### Table 6. Prediction Accuracy on Exact DDG

"... In PAGE 8: ... We compute an approximate schedule to predict register pressure from the exact DDG of Rocket before software pipelining. Table6 details the performance comparison. The columns labeled Rel Err report the relative error of the predicted register pressure from the pseudo-schedule compared to the actual register pressure from the actual schedule.... ..."

### Table 2: Attainable Recovery Times for Planned Outages

2006

"... In PAGE 5: ... MINIMIZING PLANNED AND UNPLANNED DOWNTIME WITH MAA Table 1 shows the recovery times for unplanned outages that you can attain when using the MAA best practices. Table 1: Attainable Recovery Times for Unplanned Outages Outage Type MAA Enabling Technologies Computer failure No downtime Real Application Clusters (RAC) Storage failure No downtime Automatic Storage Management (ASM) Human error lt; 30 minutes Flashback Technologies Data corruption No downtime lt; 30 seconds Hardware Assisted Resilient Storage (HARD) if applicable Data Guard Site failure lt; 30 seconds Data Guard Table2 shows the attainable recovery times for all types of planned downtime for MAA. Oracle Database 10g Release 2: Roadmap to Maximum Availability Architecture ... ..."

### Table 1: Timing Characteristics of the Real-Time Workload and its Recovery

15

"... In PAGE 1: ... S1(12) = 31, S2(12) = 31, S3(12) = 17. It is possible to observe in Table 4 that if Recovery Blocks are used to handle faults, due to the fact that the computation time of the secondary block of task 2 is equal to Cs 2 = 11 (see Table1 ), it can recover under any level of responsiveness.CL GL GE FA 31 31 17 17 Table 4: RTAB Table for Example 3 4 Conclusion A scheme was presented to provide scheduling guar- antees for a variety of fault tolerant techniques.... In PAGE 2: ... The breakdown utilization UBD of a task set with uti- lization UW = Pn i=1 Ci=Ti, including fault-induced re- quirements is given by UBD = UW maxf1 i ngminft2SigfWF i (t)=tg (6) If UW UBD then the task set is schedulable, otherwise the task set is unschedulable. Table 2 shows the fault tolerant recovery bound Uf, and the breakdown utilization UBD, for the task set used in Table1 . It can be observed that under all the recovery schemes UW gt; Uf.... In PAGE 2: ... The largest recovery time with priority i which arrives dt=TF i e in the interval [0; t] may be calculated by, LRi = maxft S i g Wi(t)=dt=TF i e (9) It follows that the largest recovery time which may be added to a set of n tasks without disturbing their schedulability is given by, LTi = minfi k ngmaxft S kg Wk(t)=dt=TF i e (10) Example 1. Considering the fault-free workload pre- sented in Table1 , the analysis for calculating the largest recovery time is given in Table 3. Table 3 shows that there is enough spare time in the schedule to satisfy the timing requirements of the fault tolerant recovery workload, except for the checkpointing model.... In PAGE 3: ... As stated in equation (2), we can compute the schedu- lability test for the fault tolerant real-time task set as, maxf1 i ngminft2SigfWF i (t)=tg 1 (4) Equation (4) gives us an schedulability condition for fault tolerant real-time task sets. Table1 shows a task set consisting of 3 periodic tasks, and its associated timing characteristics for each fault tolerant mechanism. From this requirements, in Table 2 we show the schedulability analysis for the fault tolerant models previously discussed.... ..."

### Table 1: Timing Characteristics of the Real-Time Workload and its Recovery

"... In PAGE 4: ...Table 1: Timing Characteristics of the Real-Time Workload and its Recovery Example of Slack Calculations For the real-time workload given in Table1 , a given set independent fault arrivals and their associated slack values is shown in Table 2. These values of fault arrivals have been chosen for illustration purposes only.... In PAGE 5: ... We introduce this computation via an example, as below. Example of Levels of Responsiveness Consider the workload described in Table1 . Assume that the following sequence of faults occur: (1) the rst fault occurs at time tF = 5, (2) the second at time tF = 22, and (3) the third at time tF = 52.... In PAGE 6: ... The criticality set M = fm( 1); m( 2); :::; m( n)g de nes the relative critical- ity of tasks. Let us consider the set of tasks de ned in Table1 , 1; 2; 3, with criticality values given by M = f2; 1; 3g, which are listed in decreasing order. Thus, 2 is the most critical task in the system.... ..."

### Table 1: Timing Characteristics of the Real-Time Workload and its Recovery

"... In PAGE 4: ...Table 1: Timing Characteristics of the Real-Time Workload and its Recovery Example of Slack Calculations For the real-time workload given in Table1 , a given set independent fault arrivals and their associated slack values is shown in Table 2. These values of fault arrivals have been chosen for illustration purposes only.... In PAGE 5: ... We introduce this computation via an example, as below. Example of Levels of Responsiveness Consider the workload described in Table1 . Assume that the following sequence of faults occur: (1) the rst fault occurs at time tF = 5, (2) the second at time tF = 22, and (3) the third at time tF = 52.... In PAGE 6: ... The criticality set M = fm( 1); m( 2); :::; m( n)g de nes the relative critical- ity of tasks. Let us consider the set of tasks de ned in Table1 , 1; 2; 3, with criticality values given by M = f2; 1; 3g, which are listed in decreasing order. Thus, 2 is the most critical task in the system.... ..."

### TABLE 2. Steps for computing optimal partitions with Rmax D 4

1997

Cited by 2

### Table 2: Copy Nodes in the DDG

"... In PAGE 33: ... When cloning is introduced, the degradation is reduced to around 31%. Similarly the greedy algorithm introduced on average 28 new copy instructions, and cloning reduced them to 23 (see Table2 ). When only induction variables are cloned, the degradation in performance seen with the greedy algorithm is reduced to 39% and requires an average of 26 copy instructions.... ..."

### Table 2. Other Task-Based Recovery Policies

2004

"... In PAGE 7: ... Due to lack of space, we will only give details of the Skip and Compensate recovery policies. For details about other recovery policies see Table2 and [7]. 3.... In PAGE 9: ... 5(c). Table2... ..."

Cited by 1

### TABLE 4.1. Recovery algorithm triggers and tasks.

1997

Cited by 18

### Table 1 summarizes the global SIR performance of ICA algorithms for vari- ous noise levels. Since we deal with SIR, the performance results are analyzed from the mixing matrix recovery point of view; the source denoising task is not

2006

"... In PAGE 7: ... Vrins and M. Verleysen Table1 . 100-trials empirical means and variances of global SIR performances of several ICA algorithms (global SIR is the averaged SIR computed from the individual source SIRs for a given trial); m = msharp(N).... ..."

Cited by 2