Flynn’s taxonomy classifies computer architectures based on:Memory and CPU sizeInstruction and data streamsSpeed and costCache and bandwidthB) Instruction and data streamsFlynn’s taxonomy defines SISD, SIMD, MISD, and MIMD.
Which memory model is used in shared-memory multiprocessors?NUMAUMADistributed memoryVirtual memory onlyB) UMAUniform Memory Access (UMA) provides equal access times for processors.
In parallel computing, load balancing is important to:Reduce memoryEvenly distribute tasksLimit processor countIncrease sequential partsB) Evenly distribute tasksLoad balancing avoids processor idling by distributing tasks equally.
Which of the following is NOT an advantage of parallel computing?Faster computationEnergy efficiencyEasier debuggingHandling large problemsC) Easier debuggingDebugging is harder in parallel systems due to concurrency.
Which interconnection network uses a grid-like structure?HypercubeMeshRingStarB) MeshMesh networks connect processors in a grid pattern.
Flynn’s taxonomy classifies computers based on:Instruction and data streamsMemory hierarchyCache sizeClock speedA) Instruction and data streamsFlynn classified architectures into SISD, SIMD, MISD, MIMD.
Which is an example of MIMD architecture?GPUSupercomputerALUELUB) SupercomputerSupercomputers often use MIMD, where multiple instructions operate on multiple data.
Speedup in parallel systems is limited by:Moore’s LawAmdahl’s LawLittle’s LawNoneB) Amdahl’s LawAmdahl’s Law shows parallel speedup is limited by sequential portions.
Which scheduling is used in parallel jobs?Round RobinWork stealingPriorityFIFOB) Work stealingWork stealing balances workload across multiple processors.
Parallel computing is most useful for:Word processingWeather simulationFile storageSpreadsheetsB) Weather simulationWeather models require massive parallel computations.
Amdahl’s Law is used to predict:Power consumptionSpeedup of parallel systemsCache performanceCPU cyclesB) Speedup of parallel systemsAmdahl’s Law shows limits of parallelization.
SIMD stands for:Single Instruction, Multiple DataSimple Input, Multiple DevicesSynchronized Instruction, Multi-core DesignNoneA) Single Instruction, Multiple DataSIMD executes the same instruction on multiple data simultaneously.
Race conditions occur when:Two processes access shared data unsafelyProcess runs too slowlyThreads execute sequentiallyMemory is fragmentedA) Two processes access shared data unsafelyWithout synchronization, parallel threads may corrupt data.
Barrier synchronization means:Tasks wait until all reach a checkpointOnly one thread executesThreads skip synchronizationProcesses run independentlyA) Tasks wait until all reach a checkpointBarrier ensures all processes synchronize before proceeding.
SIMD architecture is mostly useful in:Text editingImage processingFile transferDatabase queriesB) Image processingSIMD applies one instruction to multiple data streams, ideal for images.
Multithreading improves performance by:Executing multiple tasks concurrentlyIncreasing cache sizeReducing memoryLimiting context switchingA) Executing multiple tasks concurrentlyThreads allow CPU cores to work simultaneously.
Deadlock in parallel systems occurs when:Processes wait indefinitely for resourcesCPU overheatsData is replicatedCache is fullA) Processes wait indefinitely for resourcesCircular waiting causes deadlock.
Which scheduling assigns tasks to available processors dynamically?Static schedulingDynamic schedulingPreemptive schedulingRound-robinB) Dynamic schedulingTask distribution is decided at runtime.
Speedup in parallel computing is measured by:Execution time of serial vs parallelMemory usageProcessor count onlyLatency of I/OA) Execution time of serial vs parallelSpeedup = T(serial) / T(parallel).
Amdahl’s law is used to measure:Disk capacityMaximum speedup with parallelizationGPU memory usageNetwork latencyB) Maximum speedup with parallelizationIt shows diminishing returns as sequential part dominates.
False sharing in parallel computing occurs when:Caches of different processors interfere on shared memorySame code executes twiceMultiple users share file systemGPUs run out of VRAMA) Caches of different processors interfere on shared memoryCauses unnecessary invalidations, slowing performance.
Which programming model is widely used for GPUs?CUDAOpenMPMPICilkA) CUDACUDA enables parallel computing on NVIDIA GPUs.
Embarrassingly parallel tasks are:Hard to divideCannot run concurrentlyEasily divided with no inter-process communicationAlways sequentialC) Easily divided with no inter-process communicationExamples include image filtering or Monte Carlo simulations.
Parallel computing primarily improves: Data redundancyProcessing speedDisk storageLatency onlyB) Processing speedTasks are executed simultaneously to save time.
Load balancing ensures:Equal work distributionIncreased memoryReduced synchronizationFixed core usageA) Equal work distributionIt prevents any processor from being overloaded.
Amdahl’s Law relates to:Theoretical speedup limits in parallel computingCache memoryCPU pipeline stagesDisk latencyA) Theoretical speedup limits in parallel computingIt shows diminishing returns with more processors.
Message Passing Interface (MPI) is used for:Sequential computationParallel communicationGPU renderingDatabase accessB) Parallel communicationMPI allows distributed systems to communicate efficiently.
SIMD architecture means:Single Instruction Multiple DataSingle Instruction Multiple DevicesSequential Independent MachinesSimple Integrated Memory DesignA) Single Instruction Multiple DataOne operation runs on many data elements.
Amdahl’s Law calculates:Speedup from parallelizationProcessor heatData accuracy Cache sizeA) Speedup from parallelizationIt shows limits due to serial portions.
Race conditions occur whenThreads access shared data unsafelyCPUs overheat Threads idleMemory is fullA) Threads access shared data unsafelyMultiple threads interfere with updates.
GPU computing is ideal for:Sequential tasksParallel numeric operationsFile handlingText parsingB) Parallel numeric operationsThousands of GPU cores work simultaneously.
The term “load balancing” ensures:Equal work among processorsMemory allocationDisk cachingInput batchingA) Equal work among processorsIt avoids idle cores.
CUDA is used primarily for:CPU computingGPU programmingNetwork routingHardware debuggingB) GPU programmingCUDA enables parallel processing on NVIDIA GPUs.
Deadlock in parallel systems occurs whenProcessors overheatThreads wait indefinitely for resourcesCache memory is fullDisk failsB) Threads wait indefinitely for resourcesCircular dependency leads to system halt.
A system that maintains same data on multiple servers to ensure availability is called:ShardedReplicatedPartitionedQueuedB) ReplicatedReplication enhances fault tolerance and uptime.
What is a race condition?Two processes waiting foreverIncorrect results due to unsynchronized accessCPU overheatingThread starvationB) Incorrect results due to unsynchronized accessWhen threads access shared data concurrently, results can become inconsistent.
What is SIMD?Single Instruction, Multiple DataSequential Instruction, Multiple DatSingle Input, Multi DelaySimple Instruction Machine DesignA) Single Instruction, Multiple DataSIMD executes the same operation on multiple data elements simultaneously.
Which architecture supports parallel instruction execution?Von NeumannHarvard SuperscalarMicrokernelC) SuperscalarSuperscalar CPUs execute multiple instructions per clock cycle.
What helps in balancing workload across processors?Thread lockingLoad balancingMemory mappingPagingB) Load balancingIt distributes computational tasks evenly to maximize efficiency.
What is the main goal of parallel computing?Reduce cost Increase execution speed Simplify codeSave memoryB) Increase execution speedParallel computing divides tasks to run simultaneously for speed.
A system where multiple processors share memory is:Distributed memory systemShared memory systemHybrid systemCloud systemB) Shared memory systemShared memory allows all processors to access common data.
Which metric measures parallel performance improvement?Speedup DelayComplexityCostA) SpeedupSpeedup = Time(serial) / Time(parallel).
What is a key drawback of parallel systems?Fault toleranceSynchronization overhead Fast computationScalabilityB) Synchronization overheadManaging task coordination adds extra time and complexity.
The main goal of parallel computing is to:Increase program complexityReduce execution timeIncrease memory useSimplify algorithmsB) Reduce execution timeParallelism splits work among processors to improve performance.
Amdahl’s Law is used to measure:Memory latencyParallel efficiencyCache hit ratio Communication speedB) Parallel efficiencyIt defines the theoretical speedup of a task when parallelized.
SIMD architecture executes:Multiple instructions on single data Single instruction on multiple dataMultiple instructions on multiple dataSingle instruction on single dataB) Single instruction on multiple dataSIMD performs the same operation across data streams simultaneously.
Which memory is shared in multiprocessor systems?Distributed memoryLocal memoryGlobal memoryNoneC) Global memoryGlobal memory is accessible by all processors in shared-memory systems.
What is the main goal of parallel computing?To reduce memory usageTo execute multiple tasks simultaneouslyTo simplify code structureTo increase energy consumptionB) To execute multiple tasks simultaneouslyParallel computing splits a task into smaller sub-tasks executed concurrently to increase performance.
Which model divides computation into “threads” that share the same memory space?Distributed memory modelShared memory modelHybrid modelGrid modelB) Shared memory modelThreads in a shared memory model access a common memory area, improving communication efficiency.
What is Amdahl’s Law used for?Measuring CPU temperatureEstimating potential speedupPredicting cache sizeCalculating network delayB) Estimating potential speedupAmdahl’s Law estimates the maximum improvement achievable by parallelizing a task.
In GPU computing, what is a “kernel”?A memory addressA function executed on the GPU A data structureA storage moduleB) A function executed on the GPUKernels are user-defined functions that run on many GPU threads in parallel.
What limits parallel program scalability?More processorsCommunication overheadLarge memory High bandwidthB) Communication overheadToo much inter-processor communication reduces parallel efficiency.
Which component coordinates tasks among processors?SchedulerCachePipelineBusA) SchedulerThe scheduler allocates and balances workload among processors in parallel environments.
Which computing system performs one task across multiple cores simultaneously?Serial computingParallel computingDistributed computingReal-time computingB) Parallel computingParallel computing breaks one large problem into smaller tasks executed concurrently.
What is the purpose of load balancing?Reduce data redundancyDistribute tasks evenlyIncrease task priorityLimit parallel threadsB) Distribute tasks evenlyLoad balancing ensures even workload distribution for optimal performance.
What is the main goal of parallel computing?Reduce costSpeed up computationSimplify codeIncrease storageB) Speed up computationParallel computing divides tasks to run simultaneously for faster processing.
Which law predicts the theoretical speedup in parallel systems?Moore’s LawAmdahl’s Law Newton’s LawGustafson’s LawB) Amdahl’s LawIt defines the limits of speedup based on the non-parallel portion of a program.
Which is not a parallel architecture?Shared memoryDistributed memoryPipelineSerial CPUD) Serial CPUSerial processors execute one instruction at a time, not in parallel.
What does load balancing ensure in parallel computing?Equal workload among processorsFault toleranceNetwork securityData replicationA) Equal workload among processorsPrevents idle processors and improves system efficiency.
SIMD architecture stands for:Single Instruction Multiple DataSimple Instruction Many Devices Sequential Information ModelSingle Interface ModuleA) Single Instruction Multiple DataSIMD executes the same instruction on multiple data elements simultaneously.
Load balancing in parallel computing ensures:Equal CPU speedEqual distribution of work Memory optimizationFault recoveryB) Equal distribution of workIt prevents processor idling and improves efficiency.
Which memory is shared among processors?Local DistributedShared memoryCacheC) Shared memoryIn shared-memory systems, all processors access a common address space.
Flynn’s taxonomy classifies:AlgorithmsParallel architectures Software patternsOperating systemsB) Parallel architecturesIt divides computer architectures into SISD, SIMD, MISD, and MIMD categories.
A barrier in parallel programming is used to:Stop data sharingSynchronize threadsSplit memory Reduce loadB) Synchronize threadsBarriers ensure all threads reach a certain point before proceeding.
Amdahl’s Law defines:Theoretical speedup of parallel programsEnergy efficiencyCache size Data securityA) Theoretical speedup of parallel programsIt measures potential improvement based on parallelization.
Load balancing in parallel computing ensures:Equal power consumptionEqual workload among processorsThread terminationFault detectionB) Equal workload among processorsIt distributes work evenly to optimize resource use and minimize idle time.
Which type of dependency limits parallelism?Data dependency Hardware dependencyCache dependencyTime dependencyA) Data dependencyWhen tasks rely on each other’s output, they can’t run simultaneously.