SLC compute scan time

SLC 5/04
Does anyone know what the typical scan time is for the Compute instruction?
The manual states
"The execution time of a CPT instruction is longer
than a single arithmetic operation and uses more
instruction words."

I am doing simple math such as... (1|(T4:0.ACC|N7:0))*60

Thanks in advance
gm

Why do you care? If you really cared you would write your compute statement more efficiently.

Create a program that includes your CPT instruction. Also create a timer that keeps track of the time required for 1000 program scans. Compare the times for 1000 scans with the CPT statement enabled and with it disabled. Divide the difference by 1000.

Oh no - 11 posts compared to 5,083 posts (if that's any measure). Peter is one of the most advanced here mathematically - at least in my view.

I recommend that you apologize profusely then ask for help and suggestions on how to make your math better.

Wow Gary,

Peter: 5000+ posts
Gary: 11 posts

Peter has never been a member of the Dale Carnegie School of how to win friends and influence people but I think you are jumping way to conclusions with that one. He could probably write the OS in your PLC to be more efficient than AB did.

Perhaps you should pay attention and learn something before striking out at the regulars.

Didn't mean to offend anyone.


gm

Peter Nachtwey said:

Why do you care? If you really cared you would write your compute statement more efficiently.

Click to expand...

How could this be more efficiant?

Peter, if you choose to answer, be gentle. All were new and arrogant.

OK, let me minimize the pain - the first two optimizations

your statement - (1|(T4:0.ACC|N7:0))*60

Minimize constants - 60 * 1 = 60

60|(T4:0.ACC|N7:0)

multiplies are more efficient than divides

60|X|N7:0 = 60 * N7:0

(N7:0 * 60) | T4:0.ACC

Why is it that most programmers seem to be too lazy to convert even the simplest of formulas? In my schooldays it surely paid off to do so. Why don't you try if 60*N7:0|T4:0.ACC doesn't calculate faster without giving a different answer?

I can understand Peter's reaction.

Edit: Damn, Bernie beat me to it. But then I'm sure in Bernie's schooldays this brain warmup was normal day-to-day business, just as it was in my days (maybe around the same time?)

I thank my teacher, though not at the time, who insisted that the first week in high school Physics be spent on learning the slide rule. Though high school Physics was the highest math I had (1965), that was the best wek of education I had.

garym4569 said:

SLC 5/04
The manual states
"The execution time of a CPT instruction is longer
than a single arithmetic operation and uses more
instruction words."

Click to expand...

All the manual is telling you (admittedly in a roundabout way), is that a CPT, although friendly, readable, concise, etc., is always going to be slower, and use more memory, than the same maths done using sequential "standard" maths instructions.

The first thing a CPT will have to do is parse the expression to determine the evaluation order (BODMAS), then it will do the individual maths in the expression in that order, storing temporary results on the way, (for which it will have to use internal scratchpad memory locations that it may have to create first, then destroy later), to arrive at the final result.

When you code the expression as a series of simpler maths instructions, you have done all of the hard work, and the processor can just wade through them.

bernie_carlton said:

I thank my teacher, though not at the time, who insisted that the first week in high school Physics be spent on learning the slide rule. Though high school Physics was the highest math I had (1965), that was the best wek of education I had.

Click to expand...

High (very high) up on a wall at the college i went to in Rugby, England, were painted (obviously by a member of the mountaineering club), these immortal words....

Slide Rules OK

As a newbie ST compiler writer I can now say that the parse tree has been optimized thanks to Bernie. The execution time should be about half now because the number of divides have reduced in half and they are the operations that take the longest in that compute block.

Garym4569 is much further ahead now than he would have been if someone simply answered his question. Getting the best answer quickly should be the goal. A combination of Steve's and Bernie's answers are the best soltution.

I wasn't being arrogant. To be arrogant I would have to be taling about how good I am, but I didn't mention me. I even let other have a chance to shine. I could have easily optimized the compute instruction myself.

The bit about 11 posts compared to 5000+ makes no difference. If the numbers were reversed I would still make the same post if I now I am right.

I thank my teacher, though not at the time, who insisted that the first week in high school Physics be spent on learning the slide rule.

Click to expand...

This is 8th grade algebra.

However, you can't believe how hard it is to have a compiler optimize the parse tree.

daba said:

All the manual is telling you (admittedly in a roundabout way), is that a CPT, although friendly, readable, concise, etc., is always going to be slower, and use more memory, than the same maths done using sequential "standard" maths instructions.

Click to expand...

I think they are simply interpreting the test each time the compute block is execute. You have raised a very good point. It would be better yet to implement the compute block in ladder.