Dolphin ToolWorks. TotalView. Dolphin Interconnect Solutions, Inc., Framingham, Massachusetts, USA.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to the lower abstraction levels too. For example, if a graphical language was used to develop the code, then the basic debugging activities should be done at the same graphical level, but the user should have access and be allowed to debug the application s generated source code or even its assembly code [10]. 3. Integration. Multiple user interfaces (for the different abstraction levels) should be supported, either exclusively or simultaneously. If support is provided for multiple (simultaneous) debugging interfaces over the same target application, the debugging environment must provide support for ....

....[14] that only require a minimal information on the event ordering, reducing drastically the volume of the trace file(s) when comparing with the datadriven or checkpointing techniques. A discussion of further optimizations to this technique, so that even smaller event logs can be obtained is reported in [20], 16] These techniques have been applied both to shared and distributedmemory (message passing and RPC based) models. Even some well known existing parallel debuggers, like p2d2 [8] and TotalView [22] lack such a record replay facility. In [19] a replay mechanism for the PVM system is ....

[Article contains additional citation context not shown here]

Dolphin ToolWorks. TotalView. Dolphin Interconnect Solutions, Inc., Framingham, Massachusetts, USA.

....reached consumer level commodity status, it is much more a ordable than the custom networking hardware traditionally used in commercial supercomputers. Hardware of this category includes (but is not limited to) Myrinet hardware from Myricom, Inc. 3] SCI hardware from Dolphin Interconnect LLC [1], and Giganet hardware from Emulex Corporation [10] 4 There are also more choices in network software and protocols. TCP IP has long been the standard for internet communications and was easily adopted for Linux cluster use. However, TCP IP has disadvantages in some environments. Much of the ....

The Dolphin SCI interconnect white paper. Technical report, Dolphin Interconnect Solutions, Inc., February 1996.

....PCs. The typical architecture of an SCI connected PC based system is shown in fig. 1. To the already existing memory hierarchy levels of the local PC only one level is added: Remote memory, which is never cached. Nevertheless, there are different memory semantics offered by the Dolphin SCI card [2], currently the most powerful card. Remote writes can be transferred immediately to the destination node or buffered in the card Node A Node B local mapping physical memory window 32 bit 32 bit 64 Bit address space Process P2 virtual memory Process P1 virtual memory remote ....

Dolphin: The Dolphin SCI Interconnect. White Paper, Dolphin Interconnect Solutions, Olaf Helsets Vei 6, 0621 Oslo, Norway, (www.dolphinics.com).

....servers, specialized processors or data acquisition systems with the Scalable Coherent Interface as interconnect technology. High performance distributed and parallel cluster computing can be achieved with SCI hardware that is available now or in the near future. Dolphin s SBus SCI adapters [4] [3] or SCI switches can be used to set up a cluster of workstations. Additionally, a PCI SCI adapter has been announced by Dolphin. At CERN, a PCI SCI bridge has been developed which is targeted for constructing a demonstrator data acquisition system for the Large Hadron Collider experiments ....

Dolphin Interconnects Solutions A.S. 1 Gb/s SBusSCI Cluster Adapter Card. White Paper, 1995.

....if specific SCI protocol features, e.g. guaranteed data delivery, are exploited judiciously. As an example, Dolphin Interconnect Solutions, a Norwegian company offering SCI products, has demonstrated application level communications latency below 10 s in an SCI based cluster of SPARCstation20s [8]. Flexibility in terms of use or configuration. From a software point of view, there is another drawback to NOWs: poor programmability. This is due to the fact that parallel programming in a NOW almost exclusively uses message passing as the underlying communication and synchronization model, ....

....(see below) an SCI ring of two SPARCstation2s (SS2) has been set up using 25 MHz Dolphin SBus SCI adapter cards [7] The cards and cables are advertized to sustain application throughput of 10 Mbytes s 1 . With faster machines (SS20) round trip, ping pong latencies of 8 s have been reported [8]. The device driver for the SBus SCI card supports two elementary software interfaces. The raw channel interface provides a set of ioctl( calls for establishing and controlling connections between processes on different machines. Data can then be transferred across channels using write( and ....

Dolphin Interconnect Solutions: Dolphin Breaks Cluster Latency Barrier with SCI Adapter Card. Dolphin Interconnect Solutions, Press Release (1995). http://www.dolphinics.no.

....are important components to improve locality of data accesses and, thus, to increase efficiency of a DSM machine. 4 3. 2 Experimental Platform For initial software development (see below) an SCI ring of two SPARCstation2s (SS2) has been set up using 25 MHz Dolphin SBus SCI adapter cards [7]. The cards and cables are advertized to sustain application throughput of 10 Mbytes s 1 . With faster machines (SS20) round trip, ping pong latencies of 8 s have been reported [8] The device driver for the SBus SCI card supports two elementary software interfaces. The raw channel interface ....

Dolphin Interconnect Solutions: 1 Gb/s SBus-SCI Cluster Adapter Card. Dolphin Interconnect Solutions, White Paper (1994).

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