D.G. Carta TwoFast Implementations of the "Minimal Standard" Random Number Generator. CACM, 33([ January 1990.  Home/Search   Document Not in Database   Summary   ACM   TOC   Related Articles   Check

....performance counter register is writable, allowing any interrupt period up to the maximum of 64K events to be chosen. To minimize any systematic correlation between the timing of the interrupts and the code being run, we randomize the length of the sampling period by writing a pseudorandom value [Carta 1990] into the performance counter at the end of each interrupt. The default sampling period is distributed uniformly between 60K and 64K when monitoring CYCLES. 4.1.2 Attributing Events to PCs. To accurately interpret samples, it is important to understand the PC delivered to the interrupt handler. ....

CARTA, D. 1990. Two fast implementations of the "minimal standard" random number generator. Commun. ACM 33, 1 (Jan.), 87--88.

....weight functions. The first competing weight function we consider is based on the Unix system functions random and srandom in place of rand and srand, resulting in a weight function we denote W random 5 . The second function we consider uses the Minimal Standard random number generator [22, 23], resulting in the weight function: W minstd (k; S i ) 16807( 16807 Delta S i ) XOR D(k) mod (2 31 Gamma 1) Our third alternative is to modify the W rand function as follows: W rand2 (k; S i ) 1103515245( 1103515245 Delta D(k) 12345) XOR S i ) 12345) mod 2 31 ) 18) It ....

David G. Carta. Two fast implementations of the "minimal standard" random number generator. CACM, 33(1), January 1990.

....additional information for understanding the causes of stalls; see the discussion in Section 4.1.2. 6 be chosen. To minimize any systematic correlation between the timing of the interrupts and the code being run, we randomize the length of the sampling period by writing a pseudo random value [4] into the performance counter at the end of each interrupt. The default sampling period is distributed uniformly between 60K and 64K when monitoring CYCLES. 4.1.2 Attributing Events to PCs To accurately interpret samples, it is important to understand the PC delivered to the interrupt handler. ....

D. Carta. Two fast implementations of the `minimal standard' random number generator. Communications of the Association for Computing Machinery, 33(1):87--88, January 1990.

....By monitoring the different values in the avg latency array for different values of i, we can make sure that the simulation has reached a steady state. The random number function we used was Proteus fast random( which is an implementation of the ACM Minimal Standard Random Number Generator [17, 8]. We now show how our combinatorial model ties together the choice of diffracting tree parameters depth, d, and prism locations per level, L, to the number of processor, P . A diffracting tree is shown to operate optimally when P = O(dL) and L = 2 d (the number of counters) i.e. the number of ....

D.G. Carta Two Fast Implementations of the "Minimal Standard" Random Number Generator. CACM, 33(1), January 1990.

....counter register is writable, allowing any interinterrupt period up to the maximum of 64K events to be chosen. To minimize any systematic correlation between the timing of the interrupts and the code being run, we randomize the length of the sampling period by writing a pseudo random value [4] into the performance counter at the end of each interrupt. The default sampling period is distributed uniformly between 60K and 64K when monitoring CYCLES. 4.1.2 Attributing Events to PCs To accurately interpret samples, it is important to understand the PC delivered to the interrupt handler. ....

D. Carta. Two fast implementations of the `minimal standard' random number generator. CACM, 33(1):87--88, January 1990.

....list can be used to implement constant time list operations. pseudo random integer. Numerous techniques exist for generating random numbers. For example, the Park Miller generator efficiently produces high quality random numbers that are uniformly distributed between 0 and 2 31 Gamma 1 [PM88, Car90] The random number produced by fast random( is then scaled 1 to reside in the interval [0, global tickets Gamma1] which will be referred to as the ticket space. The scaled random number, winner, represents the offset of the winning ticket in the ticket space. The ticket space is then scanned ....

....and nonuniform quanta. Random Numbers An efficient lottery scheduler requires a fast way to generate uniformly distributed random numbers. The Park Miller pseudo random number generator produces high quality random numbers that are uniformly distributed between 0 and 2 31 Gamma 1 [PM88, Car90] I developed a fast, low level implementation of this generator that executes in approximately 10 RISC instructions; Figure 5 1 lists MIPS assembly language [Kan89] code for fast random( Lotteries The prototype scheduler uses a straightforward lottery implementation, similar to the one listed ....

David G. Carta. Two fast implementations of the `minimal standard' random number generator. Communications of the ACM, 33(1):87--88, January 1990.

....overhead. The statclock abstraction is available only on machines with high precision, programmable clocks. The randomized sampling intervals are generated by programming the clock s limit register with a pseudo random number. To reduce overhead, a cheap but good random number generator is used [1]. On systems without programmable clocks, statclock is called directly from hardclock, and the functionality is unchanged from the existing system. 4 Code Profiling Kernel support for user level code profiling [4] is carried out in a manner identical to CPU usage estimation. We therefore wanted ....

Carta, D. G. Two fast implementations of the "minimal standard " random number generator. Communications of the ACM 33, 1 (Jan. 1990).

.... 73, 59, 107, 108, 109, 5, 13, 62, 60, 49, 110, 50, 40] I : For implementations in commercial software and source code see [91, 43, 42, 29, 96, 9] Source code of MINSTD and MINSTD with a shuffling algorithm added are given in [93] For fast implementations using multiplier 48271 and 69621 see [11, 93]. Tests for these multipliers are given in [42] L: 57, 54, 67, 94, 42] 1.3 LCG(2 31 ; 2 16 3 = 65539; 0; 1) RANDU 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 0 0.2 0.4 0.6 0.8 1 Quote[91] Many multiplicative linear congruential generators are descendents of the infamous ....

D.G. Carta. Two fast implementations of the "minimal standard" random number generator. Comm. ACM, 33 :87--88, 1990.

No context found.

D.G. Carta TwoFast Implementations of the "Minimal Standard" Random Number Generator. CACM, 33([ January 1990.

No context found.

D.G. Carta Two Fast Implementations of the "Minimal Standard" Random Number Generator. CACM, 33(1), January 1990.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC