Scor­pio News


January–March 1987, Volume 1, Issue 1

Page 57 of 63

‘A Sideways Look at Benchmarks _ by P.D. Coker

A few years ago, PCW published a with « Lise of aleros’ on

18 An BASIC, together ‘were arranged in order

chines” versione ‘of BASIC. No Nascom or Gemini ‘one machine vas designed by Cemint

(Himi 801, which did quite well in the teste).

The benchmarks vere structured so that for testa 2 – 7, subtracting the timing for the previous test from the current tert time would give the time due to the Foutine under examination. The eighth benchmark differed from the rest, which tested che timings for arithmetic functions, GOSUBs and array handling since it dealt with transcendental functions such ae SIN, LOG and exponent

100 REM Benchmark 8 (FCW) alo Print “5”

120 Keo

a0 kek

150 BeLoceR) 160 CoSTNtK)

170 Te Ke1009 THEN 130 Leo PRINT “E*

450 END

I tried this one on the following machines (CPU and system clock frequency in brackets)

uel (6502, mate) 22.1 secs Bac B (6502, 2Ms) 513 secs Rescom 2 (280,” Utz no FDC) $0.1 seca gbBC + TORCH 286 2nd processor 30.8 sece Stascom 3 (280, amie) 330 seca Scemint MuleiBoard (.” ) 52.2 secs smap-80 Systems (7) 319 Seca SHAP=A0 Systems ( * } 50.8 secs

E¶ Using Z80 BBC BASIC and CP/M * Using MBASIC and cP/H 2.2, 4 Using MBASIC and CP/H 3.0 (CP/M Plus)

I then found that the timing for the standard BBC was 5.1 seconds according to Pow! Evidently their Beeb was cupercharged! More likely, line 170 of the program had been amended to “IF K<100 THEN 230’ since the average value of 166 Gver the 8 tests was based on a timing of 51 seconde for no. 8 Colleagues who have run this test on ocher machines mentioned in the PCW list have also commented upon inaccuracies in the reported speeds. The 260 version of BBC BASIC as used on the TORCH second procesvor war extremely fart but disk access With chis cystem is dreadfully slow 1 did not have access to the Acorn 6502 0’ 260 second processors – it would have been interesting to ree their timings

More recently, # series of Benchmerks in PASCAL was produced by the same magazine: these were 15 in number and designed to take account of the greater Fange of features found in this language, but the overall philosophy was the Sane. namely. by subtracting the timing for the previous test from the current fest’ time, some ides could be gained of the time for the routine under investigation. I was abl COMPAS v1 08, HP and ProPi

using CP/M 22 and CP/M fe to the conclusion that for BM2’ 2 PASCAL version of the BMS BASIC version, HP4 wag quite « lot faster then COMPAS, poseibly because it uses 7 rather than 11 significant figure accuracy. ‘The retulte were at follows:

This is an OCR’d version of the scanned page and likely contains recognition errors.

Page 57 of 63