Threaded Mode | Linear Mode

GeorgeOfTheJungle · (This post was last modified: 05-31-2014 01:29 PM by GeorgeOfTheJungle.)

Code:

LBL 01

1

ENTER

ENTER

ENTER

LBL 02

+

GTO 02

I hit run and then R/S after 30s and I get:

hp65: 332
hp41c: 528
hp35s: 1816

So the 41c is ~= 1.6x faster than the 65
the 35s is ~= 5.4x times faster than the 65
the 35s is ~= 3.4x times faster than the 41c

Is that right? I ask because I was expecting more difference as the hp41c came 6 years after the hp65, and the 35s almost 40 years later!

I've read somewhere that the 67 is slower than the 65, is that true?

Could somebody please run that (for 30s) in a wp34s and post the result?

Thank you,
Jorge.

Dave Britten · 05-31-2014, 12:51 PM

Average of three trials on my 34S in 4-level-stack mode, timed with my 95LX, as that's the stopwatch I have in arm's reach at the moment: 43,266

Controlling the program counter directly can also be used to great effect:

Code:

LBL 01

1

ENTER

ENTER

ENTER

+

BACK 001

END

...Yields 94,035 (just one trial); more than twice as fast.

Interestingly, this "optimization" only got me 76,250:

Code:

LBL 01

1

INC X

BACK 01

Evidently INC is slower than all the stack cloning, at least in the case of the X register.

This is all tested on firmware version 3.2 3472 on a 30b, by the way. I imagine the Windows/iOS/etc. versions score several orders of magnitude higher. Smile

I'll try my 32S later for the sake of comparison. My hunch is that it will be close to the 35S, but I can't say in which direction.

Massimo Gnerucci · 05-31-2014, 01:08 PM

There's always a previous post with the answer... Wink

GeorgeOfTheJungle · 05-31-2014, 01:23 PM

(05-31-2014 12:51 PM)Dave Britten Wrote: Average of three trials on my 34S in 4-level-stack mode, timed with my 95LX, as that's the stopwatch I have in arm's reach at the moment: 43,266

Controlling the program counter directly can also be used to great effect...
...Yields 94,035 (just one trial); more than twice as fast.

...This is all tested on firmware version 3.2 3472 on a 30b, by the way.

Wow, cool, thank you!

So the wp34s is between 130x and 280x faster. That's much better, much closer to what one would expect after 40 years, isn't it?

Is the wp34s the fastest "hp" rpn calculator ever? If it is... well, no comments.

(05-31-2014 12:51 PM)Dave Britten Wrote: I imagine the Windows/iOS/etc. versions score several orders of magnitude higher.

LOL, yes. I just clicked run then R/S in the emulator in my Mac and it went to more than half a million.

(05-31-2014 12:51 PM)Dave Britten Wrote: I'll try my 32S later for the sake of comparison. My hunch is that it will be close to the 35S, but I can't say in which direction.

Is the "original" 30b rpn programmable? If it is, could you try that too?

Thank you,
Jorge.

Dave Britten · 05-31-2014, 02:22 PM

I'm getting about 2,597 on my 32S. I was mistakenly thinking that model supported branching backwards with negative values for GTO (i), but that doesn't appear to be the case, so that's just a standard LBL/GTO approach.

(05-31-2014 01:23 PM)GeorgeOfTheJungle Wrote: Is the "original" 30b rpn programmable? If it is, could you try that too?

It is, but I don't have an "original" 30b to test, unfortunately. I only had one for about an hour or so!

I think I'll write a test for my TI-58C so I can have a laugh at how slow that thing is. I think one of the Op routines increments a register, so I don't have to bring AOS into the mix at least. I'll be surprised if it hits 200, though.

Dave Britten · 05-31-2014, 02:43 PM

TI-58C results...

Code:

000 76 LBL

001 11 A

002 00 0

003 42 STO

004 00 0

005 76 LBL

006 85 +

007 69 Op

008 20 20 (Op 20 increments reg 00)

009 61 GTO

010 85 +

I get a whopping 133. That could probably be improved with fast mode.

Using 1 SUM 00 instead of Op 20 got me 134, which I'd call equal within experimental error.

With GTO directly to an address, I only got 127. I guess with this small of a program, the label search is slightly faster than interpreting the 3-step GTO 0 05 vs. the two-step GTO +.

GeorgeOfTheJungle · 05-31-2014, 02:44 PM

(05-31-2014 02:22 PM)Dave Britten Wrote: I'm getting about 2,597 on my 32S. I was mistakenly thinking that model supported branching backwards with negative values for GTO (i), but that doesn't appear to be the case, so that's just a standard LBL/GTO approach.

7.9x, meh... :-)

(05-31-2014 02:22 PM)Dave Britten Wrote:
(05-31-2014 01:23 PM)GeorgeOfTheJungle Wrote: Is the "original" 30b rpn programmable? If it is, could you try that too?

It is, but I don't have an "original" 30b to test, unfortunately. I only had one for about an hour or so!

No wonder... :-)

In the link that Massimo has posted (btw, thanks, Massimo) it says:

HP 30b
Count: 72,517
Code: LBL 00 + GOTO 00
(as first program in memory)

And that's more or less ok, between 130x and 280x, ~= your 34s.

(05-31-2014 01:23 PM)GeorgeOfTheJungle Wrote: I think I'll write a test for my TI-58C so I can have a laugh at how slow that thing is. I think one of the Op routines increments a register, so I don't have to bring AOS into the mix at least. I'll be surprised if it hits 200, though.

Perhaps it's overclock-able...

GeorgeOfTheJungle · 05-31-2014, 02:51 PM

(05-31-2014 02:43 PM)Dave Britten Wrote: TI-58C results...

Code:

000 76 LBL 001 11 A 002 00 0 003 42 STO 004 00 0 005 76 LBL 006 85 + 007 69 Op 008 20 20 (Op 20 increments reg 00) 009 61 GTO 010 85 +

LBL + ??? I can't do that in my hp! Why do you use + instead of B?

Dave Britten · 05-31-2014, 03:30 PM

(05-31-2014 02:44 PM)GeorgeOfTheJungle Wrote: In the link that Massimo has posted (btw, thanks, Massimo) it says:

HP 30b
Count: 72,517
Code: LBL 00 + GOTO 00
(as first program in memory)

And that's more or less ok, between 130x and 280x, ~= your 34s.

Bear in mind, those were 60-second tests, with a shorter program and pre-loaded stack (label searches may be faster).

(05-31-2014 02:51 PM)GeorgeOfTheJungle Wrote: LBL + ??? I can't do that in my hp! Why do you use + instead of B?

The 58/59 used weird labeling. You had A-E corresponding to the five hot-keys at the top, and A'-E', the same hot-keys prefixed with 2nd. These would normally be used for your program's entry points. You could also use (almost) any other non-numeric key for a label, which could be called with SBR or GTO. GTO with numeric arguments would be interpreted as a branch to a step number, so that's probably why you couldn't use 0-9 as labels.

Program steps were 100% not merged, so this scheme was probably easier for the parser/state machine. Treat any key code after LBL/GTO/SBR as a label, and a single digit key-code (00-09) after GTO/SBR as a trigger to grab the next step's key code and build a 3-digit step number out of them.

GeorgeOfTheJungle · 05-31-2014, 03:58 PM

(05-31-2014 03:30 PM)Dave Britten Wrote:
(05-31-2014 02:44 PM)GeorgeOfTheJungle Wrote: And that's more or less ok, between 130x and 280x, ~= your 34s.
Bear in mind, those were 60-second tests, with a shorter program and pre-loaded stack (label searches may be faster).

Ahhh, yes, you're right, my bad. So the geniuses behind the wp34s, given the same hardware/platform, have made of it not only a better but also a faster calculator than hp? Incredible. No comment.

(05-31-2014 03:30 PM)Dave Britten Wrote:
(05-31-2014 02:51 PM)GeorgeOfTheJungle Wrote: LBL + ??? I can't do that in my hp! Why do you use + instead of B?

The 58/59 used weird labeling. You had A-E corresponding to the five hot-keys at the top, and A'-E', the same hot-keys prefixed with 2nd. These would normally be used for your program's entry points. You could also use (almost) any other non-numeric key for a label, which could be called with SBR or GTO. GTO with numeric arguments would be interpreted as a branch to a step number, so that's probably why you couldn't use 0-9 as labels.

Program steps were 100% not merged, so this scheme was probably easier for the parser/state machine. Treat any key code after LBL/GTO/SBR as a label, and a single digit key-code (00-09) after GTO/SBR as a trigger to grab the next step's key code and build a 3-digit step number out of them.

I've never had any TI in my hands. It looks nice, and quite advanced compared to the hp65! Thanks for the detailed explanation.

Cheers,
Jorge.

Dave Britten · 05-31-2014, 04:39 PM

(05-31-2014 03:58 PM)GeorgeOfTheJungle Wrote: I've never had any TI in my hands. It looks nice, and quite advanced compared to the hp65! Thanks for the detailed explanation.

Cheers,
Jorge.

Yeah, in terms of programming capability, they easily trump the 65, and can hold their own against - if not surpass - the 67. Shame about the algebraic entry which tends to complicate program logic (and you've already got unmerged steps working against you). The build quality was no match for the HPs, either, and it seems like the TIs are a bit slower (I don't have a real 67 to compare with).

On the other hand, the solid state software modules were pretty cool.

You can get them pretty cheaply on ebay. Just be prepared to clean out a leaky battery pack, and solder in a 3xAAA receptacle to fill with NiMHs.

GeorgeOfTheJungle · 05-31-2014, 05:47 PM

(05-31-2014 04:39 PM)Dave Britten Wrote: Yeah, in terms of programming capability, they easily trump the 65, and can hold their own against - if not surpass - the 67. Shame about the algebraic entry which tends to complicate program logic (and you've already got unmerged steps working against you). The build quality was no match for the HPs, either, and it seems like the TIs are a bit slower (I don't have a real 67 to compare with).

On the other hand, the solid state software modules were pretty cool.

You can get them pretty cheaply on ebay. Just be prepared to clean out a leaky battery pack, and solder in a 3xAAA receptacle to fill with NiMHs.

Most steps in the hp65 were (are!) unmerged too, it seems to predate the TI58 by 4 years (I've seen in the wikipedia) but it beats the TI by 2.5x in this benchmark... Anyway, I'm sure I'm not going get a TI... I just need to kill a bug that I caught the other day when I stumbled upon the hp65 and the 41c, in a box, in my basement... :-) I think the last time I used the 65 was, perhaps, in the very very early 80s... and the 41c was a gift I got in the early 90's, I mean, I've never been a 41c user. I'm going to be playing with them a few days, perhaps a few weeks, and then they'll surely go back to that box in the basement. Unless my wife convinces me to put them on ebay instead... :-(

Marcus von Cube · 05-31-2014, 06:13 PM

(05-31-2014 03:58 PM)GeorgeOfTheJungle Wrote: Ahhh, yes, you're right, my bad. So the geniuses behind the wp34s, given the same hardware/platform, have made of it not only a better but also a faster calculator than hp? Incredible. No comment.

I think this isn't quite correct. The design principle behind the 34S has always been accuracy over code size over speed. That is how we managed to cramp so much functionality in such a limited device (I leave it to Walter to comment on the display Wink

). In many cases, the simpler design of the original 30b firmware pays off in terms of execution speed. So it was a 30b program that won the HHC2011 RPN contest.

Massimo Gnerucci · 05-31-2014, 06:15 PM

(05-31-2014 06:13 PM)Marcus von Cube Wrote: (I leave it to Walter to comment on the display ).

Eheheh...

GeorgeOfTheJungle · 05-31-2014, 07:50 PM

(05-31-2014 06:13 PM)Marcus von Cube Wrote:
(05-31-2014 03:58 PM)GeorgeOfTheJungle Wrote: Ahhh, yes, you're right, my bad. So the geniuses behind the wp34s, given the same hardware/platform, have made of it not only a better but also a faster calculator than hp? Incredible. No comment.
I think this isn't quite correct. The design principle behind the 34S has always been accuracy over code size over speed. That is how we managed to cramp so much functionality in such a limited device (I leave it to Walter to comment on the display ). In many cases, the simpler design of the original 30b firmware pays off in terms of execution speed. So it was a 30b program that won the HHC2011 RPN contest.

So it's better, faster *and* more accurate... :-) I've ordered a 30b already. And a few wp34s overlays. And the printed owners manual... And I must kill this bug asap.

Marcus von Cube · 05-31-2014, 08:25 PM

(05-31-2014 07:50 PM)GeorgeOfTheJungle Wrote: So it's better, faster *and* more accurate... :-) I've ordered a 30b already. And a few wp34s overlays. And the printed owners manual... And I must kill this bug asap.

Don't get me wrong. In many aspects, the unmodified 30b is the faster machine. You can just do much more with the 34S which is still reasonably fast at least. And much more fun, of course Big Grin

.

Dave Britten · 05-31-2014, 08:39 PM

Didn't see the 15C LE in the list, so I gave it a try.

22,649 with LBL/GTO.
21,914 with GTO I using a negative value.

(05-31-2014 07:50 PM)GeorgeOfTheJungle Wrote: So it's better, faster *and* more accurate... :-) I've ordered a 30b already. And a few wp34s overlays. And the printed owners manual... And I must kill this bug asap.

Do you have the necessary flashing cable or USB module to solder in?

Thomas Klemm · 05-31-2014, 08:58 PM

(05-31-2014 08:39 PM)Dave Britten Wrote: Didn't see the 15C LE in the list, so I gave it a try.

You could use the following program:

Code:

001-42, 6,25   ISG I

However it must be the only step thus you have to clear all the other programs.

Cheers
Thomas

PS: I'm sure somebody already did this.

Dave Britten · 05-31-2014, 09:09 PM

(05-31-2014 08:58 PM)Thomas Klemm Wrote: You could use the following program:

Code:

001-42, 6,25 ISG I

However it must be the only step thus you have to clear all the other programs.

Getting pretty similar results, actually: 21,229.

GeorgeOfTheJungle · 05-31-2014, 09:37 PM

(05-31-2014 08:39 PM)Dave Britten Wrote: Didn't see the 15C LE in the list, so I gave it a try.

22,649 with LBL/GTO.
21,914 with GTO I using a negative value.

Hey, Dave, have you got one of each calculator ever made or what ??? :-)

That's also worse than the 34s... only ~= 1/3x !

(05-31-2014 08:39 PM)Dave Britten Wrote: Do you have the necessary flashing cable or USB module to solder in?

No, neither, but I've got some arduinos and I've read the 34s "how to flash the thing" chapter more than once... we'll see! Wish me luck!