### Table 1: Exact matrix inversion over pseudorandom and truly random input

"... In PAGE 3: ... Q-pivoting can be applied to build A?1 from scratch, using d rank-1 updates. This process is used in Table1 to evaluate the e ciency of multiple precision integer arithmetic in this context. Note that other exact arithmetic methods for solv- ing linear systems (e.... ..."

### Table 1: Types of Pseudo-Random Number Generators in PRNGlib irng Generator typ

"... In PAGE 8: ...ower triangle and to 1 on the diagonal (see Section 3.3.3). For the LF and GSR generators, the initialized mantissa l should be close to l, the mantissa of double precision oating point numbers. The generator types are numbered by the parameter irng (see Table1 ). All informa- tion about the actual status and parameter settings is stored in the integer and real work arrays iw(1:liw) and rw(1:lrw) (see Tables 2 and 3).... In PAGE 16: ... Currently PRNGlib provides the interface to the four random number generator types LFA, LFS, GSR, and MLC (cf. Table1 ). For future extension of PRNGlib by a generator typ the six routines INItyp, UNItyp, SKPtyp, PERtyp, CHKtyp, INFtyp must be provided (Package 5).... In PAGE 50: ...A Error messages Table1 1: Returned error values in PRNGlib value description subroutines 0 no error -1 error in distribution RANDIS, : : : 1 wrong number of processors npe PARBLK 2 wrong n or lstart PARBLK 10 open error RANIN, RANOUT 20 integer read/write error RANIN, RANOUT 30 double precision read/write error RANIN, RANOUT 100 idis is out of choices RANDIS 200 irng is out of choices RANCHK, : : : 300 lvec is out of choices RANUNI 1001 liw 10 RANCHK, CHKLF 1002 lrw iw(4) = r 1003 l = iw(2) not multiple of 2. 1004 pointer iw(3) lt; 0 or iw(3) iw(4) = r 1005 s = iw(5) 0 or iw(5) gt; iw(4) = r 1006 r = iw(4) gt; 9689 3001 liw 10 + 2iw(4) RANCHK, CHKGSR 3002 lrw 6 3003 = l iw(2) not multiple of 4.... ..."

### Table 9. Literalcountsfortheproposedtechniquecomparedwith pseudorandom testing.

"... In PAGE 15: ... In addition, such a special-purpose FSM wouldbespecifictoasingleCUT;ontheotherhand, the decoder DC for the proposed scheme is shared among multiple CUTs, thereby reducing overall TGC over- head. Table9 compares the overhead of the proposed de- terministic BIST scheme with the overhead of a pseu- dorandom BIST method [6] for several circuits. The... ..."

### Table 2 Comparison of performance data for reading 32-bit integers from di erent processors without virtual processing and with virtual processing when vpr = 16 on the MasPar MP-1. The number of physical processors used is nproc = 16,384. We tested two implementations of inter-processor memory accessing with virtual processing. The destination addresses were drawn from a system pseudo-random number generator in the range from 0 to nproc con ? 1 without virtual processing, and in the range from 0 to vpr nproc con

"... In PAGE 14: ...) Thus con is the expected degree of concurrency per virtual processor if it is su ciently large. The performance data is shown in Table2 . Table 2 suggested that the per- formance of the system routine for performing concurrent read on a physical processor did not grow linearly in the number of expected concurrent requests.... ..."

### Table 2: Comparison of randomness: Antirandom vs. Pseudorandom

"... In PAGE 2: ... The successive vectors in the time sequence are listed sequentially. Table2 , shows that the antirandom sequence is more random than the the pseudorandom sequences. There are several formal tests for randomness.... ..."

### Table 3: Results for pseudorandom o -line testing

1998

"... In PAGE 7: ...3 groups 4 groups 5 groups 6 groups 7 groups 0:001 0:002 0:003 0:004 0:005 c3540 c880 c2670 k 0:08 0:06 0:04 0:02 0 P(Gk) Figure 7: Dependencies between k and P(Gk) Table3 shows that the average value 3:59% for q1 is reduced to 0:2% for q2. The value qk = 0 will be achieved for at most k = 3 compacted outputs.... ..."

Cited by 5

### Table 1. Pseudo-random testability bench

"... In PAGE 3: ...number of the CUT inputs. The results of a simulation of a selected set of benchmarks are shown in Table1 . The i column shows the number of the benchmark inputs, range indicates the range of the encountered number of the test patterns to fully test the circuit (in those 1000 samples), while the statistic average value is shown in the last column.... ..."

Cited by 2

### Table 1. Pseudo-random testability bench

"... In PAGE 3: ... The number of LFSR bits was set to be equal to the number of CUT inputs. The results of a simulation of a selected set of benchmarks are shown in Table1 . The i column shows the number of the benchmark inputs (including the scan path for sequential circuits), range indicates the range of the encountered number of test patterns to fully test the circuit (in those 1000 samples), while the statistical average value is shown in the last column.... ..."