Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 11 - 20
of
20,299
Table 1: Network timing parameters for a one-way message without contention on several current com- mercial and research multiprocessors. The final two rows refer to the active message layer, which uses the commercial hardware, but reduces the interface overhead.
1993
"... In PAGE 16: ...message across H hops is given by the following. Tn28M;Hn29 = T snd + d M w e + Hr+ T rcv Table1 indicates that, for current machines, message communication time through a lightly loaded network is dominated by the send and receive overheads, and thus is relatively insensitive to network structure. Furthermore, networks with a larger diameter typically have wider links (larger w), smaller routing delays (r), and a faster cycle time because all the physical wires are short.... In PAGE 16: ... The final two rows refer to the active message layer, which uses the commercial hardware, but reduces the interface overhead. The send and receive overheads in Table1 warrant some explanation. The very large overheads for the commercial message passing machines (nCUBE/2 and CM-5) reflect the standard communication layer from the vendor.... ..."
Cited by 465
Table 1: Network timing parameters for a one-way message without contention on several current com- mercial and research multiprocessors. The final two rows refer to the active message layer, which uses the commercial hardware, but reduces the interface overhead.
1993
"... In PAGE 16: ...bit message across a5 hops is given by the following. a13 a2 a2 a3 a5 a4 a0 a13 a0 a0 a2a1 a0 a0 a4a3 a5 a3 a0 a5 a0 a0 a13 a7a6 a5 a9a8 Table1 indicates that, for current machines, message communication time through a lightly loaded network is dominated by the send and receive overheads, and thus is relatively insensitive to network structure. Furthermore, networks with a larger diameter typically have wider links (larger a3 ), smaller routing delays (a0 ), and a faster cycle time because all the physical wires are short.... In PAGE 16: ... The final two rows refer to the active message layer, which uses the commercial hardware, but reduces the interface overhead. The send and receive overheads in Table1 warrant some explanation. The very large overheads for the commercial message passing machines (nCUBE/2 and CM-5) reflect the standard communication layer from the vendor.... ..."
Cited by 465
Table 6: The execution time (in seconds) of the original and current simulation model with the global event list algorithm. The last row shows the execution times when the current simulation model passes pointers in messages instead of using message buffers.
1999
"... In PAGE 7: ... To measure the performance using the global event list protocol, the system was run with 60 and 120 servers using ring and tree topologies with duration of 576 hours. Table6 shows the execution times of the original and current simulation. The original system generally has a lower execution time, primarily because of the extra buffer copying involved in the current system.... In PAGE 7: ... To measure the effects of buffer copying we measured the execution times with the current simulation when messages use pointers. These are shown in the third row of Table6 , and are lower than the original times. The improvement is attributable to communication topology changes, as lookahead specification does not affect the global event list algorithm.... ..."
Cited by 2
Table 3: Frequency of various message types and sizes (represented by the number of data values transmitted) in the current implementation of TAM. On average, a message is sent and received every 8 TAM instructions. These statistics are sensitive to compiler optimizations and, in some sense, represent a worst case scenario.
Table 3: Frequency of various message types and sizes (represented by the number of data values transmitted) in the current implementationof TAM. On average, a message is sent and received every 8 TAM instructions. These statistics are sensitive to compiler optimizations and, in some sense, represent a worst case scenario.
Table 3: Frequency of various message types and sizes (represented by the number of data values transmitted) in the current implementation of TAM. On average, a message is sent and received every 8 TAM instructions. These statistics are sensitive to compiler optimizations and, in some sense, represent a worst case scenario.
1992
Table 3: Frequency of various message types and sizes (represented by the number of data values transmitted) in the current implementationof TAM. On average, a message is sent and received every 8 TAM instructions. These statistics are sensitive to compiler optimizations and, in some sense, represent a worst case scenario.
1992
Table 1: Control messages.
1998
"... In PAGE 6: ... They request e#0Bects processing from the fx server, con#0Cgure parameters at the broadcaster, notify the broadcaster to switch to another video source, and control video playback at the video #0Cle server. Table1 lists the control messages used in the current system. Although these messages could be unicast to the appropriate destinations, webe- lievemulticast will be more e#0Ecient when the sys- tem is integrated with other MBone tools.... ..."
Cited by 8
Table 1: Mobile Phone Requirements Incoming and edited text messages are broadcast to the display that shows the current message to the user of the mobile terminal. Based on the displayed message the user can take an appropriate action and send or modify the message by interacting with the keyboard. In a real mobile phone a lot of other features are controlled and monitored using the user interface software, but these features are beyond the scope of this example. However many difficult problems arise from the fact that different features use shared resources such as the display of the mobile phone. For more information consult the work by Lorentsen, Tuovinen, and Xu (2001). Figure 6 describes a partial context model. The domains are drawn from the problem domain model viewpoint (Figure 5).
"... In PAGE 9: ... Figure 5 shows a simplified problem domain model for the SMS messaging. In order to save space and reduce the complexity of the diagrams, we do not show the definition of the shared phenomena between the domains and Table1 shows only some of the requirements. We use the notation developed for frame diagrams by Jackson (2001).... ..."
Table 1 shows the di erent versions of the DALIB that are currently available. Please use the corre- sponding sub-directory when generating the corresponding DALIB. Directory Message Passing System
"... In PAGE 10: ... Table1 : Supported message passing systems. The DALIB version for the Paragon XP/S works also for the Intel iPSC/860.... ..."
Results 11 - 20
of
20,299
