Benchmarking

There is information on HP Calculator Benchmarking online at the HP Museum

Benchmarking

I've used the pseudoCode to test out my own calculators.

LBL <whatever>
0
STO 2
1.01234 EEX 6
STO 0
2.345
STO 1
<loop>
RCL 1
RCL 0
*
RCL 1
-
RCL 0
/
RCL 1
*
3.5
/
SQRT
1
STO+ 2
GTO <loop>

This needs tweaking for each calculator, some have alpha register names, some numeric. The handling of the <loop> is different in each calculator,too.

The PRG is run for 1 minute and the result is found in register 2 (or wherever).

LBL <whatever>
56.26
STO 0
0
STO 2
<loop>
RCL 0
SIN
ASIN
COS
ACOS
TAN
ATAN
LN
e^x
1
STO + 2
GTO <loop>

• Reference is HP9100 : Maths Score = 679, Trig Score = 40

DM42

LBL "BM"
0
STO 02
1.01234E6 
STO 00
2.345
STO 01
LBL 02
RCL 01
RCL 00
x
RCL 01
-
RCL 00
÷
RCL 01
x
3.5
÷
SQRT
1
STO+ 02
GTO 02
END

DM41X

LBL ¬BM
0
STO 02
1.01234 E6
STO 00
2.345
STO 01
LBL 02
RCL 01
RCL 00
*
RCL 01
-
RCL 00
/
RCL 01
*
3.5
/
SQRT
1
ST+ 02
GTO 02
END

HP35S

LBL X
0
STO C
1.01234E6
STO A
2.345
STO B
RCL B  <-- loop in
RCL A
x
RCL B
-
RCL A
÷
RCL B
x
3.5
÷
√x
1
STO+ C
GTO X008  --> loop in

DM15L / HP15C

42,21,15
0
42 2
1
48
0
1
2
3
4
26
6
44 0
2
48
3
4
5
44 1
42,21,2
45 1
45 0
20
45 1
30
45 0
10
45 1
20
3
48
5
10
11
1
44,40, 2
22 2
43 32

LBL E
0
STO 2
1
.
0
1
2
3
4
EEX
6 
STO 0
2
.
3
4
5
STO 1
LBL 2
RCL 1
RCL 0
x
RCL 1
-
RCL 0
÷
RCL 1
x
3
.
5
÷
√x
1
STO + 2
GTO 2
RTN

DM32

LBL A
0
STO C
1.01234 E 6
STO A
2.345
STO B
LBL B
RCL B
RCL A
X
RCL B
-
RCL A
/
RCL B
X
3.5
/
SQRT
1
STO+ C
GTO B
RTN

From HPCC Datafile V37N1 pp38-42

LBL "SVG"
"SAVAGE"
AVIEW
RAD
2499
STO 00
1
ENTER
ENTER
ENTER
STO 01
TIME
X<> 01
LBL 01
X^2
SQRT
LN
E^X
ATAN
TAN
+
DSE 00 
GTO 01
TIME
RCL 01
HMS-
HR
3600
*
END

No time calculation, it needs a stopwatch to measure the time… for calculators without TIME functions

LBL A
RAD
2499
STO A (for HP11C "STO I", for DM15L "STO 0")
1
ENTER
ENTER
ENTER
LBL B
X^2
SQRT
LN
E^X
ATAN
TAN
+
DSE A  (for HP11C "DSE", for DM15L "DSE 0" )
GTO B
RTN

On Battery:

On USB Power :

• DM41X : 58.1 seconds (FAST mode)
• DM42 : 5.56 seconds

From http://www.thimet.de/CalcCollection/CalcPerformance.html

LBL A  (For DM41X and DM42 : LBL "AA")
10
STO A  (For DM41X and DM42 : STO 00 )
LBL B
1
+
4.567E-4
-
70
+
69
-
7
x
11
/
RCL A (RCL 00)
1
-
STO A (STO 00)
X≠0?
GTO B
R↓
LOG
SIN
SQRT
SQRT
RTN


LBL C  (For DM41X and DM42 : LBL "TBM")
XEQ A  (XEQ "AA")
RCL C  (RCL 03)
1
-
STO C (STO 03)
X≠0?
GTO C (GTO "TBM")
RTN

For multiple loops store a Loop Count in C (or in Register 03)

(1000 STO C) and then XEQ C

For single loop (slow calculator) simply XEQ A

Result is the time to complete the loop(s) and deliver the answer 0.264577

Performance Index is calculated as

Loops x 34 / Time

For 1000 loops in 13 seconds:

1000 x 34 / 13 = 2615.38

Each calculator handles the calculation of functions (SIN, COS etc) differently - different algorithms, different degrees of precision etc.

One way of spotting similar underlying hard/firmware is to carry out a fixed chained calculation on each calculator and to observe the result.

One common test is a sequence of SIN COS TAN ATAN ACOS ASIN on a relatively low starting number - such as 9

9SINCOSTANATANACOSASIN gives a number (hopefully) very close to the starting value of 9 - and the variance from 9 is a fingerprint that can be used to spot similarities between different models.

Page created Thu May 26 07:58:36 2022 by John Pumford-Green

Page last updated: 06/03/25 06:49 GMT