Richard P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, 1985.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....were discovered late prevent the allocation volumes from being equal. We elected not to correct these problems, which makes the results from this benchmark a little harder to interpret. boyer is a toy term rewriting theorem prover derived from the Boyer benchmark in the Gabriel benchmark suite [14]. The program uses pairs extensively and constructs its data structures functionally. The versions used here, nboyer and sboyer, fix some bugs, are written in portable Scheme, and incorporate a problem scaling parameter [1, 3, 4, 10] nboyer and sboyer differ only in that sboyer uses a local ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. The MIT Press, 1985.

....single user workstation. Average CPU time per sentence in TEXT is 125 seconds. 6 This makes Rubinoffs system only 25 times faster than TEXT. Second, his system runs on a Symbolics 3600 in Zetalisp, while the original TEXT tactical component ran in Franzlisp on a VAX 780. Using Gabriel s benchmarks [4] for Boyer s theorem proving unification based program, which ran at 166.30 seconds in Franzlisp on a Vax 780 and at 14.92 seconds in Symbolics 3600 CommonLisp, we see that switching machines alone yields a 11 fold speed up. This means Rubinoffs system is actually only 2.3 times faster than TEXT. ....

Gabriel, R. P. Performance and Evaluation of Lisp Systems. MIT Press, Cambridge, Mass., 1985.

....preliminary benchmarks using a ABCL R3 compiler based on our framework. The first benchmarks compare the sequential execution speed of the the interpreter and our compiler to illustrate the effectiveness of compiling away the unnecessary interpretation. Sequential benchmark programs (Boyer[30] and n queens problem) are written in ABCL R3 without using parallel constructs, nor reflective operations (although side effects are employed) The programs are executed in three styles: NR) compiled without the meta level and directly executed, INT) executed by a CPS interpreter for ABCL R3, ....

Richard P Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, 1985.

.... programmed using the standard first order recursive definition; ack curried : a curried version of Ackermann function that uses higher order functions (and is almost twice as fast as the standard first order function) tak cps: the Takeuchi function in CPS, part of the Gabriel benchmark suite [15]; reverse: standard first order list reverse function; rev cps: a CPS version of reverse; split : taking a predicate and a list and returning two lists, one whose elements satisfy the predicate and another whose elements do not satisfy the predicate; fix : factorial function programmed using the ....

R. P. Gabriel. Performance and Evaluation of LISP Systems. MIT Press series in computer systems. MIT Press, Cambridge, MA, 1985

....an error occurs, it would be nice if a user can see exactly what went wrong by examining call history and variable bindings in that context, then furthermore change the function definition in question and resume execution. All these demands are realized in Lisp, summed up as these four points in [Gabriel, Page 11] 1) functions may be defined on the fly by the user, 2) functions can be compiled separately, 3) compiled and interpreted calls can be intermixed, and (4) when an error or interrupt occurs, the stack can be decoded within the context of the error. Although Gabriel calls these points ....

....four points in [Gabriel, Page 11] 1) functions may be defined on the fly by the user, 2) functions can be compiled separately, 3) compiled and interpreted calls can be intermixed, and (4) when an error or interrupt occurs, the stack can be decoded within the context of the error. Although Gabriel calls these points the Lisp philosophy , I feel that the Lisp philosophy runs much deeper, and these features of Lisp, though of crucial importance to Lisp, are but manifestations of deeper characteristics. 1.4 Language documentation Lisp s user assistance does not stop here. A readable and ....

Gabriel, Richard P. Performance and Evaluation of Lisp Systems. The MIT Press, 1985.

....of lore in the community regarding what these di erent stacks are. Some of this lore is based on seminal studies on memory performance of Prolog systems [26, 27] other on actual experience, and some on experience coming from other symbolic languages, eg, implementation of functional languages [14, 17]. We next try to summarize our own understanding of this lore. The main problem in memory management is the Heap on deterministic computations. First, we can always reclaim space in non deterministic programs, we can always reclaim local space in deterministic programs, but we cannot recover Heap ....

R. P. Gabriel. Performance and evaluation of Lisp systems. MIT Press, 1985.

....involved. We use the following allocation intensive programs as benchmarks: 1. refresh, which rst builds a tree, then repeats a process that traverses the tree and replaces each node with a new node during traversal. This program is a variation of the benchmark traverse in the Gabriel suite [9], and is enhanced with intensive allocations. 2. recruit, which is similar to refresh, but during each traversal of the tree, only the tree leaves are replaced with new nodes. This program features a large number of long lived objects. 3. revcons, which builds a list. In the course of list ....

....is obtained. 4. shrink, which rst builds a list, then builds a new list by copying every second item in the old list. Repeat this until the length of the new list becomes one. The whole process is repeated 10 times. This program is a variation of the benchmark division by 2 in the Gabriel suite [9]. These benchmarks are used to test whether avoidance of major collections can balance extra allocation overhead. Here, we explain some terminology. The residency of a program at a particular moment is the size of its live heap allocated data [14] The average residency was 12 obtained by ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, 1985.

....that were discovered late prevent the allocation volumes from being equal. I have elected not to correct these problems, and comparisons across settings must therefore be made carefully. boyer is a toy term rewriting theorem prover derived from the Boyer benchmark in the Gabriel benchmark suite [13]. The program uses pairs extensively and constructs its data structures functionally. The versions used here, nboyer and sboyer, fix some bugs, are written in portable Scheme, and incorporate a problem scaling parameter. nboyer and sboyer differ only in that sboyer uses a local tweak shared ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. The MIT Press, 1985.

....8 queens 10queens Figure 2. Meta interpreter performance The evaluation in this section focuses on the performance of meta interpreters. The rest part of a meta object is executed without interpretation. As a back end compiler, we used CMU Common Lisp 17e. The benchmark programs are the Boyer[GAB 85] and n queens problem. All programs are executed on a SUN Sparcstation 10 workstation (SuperSparc 50MHz, 128MB memory) Figure 2(a) shows the ratios of the running times of INT s executions to the PE s executions, and Figure 2(b) shows the ratios of the running times of PE s executions to the ....

GABRIEL, R. P. Performance and Evaluation of Lisp Systems. MIT Press, 1985.

....is another TS Typed Smalltalk benchmark. sieve finds all the primes between 1 and 8190 using Eratosthenes sieve algorithm and stresses integer calculations, integer comparisons, and accessing arrays of booleans. tak executes the recursive Tak benchmark from the Gabriel Lisp benchmarks [Gab85], which stresses method calling and integer arithmetic. takl performs the same algorithm, but uses lists of cons cells to represent integers, and so additionally stresses list traversals and memory allocation. Most of these micro benchmarks are 1 to 3 lines long. sieve is 8 lines long, tak is 6 ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, Cambridge, MA, 1985.

....project Apply. 4 Reversing a list with 14 elements requires the allocation of 4427042 conses. APPLY ISST VI.1 4 14 10 1993 A Conservative Garbage Collector for an EuLisp to ASM C Compiler 4 that uses only the functions cons, car and cdr but no auxiliary functions, and the traverse benchmark [ Gabriel, 1986 ] The corresponding EuLisp modules were compiled into ANSIC code and then compiled and linked with the memory management system using the GNU C compiler linker. The allocator is configured to start with 16 cards of size 4096 bytes. If after a garbage collection more than one third of the ....

R. P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, Cambridge, Massachusetts, 1986.

....of our experiments show that I of improved PMGC is higher than I of original PMGC when G is less than about 22 , and I is not different between the two PMGCs when G exceeds about 22 . Fig. 7 shows the execution time of some applications, relative to seq lisp. The running application is Boyer[1], and executed through both interpreter and compiler. The left half of the graph shows the case of GC ratio is low (about 10 ) and the right half shows the case of GC ratio is high (about 40 ) When GC ratio is high, the execution time is drastically improved by PMGC. When GC ratio is low, ....

Gabriel, Richard P.: Performance and Evaluation of Lisp systems. The MIT Press (Cambridge, Massachusetts, 1985)

....at least on universal architectures additional computing costs. Except for special cases like small integers or characters in modern Lisp systems data are represented as pointer into system s heap. To encode the type of data there are mainly two strategies tagging and BIBOP 2 [ Shaw, 1988 ] Gabriel, 1986 ] BIBOP allocates data of the same type in local heap areas and area addresses encode the types of stored data. Contrarily, tagging schemes assign type information directly to pointers e.g. utilizing unused address bits. Since Lisp data are subjected to garbage collection the Lisp heap must be ....

....application program and therefore can be used for Lisp as well as for C. To test our approach we are implementing functions like xalloc and xpalloc as library functions in C. So we can link them with arbitrary applications. At first, the concept will be tested using some of the Gabriel benchmarks [ Gabriel, 1986 ] which must be rewritten in C. Later the scheme will be linked to APPLY s core system. Besides a study of the concept s feasibility the correlation between mutator cooperation and performance of the memory management system should be investigated in detail. APPLY ISST VI.1 3 Memory Management ....

R. P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, Cambridge, Massachusetts, 1986.

.... a cons intensive version of reverse 4 (defun reverse (l) if (null l) if (null (cdr l) l (cons (car (reverse (cdr l) reverse (cons (car l) reverse (cdr (reverse (cdr l) that uses only the functions cons, car and cdr but no auxiliary functions, and the traverse benchmark [ Gabriel, 1986 ] The corresponding EuLisp modules were compiled into ANSIC code and then compiled and linked with the memory management system using the GNU C compiler linker. The allocator is configured to start with 16 cards of size 4096 bytes. If after a garbage collection more than one third of the ....

R. P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, Cambridge, Massachusetts, 1986.

....case because they are regular in size,the most common item, and are never accessed via a derived pointer. Each of these collectors was then used to run two sample programs with varying page sizes. The sample programs were the Titan Scheme compiler and repeated executions of the Boyer benchmark [6]. The page size was varied from 128 to 4096 bytes. The effectiveness of each collector was measured by observing the number of times that garbage collection took place and the amount of storage that was retained after each collection. While it is dangerous to draw too many conclusions from such a ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. The MIT Press, 1985, pages 116-135.

....in Modula 2. lines vars procs remarks ######################################################### Livermore 268 347 20 Livermore loops Whetstones 462 413 37 Floating point Linpack 814 397 31 Linear algebra [14] Stanford 1019 565 65 Hennessy s suite [18] Boyer 600 1702 98 Lisp theorem proving [15] ######################################################### sed 1751 259 31 Stream editor egrep 844 241 28 File search troff 7577 947 175 Text formatter yacc 1856 689 73 Compiler compiler metronome 4287 1240 157 Timing verifier rsim 3003 911 133 Logic simulator [39] grr 5883 1833 243 ....

Richard P. Gabriel. Performance and Evaluation of Lisp Systems, pp. 116-135. The MIT Press, 1985.

....didn t allocate any) All three versions used a page size of 1K bytes and a heap with 3 megabytes per space. The sequential version scanned from 1 to 4K bytes for every 1K bytes allocated. The concurrent version used an Allocate chunk size of 65K. We picked the Boyer benchmark from Gabriel [Gabriel 85] as our first experiment. It is a small rule rewriter designed to test the performance of Lisp systems executing theorem provers. It allocates a large amount of non trivial, fine grained data structures and then accesses those structures repeatedly. According to Gabriel, the Boyer program does ....

Richard Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, 1985.

.... programmed using the standard firstorder recursive definition; ack curried : a curried version of Ackermann function that uses higher order functions (and is almost twice as fast as the standard first order function) tak cps : the Takeuchi function in CPS, part of the Gabriel benchmark suite [14]; reverse: standard first order list reverse function; rev cps : a CPS version of reverse; split : taking a predicate and a list and returning two lists, one whose elements satisfy the predicate and another whose elements do not satisfy the predicate; fix : factorial function programmed using the ....

R. P. Gabriel. Performance and Evaluation of LISP Systems. MIT Press series in computer systems. MIT Press, Cambridge, MA, 1985

.... parsers for context free grammars em fun 490 EM clustering algorithm in functional style em imp 460 EM clustering algorithm in imperative style graphs 500 A program that counts the number of directed graphs with particular properties interpret 1000 A Scheme interpreter evaluating the takl [Gabriel 1985] benchmark lattice 200 A program that enumerates the lattice of maps between two lattices matrix 550 A program that tests whether a matrix is maximal among all matrices obtained by reordering rows and columns maze 800 A hexagonal maze generator nbody 850 A program that computes gravitational ....

Gabriel, R. P. 1985. Performance and Evaluation of LISP Systems. MIT Press series in computer systems. MIT Press.

No context found.

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press, 1985.

No context found.

Richard P. Gabriel. Performance and Evaluation of Lisp Systems. MIT Press Series in Computer Science. MIT Press, Cambridge, MA, 1985.

No context found.

Richard P. Gabriel, Performance and Evaluation of Lisp Systems, MIT Press, Cambridge, Massachusetts, 1985.

No context found.

Gabriel, Richard P. 1985. Performance and Evaluation of Lisp Systems. Cambridge, Mass: MIT Press

No context found.

AI Magazine 10 (3): 32--48. Gabriel, R. P. 1985. Performance and Evaluation of Lisp Systems. MIT Press.

No context found.

Richard P. Gabriel. Performance and Evaluation of Lisp Systems, pages 116-135. The

First 50 documents  Next 50

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