Threaded Mode | Linear Mode

pier4r · (This post was last modified: 01-18-2018 10:02 PM by pier4r.)

Added many thanks!

I got from ebay, since they are plentiful and cheap (see n1) 9860g and 9860gII , then I will see if I can replicate the reported results.
Also for possible future searches. Both can be kept alive via USB like the 50g. The USB cable is the same of the 50g (and ti89): mini B. Of course the 50g wins easily in terms of long term storage (sd card) that can be recovered even if the main system goes awry.

n1: http://www.hpmuseum.org/forum/thread-970...l#pid88145 - I do not remember where I wrote that I wonder how many functioning devices are trashed after the owner does not like them again instead of ending on ebay.

Michael de Estrada · 01-19-2018, 01:23 AM

HP 32SII

1000 loops —> 241 seconds, Result = 1395.34628770

100 loops —> 26 seconds, Result = 139.29718705

10 loops —> 3 seconds, Result = 13.71183502

Michael de Estrada · (This post was last modified: 01-19-2018 03:01 AM by Michael de Estrada.)

HP-55 (Classic)

100 loops —> 434 seconds, Result = 139.2971873

10 loops —> 43 seconds, Result = 13.71183501

Note that the times are much slower than the HP-65, which has a loop counter DSZ.

Michael de Estrada · (This post was last modified: 01-20-2018 04:22 AM by Michael de Estrada.)

Elektronika MK 54, MK 56 and MK 61

These are Soviet RPN programmable calculators from the mid 1980s to the early 1990s. They are extremely slow, so I only have results for 10 loops.

MK 54: 10 loops —> 99 seconds, Result = 13.711835

MK 56: 10 loops —> 110 seconds, Result = 13.711835

MK 61: 10 loops —> 100 seconds, Result = 13.711835

pier4r · 01-20-2018, 10:45 AM

added!

grsbanks · 01-20-2018, 10:49 AM

HP-15C: 100 loops => 313s

The DM15L does the same job 22x faster in 14s Smile

Michael de Estrada · 01-20-2018, 02:50 PM

DM42 with Firmware v3.2

It has slowed down with the new firmware, which lowered the code optimization to prevent the calculator from hanging during intensive calculations.

1000 loops powered by batteries —> 9.83 sec, Result = 1,395.34628774

10,000 loops powered by USB —> 30.48 sec, Result = 13,955.8579043

grsbanks · 01-20-2018, 03:26 PM

(01-20-2018 02:50 PM)Michael de Estrada Wrote: DM42 with Firmware v3.2

It has slowed down with the new firmware, which lowered the code optimization to prevent the calculator from hanging during intensive calculations.

1000 loops powered by batteries —> 9.83 sec, Result = 1,395.34628774

10,000 loops powered by USB —> 30.48 sec, Result = 13,955.8579043

I'm on a slightly later firmware version than you but nothing has been done to provide any acceleration in my internal test version. I get:

1000 loops on battery --> 6.45s

10000 loops on USB --> 26.57s

This is with RefLCD set to 6.

Michael de Estrada · (This post was last modified: 01-21-2018 05:18 PM by Michael de Estrada.)

(01-20-2018 03:26 PM)grsbanks Wrote:
(01-20-2018 02:50 PM)Michael de Estrada Wrote: DM42 with Firmware v3.2

It has slowed down with the new firmware, which lowered the code optimization to prevent the calculator from hanging during intensive calculations.

1000 loops powered by batteries —> 9.83 sec, Result = 1,395.34628774

10,000 loops powered by USB —> 30.48 sec, Result = 13,955.8579043

I'm on a slightly later firmware version than you but nothing has been done to provide any acceleration in my internal test version. I get:

1000 loops on battery --> 6.45s

10000 loops on USB --> 26.57s

This is with RefLCD set to 6.

I changed RefLCD from 7 to 6, and it did have a significant effect on the run times as follows:

1000 loops on battery changed from 9.83 sec to 7.08 sec

10000 loops on USB changed from 30.48 sec to 28.03 sec

They are still not as good as your times, which no doubt is due to your code being more efficient.

grsbanks · 01-20-2018, 04:09 PM

(01-20-2018 04:03 PM)Michael de Estrada Wrote: They are still not as good as your times, which no doubt is due to your code being more efficient.

For what it's worth, this is my code:

Code:

00 { 43-Byte Prgm }

01▸LBL "BM"

02 ABS

03 IP

04 RAD

05 TIME

06 →HR

07 STO 00

08 R↓

09 0

10▸LBL 00

11 RCL ST Y

12 ATAN

13 SIN

14 E↑X

15 3

16 1/X

17 Y↑X

18 +

19 DSE ST Y

20 GTO 00

21 TIME

22 →HR

23 RCL- 00

24 3600

25 ×

26 END

Michael de Estrada · 01-20-2018, 04:37 PM

(01-20-2018 04:09 PM)grsbanks Wrote:
(01-20-2018 04:03 PM)Michael de Estrada Wrote: They are still not as good as your times, which no doubt is due to your code being more efficient.

For what it's worth, this is my code:

Code:

00 { 43-Byte Prgm } 01▸LBL "BM" 02 ABS 03 IP 04 RAD 05 TIME 06 →HR 07 STO 00 08 R↓ 09 0 10▸LBL 00 11 RCL ST Y 12 ATAN 13 SIN 14 E↑X 15 3 16 1/X 17 Y↑X 18 + 19 DSE ST Y 20 GTO 00 21 TIME 22 →HR 23 RCL- 00 24 3600 25 × 26 END

Yes, your code is better, since you are doing everything in the stack, whereas I am using variables. I didn’t realize you could access the stack with DSE and ISG, which really helps. My new times using your code are 6.83 sec with 1000 loops on batteries and 27.01 sec with 10000 loops on USB.

TheKaneB · (This post was last modified: 01-21-2018 10:23 PM by TheKaneB.)

Sharp EL-9300 graphing calculator, exact same code that I wrote for the EL-5120 Solver, got 22 seconds for 100 iterations.

Guenter Schink · (This post was last modified: 01-22-2018 10:09 PM by Guenter Schink.)

Epson HX-20 an early "laptop" Smile

Basic
single precision <= 7 digits
10 cycles: 2 sec; Result: 13.7118
100 cycles: 18 sec; Result: 139.297
1000 cycles: 173 sec Result:1395.36

double precision <= 16 digits
10 cycles: 2 sec; Result: 13.71183562278748
100 cycles: 18 sec; Result: 139.297193646431
1000 cycles: 173 sec; Result: 1395.346369147301

No apparent difference of runtime between single and double precision.

Günter

pier4r · 01-23-2018, 07:33 PM

added all. The dm42, as far as I understood, did not change significantly the timings using the code of grsbanks.

Interesting the epson, that is as large as a 9825/9100 as far as I understood.

Also, what about 91xx/98xx series?

Thomas Okken · (This post was last modified: 01-23-2018 08:03 PM by Thomas Okken.)

(01-23-2018 07:33 PM)pier4r Wrote: Interesting the epson, that is as large as a 9825/9100 as far as I understood.

No, no! The HX20 has less than a third of the footprint of the 9825, and it's lower, weighs 3.5 lbs instead of 26, and runs on batteries. There's no comparison, really.

Cool machine, but I couldn't have afforded it back then. (It wasn't as far out of reach as the HP machines, though!)

TheKaneB · 01-23-2018, 08:19 PM

Speaking of little laptop-thingy. I unleashed my trusty Olivetti M10 (similar to the Tandy Model 100).

With 10 iterations I got about 9.5 seconds
With 100 iterations about 91 seconds

It has 14 digits, but I used a 12 digits approximation of "e" (copied from my Prime), so the result is a bit off.

This thing has a 8085 CPU and runs Microsoft Basic

Guenter Schink · 01-23-2018, 09:10 PM

(01-23-2018 07:49 PM)Thomas Okken Wrote:
(01-23-2018 07:33 PM)pier4r Wrote: Interesting the epson, that is as large as a 9825/9100 as far as I understood.

No, no! The HX20 has less than a third of the footprint of the 9825, and it's lower, weighs 3.5 lbs instead of 26, and runs on batteries. There's no comparison, really.

Cool machine, but I couldn't have afforded it back then. (It wasn't as far out of reach as the HP machines, though!)

Nor couldn't I. Got it in 2012 for 40€ for my collection. It's including the microcassette tape drive. Rock solid - keyboard like on an electrical typewriter Smile

The footprint is close to an A4 paper and the height goes from 33mm to 60mm (front to rear)

Günter

**Gene** · 01-23-2018, 09:43 PM

(01-22-2018 10:06 PM)Guenter Schink Wrote: No apparent difference of runtime between single and double precision.

Günter

Gene: Now THAT I find really interesting. Wonder what it is doing under the hood that is not faster for single precision or not slower for double?

Dieter · 01-23-2018, 09:48 PM

(01-23-2018 08:19 PM)TheKaneB Wrote: It has 14 digits, but I used a 12 digits approximation of "e" (copied from my Prime), so the result is a bit off.

?!? – why do you need a value for e ? Is there really no EXP() function?
Especially since you mention...

(01-23-2018 08:19 PM)TheKaneB Wrote: This thing has a 8085 CPU and runs Microsoft Basic

...and an exponential function is an absolutely ...basic feature of BASIC.

Dieter

TheKaneB · 01-23-2018, 10:43 PM

(01-23-2018 09:48 PM)Dieter Wrote:
(01-23-2018 08:19 PM)TheKaneB Wrote: It has 14 digits, but I used a 12 digits approximation of "e" (copied from my Prime), so the result is a bit off.

?!? – why do you need a value for e ? Is there really no EXP() function?
Especially since you mention...

(01-23-2018 08:19 PM)TheKaneB Wrote: This thing has a 8085 CPU and runs Microsoft Basic

...and an exponential function is an absolutely ...basic feature of BASIC.

Dieter

I think you know it better than me Smile

Tomorrow I'll try again and see if it has the EXP() function and redo the benchmark.
I used the standard ^ operator for powers.