Welcome to the MacNN Forums.

alligator · Oct 11, 2003, 03:11 PM

I'm using the calculator on my G5, and I subtracted 87.376-87.143 and it returns a value of 0.2330000000000041.

Why? A bug in the G5 chip?

Thinine · Oct 11, 2003, 03:16 PM

No, just the precision (or lack thereof) of Apple's calculator.

Scotttheking · Oct 11, 2003, 03:26 PM

Just floating point precision error.

olePigeon · Oct 11, 2003, 05:13 PM

Originally posted by Scotttheking:
Just floating point precision error.

Try deviding by zero, see if your CPU bursts into a puff of smoke.

Eug · Oct 12, 2003, 02:12 AM

Are you using X.2? I had these errors in X.2, but they seem to be fixed in Panther.

I had a similar thread way back when, but I can no longer replicate these errors in X.3.

And for the record: 87.376 - 87.143 = 0.233 in Panther.

Try deviding by zero, see if your CPU bursts into a puff of smoke.

Actually, the calculator now gives an answer of "Infinity". Cool.

sideus · Oct 12, 2003, 02:19 AM

Originally posted by olePigeon:
Try deviding by zero, see if your CPU bursts into a puff of smoke.

:XI: · Oct 12, 2003, 08:11 AM

Originally posted by Eug:
Are you using X.2? I had these errors in X.2, but they seem to be fixed in Panther.

I had a similar thread way back when, but I can no longer replicate these errors in X.3.

And for the record: 87.376 - 87.143 = 0.233 in Panther.

Actually, the calculator now gives an answer of "Infinity". Cool.

87.376 - 87.143 = 0.233 in 10.2.8 on a G4

Divide by zero also gives 'Inf'

Boochie · Oct 12, 2003, 08:57 AM

I get what appears to be the correct answer when the calculator is in "basic" mode, but the roundoff error is visible in "advanced" mode. Weird. G4/867 with 10.2.8.

Originally posted by :XI::
87.376 - 87.143 = 0.233 in 10.2.8 on a G4

Divide by zero also gives 'Inf'

Eug Wanker · Oct 12, 2003, 10:24 AM

Originally posted by Boochie:
I get what appears to be the correct answer when the calculator is in "basic" mode, but the roundoff error is visible in "advanced" mode. Weird. G4/867 with 10.2.8.

In Panther in basic mode there isn't a problem either.

cambro · Oct 12, 2003, 06:20 PM

Originally posted by Scotttheking:
Just floating point precision error.

Yup, what he said.

You all know that floating pt. number addition/subtraction posses some interesting problems in a binary system, right? It is impossible to do without very small error.

That said, Apple shouldn't let you see this in the Calc. App. But somehow I don't think this is a deal breaker for very many users

Cadaver · Oct 12, 2003, 08:00 PM

Originally posted by Boochie:
I get what appears to be the correct answer when the calculator is in "basic" mode, but the roundoff error is visible in "advanced" mode. Weird. G4/867 with 10.2.8.

The error is also displayed when "paper tape" is visible (at least on this G4, 10.2.8 vers 2; I'll check my G5 when I get home, but I'm sure it's there). This "bug" has been around since 10.2.0 (and perhaps 10.1 I'm thinking). It's a round-off error in Calculator.app. I've seen freeware calc apps for the PC do this kind of thing, too.

alligator · Oct 13, 2003, 05:48 PM

I seem to recall a huge uproar when the Pentium III chip had flaws and couldn't calculate right. Why is this glossed over from Apple? To me this is completely unacceptable by today's standards. Sure, it might fly in 1975, but this is 2003. Come on, how am I supposed to calculate Pi to 10 trillion decimal places when the first 8 numbers are messed up??????

HodyOne · Oct 13, 2003, 06:27 PM

Originally posted by alligator:
I seem to recall a huge uproar when the Pentium III chip had flaws and couldn't calculate right. Why is this glossed over from Apple? To me this is completely unacceptable by today's standards. Sure, it might fly in 1975, but this is 2003. Come on, how am I supposed to calculate Pi to 10 trillion decimal places when the first 8 numbers are messed up??????

Because Apple's error is an algorithmic software error while the pentium error was a hardware error.

If you want to calculate Pi to 10 trillion decimal places get Mathematica.

piracy · Oct 13, 2003, 10:58 PM

Originally posted by alligator:
I seem to recall a huge uproar when the Pentium III chip had flaws and couldn't calculate right. Why is this glossed over from Apple? To me this is completely unacceptable by today's standards. Sure, it might fly in 1975, but this is 2003. Come on, how am I supposed to calculate Pi to 10 trillion decimal places when the first 8 numbers are messed up??????

Has nothing to do with the chip.

Take a few CS courses, learn about how floating point arithmetic works, and come back and chat with us.

Eug · Oct 14, 2003, 09:37 AM

Sure, it might fly in 1975, but this is 2003.

Apple shouldn't let you see this in the Calc. App.

I will point out once more that the algorithm has already been updated in Panther's Calculator.

lacrymology · Oct 14, 2003, 10:03 AM

Originally posted by alligator:
I seem to recall a huge uproar when the Pentium III chip had flaws and couldn't calculate right. Why is this glossed over from Apple? To me this is completely unacceptable by today's standards. Sure, it might fly in 1975, but this is 2003. Come on, how am I supposed to calculate Pi to 10 trillion decimal places when the first 8 numbers are messed up??????

I am not aware of PIII calculation bugs (there was a boot bug), but perhaps you are refering to the famous Pentium f00f bug that would instantly crash the system, regardless of OS? That is a far more serious issue than a typical floating point precision error.
-m

Mattmac21 · Oct 14, 2003, 10:19 AM

Originally posted by alligator:
I'm using the calculator on my G5, and I subtracted 87.376-87.143 and it returns a value of 0.2330000000000041.

Why? A bug in the G5 chip?

87.376-87.143 = .233 on a my machine; 1.8Ghz G5

HodyOne · Oct 14, 2003, 12:24 PM

Originally posted by lacrymology:
I am not aware of PIII calculation bugs (there was a boot bug), but perhaps you are refering to the famous Pentium f00f bug that would instantly crash the system, regardless of OS? That is a far more serious issue than a typical floating point precision error.
-m

The Pentium bug that caused an uproar in the early '90's was a calculation error - it would not crash the system. To most people, this is FAR more serious than a crashing bug. With a crashing bug at least you know that something is wrong. The calculation bug would cause programs to merrily go on their way producing incorrect values. There's really no comparison.

ZildjianKX · Oct 14, 2003, 01:06 PM

Originally posted by piracy:
Has nothing to do with the chip.

Take a few CS courses, learn about how floating point arithmetic works, and come back and chat with us.

Take some computer engineering courses and come back and talk to us... not all errors or programming related.

An Intel Engineer left 5 spaces blank in their SRT division table... so when any of those places were used for a division, it returned zero. Cost intel about 1 billion to replace all those chips...

lacrymology · Oct 14, 2003, 01:23 PM

Originally posted by HodyOne:
The Pentium bug that caused an uproar in the early '90's was a calculation error - it would not crash the system. To most people, this is FAR more serious than a crashing bug. With a crashing bug at least you know that something is wrong. The calculation bug would cause programs to merrily go on their way producing incorrect values. There's really no comparison.

The bug I was refering to was:

http://x86.ddj.com/errata/dec97/f00fbug.htm

-m

Eug · Oct 14, 2003, 01:50 PM

Originally posted by ZildjianKX:
Take some computer engineering courses and come back and talk to us... not all errors or programming related.

An Intel Engineer left 5 spaces blank in their SRT division table... so when any of those places were used for a division, it returned zero. Cost intel about 1 billion to replace all those chips...

He was referring to the issue with Calculator, which is a software one, and it has already been corrected. And apparently, if you used other software with the G5/G4, it was not an issue.