Prev: I don't get it... HP 50G
Next: New open source OS for ARM based HP calculators
From: Mike on 22 Jul 2006 13:55
I would really appreciate it if one of the owners of the new hp50g
would report the results of a 2500-iteration Savage Benchmark on it.

The program I use on my hp48gx and hp49g+ should run on the hp50g:

<< RAD TICKS 1 1 2499 START sqrt SQ LN EXP ATAN TAN 1 + NEXT
TICKS 3 ROLL - b->r 8192 / 400 2 BEEP DEG >>

where sqrt = squart root
b->r = binary to real conversion.

No arguements are required, and the program results are a numerical
value close to 2500 (would be 2500 exactly if calcs were ideal), and a
performance time in seconds.

On my hp48gx (rev. R) performs this in 117 seconds.

My hp49g+ version 2.06 performs this in 107 seconds (surprinsingly
slow).

My hp49g+ when it had the original ROM version 1.20 performed this in
97 seconds.

Thanks,
Mike

From: Heiko Arnemann on 22 Jul 2006 14:33
Hello Mike,
thank you for your benchmark
news:1153590931.409553.68200(a)m79g2000cwm.googlegroups.com...

> << RAD TICKS 1 1 2499 START sqrt SQ LN EXP ATAN TAN 1 + NEXT
> TICKS 3 ROLL - b->r 8192 / 400 2 BEEP DEG >>

First, the program for the 49g should be with reals instead of
zints, after ASCII transfer it shoul look like this:

%%HP: T(3)A(D)F(.);
\<< RAD TICKS 1. 1. 2499.
START \v/ SQ LN EXP ATAN TAN 1. +
NEXT TICKS 3. ROLL - B\->R 8192. / 400. 2. BEEP DEG
\>>

> On my hp48gx (rev. R) performs this in 117 seconds.
>
> My hp49g+ version 2.06 performs this in 107 seconds (surprinsingly
> slow).
>
> My hp49g+ when it had the original ROM version 1.20 performed this in
> 97 seconds.

The result on hp 49g+ depens on flag settings, too, I assume.

my flags:
%%HP: T(3)A(D)F(.);
{ # 4071257920090719220d # 0d # 9230128673280131992d # 0d }

my result ROM 2.05-4:
within 64 s.

By the way, your prog modifies the flag settings (RAD.. DEG) :-)

Best regards
Heiko


From: Mike on 22 Jul 2006 19:14
Heiko Arnemann wrote:

> First, the program for the 49g should be with reals instead
> of zints,
> ...
> my result ROM 2.05-4:
> within 64 s.

Hi Heiko,

Thanks very much for the reminder about zints on the hp49g+. I'd
forgotten all about that!

After adding the decimal points to the appropriate values to change
from zints to reals (which will be done automatically if the hp49g+ is
in "approximate" mode before writing and saving the program), my Savage
Benchmark now reads:

<< RAD TICKS 1.
1. 2499. START sqrt SQ LN EXP ATAN TAN 1. + NEXT
TICKS 3. ROLL - b->r 8192. /
400. 2. BEEP DEG >>

where sqrt = square root, b->r = convert binary to real.

Now my hp49g+, ROM version 2.06, performs the benchmark in just a
little more than 64 seconds, just as you reported. That's much better,
but it's still only about 1.8 times faster than an old hp48gx.

Typical results: 2499.99948647, 64.28 seconds.

It makes no difference in results if the hp49g+ is in exact or
approximate mode, if the program has been written to use real values
and not zints.

It will be interesting to read how the hp50g compares to the hp49g+, if
someone will post that info.

Best regards!
Mike

From: ttaylor on 22 Jul 2006 19:21
2499.99948647
64.4040527344

in approx mode

VER

4.20060106

VERSION

"Version HP50-C
Revision 2.08"

From: Mike on 22 Jul 2006 19:58
ttaylor(a)elp.rr.com wrote:

> 2499.99948647
> 64.4040527344
> ...
> "Version HP50-C
> Revision 2.08"

Thanks! It appears that the hp50g and the hp49g+ operate at exactly
the same speed. I was hoping that more of the internal programming of
the hp50g would have been converted to run "native" on the ARM
processor for a better speed advantage.

Best regards!
Mike

 |  Next  |  Last
Pages: 1 2 3
Prev: I don't get it... HP 50G
Next: New open source OS for ARM based HP calculators