### TABLE III NUMERICAL EXPERIMENTS FOR MULTISERVICE RESOURCE-SHARING SYSTEMS: 75 61525967 615048

### Table 3 :Results for resource-sharing example

"... In PAGE 4: ... Therefore, the number of matches is 3 times the number of blocks. Table3 presents the reduced graph size and CPU time spent by QUEST in the detection of all the resource sharing opportunities in the designs consisting of (8-64) replications of the basic block which consists of (72-576) architectural components. 4.... ..."

### Table 2: Experiment Parameter Settings advantages of resource sharing are most evident for resource-limited situations (i.e., small parallel systems). Nevertheless, for su ciently large values of f, our algorithm outperformed its one-dimensional adversary in the entire range of system and query sizes.

1996

Cited by 33

### Table 8 Average IPC, Symbiosis factor, and Standard Deviation in IPC across frames for all pairs of multimedia applications with the dynamic resource sharing policy.

2002

"... In PAGE 12: ... Tullsen et al.[23] quantify the notion of goodness of a co-schedule consisting of n tasks using a mea- sure called symbiosis factor defined as follows: D7DDD1CQCXD3D7CXD7 CUCPCRD8D3D6 BP C8D2 CXBPBD(realized IPC of CYD3CQCX/single-threaded IPC of CYD3CQCX) Table8 gives the symbiosis factors for all 2-task co-schedules for the set of multimedia applications and base system we consider in this study. The higher the symbiosis factor for a set of threads, the more efficiently an SMT processor is being exploited.... In PAGE 15: ...BE The intuition for this is similar to that for our previous observation on IPCs on single threaded processors [10]. Table8 gives the standard deviation in IPC for each multimedia application we consider in this study across its different frames for all possible 2-task co-schedules, with the dynamic resource sharing policy. Therefore, for each frame type of each application, we profile one frame of that type with all possible co-schedules to obtain IPC estimates for all tasks for all the co-schedules.... In PAGE 30: ...2, and compare these results to those from the evaluations of synthetic workloads. Table8 gives the average IPC and standard deviation in IPC across frames for each multimedia applica- tion we consider in this study, for all possible 2-task co-schedules, with the dynamic resource sharing policy. It also gives the average total IPC and the symbiosis factor for each co-schedule.... ..."

Cited by 20

### Table 5 Resource Sharing Model

"... In PAGE 16: ...Table 5 Resource Sharing Model The results presented in Table5 illustrate the substantial gain in CPU time as the number of automata is reduced, and this with relatively little impact on memory re- quirements. Furthermore, this is seen to be true even when the state space within the grouped automata are not reduced.... ..."

### Table 5: Resource Sharing Model

"... In PAGE 20: ... This reduction is not possible in models with N ? 1 resources. The results presented in Table5 illustrate the substantial gain in CPU time as the number of automata is reduced, and this with relatively little impact on memory requirements. Furthermore, this is seen to be true even when the state space within the grouped automata are not reduced.... ..."

Cited by 9

### Table 1: Resource Sharing Model

"... In PAGE 32: ... Furthermore, 50% of the transitions are functional, which is quite high. Table1 presents information concerning the size of the state space (both the complete product state space and the reachable state space, n); the number of nonzero elements, nz, in the sparse matrix representation; and various timing results under a variety of model parameter values. It may be observed that the reachable state space varies according to P (the number of resources that may simultaneously access the resource), from almost nothing to one less than the size of the product state space.... In PAGE 33: ...76 of the time required by the SUN. The numbers in bold type in Table1 are 0.76 times the number of seconds obtained on the SUN.... ..."

### Table 4. Resource Sharing Method Matrix

2004

"... In PAGE 17: ...acility sharing (FS), and even setup (SS) as a result of the primary sharing method (e.g., tooling sharing (TS)). To clarify the cause-and-effect relationships in the sharing methods, a resource sharing method matrix is utilized (see Table4... ..."

### Table 2: Effect of Resource Sharing on Partitioning

1990

"... In PAGE 3: ... CPU run times are reported in seconds DecStation 3100 with 16 MBytes of memory. Table2 shows the effect of resource sharing using an example of fifth order digital wave filter[14 scription was translated into HardwareC from an ISP description consists of 26 add operations and 8 multiply In addition, the design also contains 15 I/O operations. description used in Table 2 the multiply by a constant filter description is replaced by combinational shift synthesis phase by Hercules.... In PAGE 3: ... Table 2 shows the effect of resource sharing using an example of fifth order digital wave filter[14 scription was translated into HardwareC from an ISP description consists of 26 add operations and 8 multiply In addition, the design also contains 15 I/O operations. description used in Table2 the multiply by a constant filter description is replaced by combinational shift synthesis phase by Hercules. As shown in the Table elliptic filter without any resource sharing is 6458.... In PAGE 3: ... On partitioning the overall size increases increases to 17 cycles. Table2 also compares running for four different heuristics: simulated annealing (SA), cost function consisting of only area and pin-out costs... In PAGE 4: ... Table 3 illustrates the effect of latency, area and pin-out constraints on partitioning results for the elliptic filter containing 26 add and 8 multiply resources. In contrast to the filter considered in Table2 , the multiply oper- ations are now not restricted to be by 2 only. The multiply operations are instead modeled by calls to hardware blocks requiring two cycles per oper- ation.... In PAGE 4: ... The multiply operations are instead modeled by calls to hardware blocks requiring two cycles per oper- ation. Therefore, total size of the unpartitioned filter is 12542 considerably bigger than the filter description used in Table2 . The latency of the unpar- titioned filter in this case is 20 cycles.... ..."

Cited by 1

### Table 1: The objective functions for the Resource Sharing game.

"... In PAGE 7: ... Work cannot begin until all players agree on a way to apportion the auxiliary power. Table1 displays these quantities for a situation... ..."