### Table 2: Technology Mapping results

"... In PAGE 8: ... The results show that the Boolean approach reduces the number of matching algorithm calls, nd smaller area circuits in better CPU time, and reduces the initial network graph because generic 2-input base function are used. Table2 presents a comparison between SIS and Land for the library 44-2.genlib, which is distributed with the SIS package.... ..."

### Table 1: A Template for Incremental AC Algorithms. 1: Input: Randomized parameterized policy ( j ; ), Value function feature vector f(s). 2: Initialization: Policy parameters = 0, Value function weight vector v = v0, Step sizes = 0; = 0; = c 0.

"... In PAGE 5: ... 4 Actor-Critic Algorithms We present four new AC algorithms in this section. These algorithms are in the general form shown in Table1 . They update the policy parameters along the direction of the average reward gradient.... ..."

### Table 1: A Template for Incremental AC Algorithms. 1: Input: Randomized parameterized policy ( j ; ), Value function feature vector f(s). 2: Initialization: Policy parameters = 0, Value function weight vector v = v0, Step sizes = 0; = 0; = c 0.

"... In PAGE 5: ... 4 Actor-Critic Algorithms We present four new AC algorithms in this section. These algorithms are in the general form shown in Table1 . They update the policy parameters along the direction of the average reward gradient.... ..."

### Table 3 : Execution times 9 Conclusion Several new variants of scalar product algorithms have been implemented which in general are faster than those known before. Their accuracy is very high and guaran- teed, so each might be used for veri ed accurate computation. But only the algorithm KDPACC which uses the large table for preprocessing and then adds the values into the long accumulator provides optimal accuracy for all relevant roundings. It also facilitates accumulated calculation of dotproducts and supports the data type dot precision. Its runtime for large vectors is about a factor of 3 faster than the long accumulator algorithm and nearly as fast as the less accurate variants. We therefore suggest to use this algorithm for vectors with many components (more than 50) and prefer the long accumulator algorithm for fewer components.

### Table 2. DESCRIPTION THE GENERAL CASE OF THE NEW IMPLEMENTATION:

"... In PAGE 13: ... After the initialization at the beginning of the body of the first main loop for t = 1, we have unit = 10, current_mask = 3, init = 20, prev_diag = prev_diag_shift = 0, curr_diag = 1. Table2 contains a trace of some of the important variables of the algorithm. The variables e, p1, and p2 are displayed as masked by m1 or m2 to show the meaningful bits of each piece.... ..."

### Table 4 lists the speed of the RISC optimized code for a variety of problem sizes when run on the SGI Origin 2000 (R12000), the SGI Origin 3000 (R12000), the SUN HPC 10000, and the HP Superdome (HP PA 8600). The SGI Origin 2000 was equipped with 128, 300-MHz R12000 processors with 8-MB secondary caches and 2 GB of memory per 2-processor node. The SGI Origin 3000 was equipped with 256, 400-MHz R12000 processors with 8-MB secondary caches and 4 GB of memory per 4-processor node. The SUN HPC 10000 was equipped with 64, 400-MHz Ultra SPARC II processors with either 4 or 8 MB of secondary caches (one of our systems was upgraded before the series of runs was finished) and 4 GB of memory per 4-processor node. There was insufficient memory to run the 206-million grid point test case on the SUN HPC 10000. The HP Superdome was configured with 48, 552-MHz HP PA 8600 processors with 1 MB of on-chip data cache and 1 GB of memory per processor. For some unknown reasons, we could not run several of the cases on this system, even when there was more than enough memory to run the job.

2002

"... In PAGE 8: ...rocessor runs for the 3-million grid point test case. ........................................... 6 Table4 . The performance of the RISC optimized version of F3D for single processor runs on the SGI Origin 2000, the SGI Origin 3000, the SUN HPC 10000, and the HP Superdome for a range of test cases.... In PAGE 16: ... Table4 . The performance of the RISC optimized version of F3D for single processor runs on the SGI Origin 2000, the SGI Origin 3000, the SUN HPC 10000, and the HP Superdome for a range of test cases.... ..."

### Table 1: Number of iterations required to nd an optimal solution using the previous solution as a starting point and using the vector (0; : : : ; 0) as a starting point.

2001

"... In PAGE 24: ... When a new user enters the system, running the MFVA algorithm using as a starting point the optimal solution for the problem prior to the new user apos;s arrival typically results in substantial computational savings. Table1 shows results from a power control problem involving a system of ten by ten cells and approximately nine hundred mobile users. The number of iterations required to nd the optimal solution for an initial problem is given, along with the number of iterations required to nd the optimal solution when additional users enter the system.... ..."

Cited by 10

### Table 5. Results obtained for the two-loop network in previous research

1998

Cited by 4

### Table 1. Comparison of results for various approaches.

"... In PAGE 8: ... 4. Numerical Results Table1 compares the balance and uniformity (t,s) of (n,2) de Bruijn sequences... In PAGE 9: ... In the case of Algorithm II, the characteristics of the sequences obtained by the optimal mappings with respect to both balance and uniformity criteria are shown. ------------------------- Table1 goes here ------------------------- In Table 1, we observe that: 1. Although Algorithm I generates sequences with optimal uniformity (minimum s), the corresponding balance criterion t is rather large.... In PAGE 9: ... In the case of Algorithm II, the characteristics of the sequences obtained by the optimal mappings with respect to both balance and uniformity criteria are shown. -------------------------Table 1 goes here ------------------------- In Table1 , we observe that: 1. Although Algorithm I generates sequences with optimal uniformity (minimum s), the corresponding balance criterion t is rather large.... ..."