L. Prylli and B. Tourancheau, BIP: a new protocol designed for high performance networking on myrinet, Proceedings of IPPS/SPDP98, 1998.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....via some form of shared memory must use exactly the same mechanisms as message passing applications use. However, we claim that shared memory systems have several characteristics that can be used in improving the messaging performance even further. In order to support this claim, we focus on MSAs [15, 5, 8, 1, 13, 7, 14, 6, 11, 3] that have been proposed and implemented for the Myrinet LAN and on applications running on top of pagebased software distributed shared memory systems (software DSMs) More specifically, we support our claim by studying the communication behavior of several applications running on top of the ....

L. Prylli and B. Tourancheau. BIP: A New Protocol Designed for High-Performance Networking on Myrinet. In Proceedings of IPPS/SPDP98, 1998.

....Systems Inc. Pacific Northwest National Laboratory 1. Introduction In the past five years, the Myrinet network has received much attention in the literature on high performance communication. Multiple projects have focused on developing efficient messaging middleware (e.g. AM[1] PM[2] BIP[3], and HPVM FM [4] by exploiting the programmable network interface card (NIC) of Myrinet. The most common use of these interfaces has been support of MPI and other internal research projects. The HPVM project used its FM system to implement MPI and other programming interfaces including two ....

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In Workshop PC-NOW, IPPS/SPDP98, Orlando, USA, 1998.

....bridges the performance gap between remote memory operations that target registered and those that use regular memory. Our approach relies on the standard unmodified system software and drivers for Myrinet and cLAN rather than on custom alternative drivers and interfaces (e.g. AM [1] PM [2] BIP [3], and FM [4] interfaces that replace the standard Myrinet Control Program (MCP) on the network interface card. The paper makes several contributions to the field. First it presents a software architecture that supports efficiently a complete set of remote memory operations including remote copy, ....

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In Workshop PC-NOW, IPPS/SPDP98, 1998.

....the operating system in the critical path of communication significantly reduces performance due to the latency of context switching and other overhead. This has fueled the development of userlevel communication protocol systems such as AM [10] VMMC [8] FM [14] U Net [18, 19] LAPI [16] and BIP [15]. GM [3] is a user level communication protocol that runs over the popular Myrinet interconnect [9] It gives reliable in order delivery of packets with very low latency, and high bandwidth. Thus, it is interesting to analyze whether Myrinet interconnect with GM communication layer is suitable for ....

L. Prylli and B. Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. In Proceedings of the International Parallel Processing Symposium Workshop on Personal Computer Based Networks of Workstations, 1998. http://lhpca.univ-lyon1.fr/.

....a cluster of dual INTEL PENTIUM II 450 MHz PC nodes with 128 MB of RAM and a 33 MHz 32bit PCI bus running LINUX Kernel 2.2.13. The cluster interconnection networks are 100 Mb s FAST ETHERNET for TCP, DOLPHIN SCI ( 8] 7] D310 NICs) for SISCI and MYRINET (LANai 4. 3, 32bit bus, 1 MB SRAM) for BIP [10]. Results were obtained by ping pong tests (half round trip) Raw Performances Raw performance results for Madeleine III over TCP Fast Ethernet, SISCI SCI and BIP Myrinet are displayed in figures 6, 7 and 8. The performance of Madeleine III over SISCI SCI shows that the inherent overhead of ....

L. Prylli and B. Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472--485. Held in conjunction with IPPS/SPDP 1998.

....a HPF version and an OpenMP. These parallel codes have been executed on a Compaq Proliant 6000 (4 processors) a Cray T3E 750 (300 processors) a HP class V (16 processors) a SGI Origin2000 (32 processors) a cluster of PCs (8 BiPentium III 800 MHz under Linux equipped with Myrinet devices[3]) and a COMPAQ SC 232 (232 processors) This Section presents the analysis we have conducted and the experimental results we have obtained with the HRM1D code. These experimental results are discussed according to the following criteria : efficiency, scalability, maintainability, developing costs ....

Loic Prylli, Bernard Tourancheau. Bip : a new protocol designed for high performance networking on myrinet. In Workshop PC-NOW, IPPS/SPDP98 (www-bip.univ-lyon1.fr/software/bip-manual.ps)

....Mb s Fast Ethernet for TCP, Dolphin SCI ( 8] 7] D310 NICs) for 10 100 1000 10000 100000 Transfer time over Madeleine III TCP 0 5 10 Bandwidth over Madeleine III TCP Figure 6: Latency and bandwidth over TCP Fast Ethernet SISCI and Myrinet (LANai 4. 3, 32bit bus, 1 MB SRAM) for BIP [10]. Results were obtained by ping pong tests (half round trip) Raw Performances Raw performance results for Madeleine III over TCP Fast Ethernet, SISCI SCI and BIP Myrinet are displayed in gures 6, 7 and 8. The performance of Madeleine III over SISCI SCI shows that the inherent overhead of ....

L. Prylli and B. Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472485. Held in conjunction with IPPS/SPDP 1998.

....port opening and closing, memory mapping, interrupt handling and loading the MCP. The ability to change the network behavior through the MCP has made possible the implementation and testing of a number of communication protocols on Myrinet, including Active Messages [4] Fast Messages [12] and BIP [13]. Myricom s own software, called GM, borrows many features from these protocols and is currently the most widely used and preferred software for Myrinet. All these protocols are low overhead protocols that avoid operating system intervention by providing a zero copy mode of operation directly from ....

L. Prylli and B. Tourancheau. BIP: A new protocol designed for high performance networking on myrinet. Lecture Notes in Computer Science, 1388:472-480, 1998.

....One of the goal of Madeleine II is to support multi modal protocols such as VIA [5] or SCI [4] Such protocols provide several data transfer methods, namely a regular transfer mode and a DMA (Direct Memory Access) mode. Moreover, it should be able to easily take into account protocols such as BIP [8] which makes a difference between short buffers and long buffers. As a consequence, Madeleine II features specific modules to encapsulate each of these sub protocols. These modules are called Transmission Modules (TM) Table 2 displays the generic TM interface (note that some functions may not be ....

....user data (it is built on the fly) and would then be sent after the message body, generating additional (unaligned) and expensive bus transactions. 5.2. 4 BIP BIP (Basic Interface for Parallelism) is a low level communication interface specifically designed for the MYRINET network architecture [8]. The most interesting functionality of BIP is to provide communication control in user space: the application may interact directly with the network interface card. The BIP interface makes a distinction between short messages ( 1 kB) and long messages. Consequently, the BIP driver of Madeleine ....

L. Prylli, B. Tourancheau, BIP: a new protocol designed for high performance networking on Myrinet, in: 1st Workshop on Personal Computer based Networks Of Workstations (PCNOW '98), Vol. 1388 of Lect. Notes in Comp. Science, Springer-Verlag, 1998, pp. 472--485.

....library in the multi cluster management department. For explanation purpose we suppose that two three nodes clusters at are our disposal: the foo cluster with a SCI interconnection network and the goo cluster with a Myrinet interconnection network. The protocols used are SISCI and BIP ([13]) respectively. 3.1.1 Channels Building The different channel mechanisms provided by Madeleine constitute a practical means to virtualize a cluster of clusters configuration. All the relevant pieces of information are contained in two configuration files written by the user: a network ....

....oh MPICH over the Madeleine communication library, no comparisons with other existing solutions are presented (such comparisons can be found in [2] The figure will be shown for two types of high performance networks: SCI and Myrinet. The protocols used are respectively SISCI and BIP [13]. The hardware used for the experiments are Bi Pentium II nodes 450 MHz, with 128 Mbytes memory and 32 bitswide PCI bus. As for the network interface cards, Dolphin s D310 and Myricom s Lanai 9.xx were at our disposal. The test is a ping pong and we measured both transfer time and throughput. As ....

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on myrinet. In Parallel and Distributed Processing, IPPS/SPDP'98, volume 1388 of Lecture Notes in Computer Science, pages 472--485. Springer-Verlag, April 1998.

....device handles intra process communication. The smp plug device handles intra node communication (case of SMP nodes in the configuration) The ch mad device handles any inter node communication. Since the ch self and smp plug devices are both parts of the SMP implementation of MPI BIP ( 9] [16]) detailing their performance and internals is not relevant for this paper. As a consequence they won t be further discussed Network heterogeneity is dealt with by the Madeleine software layer as shown by Figure 2. This heterogeneity is handled through the use of the channels objects of ....

L. Prylli and B. Tourancheau. BIP: a new protocol designed for high performance networking on myrinet. In Parallel and Distributed Processing, IPPS/SPDP'98, volume 1388 of Lecture Notes in Computer Science, pages 472--485. Springer-Verlag, April 1998.

....naturally developped in close collaboration with Myricom hardware developpers. In addition to a pool of hardware constants and small subroutines, GM has been in used in this project as the reference to solve any documentation hole or inaccuracy that has arisen during developpment. 2.3. 2 BIP BIP [PT98] was the first communication library bringing the full wire performance to the user application. As it has less high level features and supports less host architectures than GM, its volume is much lower and so it seems at first much quicker to grasp. But its highly tuned approach and hacking ....

L. Prylli and B. Tourancheau. BIP: A new Protocol Designed for High Performance Networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472--485, Springer-Verlag, April 1998.

....of our implementation on a 32 node cluster with various network configurations are examined and reported in this paper. 1 Introduction The performance problem related to the communication software has been an active research issue for the past decade. Currently, lightweight messaging systems [7, 18, 19, 25, 32, 35, 36] offer the best communication performance, as they create a fast communication path that bypasses the traditional in kernel messaging protocol stack (e.g. TCP IP) which is a serious obstacle in exploiting the high performance of modern network [1, 15, 16] Although most of these lightweight ....

Loic Prylli and Bernard Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. In IPPS/SPDP Workshops, pages 472--485, 1998.

....of workstations. Such commodity clusters can either use commodity communication hardware, such as switched Fast Ethernet, or hardware designed especially for highperformance computing, such as Myrinet or switches based on the VI Architecture standard. Significant research has been carried out [10] in improving communication speeds using such dedicated hardware. Although such hardware improves communication speeds dramatically, at this point in time it is considerably more expensive. The large number of PCs on peoples desks makes PC This work was supported by the National Institutes of ....

....(one way) and bandwidth for the ping pong benchmark on the ASCI Red. On PC clusters, Myrinet based communication systems (such as FM [8] which was one of the earliest) provide a much higher performance compared with our switched Ethernet implementation. One of the fastest, for example, is BIP [10], which supports a less than 5sec. latency, and 120 MB sec bandwidth. So, while it is still important to try improving the performance on the commodity hardware, it is worthwhile asking whether this big a disparity in communication performance renders parallel applications infeasible. To answer ....

L. Prylli and B. Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. Lecture Notes in Computer Science, 1388:472--??, 1998.

....2.3 parallel benchmark. 4.1 Platform hardware and software The platform contains a Myrinet network with four ports. We use 4 Pentium II 300 Mhz biprocessor nodes for the experiments. Each Myrinet PCI interface has a 1MB local memory. The software environment includes Linux 2.0.33, the BIP 0. 94c [5] version of MPI library, the F77 PGI 1.7 programming environment and the Linux Pthread library. BIP raw performances on Myrinet connected PC is a latency of 5 us and a bandwidth of 1 Gbit s. MPI BIP reach 20 us (latency) and 1 Gb s (bandwidth) All benchmarks have been compiled with the 02, unroll ....

L. Prylli and B. Tourancheau. Bip: a new protocol designed for high performance networking on myrinet. In Workshop o Persocial Computers based Networks Of Workstatios, 1998.

....ports et de biprocesseurs de la famille x86 de trois gnrations diffrentes: des Pentium Pro 200 Mhz, des Pentium II 300 Mhz et des Pentium II 400 Mhz. Chaque interface du rseau Myrinet contient 1 Mo de mmoire locale. L environnement logiciel comprend le systme Linux 2.0.33, la version 0. 95d de BIP [15], MPI BIP, l environnement de programmarion PGI F77 1.7 et la librairie Pthread de Linux. Les performances asymptotiques de BIP sur Myrinet sont une latence d environ 5 us et un dbit de 1 Gbits s. Avec MPI BIP, la latence atteint environ 20 us et le dbit reste voisin de 1 Gbits s. Tousles ....

L. Prylli and B. Tourancheau. Bip: a new protocol designed for high performance networking on myrinet. In Workshop o Persocial Computers based Networks Of Workstatios, 1998.

....to unify the view of the memory model. The main drawback of the SMM approach is the performance cost of the unifying mechanisms. 1.1 Message passing SMM [3] 4] and [2] have developed a version of their message passing libraries to work in SMP platforms. We also have ported a version of the BIP [5] library for SMP PC. The BIP library had been initially developed for the Myrinet communication network. The cost of message passing in shared memory has three components: First, the system bus will be shared for intra node communication. So, colliding communications must be sequentialized. This ....

L. Prylli and B. Tourancheau. Bip: a new protocol designed for high performance networking on myrinet. In Workshop on Personal Computers based Networks Of Workstations, 1998.

....several forms of active messages, asynchronous remote copy, distributed shared memory, and even optimized versions of the traditional send receive paradigm, have been proposed. Combined, all these initiatives left the giga bit per second, application to application bandwidth barrier behind [7]. Nevertheless, good communication performance is hard to obtain when dealing with anything but the test applications supplied by the communication package developers. Real applications, not seldom, present disappointing performance. We believe many performance losses to have a common root: the ....

L. Prylli and B. Tourancheau. BIP: a New Protocol Designed for High Performance Networking on Myrinet. In Proc. PC-NOW'98, April 1998.

....whose goal is to efficiently exploit the resources provided by the underlying communication networks used in this kind of architecture. The most popular example of such communication networks today is Myrinet [7] From this effort, a great number of low level communication protocols such as BIP [20] and GM [13] have arised with the goal of providing a set of primitives to allow the user to specify the communication behaviour of his application, taking advantage of the low latency and high bandwidth rates provided by the communication networks. Besides, several implementations of the MPI ....

....B. DECK SISCI DECK (Distributed Execution and Communication Kernel) 5, 6] is a parallel programming environment developed at the GPPD UFRGS, intended for the programming of clusters. Early versions of DECK have been implemented, first based on Unix sockets, and then on BIP for Myrinet [20, 4]. With the acquisition of a SCI cluster by the group, a SCI implementation has been provided [8] Communication in DECK is realised by means of two kinds of abstractions: messages and mail boxes. Messages are containers in which data can be packed and unpacked; mail boxes, in turn, are temporary ....

Loc Prylli and Bernard Tourancheau. BIP: A new protocol designed for high performance networking on Myrinet. In Jose Rolim, editor, Proc. of PC-NOW'98, volume 1388 of Lecture Notes in Computer Science, pages 472--485. Berlin, Springer, 1998.

....grant no. SCHR 603 1 1. messages, asynchronous remote copy, distributed shared memory, and even optimized versions of the traditional send receive paradigm, have been proposed. Combined, all these initiatives left the giga bit per second, application to application bandwidth barrier behind [7]. Nevertheless, good communication performance is hard to obtain when dealing with anything but the test applications supplied by the communication package developers. Real applications, not seldom, present disappointing performance. We believe many performance losses to have a common root: the ....

L. Prylli and B. Tourancheau. BIP: a New Protocol Designed for High Performance Networking on Myrinet. In Proc. PC-NOW'98, April 1998.

....descriptor, therefore enabling receiving process to access to data. The use of these memory mapped registers allows the transfer to proceed without the use of interrupts, that represent a major penalty for modern super scalar CPUs [2] Although the concepts behind the VIA standard are not new [3 8], its main advantage is that it has been carefully designed to be operating system and processing architecture independent, therefore sharing a significant goal with the MPI standard [10] and allowing programmers to write fully portable applications. In order to design the communication suitable ....

L. Prylli and B. Tourancheau, "BIP: A New Protocol Designed for High Performance Networking on Myrinet," Lecture Notes in Computer Science, vol. 1388, pp. 472--485, Mar. 1998.

....There has been extensive work in examining protocol layer costs over other user level communication abstractions for high performance systems [9, 21, 23, 39] and from these we draw much of our methodology. Additionally, we draw upon the lessons of related highperformance network architectures [15, 31, 32] and the benchmark techniques developed by Culler et al. in [13, 14] to analyze our implementations. Specific to the VI Architecture, Speight et al. 34] conducted a benchmark study of two commercial VI implementations and compared the results to TCP IP on gigabit ethernet. Madu et. al [25] ....

L. Prylli and B. Tourancheau. BIP: A new protocol designed for high performance networking on myrinet. In Proceedings of the International Parallel Processing Symposium Workshop on Personal Computer Based Networks of Workstations, April 1998. 15

....Fellowship award, and grants from the Ohio Board of Regents. for degradation of the communication subsystem performance as seen by the applications. The research community has proposed a group of communication systems [12] such as AM [21] VMMC [8] FM [14] U Net [20, 22] LAPI [17, 6] and BIP [15] to address these issues. All of these communication systems use much simpler communication protocols in comparison with legacy protocols such as the TCP IP. The role of the operating system has been much reduced in these systems and in most cases user applications are given direct access to the ....

Loc Prylli and Bernard Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. In Proceedings of the International Parallel Processing Symposium Workshop on Personal Computer Based Networks of Workstations, 1998. http://lhpca.univ-lyon1.fr/.

....[1] often allow this bandwidth to be accessible between user level processes. While message latency has not totally kept pace with bandwidth, the use of memory mapped interface [3, 12] has allowed hardware latencies to also be reduced dramatically. For example, the BIP software developed at INRIA [22] allows round trip latencies of small messages between user level applications on the order of 10 secs over Myrinet hardware. Unfortunately, this improved performance is not uniformly available to different types of applications. High performance communication infrastructures built on top of such ....

....modeling interrupt based approaches, the incoming message queue is checked after each basic block. We simulated three different network configurations, chosen to reflect current UDP over ATM connections ( slow ) an OC 12 network ( medium ) and gigabit Myrinet [4] running the very low latency Bip [22] software package ( fast ) Table 1 shows the constants used throughout the paper. These constants were chosen to reflect the performance of CVM running on Linux 2.0.32 on 200 MHz Pentium Gateway machines. The applications used in this study are Barnes, FFT, Water, and Spatial from SPLASH 2 ....

L. Prylli and B. Tourancheau, "BIP: a new protocol designed for high performance networking on myrinet," in Workshop PC-NOW, IPPS/SPDP98, 1998.

....section. We have run it on the popc machine of LHPC. The popc (pile of PC) is a high performance cluster based on a modular architecture featuring Intel Pentium Pro processors (200 Mhz, 256 KB of cache L2 memory) interconnected by a fast MYRINET network. We use BIP as communication protocol. BIP [19](Basic Interface Protocol) allows the network to reach a 10 s latency and a 100 Mbytes s bandwidth. Size of the matrix 1000 2000 2500 Number of tasks (n 2 =2 3n=2 Gamma 2) 501498 2 002998 3 128748 Number of edges (n 2 n Gamma 4) 1 000996 4 001996 6 252496 Number of float op ( n 3 ....

L. Prylli and B. Tourancheau. Bip: a New Protocol Designed for High Performance Networking on Myrinet. In Workshop PC-NOW IPPS/SPDP'98, Orlando, USA, Apr. 1998.

....message in the network, but indefinite sequence of multipackets in the network really tests the communication libraries performance on a finite bandwidth network. These factors greatly reduce the performance of communication operations when compared to the performance of bare network hardware [48]. Experiments conducted by Dillon [16] comparing the performance of different MPI implementations and PVM revealed that the performance of libraries degrade with heterogeneous networks. The communication across heterogeneous networks needs data encoding and decoding operations at the sender and ....

....with user level communication are multiplexing the network among processes, providing protection among processes, managing resources without the help of a kernel, and effi cient programming interface to the network for versatile use by the user applications. The communication library named BIP [48] has a user level communication facility. This has helped in reducing the software overheads [48] 4.2.4 True Zero Copy Copying of messages from the application data structures to the kernel space data structures has a larger share in the software overheads during communication. Reducing ....

[Article contains additional citation context not shown here]

Prylli, L., and Tourancheau, B. BIP: A new protocol designed for high performance networking on myrinet. In "10 IPPS/SPDP Workshop on Parallel and Distributed Processing" (1998), pp. 472--485.

....greatly reducing the communication overhead. The network interface must have a certain degree of intelligence to enforce the protection boundaries established by the operating system. Several academic research efforts in high performance communications have demonstrated successful ULN solutions [5, 6, 7, 8, 9]. By standardizing the interface between the application and the networking hardware, the Virtual Interface Architecture [1] may bring fast ULN into the mainstream and enable its use by a wide variety of important, demanding applications. Already, some initial proof of concept studies have been ....

....as the operating system is involved in creating the required mappings. VMMC, however, exposes no queues and has very basic primitive operations (some of which may be a hardware load or store, in fact) User applications typically build message queues with the communication regions. BIP [8], the Basic Interface for Parallelism, is an extremely simple ULN, developed for Myrinet, with a single send queue and receive queue for one process on each node. The basic bip send( and bip recv( calls are passed direct pointers to data buffers. While queues are used, they are not directly ....

L. Prylli and B. Tourancheau. BIP: a New Protocol Designed for High Performance Networking on Myrinet., Workshop PC-NOW, IPPS/SPDP98, Orlando, USA, 1998.

....using a cache mechanism that reduces memory allocations. As a result, the latency of a typical thread migration (default stack) is very small: about 1 millisecond on an IBM SP1 (using PVM for communications) and only 78 microseconds on a Myrinet cluster of 200MHz PentiumPro machines (using BIP [30], 27] C. Related work Distributed multithreaded environments are nowadays numerous. On the one hand, some of them propose a programming model close to the PM 2 one (Nexus [14] Athapascan 0 [16] especially as far as the LRPC decomposition operator is concerned. However, Their aim is not to ....

....Therefore, it is crucial that the underlying communication layer is able to ensure the delivery of messages with no intermediate extra copy of data. Many existing communication libraries provide such a functionality, including those providing a very low level of network abstraction [37] 28] [30], 32] but also those such as MPI providing a rich set of high level communication functionalities. However, the problem of realizing an RPC operation without any extra copy of data is more complex than realizing a zero copy message transmission. When a RPC request is sent to a server, the ....

[Article contains additional citation context not shown here]

Prylli, L., and Tourancheau, B. BIP: A New Protocol designed for High-PerformanceNetworking on Myrinet. In Proc. of PC-NOW IPPS-SPDP98 (Orlando, USA, March 1998). Available at http://lhpca.univ-lyon1.fr/bip.html.

....programs. 3 Performance evaluations 3.1 Experimentation platform The platform contains a Myrinet network with four ports. 2 400 Mhz Pentium II biprocessor nodes and 2 300 Mhz Pentium II biprocessor nodes are used for the experiments. The software environment includes Linux 2.0.36, BIP 0. 95c [13] and Linux Pthread library. BIP raw performances on Myrinet is a latency of 5 us and a bandwidth of 1 Gbit s. Every tests use 2 300 Mhz PII and 1 400 Mhz PII. Client and server use 300 Mhz nodes and the broker runs on a 400 Mhz node. 3.2 Data transfers performance: Put,Get) We measure the ....

L. Prylli and B. Tourancheau. Bip: a new protocol designed for high performance networking on myrinet. In Workshop on Personal Computers based Networks Of Workstations, 1998.

....Forschungsgemeinschaft grant no. SCHR 603 1 1. remote copy, distributed shared memory, and even optimized versions of the traditional send receive paradigm, have been proposed. Combined, all these initiatives left the giga bit per second, application to application bandwidth barrier behind [4]. Nevertheless, good communication performance is hard to obtain when dealing with anything but the test applications supplied by the communication package developers. Real applications, not seldom, present disappointing performance. We believe many performance losses to have a common root: the ....

L. Prylli and B. Tourancheau. BIP: a New Protocol Designed for High Performance Networking on Myrinet. In Proc. PC-NOW'98, April 1998.

....mapcoloring problem, using serial C, serial Java, and multithreaded Java. These programs have rst been run on a eight node cluster of 200 MHz Pentium Pro processors, running Linux 2. 2, interconnected by a Myrinet network and using MPI implemented on top of the BIP communication interface [21]. We have also executed the programs without any modi cation on a four node cluster of 450 MHz Pentium II processors running Linux 2.2, interconnected by a SCI network using the SISCI communication interface. The serial C program is compiled using the GNU C compiler, version 2.7.2.3 with O6 ....

L. Prylli, B. Tourancheau, BIP: A new protocol designed for high performance networking on Myrinet, in: Proc. of First Workshop on Personal Computer Based Networks Of Workstations (PC-NOW '98), Vol. 1388 of Lect. Notes in Comp. Science, Springer-Verlag, 1998, pp. 472485.

....descriptor, therefore enabling receiving process to access to data. The use of these memory mapped registers allows the transfer to proceed without the use of interrupts, that represent a major penalty for modern super scalar CPUs [2] Although the concepts behind the VIA standard are not new [3 8], its main advantage is that it has been carefully designed to be operating system and processing architecture independent, therefore sharing a significant goal with the MPI standard [9] and allowing programmers to write fully portable applications. In order to design the communication suitable ....

L. Prylli and B. Tourancheau, "BIP: A New Protocol Designed for High Performance Networking on Myrinet," Lecture Notes in Computer Science, vol. 1388, pp. 472--485, Mar. 1998.

No context found.

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on myrinet. In Workshop PC-NOW, IPPS/SPDP98, Orlando, USA, 1998.

No context found.

L. Prylli and B. Tourancheau. BIP: A new Protocol Designed for High Performance Networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472-485, Springer-Verlag, April 1998.

....such as VIA [4] or SISCI SCI [7] Such interfaces provide several data transfer methods. For instance, regular Processor IO (PIO) and Direct Memory Access (DMA) are available for Dolphin SCI NICs. Moreover, it should be able to easily take into account interface implementations like BIP Myrinet [13] which make a difference between short buffers and long buffers. As a consequence, Madeleine II features specific modules to encapsulate each of these sub interfaces. These modules are called Transmission Modules (TM) Table 2 shows the common interface of the TMs (note that some functions may ....

....in the current version, because of the poor performance of the SCI DMA: we have not been able to get more than 35 MB s with Dolphin SCI D310 NICs. 5.2. 2 BIP Myrinet BIP (Basic Interface for Parallelism) is a low level communication interface specifically designed for the Myrinet network protocol [13]. The main advantage of BIP is to provide communication control in user space: the application may directly interact with the network interface card. The BIP interface makes a distinction between short messages ( 1 kB) and long messages. Short messages are temporarily stored into internal buffers ....

[Article contains additional citation context not shown here]

L. Prylli and B. Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472--485. Springer-Verlag, Apr. 1998.

No context found.

L. Prylli and B. Tourancheau, BIP: a new protocol designed for high performance networking on myrinet, Proceedings of IPPS/SPDP98, 1998.

No context found.

L. Prylli and B. Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. In Proceedings of the International Parallel Processing Symposium Workshop on Personal Computer Based Networks of Workstations, 1998. http://lhpca.univ-lyon1.fr/.

No context found.

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In Workshop PC-NOW, IPPS/SPDP98, Orlando, USA, 1998.

No context found.

L. Prylli and B. Tourancheau, BIP: A New Protocol Designed for High Performance Networking on Myrinet, PC-NOW'98, Vol. 1388 of Lect. Notes in Comp. Science, April, 1998, 472--485.

No context found.

Prylli, L., Tourancheau, B.: BIP: a new protocol designed for high performance networking on Myrinet. In: Proc. 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98). Volume 1388 of Lect. Notes in Comp. Science., Springer-Verlag (1998) 472--485

No context found.

Prylli, L., Tourancheau, B.: BIP: a new protocol designed for high performance networking on Myrinet. In: Proc. 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98). Volume 1388 of Lect. Notes in Comp. Science., Springer-Verlag (1998) 472485

No context found.

L. Prylli and B. Tourancheau. BIP: a new protocol designed for high performance networking on myrinet. In the PC-NOW workshop, IPPS/SPDP 1998, Orlando, USA, 1998.

No context found.

Prylli, L., Tourancheau, B.: BIP: A new protocol designed for High-Performance networking on Myrinet. In: 1st Workshop on Personal Computer based Networks Of Workstations (PCNOW '98). Volume 1388 of Lecture Notes in Computer Science., Orlando, USA, SpringerVerlag (1998) 472--485

No context found.

Prylli, L., Tourancheau, B.: BIP: a new protocol designed for high performance networking on Myrinet. In: Proc. 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98). Volume 1388 of Lect. Notes in Comp. Science., Springer-Verlag (1998) 472--485

No context found.

L. Prylli and B. Tourancheau, "BIP: a new protocol designed for high-performance networking on Myrinet," Proceedings IPPS/SPDP '98,Springer-Verlag, Orlando, FL, USA, 1998.

No context found.

L. Prylli and B. Tourancheau. BIP: A New Protocol Designed for High Performance Networking on Myrinet. In Proceedings of the International Parallel Processing Symposium Workshop on Personal Computer Based Networks of Workstations, 1998. http://lhpca.univ-lyon1.fr/.

No context found.

Loic Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In Proceedings of the PC-NOW '98 International Workshop on Personal Computer based Networks Of Workstations, Orlando, Florida, April 1998. Available from http://www-bip.univ-lyon1.fr/ PUBLICATIONS/pub/PT97pcnow.ps.gz.

No context found.

Loc Prylli and Bernard Tourancheau. BIP: a new protocol designed for high performance networking on Myrinet. In 1st Workshop on Personal Computer based Networks Of Workstations (PC-NOW '98), volume 1388 of Lect. Notes in Comp. Science, pages 472-485. Held in conjunction with IPPS/SPDP 1998.

No context found.

L. Prylli and B. Tourancheau. "BIP: a new protocol designed for high performance networking on myrinet". In the PC-NOW workshop, IPPS/SPDP 1998, Orlando, USA, 1998.

No context found.

Loc Prylli and Bernard Tourancheau. BIP: a New Protocol Designed for High Performance Networking on Myrinet. In Parallel and Distributed Processing, IPPS/SPDP'98, volume 1388 of Lecture Notes in Computer Science, pages 472-485. Springer-Verlag, April

First 50 documents  Next 50

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC