Benchmarks

The following are a collection of benchmarks that have been gathered from around the internet and translated into PL/I.

Tower of Hanoi
Source:	hanoi.pli
Description:	This benchmark was originally written in C and appeared in Albert Aburto's collection here. Due to increasing processor speed, rather than compute the time taken to complete the Tower Of Hanoi puzzle this benchmark computes the average number of moves in a 25μs slice.
Build Instructions:	After downloading the source module, as well as the timer module below, this software can be built using the following DCL commands: $ PLI HANOI+TIMER $ LINK HANOI $ RUN HANOI
Results:	The below table shows gathered results for the PL/I version of the benchmark. Results for the C version can be found here.
	System	OS, Compiler	CPU	MHz	Moves/25μs
	Digital VAXstation 4000-90	OpenVMS V7.3, Kednos PL/I 3.8	VAX KA49-A	70	29.689
	Compaq PWS XP1000	OpenVMS V8.2, Kednos PL/I 4.4G	Alpha EV6 (21264)	500	528.974
	Compaq AlphaServer ES45	OpenVMS V8.3, Kednos PL/I 4.4G	Alpha EV68CB (21264C)	1000	1058.281
	HP Integrity rx2600 Server	OpenVMS V8.3-1H1, Kednos PL/I 4.4G and PLI_RTL_AV T4.5-B^*	Itanium 2 Madison 1.4GHz/1.5MB	1400	47.345

Fibonacci Numbers
Source:	fib.pli
Description:	This benchmark was originally written in C and appeared in Albert Aburto's collection here. Its purpose is to calculate the first 40 Fibonacci numbers. The last of which is 102334155. This benchmark was used to determine the usefulness of an address lookup cache in the OpenVMS I64 Translated Image Environment (TIE).
Build Instructions:	After downloading the source module, as well as the timer module below, this software can be built using the following DCL commands: $ PLI FIB+TIMER $ LINK FIB $ RUN FIB
Results:	The table below shows results gathered for the PL/I version of the Fibonacci Numbers benchmark. Results for the original C version can be found here.
	System	OS, Compiler	CPU	MHz	Run Time (sec)
	Digital VAXstation 4000-90	OpenVMS V7.3, Kednos PL/I V3.8	VAX KA49-A	70	90.047
	Compaq PWS XP1000	OpenVMS V8.3, Kednos PL/I V4.4G	Alpha EV6 (21264)	500	3.766
	Compaq AlphaServer ES45	OpenVMS V8.3, Kednos PL/I V4.4G	Alpha EV68CB (21264C)	1000	1.828
	HP Integrity rx2600 Server	OpenVMS V8.3-1H1, Kednos PL/I 4.4G and PLI_RTL_AV T4.5-B^*	Itanium 2 Madison 1.4GHz/1.5MB	1400	70.695

Prime Numbers
Source:	prime.pli
Description:	This benchmark determines how long it takes to calculate the specified number of prime numbers. This benchmark is suited to testing raw number crunching, in particular integer division.
Build Instructions:	After downloading the source module, as well as the timer module below, this software can be built using the following DCL commands: $ PLI PRIME+TIMER $ LINK PRIME $ RUN PRIME
Results:	The table below shows results gathered for the Prime Numbers benchmark. In each of these tests the number of primes calculated was 55,000. The last of which is 679,277.
	System	OS, Compiler	CPU	MHz	Run Time (sec)
	Digital VAXstation 4000-90	OpenVMS V7.3, Kednos PL/I 3.8	VAX KA49-A	70	988.617
	Compaq PWS XP1000	OpenVMS V8.2, Kednos PL/I 4.4G	Alpha EV6 (21264)	500	158.883
	Compaq AlphaServer ES45	OpenVMS V8.3, Kednos PL/I 4.4G	Alpha EV68CB (21264C)	1000	79.102
	HP Integrity rx2600 Server	OpenVMS V8.3-1H1, Kednos PL/I 4.4G and PLI_RTL_AV T4.5-B^*	Itanium 2 Madison 1.4GHz/1.5MB	1400	3730.970

Timer Module
Source:	timer.pli
Description:	This module is not a benchmark. It is the timer code used by most of the other benchmarks to determine the time taken to complete. Currently there is only a module for OpenVMS (all versions). If you have a PL/I compiler on your non-OpenVMS system and you have run these benchmarks, please submit your timer module (plus instructions) so it can be published here for others to use.

^*These results were obtained by translating an Alpha version of the benchmark using an experimental release of the OpenVMS Migration Software for Alpha to IPF Systems (OMSAI) binary translator from HP. Run-time support was provided by the Kednos Native IPF PL/I Run-Time Library for Translated Images.