<<< DUA0:[NOTES$LIBRARY]LEO.NOTE;1 >>> -< SIG Langages Et Outils >- ================================================================================ Note 81.0 Dhry/Whetstone 11 replies DECUSF::DUFLOS_D 4 lines 7-NOV-1989 14:18 -------------------------------------------------------------------------------- Quelqu'un dans l'assistance honorable sait-il ou trouver les sources du benchmark "dhrystones" en C ou en PASCAL. Merci beaucoup d'avance. ================================================================================ Note 81.1 Dhry/Whetstone 1 of 11 DECUSF::PILKIEWICZ_C "Claude PILKIEWICZ/CREDINTRANS" 8 lines 8-NOV-1989 13:23 -< Ils arrivent >- -------------------------------------------------------------------------------- Ils seront, en C et en ADA, dans la directory VMS: sous le nom BENCH.bck. Les versions ADA proviennent de l'ADA repository, les versions C de ?????. Bien sur sans aucune garantie sur le fait que ce sont les bons. ================================================================================ Note 81.2 Dhry/Whetstone 2 of 11 DECUSF::PILKIEWICZ_C "Claude PILKIEWICZ/CREDINTRANS" 11 lines 8-NOV-1989 13:57 -< Remarques >- -------------------------------------------------------------------------------- Je tiens a faire deux remarques sur ces programmes: 1) la considerable difference de resultats donnee, sur une meme machine par ces programmes suivant les langages. Ils permettent clairement de mettre en evidence les differences de qualite des compilateurs et la puissance des concepts comme l'inline. 2) le fait que ces programmes utilisent des sous-programmes de tailles anormalement petites. Ils ne rendent pas compte, a mon avis, des performances reelles des machines dans une utilisation normale. ================================================================================ Note 81.3 Dhry/Whetstone 3 of 11 DECUSF::DUFLOS_D 1 line 8-NOV-1989 14:37 -< Merci !!! >- -------------------------------------------------------------------------------- ================================================================================ Note 81.4 Dhry/Whetstone 4 of 11 DECUSF::WERZ_P "Pascal WERZ, MagneTech, ORSAY" 9 lines 8-NOV-1989 18:35 -< no privilege.... >- -------------------------------------------------------------------------------- Malheureusement, je n'ai pas d'accès READ a VMS:BENCH.BCK!!!! (comme a beaucoup d'autres fichiers dans VMS:, d'ailleurs). Pouvez-vous être assez gentil, en temps qu'OWNER, pour modifier la protection pour un acces W:R? Ce serait gentil et permettrait à tout le monde d'en profiter! Un grand merci. Pascal WERZ. ================================================================================ Note 81.5 Dhry/Whetstone 5 of 11 DECUSF::PILKIEWICZ_C "Claude PILKIEWICZ/CREDINTRANS" 8 lines 9-NOV-1989 13:37 -< G:RE,W:RE >- -------------------------------------------------------------------------------- C'est fait! Par ailleurs, je ne comprends pas pourquoi vous n'avez pas d'acces a ses fichiers. Ils ont la protection g:re, et j'avais cru comprendre que nous etions tous dans le groupe DECUS. Les gestionnaires de serveur ont sans doute une explication ??? ================================================================================ Note 81.6 Dhry/Whetstone 6 of 11 DECUSF::GERARD_G "G. Gerard ENST centre de calcul" 10 lines 9-NOV-1989 15:11 -< identifier+resource=aspirine >- -------------------------------------------------------------------------------- c'est un pb d'evaluation de droits d'acces: Les fichiers de VMS: appartiennent a l'identifier DECUS, qui est declare avec l'attribut resource, et qui est donne a tous les gens de decus. L'evaluation du droit d'acces, hors acl, se fait avec l'uic du demandeur, et ne donne donc pas l'acces au fichier si celui-ci n'est pas accessible par le world. Pour autoriser l'acces par defaut, il faudrait une acl sur les fichiers autorisant les gens possedant l'identifier DECUSF de lire lesdits fichiers. Soit une acl de protection par defaut, soit une acl par defaut, sur le directory VMS: Ouf. ================================================================================ Note 81.7 Dhry/Whetstone 7 of 11 DECUSF::DANA_M "Michel DANA, Telecom PARIS, Responsa" 5 lines 9-NOV-1989 17:54 -< c'est deja fait, mais ca marche "moyen" >- -------------------------------------------------------------------------------- il y a tout ca sur VMS... le drame, c'est que lorsqu'on fait rename, ca n'est pas pris en compte... mais je vais faire un truc qui mettra automatiquement les bons droits aux fichiers tous les jours ================================================================================ Note 81.8 Dhry/Whetstone 8 of 11 DECUSF::WERZ_P "Pascal WERZ, MagneTech, ORSAY" 4 lines 9-NOV-1989 18:11 -< Tous? >- -------------------------------------------------------------------------------- Même sur les fichiers qui y sont depuis un moment, il y en a quelques uns qui m'interessent???? P.WERZ ================================================================================ Note 81.9 Dhry/Whetstone 9 of 11 DECUSF::THONON_D "Daniel Thonon SEMA-GROUP Meylan" 9 lines 10-NOV-1989 10:56 -< Bravo VAXC >- -------------------------------------------------------------------------------- Resultat des courses sur VS3100: Compilateur Whetstone VAXC 3.0 sans Mathlib: 3450 KWIPS Ada 1.5 WhetAdaa: 4350 KWIPS Ada 2.0 WhetAdaa: 5150 KWIPS Sans commentaires! ================================================================================ Note 81.10 Dhry/Whetstone 10 of 11 DECUSF::ALBERT_JN 12 lines 10-NOV-1989 18:58 -< Commentaires demandes >- -------------------------------------------------------------------------------- Je n'y connais pas grand chose en Whestone & co. Que faut-il faire pour avoir les memes estimations pour VAX-FORTRAN (et autres) ? Ces resultats sont-ils ``significatifs'' ? Il me semble qu'on choisit un langage en fonction de son efficacite pour traiter le probleme a resoudre. Il peut donc y avoir des variations importantes. Ces differences se retrouvent-elles pour d'autres implémentations de ces langages, ou sont-elles liees a l'implementation par DEC du langage C ? Mais c'est certainement un bon point de depart pour une discussion. ================================================================================ Note 81.11 Dhry/Whetstone 11 of 11 DECUSF::THONON_D "Daniel Thonon SEMA-GROUP Meylan" 195 lines 14-NOV-1989 11:37 -< DOCS >- -------------------------------------------------------------------------------- Ci-joint les explications des benchmarks ADA. ------------------------------------------------------------------------- USERS MANUAL version 1.0 for BENCHMARKING COMPILERS by Doris Sayon and Bonnie Burkhardt May 9, 1985 Western Division GTE Government Systems Group 100 Ferguson Drive Mountain View, CA 94039  CONTENT 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . 3 2 WHETSTONE BENCHMARK . . . . . . . . . . . . . . . . 3 2.1 Description: . . . . . . . . . . . . . . . . . . . 3 2.2 Usage Instructions: . . . . . . . . . . . . . . . 3 3 DHRYSTONE BENCHMARK . . . . . . . . . . . . . . . . 5 3.1 Description: . . . . . . . . . . . . . . . . . . . 5 3.2 Usage Instructions: . . . . . . . . . . . . . . . 5  Page 3 1 INTRODUCTION The suite of tests were gathered and developed at GTE Government Systems to provide a benchmarking capability to determine which ADA compiler would be suitable to be used for our needs. As a measurement of the efficiency of the compiler, the Whetstone test was written in FORTRAN, 'C', and Pascal as well as ADA to give a relative measurement across languages as well as compilers. 2 WHETSTONE BENCHMARK 2.1 Description: The first program in the suite is the Whetstone benchmark, which measures processor speed. Included in the Whetstone data generated from the test is some information gathered from package SYSTEM which tells of the environment in which the test is taking place. In order to compare the processor speed with/without a math library and with/without a full set of checks, four variations of the Whetstone program were created. These versions were called A, B, C, and D. The files are also labelled according to the version, Whetadaa, Whetadab, Whetadac, and Whetadad. The versions are defined as follows: A stubbed math library, no checks B stubbed math library, full checks C full math library, no checks D full math library, full checks 2.2 Usage Instructions: In order to use the Whetstone benchmark, you must first create a file called ENVIRON.INF in the directory in which the Whetstone program will be running. The information in ENVIRON.INF will be of the following format: Machine: xyz Op. Sys: 4.xxx Physical Mem: yy Compiler Tested: compiler 1.o of xyz company This information will be copied into the data file which will be created from running the whetstone test. This will help keep track of what compiler and machine the test was run on. If the ENVIRON.INF file does not exist, there will be a message "Something wrong with the existence of the ENVIRON.INF file"  Page 4 written into the file and the tests results will still be recorded. Compile the files in the following order: 1. MATHSTUB.ADA should be compiled first because it is a package which needs to be resolved in the WHETADA.ADA file 2. HELPTOOLS.ADA should be compiled next. 3. WHETADA.ADA (all versions) should be compiled last. After successful compilation of the above packages, the program should be ready to run. When the program is run, the results will appear both on the terminal and in a data file. The name of the data file has following format: Wnnddhhmm.DAT when nn: two character sequence which represents the month (i.e. ja,fe,mr,ap,my,jn,jl,ag,se,oc,nv,dc) dd: two number sequence for day (01 thru 31) hh: two number sequence for hour (00 thru 23) mm: two number sequence for minutes (00 thru 59) Example: a test run at 2:47 pm on Jan 8 will have the file name WJA081447.DAT The encoding of the date into the name of the file will help distinguish between several runs of the same test. If compile time and link time is desired, you can compile and link the Whetstone programs by executing the command file WHETADA.COM. This command file displays the current time before and after each compile and link to SYS$OUTPUT (defaulted to the terminal screen.) The user can then determine the compile and link time by computing the time difference between the start and stop of each compile and link. The Whetstone test can also be used to compare execution speed and size of the executable image when using various compile options, such as OPTIMIZE=TIME, OPTIMIZE=SPACE, NOOPTIMIZE. This can be accomplished by compiling the Whetstone program with the desired compile option, linking, and executing the program as described previously. The command file WHETADA.COM also compiles, links, and executes all four version of the Whetstone with 3 different compile options, OPTIMIZE=TIME, OPTIMIZE=SPACE, NOOPTIMIZE. Size of the executable code can be found by executing the command: $DIR/SIZE *.EXE  Page 5 The size of the executable version will be written to SYS$OUTPUT (defaulted to the terminal screen.) The user can then determine the size of the executable code. This process is then repeated for the Whetstone test written in the other languages. Since there is no way in 'C' to designate which checks are to be included in the executable code, the 'C' Whetstone only has 2 versions, A and B. Version A has a stubbed math library and Version B has a full math library. 3 DHRYSTONE BENCHMARK 3.1 Description: The next program in the suite is the Dhrystone benchmark. The Dhrystone benchmark measures statement execution per unit time. Included in the Dhrystone data which is generated from the test, there is also some information gathered from package SYSTEM which tells of the environment in which the test is taking place. 3.2 Usage Instructions: In order to use the Dhrystone benchmark, you must first make sure the HELPTOOLS package has been compiled and is in the library to be referenced for resolution of packages. Next, DHRYADA should be compiled after helptools has compiled. If the compilation was successful, execute the Dhrystone benchmark. The data should appear both on your terminal and in a data file. The name of the data file is encoded the same was as the data file for the Whetstone data except that the filename starts with a "D" instead of "W". The existence of ENVIRON.INF should exist for the same reasons as expressed in the information above for the Whetstone benchmark.