640909.html,.
Abstract.Distributed protocols that run in dynamic environments such
as the Internet are often not able to use an upper bound on the num-
ber of potentially participating processes.In these settings adaptive and
uniform algorithms are desirable where the step complexity of all opera-
tions is a function of the number of concurrently participating processes
(adaptive)and the algorithm does not need to know an upper bound on
the number of participating processes(uniform).Adaptive algorithms,
however,are generally not adaptive with respect to their memory con-
sumption-if no upper bound on the number of participating processes
is known in advance-they require unbounded MWMR registers and an
unbounded number of such registers(even if only?nitely many distinct
processes appear),making them impractical for real systems.In this pa-
per we ask whether this must be the case:Can adaptive algorithms where
no upper bound on the number of participating processes is known in
advance be uniformly implemented with?nite memory(if only?nitely
many distinct processes keep reappearing)?We will show that in the dy-
namic setting it is impossible to implement long-lived adaptive splitters,
collect and renaming with in?nitely many bounded MWMR registers,
making such adaptive algorithms impractical in dynamic settings.On the
positive side we provide algorithms that implement a long-lived uniform
adaptive splitter if unbounded registers are available and that implement
a non-uniform adaptive splitter with?nitely many bounded registers if
an upper bound on the number of participating processes is known in
advance.
1Introduction
Many well known and important distributed algorithms such as atomic snapshot or renaming,require processes to gather information about each other.For ex-ample,in the renaming problem,before choosing a unique new name,processes need to know which names other processes have already chosen.To communicate this information an array of Single-Writer Multi-Reader(SWMR)registers can be used.Each process has a unique array entry assigned to it and only a?xed process is allowed to write to each array location while all processes can read Research supported by the Dept.of Transportation under METRANS USC-111699.
them.A process can update information about itself by writing into its entry and it can then collect information about the other processes by reading all entries in an arbitrary order.
Such a collect algorithm with step complexity O(N),however,where N is the total number of processes in the system,is possibly ine?cient or impractical if only few of the N processes are actually participating or if no upper bound on the number of participating processes is known in advance.This and other similar issues motivated researchers to look for adaptive and uniform algorithms whose step complexity only depends on the number of concurrently participating processes(adaptive)and that don’t need to know an upper bound on the number of participating processes in advance(uniform).
Adaptive algorithms have a worst case step complexity that is bounded by a function of the number of concurrently participating,or actually active pro-cesses[3].Motivated by Lamport’s MX algorithm[18],many such adaptive al-gorithms have since been designed[1,3–5,10,11,14,19].The strongest forms of adaptiveness in the read/write shared memory model have been de?ned and achieved in recently presented long-lived adaptive collect[6]and renaming[1,12] algorithms.In these algorithms,called adaptive to point contention the number of steps taken by a process executing an operation is a function of the maxi-mum number of processes that were active simultaneously at some point in time during this operations execution interval.Algorithms adaptive to interval con-tention have a slightly weaker level of adaptiveness.Here the number of steps taken during an operation is a function of the total number of di?erent processes active during this operations execution interval.Finally,an algorithm is adaptive to total contention if the number of steps taken by a process is a function of the total number of processes active since the beginning of the execution.
In all these algorithms[1,3–7,11–13,15,16,19,20],however,the memory con-sumption is a function of the upper bound N on the number of processes that might participate in the algorithm.Moreover these algorithms usually assume unbounded MWMR registers.They are not concerned with a distributed sys-tem such as the Internet,where we have a potentially huge number of processes that might participate in some protocol but it is known that with very high probability only a small number of processes will be active at any given time or participate.It is unrealistic and wasteful for such a system to provide a huge number of shared memory registers for the operation of such a protocol.Also in a real system unbounded MWMR registers are not available.Algorithms that operate in a dynamic setting are not able to use a priori knowledge about a?nite upper bound on the number of processes in the system and are called uniform algorithms.Aguilera,Englert and Gafni[8]showed that there are single shot tasks such as generalized weak test and set[9]that cannot be solved uniformly with a?nite number of MWMR registers.In other words a protocol solving this task with?nitely many MWMR registers must know the number of participating processes in advance.Since generalized weak test and set is a one-shot algorithm, this implies that the long lived nature of test and set and the requirement that the step complexity adapt to interval contention are not the only requirements that preclude a solution in?nite space.
So far all known adaptive algorithms(e.g.[1,3–7,11–13,15,16,19,20])re-quire knowledge of an upper bound on the number of participating processes. This upper bound must be hard coded into the algorithm,making them non-uniform.They are not adaptive with respect to their memory consumption.To summarize,this lack of adaptiveness and non-uniformness has two aspects.
1.Unbounded MWMR registers are needed,greatly limiting their usability in
real systems.
2.The algorithms must a priori know N,an upper bound on the number of
participating processes.
These algorithms are non-uniform since for each possible number of participating processes a”di?erent”algorithm is required.
It would be desirable to have uniform algorithms that can”on the?y”adjust to the number of participating processes in such a way that no matter how many are actually participating,as long as the number is?nite,the algorithm-even in an unbounded execution will use only?nitely many bounded MWMR registers. To our knowledge,no such algorithm exists.
So we ask if this apparent lack of uniformity and reliance on unbounded memory is an inherent property of adaptive algorithms.What is the relationship between uniform,adaptive algorithms and bounded memory?To get an answer to this question we will investigate whether it is possible to have truly adaptive algorithms using bounded memory in a setting where no upper bound N on the number of participating processes is known.In an execution where only?nitely many distinct processes appear can we implement adaptive algorithms using ?nitely many bounded MWMR registers?
In other words are there uniform adaptive algorithms that can use bounded memory in in?nite executions?This is an important question since its answer has strong implications on the practicality of adaptive algorithms in dynamic settings.
Implementation possible ∞-many bounded reg.,uniform impl.No
Fin.many bounded reg,N known,non-uniform impl.Yes ∞-many unbounded reg.,uniform impl.Yes
Table1:Implementability of long-lived,adaptive splitter,collect,renaming.
We will?rst show that even if we provide in?nitely many bounded MWMR registers it is impossible to implement weak test&set in an execution where N is unknown.By reduction this implies that the use of at least?nitely many unbounded registers is a necessary ingredient for uniform and adaptive imple-mentations of long lived and adaptive splitter,collect and renaming.In other words,adaptive algorithms that want to be uniform must either use unbounded registers as a black box or have some a priori knowledge about the upper bound of possibly participating processes.This makes them impractical in dynamic settings.As it turns out they require a closed and controlled environment to be
truly e?ective.On the positive side we present a long-lived,uniform and adaptive to interval contention implementation of a splitter that in an unbounded exe-cution(where only?nitely many distinct processes appear)uses in?nitely many unbounded MWMR registers.We?nally show how this algorithm can be modi-?ed to use bounded memory if an upper bound on the number of participating processes is known in advance(making the resulting algorithm non-uniform). The results are summarized in Table1.
1.1Adaptive splitter
The splitter that we will implement is an essential building block of many of these adaptive renaming,snapshot and collect algorithms(e.g.[1,3–5,10,11,14, 19]).Splitters were?rst introduced by Moir and Anderson[20].A splitter is a variant of mutual exclusion.If one process accesses the splitter alone it captures the splitter and the resource(name)associated with it.When several processes access the splitter concurrently they may all fail leaving the splitter un-captured. The splitter wait-freely partitions the processes that fail into two groups,right and down.At the same time the splitter ensures that not all contending processes go down and not all go to the right.If splitters are put into a grid then processes that begin in the top left are split into smaller and smaller groups as they traverse the splitter until individual processes access a splitter on their own,guaranteeing them success.Our splitter will have fewer properties than the Moir Anderson[20] splitter but it has a long-lived and adaptive implementation.
1.2Related Work
Englert and Goldstein[17]recently considered protocols that they called memory-adaptive.In such protocols processes are at all times only allowed to write to MWMR registers whose index is a function of the contention during the previous shared memory write operation.They showed that in a system with in?nitely many MWMR registers and in?nitely many SWMR registers,for any constant d,there exists a number N d such that if N d processes are allowed to participate in a memory-adaptive(to POINT contention)execution of the protocol,then at least one does not make a single uncovered write to a shared register in d writes.In other words they showed that under these conditions pro-cesses cannot memory-adaptively store a value in shared memory.This implies that any time-adaptive collect or renaming algorithm in this setting that uses only?nitely many of the in?nitely many MWMR registers(if such an algorithm exists)cannot be built from memory-adaptive building blocks alone.
Note that our model used in this paper is more general than the model in[17]. We allow processes to write to registers with any index.We only require that in an unbounded execution-as long as only?nitely many processes appear-only?nitely many such registers are actually written to.Hence the results pre-sented in[17]do not answer the question whether there exists a long-lived,step complexity-adaptive(not necessarily memory-adaptive)and uniform renaming or collect protocol that uses only?nitely many bounded MWMR registers.
1.3Contributions and Paper Organization
In Section2we introduce our model.Our contributions are as follows.
I.In Section3we show that using only bounded shared MWMR registers there
is no adaptive,uniform implementation of long-lived splitter,collect and renaming in an unbounded execution with?nitely many processes appearing. II.In Section4we provide a uniform algorithm adaptive to interval contention that implements a splitter and uses in?nitely many unbounded MWMR registers in an unbounded execution.
III.In Section5we illustrate that using a previously presented algorithm[7] and given a bound on N the largest id of participating processes we can implement a non-uniform,long-lived,adaptive splitter with?nitely many bounded MWMR registers(in an execution where only?nitely many pro-cesses participate).
We conclude with some?nal remarks in Section6.
2Model and Preliminaries
Our algorithms assume an asynchronous read/write shared memory model.This model consists of a set of N asynchronous processes p0,p1,...,p N?1and a set of registers shared by the processes.The processes communicate only through the registers which provide two atomic operations,read and write.We assume single-writer,multi-reader(SWMR)and multi-writer,multi-reader registers(MWMR).
A bounded register of size M is a register that can hold at most M distinct values. An unbounded register is a register for which no such bound exists.A protocol is called long-lived if processes are able to execute it in?nitely often.Algorithms that are not able to use a priori knowledge about a?nite upper bound on the number of processes in the system are called uniform algorithms.
Letαbe an execution of a long-lived algorithm A andα a pre?x ofα. Process p i is participating at the end ofα ,ifα includes an invocation of some operation of A by process p i without the matching response.The active processes at the end ofα ,are denoted Cont(α ).It is the set of processes participating at the end ofα .Given a subsequenceβofα,letα βbe the shortest pre?x ofαthat containsβ.We de?ne the interval contention denoted IntCont(β)and point contention denoted PntCont(β)as follows:
IntCont(β)=|
Cont(α β )|
α β pre?x ofα β
PntCont(β)=max|Cont(α β )|
α β pre?x ofα β
Intuitively the interval contention of a subsequenceβis the number of dif-ferent processes that were participating duringβ,while the point contention is the maximum number of processes active at any point in time duringβ.Clearly PntCont(β)≤IntCont(β).
We de?ne the total contention of any executionαas the number of processes that took steps inα.
Let op be an operation.We de?ne the execution interval of op,denoted β(op)as the subsequence ofαstarting at the invocation of op and ending at the completion of op.The interval contention of an operation op is de?ned as IntCont(β(op)).In the rest of the paper,k denotes IntCont(β(op))for some oper-ation op.The step complexity of an algorithm is adaptive to interval contention if there is a bounded function S,such that the number of steps performed by any process p i in any execution interval of an operation op i of A is at most S(IntCont(β(op i))).Clearly the contention of an execution interval is bounded by N,the total number of processes participating in the algorithm.We will assume that this bound is a priori unknown,that is that our algorithms have no knowledge of this bound in advance.Hence,in an adaptive algorithm with bounded concurrency any operation op i(by p i)terminates with a bounded num-ber of steps,regardless of the actions of any other processes(than p i).So any adaptive algorithm is by de?nition wait-free.
Informally a splitter is a weak mutual exclusion primitive.Processes access the splitter through an invoke operation.This operation has three possible re-sponses:stop,right or down.A process that receives a stop response has captured the splitter.Such a process releases the splitter by invoking a release operation that has only one possible response:done.
In this paper we will only consider well-formed operations where no process has more than one pending operation at any given point in time and where a process only invokes a release operation if and only if its last event was a stop response.
For every process p we de?ne the p-active intervals with respect to the splitter execution to be:1.From an acquire invocation of the splitter by p until the corresponding response if the splitter was not captured by p.2.From an acquire invocation of the splitter by p until the done response to the corresponding release invocation,given that the splitter was captured by p.
A process that has no acquire operations cannot be p-active and is hence called p-idle.We will implement an adaptive splitter with the following proper-ties:
1.At any point in time the adaptive splitter is captured by at most one process.
(Mutual exclusion)
2.If the pre?x of a busy period is only an invocation of acquire and its response,
then the response must be a stop.(Processes that access a”new”splitter by themselves must capture it.)
3.Not all responses to the acquire invocations during a busy period are right.
4.Not all responses to the acquire invocations during a busy period are down.
5.No busy period of the adaptive splitter that contains in?nitely many events
where the events are only acquire invocations and down responses(i.e.where no process captures it or goes right).
As always with such splitters,if two or more processes access it concurrently it is possible that none of them captures it.But not all acquire attempts return down and not all return right.
The key di?erence between our splitter and the Moir Anderson[20]splitter are the following two properties:1.At the point in time where a process is guaranteed to go right there is at least another process active in the splitter that might either stop or go down.2.At the point in time where a process is guaranteed to go down there is at least another process active in the splitter that is guaranteed not to go down or is undecided(might go down,right or stop). 3Impossibility of uniform,adaptive splitter,collect and renaming with in?nitely many bounded registers
We now show that it is impossible to uniformly implement adaptive long-lived splitter,collect or renaming using in?nitely many bounded MWMR registers in an unbounded execution where at most?nitely many distinct processes may appear.To do so we use a so called weak test and set object[2].
3.1Weak test and set
We model the behavior of a weak test and set object with the following program (Figure1).Each process is in one of four possible states:thinking,WT&Set, eating and RESET.
TS:object of type WT&S
Process p:
repeat forever
thinking section p
tbit:=WT&SET p(TS)
if tbit=0{
eating section p
RESET p(TS)}
end repeat forever
Fig.1.Weak Test&Set algorithm.
A WT&S object satis?es the following two properties:
–Exclusion:At most one process is eating at any system state of the execu-tion.
–If a process becomes hungry,that is leaves the thinking state,while all other processes are thinking and it only takes steps then it must eventually start eating.
Clearly if we can implement a long-lived adaptive splitter,we can implement WT&SET.The reduction is straightforward,we simply use the adaptive splitter to implement the WT&SET object.A process that captures the splitter enters the eating section.Processes that fail to capture the splitter do not eat.This algorithm has the required properties,it implements WT&SET.Moreover as was shown in[2]both a long-lived adaptive collect or a long-lived adaptive renaming algorithm can be used to implement the WT&SET object.Hence it su?ces to show that we cannot implement an adaptive WT&SET object using only register reads and writes in an unbounded execution where?nitely many distinct processes may appear and in?nitely many bounded registers are available.
Theorem1.There does not exist a uniform,long-lived,adaptive implementa-tion of a Weak-Test&Set object using only read and write operations,in?nitely many bounded MWMR registers and in?nitely many bounded SWMR registers in an execution where only?nitely many distinct processes appear.
We?rst restate a combinatorial lemma proved in[17].
Lemma1.Let N be the set of all integers.Assume that you are given an in?nite collection of sets S i?N such that|S i|≤k for some nonnegative constant k. Then?X?N such that|X|=∞and such that?x,y∈X,y=x?x∈S y.
The proof of this lemma can be found in[17].Processes are potentially able to write information about themselves into their own SWMR registers.This information could then subsequently be read by other processes”helping”them in their execution of the WT&SET algorithm.The sets S y here represent the ids of processes corresponding to the SWMR registers that process y is going to read.Hence informally speaking the lemma says that there is an in?nite set of processes X such that no two processes in X read each others SWMR registers when executing the WT&SET algorithm while believing they are running alone.
Using Lemma1,the idea of the proof of Theorem1is to show that we can get at least two carefully selected processes p and q to execute the algorithm together in such a way that these two processes will not be able to distinguish each others writes.In other words p will interpret a write of q it reads as one of its own writes and vice versa.As a result they will both capture the WT&SET bit(eat)at the same time,a contradiction.
We?rst show that we can?nd an in?nite set of processes that will all write to the same MWMR register.Note that we will not require that all these in?nitely many processes execute the algorithm.They will simply provide us with the collection of processes from which we will be able to carefully select the?nitely many processes with the desired properties.
Lemma2.There exists an in?nite set S of processes that write to the same MWMR register and that do not read any register to which only?nitely many of the processes in S wrote to.
Proof Sketch:By Lemma1,?X0?N such that|X0|=∞and such that
.This implies that there exists an in?nite set of ?x,y∈X0,y=x?x∈S(0)
y
processes X0such that no process reads the single-writer register of any other process in the set.Assume now that the processes in X0write to in?nitely many distinct MWMR registers in their?rst write.Then we can intuitively think of these in?nitely many MWMR as SWMR registers(where the last process writing to a register is the single writer)and apply Lemma1one more time-we get an in?nite set of processes X1?X0with the same properties as X0(no process reads information of any other participating process).If processes continue to write to in?nitely many distinct MWMR registers we can inductively continue this construction.Since our algorithm must be adaptive,a process that believes to execute the algorithm alone must terminate within a constant number k of
steps.Hence we obtain an in?nite set of processes X k that all?nish the execution of the algorithm believing they ran solo and hence capturing the WT&SET bit, a contradiction.Hence in?nitely many processes will eventually have to write to?nitely many MWMR registers.So there must exist an in?nite set S of processes that writes to the same MWMR register and where no process in S reads information about any other process in S in any other register.
Using induction we now?nish the proof of the Theorem:
We will construct an in?nite set of processes that when executing the algo-rithm at the same time,write the same values to the same MWMR registers in the same order and do not read each others SWMR registers.
By Lemma2there must be an in?nite set of processes X0,where no two processes read each others SWMR registers and that write to the same bounded MWMR register r1.By the boundedness of r1and the pigeonhole principle it follows that there is an in?nite set of processes X1?X0such that none of the processes in X1is able to distinguish its write to r1from a write from one of the other processes in X1.If the processes in X1?nish their execution after writing to r1we are done.Otherwise by Lemma2there exists an in?nite subset X2of X1of processes such that they all write to the same register r2,do not read each others SWMR registers(by Lemma2)and hence are not aware of each other.Since each of these processes believes it executes the algorithm on its own and since the algorithm is adaptive,there exists a k such that after k writes each of these processes must terminate its execution.Given k we can inductively continue this construction until we obtain X k,an in?nite set of processes that when executing the algorithm at the same time,write the same values to the same MWMR registers in the same order and do not read each others SWMR registers or any other register that could provide them information about each other.We now simply select two processes p and q from X k and let them execute the algorithm at the same time.Since both p and q are members of X k,that is,they write the same values to the same registers in the same order they will both capture the WT&SET bit at the same time,a contradiction. 4Adaptive,long-lived and uniform splitter using unboundedly many unbounded MWMR registers
We now present an algorithm that implements an adaptive,long-lived splitter in a system where no upper bound on the number of participating processes is known.The algorithm assumes that at most?nitely many processes will partic-ipate in the algorithm and uses in?nitely many unbounded MWMR registers.It is based on[7].
We present the implementation of the long-lived adaptive splitter in Figure2. Note that in the given construction a process that comes alone captures the adaptive splitter in O(1)steps.It also releases the splitter in one step by setting its status to idle.However when k processes access the splitter concurrently some process may perform k steps.For example,let process p and q execute in lock-step until both are about to write to X.Assume that p now writes to
acquire()for process p returns stop,down or right.
Type:pid=process id,0,....
Shared:
X[0...],In?nitely many unbounded atomic registers of type pid each initialized to0.
Y[0...],In?nitely many unbounded atomic registers of type(pid each initialized to0.
Z[0...],In?nitely many unbounded atomic registers of type pid each initialized to0.
status[0...],In?nitely many atomic registers of type{start,active,idle}initialized to idle.
I[0...],In?nitely many unbounded atomic registers consisting of i,
where i is an integer0≤i initialized to?1.
Master,an integer in the range0,...initialized to0.
Code:
1status[p]:=start;//indicate status
2m:=Master;//Master is supposed to be last process to update index
3c:=I[m];//find last copy of splitter used
4current:=c+1;//try next copy of splitter
5if((status[(X[c])]=active or(status[Y[c]]=active or(status[(Z[c])]=active then
status[p]:=idle;return right
6status[p]:=active;//set state in single-shot splitter.
7X[current]:=p;//emulate single shot splitter.
8if Y[current]=0then
status[p]:=idle,return right
9I[p]:=c;//Once p writes in Y[current],another process q may read it later.If I[p]is not updated this process will loop. 10Y[current]:=p,nextDB;
11if(X[current]=p)then
update(current),status[p]:=idle,return down
12Z[current]:=p;//make sure to be seen after capturing splitter.
13if(X[current]=p)then
update(current),status[p]:=idle,return down
14update(current)
15return stop
Procedure release()for process p
16status[p]:=idle
17return;
Procedure update(c)for process p.
18I[p]:=c;
19Master:=p;
do forever
20if(Master=p)then return//another process will update pointer,p can return.
21c:=c+1;
22if(Y[c]=0then return;//fresh copy found,p can return.
23q:=Y[c];
24if(I[q]=(i?1)then c:=I[q],I[p]:=c;//only update I[p]if it is smaller than I[q].
od;
Fig.2.Implementation of adaptive and long-lived splitter.
X and that then q immediately overwrites p in X.Let them then execute in
lock-step again until they read X.Assume that p now reads X and?nds the condition to fail and stops right before performing update(c).Process q can now continue and either capture the splitter or leave.While p is still active but not performing any steps k di?erent processes can capture and release single shot copies of the splitter.Hence to?nish its execution p will now have to iterate k
times through the loop in update(),Figure2.Since k is at most N?1,however
the step complexity of p will still be O(N).
We say that a process captures the adaptive splitter object(Figure2)if it reaches line14of the algorithm.We?rst show that if a process reaches line14,no
other process will reach this line until the?rst process?nished its execution.I.e.
we show that our splitter has a standard mutual exclusion property.The proof
is based on the correctness proof by Afek et al.[7]of their adaptive splitter
implementation.Out of lack of space we simply quote the main result.We begin by showing that no more than one process at a time can capture the splitter. Lemma3.No two processes p and q reach line14of the code concurrently. Proof.We leave the proof to the full version of the paper
Lemma4.Any process executing the algorithm in exclusion must reach line15. Proof.Same as in[7].
Theorem2.For a system with?nitely many active processes there exists a long-lived adaptive uniform splitter implementation using in?nitely many un-bounded MWMR registers.
5Adaptive and long-lived splitter implementation using bounded registers,given N
Given N,an upper bound on the id’s of processes that will appear during an in-?nite execution of the long-lived adaptive splitter,the previous algorithm can be modi?ed so that it uses only?nitely many bounded registers.Moreover,bounded MWMR and SWMR registers can be used instead of unbounded registers by es-timating the size of registers required based on N.The resulting algorithm is the same as presented in[7].Its main feature is that by assigning unique registers to processes in advance it allows us to reuse registers that processes wrote to before. Every time before accessing a new copy of the adaptive splitter,a process checks whether the process assigned to this copy is currently active.If so,the process does not write to this copy but simply”walks away”,i.e.returns”right”.This prevents a process p from writing to copies of the splitter where active processes are currently about to write to one of the registers,hence possibly erasing all traces of this process p.We are only able to do this since we know an upper bound on process id’s in advance,hence are able to assign three unique MWMR registers to all possibly participating processes.Since moreover processes write only their id’s and smaller values to all registers we are also able to bound the size of the registers in advance.
6Discussion
We showed that any uniform,long-lived adaptive implementation of a splitter, collect or renaming cannot use only bounded MWMR registers if no a priori up-per bound on the number of participating processes is known,even if only?nitely many processes participate in the algorithm.This means that any such adaptive algorithm must use unbounded MWMR registers.Hence even if?nitely many such registers would be su?cient unbounded memory is needed,making these step complexity adaptive algorithms impractical in dynamic settings.This shows that adaptive algorithms that run on real systems are inherently non-uniform. They are inherently designed for closed and controlled settings:Designers must ?nd a way to realistically limit the size of the system so they can hard code an upper bound on the size of the system into the algorithm.
References
1.Y.Afek,H.Attiya,A.Fouren,G.Stupp and D.Touitou.Long-Lived Renaming
made adaptive.Proc.of18th ACM Symp.on Principles of Distributed Computing (PODC):91-103,1999.
2.Y.Afek,P.Boxer and D.Touitou.Bounds on the shared memory requirements
for long-lived and adaptive objects.Proc.of the19th ACM Symp.on Principles of Distributed Computing(PODC):81-89,2000.
3.Y.Afek,D.Dauber and D.Touitou.Wait-free made fast.Proc.of the27th Ann.
ACM Symp.on Theory of Computing:538-547,1995.
4.Y.Afek and M.Merritt.Fast,wait-free(2k?1)-renaming.In Proc.of the18th
Ann.ACM Symp.on Principles of Distributed Computing:105-112,1999.
5.Y.Afek,M.Merritt,G.Taubenfeld and D.Touitou.Disentangling multi-object
operations.In Proc.of16th Annual ACM Symp.on Principles of Distributed Com-puting:111-120,1997.
6.Y.Afek,G.Stupp and D.Touitou.Long-lived adaptive collect with applications.
Proc.of the40th Ann.Symp.on Foundations of Computer Science:262-272,1999.
7.Y.Afek,G.Stupp and D.Touitou.Long Lived Adaptive Splitter and Applications.
Distributed Computing,15(2):67-86,2002.
8.M.Aguilera,B.Englert and E.Gafni.Uniform Solvability with a?nite number of
MWMR registers.In Proc.17th International Conference DISC2003:16-30,2003.
9.J.Anderson and J-H.Yang.Time/contention trade-o?s for multiprocessor syn-
https://www.wendangku.net/doc/cd1640909.html,rmation and Computation,124(1):68-84,1996.
10.H.Attiya and E.Dagan.Universal operations:Unary versus binary.In Proc.15th
ACM Symp.on Principles of Distributed Computing:223-232,1996.
11.H.Attiya and A.Fouren.Adaptive wait-free algorithms for lattice agreement and
renaming.In Proc.17th ACM Symp.on Principles of https://www.wendangku.net/doc/cd1640909.html,p.:277-286,1998.
12.H.Attiya and A.Fouren.Algorithms adaptive to point contention.In J.ACM,
50(4):444-468,July2003.
13.H.Attiya,A.Fouren and E.Gafni.An adaptive collect algorithm with applications.
Distributed Computing,15(2):87-96,2002.
14.H.Attiya,F.Kuhn,M.Wattenhofer and R.Wattenhofer.E?cient Adaptive Col-
lect using Randomization.Proc.18th Conference on https://www.wendangku.net/doc/cd1640909.html,p.(DISC),2004.
15.H.Attiya and I.Zach.Fully adaptive algorithms for atomic and immediate snap-
shots.www.cs.technion.ac.il/hagit/pubs/AZ03.pdf,2003.
16. B.Englert and E.Gafni.Fast Collect in the Absence of Contention.Proc.22nd
IEEE Intern.Conference on https://www.wendangku.net/doc/cd1640909.html,p.Systems(ICDCS):537-543,2002. 17. B.Englert and D.Goldstein.Can Memory be used adaptively by Uniform Algo-
rithms?Proc.9th Intern.Conf.on Principles of Distr.Systems(OPODIS),2005.
https://www.wendangku.net/doc/cd1640909.html,mport.A fast mutual exclusion algorithm.ACM Transactions on Computer
Systems,5(1):1-11.February1987.
19.M.Merritt and G.Taubenfeld.Speeding Lamport’s fast mutual exclusion algo-
https://www.wendangku.net/doc/cd1640909.html,rmation Processing Letters,45:137-142,1993.
20.M.Moir and J.Anderson.Wait-free algorithms for fast,long-lived renaming.Sci-
ence of Computer Programming,25(1):pp.1-39,1995.
库库尔坎金字塔
地理位置 位于墨西哥大学城以南的库库尔坎（Kukulcan），是玛雅文化前古典期晚期（公元前800年——公元前200年）中部高原文化的重要文化遗址之一，“库库尔坎”的原意是“舞蹈唱歌的地方”，或表示“带有羽毛的蛇神”。 建筑结构 “库库尔坎”金字塔是我们现在看到的早期墨西哥金字塔是一座用土筑成的 九层圆形祭坛，高29米 ，周边各宽55米多，周长250米左右，最高一层建有一座6米高的方形坛庙，库库尔坎金字塔高约30 米，四周环绕91 级台阶，加起来一共364级台阶，再加上塔顶的羽蛇神庙，共有365阶，象征了一年中的365 天。台阶的两侧宽达1米的边墙，北边墙下端，有一个带羽行的大蛇头石刻，蛇头高1.43米，长达1.80米，宽1.07米，蛇嘴里吐出一条大舌头，颇为独特。 这座古老的建筑，在建造之前，经过了精心的几何设计，它所表达出的精确度和玄妙而充满戏剧性的效果，令后人叹为观止：每年春分和秋分两天的日落时分，北面一组台阶的边墙会在阳光照射下形成弯弯曲曲的七段等腰三角形，连同底部雕刻的蛇头，宛若一条巨蛇从塔顶向大地游动，象征着羽蛇神在春分时苏醒，爬出庙宇。每一次，这个幻像持续整整3 小时22 分，分秒不差。这个神秘景观被称为“光影蛇形”。 每当“库库尔坎”金字塔出同蛇影奇观的时候，古代玛雅人就欢聚在一起，高歌起舞，庆祝这位羽毛蛇神的降临。 库库尔坎金字塔，是玛雅人对其掌握的建筑几何知识的绝妙展示，而金字塔旁边的天文台，更是把这种高超的几何和天文知识表现得淋漓尽致。 建造工艺 在没有金属工具、没有大牲畜和轮车的情况下，古代玛雅人却能够开采大量重达数十吨的石头，跋山涉水、一路艰辛地运到目的地，建成一个个雄伟的金字塔。金字塔最高的可达70米，其规模之巨大、施工难度之高，令人吃惊。 古代玛雅的金字塔和古埃及的金字塔在建筑形式上有着明显的不同。埃及的金字塔的塔顶是尖的，而玛雅金字塔却是平顶，塔体呈方形，底大顶小，层层叠叠，塔顶的台上还建有庙宇；在用途上也不一样：埃及金字塔是法老的陵墓，而玛雅的金字塔除个别外，一般是用来祭祀或观察天象的。除金字塔外，玛雅人还兴建了不少功能性强，技艺高的民用建筑，主要有：房屋（包括庙宇、府邸、民居等）、公共设施（球场、广场、集市等）、基础设施（桥梁、大道、码头、堤坝、护墙等）和水利工程（水渠、水库、水井、梯田）等。
库库尔坎金字塔位于墨西哥
库库尔坎金字塔位于墨西哥尤卡坦半岛北部，是曾经古玛雅人留下的文明遗址。 库库尔坎在玛雅文化中是带羽蛇神的意思，被誉为是太阳的化身，即太阳神。羽蛇神主宰着晨星、发明了书籍、立法，而且给人类带来了玉米，还代表着死亡和重生，也是祭司们的保护神。古典时期，玛雅“真人”手持权杖，权杖一端为精致小形、中间为小人的一条腿化作蛇身、另一端为一蛇头。到了后古典时期，出现了多种变形，但基本形态完全变了，成为上部羽扇形、中间蛇身下部蛇头的羽蛇神形象。 羽蛇神与雨季同来，而雨季又与玛雅人种玉米的时间相重合，因而羽蛇神又成为玛雅农人最为崇敬的神明。在玛雅人眼中羽蛇神与犬神有同样高的地位 在现今留存的最大的玛雅古城奇岑－伊扎，库库尔坎金字塔屹立其中。它的每个面有91级台阶，4个面象征一年四季，共364级台阶，加上塔顶的神殿共365级，象征一年365天。4个边缘分别代表春分、秋分、夏至和冬至。在金字 塔的北面两底角雕有两个蛇头。每年春分、秋分两天，太阳落山时，可以看到蛇头投射在地上的影子与许多个三角形连套在一起，成为一条动感很强的飞蛇。好像从天空中游向大地，栩栩如生。春分蛇由上至下出现，仿佛从神庙中爬出来一样，秋分则由下至上出现，像是爬进去一样。只有每年的这两天里才能看到这一奇景。所以，现在它已经成为墨西哥的一个著名旅游景点。而在当年，玛雅人可以借助这种将天文学与建筑工艺精湛地融合在一起的直观景致，准确把握农时。与此同时，也准确把握崇拜羽蛇神的时机。考古学家在玛雅神庙下方发现神秘的坑洞，进一步的考察发现，坑洞是自然形成的，延伸大约34米，深度为20米 左右。最新的消息指出，神秘下方的洞穴可连接到其他洞穴或者地下湖泊，同时具有深层次的宗教意义，在玛雅文化中处于非常重要的地位。这座寺庙名称为Kukuerkan，即库库尔坎金字塔，这是中美洲最著名的古文明遗址之一，是玛雅文明的巅峰之作。那么金字塔下方的洞穴为何会存在呢，具有何种宗教意义？ 考古学家发现玛雅人在1000年前建造了金字塔，位于一个巨大的天然深坑 上，他们相信建立在深坑顶部的建筑是宗教信仰的一部分，因为库库尔坎金字塔内有一尊玛雅蛇神，或者认为是有羽毛的蛇，它是玛雅人非常崇拜的蛇神。由于蛇神需要水，因此神殿下方是一个被填满水的深坑，水流的运行由北向南。考古学家通过一种新的探测技术对神殿下方的结构进行研究，发现了深坑的存在，探测器能够探测到岩石结构走向，分辨出水流和岩石的相对位置。
2019年(春)六年级美术下册 第1课《追寻文明的足迹》说课教案 人美版
2019年(春)六年级美术下册第1课《追寻文明的足 迹》说课教案人美版 教材分析： 本课是“欣赏`评诉”领域的内容。世界文化遗产是人类进步的闪光足迹，是全世界共同的财富。通过本课的欣赏活动，让学生了解人类的智慧和壮举，了解不同历史时期的文化现象，从而引发学生对人类文化遗产的关注，激发学生的求知欲。 教学内容与目标： 认知： 1、知道什么叫文化遗产，能说出7处-10处世界文化遗产。 2、能够对世界文化遗产的保护有正确的认识。 情意： 1、让学生认识到文化遗产是人类共同的财富，增强保护文化遗产的意识和方法。 2、学习了解各个国家不同时期的文明足迹，尊重不同文化背景下的文化现象，接纳多元文 化，拓展国际视野。 能力： 1、能够运用美术知识制作宣传栏，传播世界文化遗产知识。 2、能够收集世界文化遗产的相关资料，初步认识、了解自己感兴趣的文化遗产。 教学重点与难点： 重点： 1、通过资料阅读、课上交流等途径，认识7处—10处世界文化遗产，较细致的了解2处— 3处世界文化遗产。 2、能够从文化遗产的历史背景、审美价值、社会影响、经典特征等方面，介绍2处—3处 世界文化遗产。 难点：有步骤、有层次的介绍2处—3处世界文化遗产。 学习材料： 资料、剪刀、彩纸等 教学过程： 一、了解不同背景的世界文明遗迹 1、课前将学生分成四个组：阿布辛拜勒神庙（非洲）、库库尔坎金字塔（美洲）、法隆寺 （亚洲）、奥林匹亚遗址（欧洲） 每个学生根据书中的问题，查找最下边的 一行图片资料。 （1）《掷铁饼者》、《竞技表演》是哪方面的内容？用的是什么手法？ （2）阿布辛拜勒神庙建造于什么地方、什么时代、有什么特点？ （3）库库尔坎金字塔的所在地、建造时间、艺术特色。说一说它和埃及金字塔有什么相同与不同之处？ （4）法隆寺的建筑特点，它和中国的塔有什么相同与不同之处？ 教学意图：培养学生收集信息、整理信息的能力，以及小组合作、探究的意识。 二、欣赏分析多元的世界文化遗产
(完整word版)人美版六年级美术试题
(完整word版)人美版六年级美术试题 亲爱的读者： 本文内容由我和我的同事精心收集整理后编辑发布到文库，发布之前我们对文中内容进行详细的校对，但难免会有错误的地方，如果有错误的地方请您评论区留言，我们予以纠正，如果本文档对您有帮助，请您下载收藏以便随时调用。下面是本文详细内容。 最后最您生活愉快 ~O(∩_∩)O ~
六年级美术试 一、填空。. 1、《哪吒闹海》是一幅__壁_画，作者张仃在这幅画中运用了中国传统壁画表现手法，又借鉴了构图的理念，在色彩、造型等方面都进行了成功的探索。 2、拉斯科洞窟的岩画是人类最早的绘画。 3、法隆寺是世界上最古老的木建筑之一。 4、装饰画偏重表现形式的装饰画偏重表现形式的装饰性，不强调真实光影和透视关系，注重色彩的象征性及整体的和谐，多以夸张、变形、简洁的手法，给人以明快、强烈的艺术美感。 5、农民画家敢于把牛画的艳丽，显示了他们对色彩生活丰富的想象力和对的无限热爱。 6、雕版印刷三大特点:主板线稿多块色板鬃毛刷。 7、我们曾学过纸版画对印版画粉印版画拓印版画等许多版画制作方法。 8、肌理是指由于材料的不同配列、组成和构造而使人得到的触觉质感和质地。手感触感织法性质纹理等说法都可包含在肌理之中。 9、第一部动画片是《一张滑稽面孔的幽默姿态》，第一部有声动画片是《威利号汽船》，第一部彩色动画片《白雪公主，第一部三维动画片是《玩具总动员》。 10、传统民间门神形象中取自《封神演义》的是赵公明和燃灯道人。 11、许多文化遗产是、文字、绘画、建筑、遗址、口传方式记录下来的，非物质文化遗产是不用文字以民间戏曲为主的活态文化，如各民族的文字绘画建筑遗址口传方式民间戏曲舞蹈神话传说节日礼仪刺绣玩具剪纸面花皮影等。 12、抓髻娃娃是我国北方民间剪纸中一个典型的形象，它既象征着性生命
人美版六年级美术下册教案
六年级 美术(电子备课) 于颖 目录: 六年级下册: 第1课追寻文明的足迹 第2课探访自然的器官 第3课装饰色彩 第4课装饰画 第5课彩球的设计 第6课城市雕塑 第7课各种材料来制版 第8课装饰柱 第9课精彩的戏曲 第10课戏曲人物 第11课画故事 第12课动画片的今夕 第13课拟人化的动漫形象 第14课留给母校的纪念 第15课我的成长记录 第16课剪纸中的古老记忆 第17课绣在服装上的故事
第18课复制与传播 学情分析: 本学期我教的就是六年级三个班的美术课,随着学生年龄的增长、生理、身心健康的不断发展,学生们对美术的要求有所不同,对美术知识、技能的掌握程度也不同。她们更多地追求现实的、真实的东西或喜欢目前流行的“动漫”,对此进行大量的临摹在绘画技巧上显得成熟,线条也十分流畅,但画风呆板,表现欲下降,缺少童趣。但她们对色彩有辨别能力,把握物体形状的能力很强,乐于动手,对手工制作课充满了极高的热情。对美具有了较高的欣赏能力。 这学期就是她们迈入初中的过渡时期,既要努力响应学生目前的兴趣爱好,又要深入浅出地灌输客观的美术学科理论知识,与初中知识进一步接轨,将平时教给学生的初中知识更加明朗化介绍给学生,使她们从儿童绘画到少年绘画的转型,更重视有欣赏内容与工艺设计教学,即学生的创造能力上的培养,其实就就是对儿童想象力的深化,转变为更理论性含量的创造能力。 知识要点: 该教材贴近学生的心理,符合学生的年龄特点,每课的主题都有美术带来的有趣的惊喜。六年级美术教材一共分为19课。每一课都侧重于“造型·表现”、“设计·应用”、“欣赏·评述”、“综合·探索”四个学习领域中的一个方面,同时也有不少系列的课,教学内容前后联系比较紧密,非常适合学生较深入地做一个专题。并且有些教学内容还设计成了案例学习、问题学习与项目学习,这就要求我们老师应多引导学生进行主动的探究学习,增加学生集体合作的机会。其次本册教材继续体现了人美版教材的可操作性强、贴近生活、关注美术文化的渗透与注重学生学习兴趣的培养等特点。 “造型·表现”:在教学中,让学生运用形、色与肌理等美术语言,选择合适自己的各种工具、材料,记录与表现所见所闻、所感所想的事物,发展美术创作能力,传递自己的思想与情感。 “设计·应用”:运用对比与与谐、对称与均衡、节奏与韵律等组合原理,了解一些简单的创意、设计方法与媒材的加工方法,进行设计与装饰,美化身边的环境。“欣赏·评述”:欣赏认识自然美与美术作品的材料、形式与内容等特征,通过描述、分析与讨论等方式,了解美术表现的多样性,能用一些简单的美术术语,表达自己对美术作品的感受与理解。 “综合·探索”:结合学校与社区的活动,以美术与科学课程与其她课程的知识、技能相结合的方式,进行策划、制作、表演与展示,体会美术与环境及传统文化的关系。 【授课内容】1、追寻文明的足迹 【课时建议】1课时 【教程目标】 知识与技能:知道什么叫文化遗产,能说出7-10处世界文化遗产。能够对世界文化遗产的保护有正确的认识。 过程与方法:通过资料阅读、课上交流等途径,认识了解2处-3处世界文化遗产;能够从文化遗产的历史背景、审美价值、社会影响、经典特征等方面,介绍2-3处世界文化遗产。 情感、态度与价值观:通过本课学习,认识到文化遗产就是人类共同的财富,增强保护文化遗产
科潘玛雅古迹遗址
科潘玛雅古迹遗址 科潘玛雅古迹遗址位于洪都拉斯西部边境，遗址面积数十平方公里，1980年入选《世界遗产名录》。科潘属于玛雅文明古典期（公元250—900年），为玛雅文明最盛时期。金字塔式寺庙较为精美，台顶神殿有大量浮雕。科潘是玛雅文化的学术中心，不少建筑遗址与天文、历法等学术活动有关。这里保存了玛雅文化中最长的铭刻，多达2500字，号称“象形文字的梯道”。科潘玛雅遗址是玛雅文明最重要的地区之一，有着宏大的建筑，还有丰富的象形文字，是极少数起源于热带丛林的文明的例证。这些建筑表明科潘的玛雅人有高度发展的经济和文化。 公元1839年，美国探险学家史蒂芬斯(JohnlioydStephens)和卡瑟伍德(FrederikCatherWood)受到一个古老传说的暗示，披荆斩棘，深入浓荫蔽日的雨林之中，然而，他们没有找到被巫师催眠的美丽公主，却发现了一座已荒废千年的古代城市遗址。在这座叫作科潘的旧城废墟上，高大的纪念碑被藤条缠绕，湮没在荆棘之中；雄伟的金字塔上长满了粗壮的树木，变成一座座荒丘。史蒂芬斯他们被眼前的这一切惊呆了，这些遗迹所代表的就是辉煌灿烂的玛雅文明。 玛雅文明是世界著名的古代文明之一，也是唯一诞生在热带丛林而非大河流域的文明。玛雅人具有的抽象思维能力让同时代的旧大陆文明相形见绌。他们创造了精确的数学体系和天文历法系统，以及至今仍有待我们去破译的象形文字系统。玛雅人最重视对太阳和月亮的观测，他们能算出日蚀和月蚀出现的时间，并已将七大行星都列入了研究范围。他们对金星运行周期的计算和现代科学实测结果完全一致。玛雅历法体系由"神历"、"太阳历"和"长纪年历"组成。玛雅人有一个独特的数学体系，在这个体系中，最先进的部分便是"0"这个符号的使用，它的发明和使用比欧洲大约早了800年。玛雅的数学体系的适用性和科学性，使他们能在许多科学和技术活动中解决各种难题。在世界各古代文明中，除了起源于印度的阿拉伯数字之外，玛雅数字要算是最先进的了。但非常可惜，有关玛雅数学的图书或文献一本也没有流传下来。玛雅的象形文字对现代人来说真是一部天书，它的谜底直到今天仍未解开。玛雅象形文字以近似圆形或椭圆为主，字符的线条更多地依随图形起伏变化、圆通流畅。 科潘是玛雅象形文字研究最发达的地区，它的纪念碑和建筑物上的象形文字符号书写最美、刻制最精、字数最多，例如，在科潘遗址中，有一条六七十级的梯道，用2500多块加工过的方石砌成，这是一座纪念性的建筑物，梯道建在山坡上，直通山顶的祭坛。宽10米，两侧各刻着一条花斑巨蟒，蟒尾在山丘顶部：梯道的每块方砖上都刻着象形文字，每个形文字的四周均雕有花纹，梯道共刻了2000多个象形文字符号，它是玛雅象形文字最长的铭刻，也是世界题铭学上少见的珍贵文物，由此被称为"象形文字梯道。" 不仅如此，科潘的经济与政治实力仅次于蒂卡尔而远远超过其他城邦，在文化上则完全可以和蒂卡尔并肩而立，甚至还略有超越，有学者认为科潘的重要意义决不在蒂卡尔之下，它们如双峰并立，是玛雅文明两座最伟大的灯塔。确实，从考古发掘的城市遗址看，科潘在规模上可能略逊于蒂卡尔，但美丽却有过之而无不及。据记载公元805年以后，玛雅人突然弃科潘城北迁，科潘城随之变成一片废墟。 奇琴伊察玛雅城邦遗址曾是古玛雅帝国最大最繁华的城邦。遗址拉于尤卡坦半岛中部。始建于公元514年。城邦的主要古迹有：千柱广场、武士庙及庙前的斜倚的两神石像。9层，高30米的呈阶梯形的库库尔坎金字塔以及圣井(石灰岩竖洞)和筑在高台上呈蜗形的玛雅人古
六年级美术下册--《追寻文明的足迹》说课教案(人美版)
六年级美术下册--《追寻文明的足迹》说课教案（人美版） 教材分析： 本课是“欣赏`评诉”领域的内容.世界文化遗产是人类进步的闪光足迹，是全世界共同的财富.通过本课的欣赏活动，让学生了解人类的智慧和壮举，了解不同历史时期的文化现象，从而引发学生对人类文化遗产的关注，激发学生的求知欲. 教学内容与目标： 认知： 1、知道什么叫文化遗产，能说出7处-10处世界文化遗产. 2、能够对世界文化遗产的保护有正确的认识. 情意： 1、让学生认识到文化遗产是人类共同的财富，增强保护文化遗产的意识和方法. 2、学习了解各个国家不同时期的文明足迹，尊重不同文化背景下的文化现象，接纳多元文化， 拓展国际视野. 能力： 1、能够运用美术知识制作宣传栏，传播世界文化遗产知识. 2、能够收集世界文化遗产的相关资料，初步认识、了解自己感兴趣的文化遗产. 教学重点与难点： 重点： 1、通过资料阅读、课上交流等途径，认识7处—10处世界文化遗产，较细致的了解2处—3 处世界文化遗产. 2、能够从文化遗产的历史背景、审美价值、社会影响、经典特征等方面，介绍2处—3处世 界文化遗产. 难点：有步骤、有层次的介绍2处—3处世界文化遗产. 学习材料： 资料、剪刀、彩纸等 教学过程： 一、了解不同背景的世界文明遗迹 1、课前将学生分成四个组：阿布辛拜勒神庙（非洲）、库库尔坎金字塔（美洲）、法隆寺（亚 洲）、奥林匹亚遗址（欧洲）
每个学生根据书中的问题，查找最下边的 一行图片资料. （1）《掷铁饼者》、《竞技表演》是哪方面的内容？用的是什么手法？ （2）阿布辛拜勒神庙建造于什么地方、什么时代、有什么特点？ （3）库库尔坎金字塔的所在地、建造时间、艺术特色.说一说它和埃及金字塔有什么相同与不同之处？ （4）法隆寺的建筑特点，它和中国的塔有什么相同与不同之处？ 教学意图：培养学生收集信息、整理信息的能力，以及小组合作、探究的意识. 二、欣赏分析多元的世界文化遗产 1、教师讲述关于拉斯科岩画的故事. （1）这幅画表现的什么？让你联想到什么？ （2）猜猜当时的人用什么工具创作的？ （3）他们为什么要画这些壁画？ 2、2008年8月8日在北京将要发生一件非常重要的事情，是什么？有谁知道这界奥运会是 第几届？第一届奥运会在哪里召开的？这个国家在哪个洲？ （1）教师介绍希腊奥运之城. （2）请学生代表介绍查找到的关于《竞技表演》、《掷铁饼者》的资料 小结：古代的建筑、雕塑和体育已经完美的结合了，说说你还知道奥运会的那些项目？书中还有哪些建筑在欧洲？ 3、学生介绍有关阿布辛拜勒神庙资料. 小结：由此我们知道建筑、雕塑和宗教也有着亲切的联系，你还知道埃及有什么建筑？他们是哪个洲的文明遗迹？说说埃及金字塔和库库尔坎金字塔有什么不同？在美洲还有什么文明遗迹，请你说说. 4、对于美洲古代文明遗迹你有什么感想？ 5、法隆寺的建筑特点，它和中国的塔有什么相同与 不同之处？书中还有哪个是亚洲的建筑？ 6、亚洲还有什么文明遗迹？ 教学意图：
奇琴伊查库库尔坎金字塔英文介绍
The stepped pyramids, temples, columned arcades, and other stone structures of ChichénItzá were sacred to the Maya and a sophisticated urban center of their empire from A.D. 750 to 1200. Viewed as a whole, the incredible complex reveals much about the Maya and Toltec vision of the universe—which was intimately tied to what was visible in the dark night skies of the Yucatán Peninsula. The most recognizable structure here is the Temple of Kukulkan, also known as El Castillo. This glorious step pyramid demonstrates the accuracy and importance of Maya astronomy—and the heavy influence of the Toltecs, who invaded around 1000 and precipitated a merger of the two cultural traditions. The temple has 365 steps—one for each day of the year. Each of the temple’s four sides has 91 steps, and the top platform makes the 365th. Devising a 365-day calendar was just one feat of Maya science. Incredibly, twice a year on the spring and autumn equinoxes, a shadow falls on the pyramid in the shape of a serpent. As the sun sets, this shadowy snake descends the steps to eventually join a
库库尔坎金字塔
库库尔坎金字塔 库库尔坎金字塔 库库尔坎金字塔在墨西哥尤卡坦半岛北部，有座著名的古建筑称为库库尔坎金字塔，是曾经的古代玛雅人留下的文明遗址。库库尔坎在玛雅文中是带羽毛之神蛇的意思，被誉为是太阳的化身。 基本认识 地理位置 位于墨西哥大学城以南的库库尔坎（Kukulcan），是玛雅文化前古典期晚期（公元前800年——公元前200年）中部高原文化的重要文化遗址之一，“库库尔坎”的原意是“舞蹈唱歌的地方”，或表示“带有羽毛的蛇神”。 建筑结构 “库库尔坎”金字塔是我们现在看到的早期墨西哥金字塔是一座用土筑成的九层圆形祭坛，高29米 库库尔坎金字塔 ，周边各宽55米多，周长250米左右，最高一层建有一座6米高的方形坛庙，库库尔坎金字塔高约30 米，四周环绕91 级台阶，加起来一共364级台阶，再加上塔顶的羽蛇神庙，共有365阶，象征了一年中的365 天。台阶的两侧宽达1米的边墙，北边墙
下端，有一个带羽行的大蛇头石刻，蛇头高1.43米，长达1.80米，宽1.07米，蛇嘴里吐出一条大舌头，颇为独特。这座古老的建筑，在建造之前，经过了精心的几何设计，它所表达出的精确度和玄妙而充满戏剧性的效果，令后人叹为观止：每年春分和秋分两天的日落时分，北面一组台阶的边墙会在阳光照射下形成弯弯曲曲的七段等腰三角形，连同底部雕刻的蛇头，宛若一条巨蛇从塔顶向大地游动，象征着羽蛇神在春分时苏醒，爬出庙宇。每一次，这个幻像持续整整3 小时22 分，分秒不差。这个神秘景观被称为“光影蛇形”。每当“库库尔坎”金字塔出同蛇影奇观的时候，古代玛雅人就欢聚在一起，高歌起舞，庆祝这位羽毛蛇神的降临。库库尔坎金字塔，是玛雅人对其掌握的建筑几何知识的绝妙展示，而金字塔旁边的天文台，更是把这种高超的几何和天文知识表现得淋漓尽致。[1] 建造工艺 库库尔坎金字塔资料图(5张) 在没有金属工具、没有大牲畜和轮车的情况下，古代玛雅人却能够开采大量重达数十吨的石头，跋山涉水、一路艰辛地运到目的地，建成一个个雄伟的金字塔。金字塔最高的可达70米，其规模之巨大、施工难度之高，令人吃惊。古代玛雅的金字塔和古埃及的金字塔在建筑形式上有着明显的不同。埃及的金字塔的塔顶是尖的，而玛雅金字塔却是平顶，塔体呈方形，底大顶小，层层叠叠，塔顶的台上还建有庙宇；在用途上也不一样：埃及金字塔是法老的陵墓，而玛雅的金字塔除个别外，一般是用来祭祀或观察天象的。除金字塔外，玛雅人还兴建了不少功能性强，技艺高的民用建筑，主要有：房屋（包括庙宇、府邸、民居等）、公共设施（球场、广场、集市等）、基础设施（桥梁、大道、码头、堤坝、护墙等）和水利工程（水渠、水库、水井、梯田）等。 历史溯源 玛雅文化的历史
玛雅遗址的建筑历史详细介绍
玛雅遗址的建筑历史详细介绍 蒂卡尔是玛雅文明的文化和人口中心之一。这里最早的纪念碑建造于公元前四世纪。城市在玛雅古典时期——大约公元200年到公元850年左右，达到顶峰。 蒂卡尔的历史可以追溯到公元前1世纪，起先她只是个二流城邦，奉北部城邦米拉多为盟主，蒂卡尔可能从公元二世纪开始奉行与中美洲当时的大国提奥提华坎友好的政策，并取得成功使自己国力大涨，从而成为玛雅世界最强大的城邦之一。 公元292年，Balam Ajaw(传统上译为“Decorated Jaguar”，被装饰过的美洲虎)继位为蒂卡尔国王，开始了强大的蒂卡尔王朝。在一块刻有他形象的石碑上记录有公元292年7月8日的日期。这个时间也通常被历史学家当作玛雅古典文明的开始之日。 公元360年，Chak Toh Ich’ak I 美洲虎之爪一世(Jaguar Paw I)为国王之时，蒂卡尔空前强大，“美洲虎之爪一世”的弟弟“吸烟的青蛙”于公元378年1月16日征服邻邦乌夏克吞(Uaxactun)，并成为乌夏克吞的君主。
美洲虎之爪一世以后的蒂卡尔国王叫做Nun Yax Ayin，(King Curl-Nose 蜷鼻王) 是一个来自提奥提华坎的贵族，于379年继位，蜷鼻王虽来自提奥提华坎，但却也是个地道的蒂卡尔国王，保持着蒂卡尔的强大与繁荣。 公元411年至公元456年，Siyah Chan K'awil II 暴风雨天空二世(Stormy Sky II) 为国王。从美洲虎之爪一世到暴风雨天空二世的一百年间是蒂卡尔历史上第一个强大的时期。 此后蒂卡尔以北的城邦卡拉克穆尔强大，蒂卡尔出现了一个衰落时期，直到公元7世纪，Hasaw Chan K’awil 阿赫卡王(双月，巧克力领主Double Moon or Lord Chocolate 682年-734年在位)，蒂卡尔重新强大，打败卡拉克穆尔，使得蒂卡尔重新成为玛雅中部地区的霸主。以后是Yik’in Chan Kawil 雅克金王(734年-766年在位)和Yax Nuun Ayiin II 奇坦王(Chitam 768年-790年在位)，在这一百年间是蒂卡尔第二次鼎盛时期。蒂卡尔壮丽的遗迹主要也是这三位国王在位期间筑成。 奇坦王之后，与同一时期其他玛雅城邦一样，蒂卡尔迅速衰落，已知可考的最后一位国王是Jasaw Chan K'awiil II (869年-889年在位) 在这之后，没有建造什么主要的纪念碑。一些宫
人美版六年级下册美术练习题
六年级美术练习题 一、填空1、《哪吒闹海》是一幅有民族气派的现代（壁画），画家（张仃）在画中运用了传统壁画的表现手法，又借鉴了现代装饰画的理念，在（构图）（色彩）（造型）等方面都进行了成功的探索，（构图）新颖、色彩绚丽，颇有（旋转感）和装饰意趣。 2、《拉斯科洞窟的岩画》被称为“史前的卢浮宫”，产生于1.5万年前的旧石器时代，这是人类最早的绘画。是用蕨草、羽毛等工具，木炭、矿砂等颜料画成的。书中画的是（马）。 3、墨西哥的《库库尔坎金字塔》与埃及金字塔不同，埃及金字塔外表的四个面是（平）的，顶端是（尖）的；《库库尔坎金字塔》外表呈（阶梯）形，塔顶是（平）的，是羽蛇神庙，它产生于2000多年前的玛雅文明，由9层晒台91级台阶组成。 阿布辛拜勒神庙前的拉美西斯二世雕塑始建于公元前1300年——公元前1233年。它是埃及最强大的国王拉美西斯二世时期建造的。 4、日本的《法隆寺》建于1400年前，是最古老的木建筑，它与中国传统的塔造型相近，但比中国的塔纤秀、飘逸一些，檐翼宽大，塔顶尖细。 5、奥林匹克运动会起源于公元前776年古希腊的运动竞赛，由古希腊的一个地名奥林匹亚而得名。古希腊雕塑是世界雕塑史上的一座丰碑，健美的运动员，造型高度写实，姿态优雅，容貌端正，反映古希腊人对美的追求。（掷铁饼者公元前450年、竞技表演公元前510年）哈夫拉金字塔埃及公元前2500年、巨石阵英国公元前3100-公元前1100年、毛阿伊石像智利公元700-800年、比萨斜塔意大利1173-1350年、圣瓦西里大教堂俄罗斯1561年。 6、世界第一峰是珠穆朗玛峰，海拔8844.43米，藏语中，其含义是（圣母）的意思。尼泊尔人称它是（“萨加玛塔峰”），意思是“摩天峰”。 7、《巨人之路》位于英国，它的四万根石柱，完全是（大自然）的杰作。 大堡礁位于澳大利亚东北部，由2900个珊瑚礁组成，是世界七大自然奇景之一。 自然美是一个整体，天空、大地、山峦、河流、树木……构成了和谐之美。自然美不是精致的，而是动态的，有四季、气候、生命的变化……我们可以亲身感受他们的芬芳、声音、冷暖……自然美的特征有a 自然本身的质地、材料、色彩、形状等。B 自然偏重于形式。C 自然具有联想性等。 维多利亚瀑布非洲津巴布韦，加兰巴国家公园非洲刚果民主共和国，科罗拉多大峡谷美洲美国，罗斯格拉西亚雷斯冰川国家公园莫雷罗冰川美洲阿根廷，马埃谷地自然保护区非洲塞舌尔，乌卢鲁国家公园的艾尔斯巨石大洋洲澳大利亚吴哥古迹柬埔寨亚洲、科罗拉多大峡谷美国美洲
世界著名旅游景点介绍(doc 18页)(精美教案)
世界著名旅游景点一览秘鲁印加遗址——Machu Picchu – Peru 20世纪初，人们传说在秘鲁安第斯山脉的崇山峻岭中有一座神秘的古城。西班牙人在长达300年的殖民统治期间对它一无所知，秘鲁独立后100年间也无人涉足。400多年的时光，只有翱翔的山鹰一睹古城的雄姿。它就是今天的马丘比丘印加遗址。
墨西哥玛雅古迹——Chichen Itza - Mexico 奇琴伊察玛雅城邦遗址曾是古玛雅帝国最大最繁华的城邦。遗址拉于尤卡坦半岛中部。始建于公元514年。城邦的主要古迹有：千柱广场，它曾支撑巨大的穹窿形房顶。可见此建筑物之大。武士庙及庙前的斜倚的两神石像。9层高30米的呈阶梯形的库库尔坎金字塔。以及圣井(石灰岩竖洞)和筑在高台上呈蜗形的玛雅人古天文观象台，称“蜗台”。
印度泰姬陵——Taj Mahal 泰姬陵是莫卧儿王朝第五代君主沙杰罕为宠姬泰吉·玛哈尔修筑的陵墓，整个建筑通体用白色大理石砌成，外形端庄华美，无懈可击，寝宫门窗及围屏都用白色大理石镂雕成菱形带花边的小格，墙上用翡翠、水晶、玛瑙、红绿宝石镶嵌着色彩艳丽的藤蔓花朵，光线所至，光华夺目，璀璨有如天上的星辉。几百年来，不知有多少文人墨客为泰姬陵折腰，写下无数
动人的诗篇。但是当你真正站在她面前时，才会深深体会到，泰姬陵之美，原来是任何文字都无法书写的。绝代有佳人，遗世而独立。俏立于亚穆纳河畔那个洁白晶莹、玲珑剔透的身 影，秀眉微蹙，若有所思。 好望角——Cape Town 非洲西南端的岬角。1488年葡萄牙航海家迪亚士在寻找欧洲通向印度的航路时到此，因多风暴，取名风暴角。但从此通往富庶的东方航道有望，故改称好望角。苏伊士运河通航前，来往于亚欧之间的船舶都经过好望角。现特大油轮无法进入苏伊士运河，仍需取此道航行。好望角多暴风雨，海浪汹涌，位于来自印度洋的温暖的莫桑比克厄加勒斯洋流和来自南极洲水域
人美版六年级美术下册《追寻文明的足迹》
《追寻文明的足迹》创新教案 学习目标： 1能够收集世界文化遗产的相关资料，初步认识、了解自己感兴趣的文化遗产。 2通过本课学习，认识到文化遗产是人类共同的财富，增强保护文化遗产的意识及方法。 3学习了解各个国家不同时期的文明足迹，尊重不同文化背景下的文化现象，接纳多元文化，拓展国际视野。 教学重、难点： 重点：1通过资料阅读、课上交流等途径，认识7处~10处世界文化遗产，较细致地了解2处~3处世界文化遗产；2能够从文化遗产的历史背景、审美价值、社会影响、经典特征等方面，介绍2处~3处世界文化遗产。 难点：有步骤、有层次地介绍2处~3处文化遗产。 课前准备： 教师：文化遗产高清组图，视频资料等； 学生：搜集感兴趣的文化遗产资料（格式不限）。 教学过程： 一、我们一起来了解一下奥运会的发祥地——希腊。 1、欣赏图片，了解奥林匹亚遗址及有关奥运会知识。 希腊属于欧洲，首都是雅典。奥林匹亚是奥林匹克运动的发祥地。这里气候宜人，景色优美，到处都是橄榄树、桂树和柏树。古奥运会起源于公元前776年，终止于公元394年，共举行293届，绵延1170年。1896年4月6日至15日，第一届现代奥运会终于在雅典隆重举行。奥林匹克运动从此开始了一个崭
新的纪元。现代奥林匹克运动会虽然改在各国轮流举行，但仍然沿用这一名称，并且在这里点燃各届奥运会的圣火。奥林匹亚有世界上最古老的运动场。 2、欣赏《掷铁饼者》，了解相关内容。 在古代奥运会上获一次冠军者，可以在运动场墙壁上镌刻自己的名字；三次夺冠，雕塑家就会找上门来，为你塑像。我们看到的《掷铁饼者》就是这样一尊雕塑，它属于圆雕作品。大家看到这尊雕塑有什么感受？ 刻画的是一名强健的男子掷铁饼的过程。在作品中创造了一个出色的健美而又富于力感的运动员形象。 雕塑选择的是将铁饼摆到最高点、即将抛出的一刹那，有着强烈的“引而不发”的吸引力。虽然是一件静止的雕塑，但艺术家把握住了从一种状态转换到另一种状态的关键环节，达到了使观众心理上获得“运动感”的效果，成为后世艺术创作的典范。这尊雕像被认为是“空间中凝固的永恒”，直到今天仍然是代表体育运动的最佳标志。 此雕塑由米隆创作于约公元前450年，被公认为体育运动和健美体魄的象征。这是雕刻家从实际生活中观察得来的真实形象。我们看到的大理石雕复制品，高约152厘米，罗马国立博物馆、梵蒂冈博物馆、特尔梅博物馆均有收藏，原作为青铜材质。 3、欣赏《竞技表演》。 《竞技表演》也是一座以表现奥林匹克运动为内容的雕塑作品，它属于浮雕。从图片上大家可以看出哪些内容？（一些运动员的动作，各种运动项目。）从这些人物的表现上你感觉这些运动员身体素质怎样？（非常健美，有力量的感觉。）这座浮雕通过表现竞技场上的运动员们在精神上与肉体上的坚强有力和美感，向我们传达出生命的健美、庄重与和谐，给我们带来美的享受。 二、了解其它世界文化遗产。