• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.30B no.2
• /
• pp.33-43
• /
• 1993

Lock mechanism is essential for synchronization on the multiprocessor systems. Lock mechanism needs to reduce the time for lock operation in low lock contention. Lock mechanism must consider the case of the high lock contention. The conventional lock control scheme in memory results in the increase of bus traffic and memory utilization in lock operation. This paper suggests a lock scheme which stores the lock data in cache and manages it efficiently to reduce the time spent in lock operation when the lock contention is low on a multiprocessor system built on HiPi-bus(Highly Pipelined bus). This paper also presents the design of the HIPi-CLOCK (Highly Pipelined bus Cache LOCK mechanism) which transfere the data from on cache to another when the lock contention is high. The designed simulator compares the conventional lock scheme which controls the lock in memory with the suggested HiPi-CLOCK scheme in terms of the RMW(Read-Modify-Write) operation time using simulated trace. It is shown that the suggested lock control scheme performance is over twice than that of the conventional method in low lock contention. When the lock contention is high, the performance of the suggested scheme increases as the number of the shared lock data increases.

• Proceedings of the IEEK Conference
• /
• 1999.11a
• /
• pp.421-425
• /
• 1999

As the market size of multiprocessor systems for commercial applications, parallel systems, especially cache-coherent shared-memory multiprocessors that are conventionally designed for scientific applications need to be tuned in different fashion to achieve the best performance for new application area. In this paper, indepth investigation on the memory behavior which is the primary cause for performance changes were made. We chose representative benchmarks in scientific and commercial application areas. After running execution-driven simulation for bus-based cache-coherent shared-memory multiprocessors, we experienced significant differences and conclude that the systems must be carefully and differently designed to achieve the best performance when they are built for distinct applications.

• Journal of the Institute of Electronics Engineers of Korea CI
• /
• v.37 no.6
• /
• pp.1-9
• /
• 2000

This paper presents a Cluster Oriented Multistage Interconnection Network called COMR. COMR can be constructed suitable for the parallel application with localized communication by providing the shortcut path inside the processor-memory cluster which has frequent data communication. We evaluate the performance of COMR with respect to probability of acceptance, bandwidth, cost-effectiveness and average distance under varying degrees of localized communication. According to the result of analysis for performance evaluation, COMR shows higher performance than the regular MINs of the same network size in the highly localized communication. In the worst case, the diameter of an N$\times$N COMR is only n+1 which has only one stage more as compared the MIN with the same network size. Therefore COMR can be used as an attractive interconnection network for parallel applications with not only the localized communication distribution but also the uniform distribution in shared-memory multiprocessor system.

• Journal of KIISE:Computer Systems and Theory
• /
• v.33 no.9
• /
• pp.699-708
• /
• 2006

Bus splitting technique reduces bus energy by placing modules with frequent communications closely and using necessary bus segments in communications. But, previous bus splitting techniques can not be used in MPSoC platform, because it uses cache coherency protocol and all processors should be able to see the bus transactions. In this paper, we propose a bus splitting technique for MPSoC platform to reduce bus energy. The proposed technique divides a bus into several bus segments, some for private memory and others for shared memory. So, it minimizes the bus energy consumed in private memory accesses without producing cache coherency problem. We also propose a task allocation technique considering cache coherency protocol. It allocates tasks into processors according to the numbers of bus transactions and cache coherence protocol, and reduces the bus energy consumption during shared memory references. The experimental results from simulations say the bus splitting technique reduces maximal 83% of the bus energy consumption by private memory accesses. Also they show the task allocation technique reduces maximal 30% of bus energy consumed in shared memory references. We can expect the bus splitting technique and the task allocation technique can be used in multiprocessor platforms to reduce bus energy without interference with cache coherency protocol.

• Journal of Korea Society of Industrial Information Systems
• /
• v.7 no.5
• /
• pp.121-128
• /
• 2002

Although multiprocessor systems are widely used in recent years to run commercial workloads, data sharing patterns are rarely explored due to several difficulties. In this paper, we made in-depth sharing pattern analysis for a representative OLTP application, the TPC-B benchmark, running on a cache-coherent shared-memory multiprocessor system. In addition, to illustrate their effects on the performance, the number of cache misses were measured for various numbers of processors, cache sizes and cache block sizes. From these measurements, we found out the shared data in TPC-B largely bear quite different sharing characteristics from those in scientific applications.

• Journal of the Korean Institute of Telematics and Electronics C
• /
• v.34C no.12
• /
• pp.9-19
• /
• 1997

This paper describes the design and FPGA implementation of a system control unit within a multiprocessor chip which can be used as a node processor ina massively parallel processing (MPP) caches, memory management units, a bus unit and a system control unit. Major functions of the system control unit are locking/unlocking of the shared variables of protected access, synchronization of instruction execution among four integer untis, control of interrupts, generation control of processor's status, etc. The system control unit was modeled in very high level using verilog HDL. Then, it was simulated and verified in an environment where trap handler and external interrupt controller were added. Functional blocks of the system control unit were changed into RTL(register transfer level) model and synthesized using xilinx FPGA cell library in synopsys tool. The synthesized system control unit was implemented by Xilinx FPGA chip (XC4025EPG299) after timing verification.

• Journal of the Korea Society of Computer and Information
• /
• v.12 no.3
• /
• pp.51-58
• /
• 2007

The shared-nothing multiprocessor architecture has advantages in scalability, this architecture has been adopted in many multiprocessor database system. But, if the data are not uniformly distributed across the processors, load will be unbalanced. Therefore, the whole system performance will deteriorate. This is the data skew problem, which usually occurs in processing parallel hash join. Balancing the load before performing join will resolve this problem efficiently and the whole system performance can be improved. In this paper, we will present an algorithm using merit of very large memory to reduce disk access overhead in performing load balancing and to efficiently solve the data skew problem. Also, we will present analytical model of our new algorithm and present the result of some performance study we made comparing our algorithm with the other algorithms in handling data skew.

• Journal of the Korea Society of Computer and Information
• /
• v.11 no.2 s.40
• /
• pp.35-42
• /
• 2006

In this paper we implement a new simulator for performance analysis of a parallel digital signal processing distributed shared memory multiprocessor systems. using Parsec The key idea of this simulator is suitable in simulation of system that uses DMA function of TMS320C6701 DSP chip and local memory which have fast access time. Also, because correction of performance parameter and reconfiguration for hardware components are easy, we can analyze performance of system in various execution environments. In the simulation, FET, 2D FET, Matrix Multiplication. and Fir Filter, which are widely used DSP algorithms. have been employed. Using our simulator, the result has been recorded according to different the number of processor, data sizes, and a change of hardware element. The performance of our simulator has been verified by comparing those recorded results.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.7
• /
• pp.2219-2228
• /
• 2000

Caches in a multiprocessing environment introduce the cache coherence problem. When multiple processors maintain locally cached copies of a unique shared-memory location, any local modification of the location can result in a globally inconsistent view of memory. Cache coherence protocols are important to operate a shared-memory multiprocessor system with efficiency and correctness. Since random testing and simulations are not enough to validate correctness of protocols, it is necessary to develop efficient and reliable verification methods. In this appear we present our experience in using VIS (Verification Interacting with Synthesis), a tool of formal method, to analyze a number of property of a cache coherence protocol, RACE (Remote Access Cache coherent Enforcement).

• Journal of KIISE:Computer Systems and Theory
• /
• v.32 no.7
• /
• pp.349-358
• /
• 2005

False sharing is a result of co-location of unrelated data in the same unit of memory coherency, and is one source of unnecessary overhead being of no help to keep the memory coherency in multiprocessor systems. Moreover. the damage caused by false sharing becomes large in proportion to the granularity of memory coherency. To reduce false sharing in a page-based DSM system, it is necessary to allocate unrelated data objects that have different access patterns into the separate shared pages. In this paper we propose call-site tracing-based shared memory allocator. shortly CSTallocator. CSTallocator expects that the data objects requested from the different call-sites may have different access patterns in the future. So CSTailocator places each data object requested from the different call-sites into the separate shared pages, and consequently data objects that have the same call-site are likely to get together into the same shared pages. We use execution-driven simulation of real parallel applications to evaluate the effectiveness of our CSTallocator. Our observations show that by using CSTallocator a considerable amount of false sharing misses can be additionally reduced in comparison with the existing techniques.