Re: calc VAL field not updating from bi VAL field

[Eric Norum <eric@norum.ca>]

Perhaps, but I've never seen a machine where the following code snippet would print…. (even if the variables were made volatile to ensure that the loads and stores actually occurred)….


	int a, b;
	double x;

	a = 1;
	x = a;
	b = x;
	if (a != b) printf("THIS IS A REALY BAD MACHINE!\n");


On May 30, 2012, at 12:43 PM, Maren Purves wrote:

> Eric Norum wrote:
>> On May 30, 2012, at 10:25 AM, Tim Mooney wrote:
>> 
>>>> …..
>>> 
>>> It seems that the integer to double conversion really is not exact
>>> for 1.
>> 
>> That seems like a pretty huge flaw in the architecture!  I've never
>> encountered a machine with arithmetic that bad!
> 
> I have experienced very bad rounding errors (in a gensub, in C code)
> on a ppc604/mv5100 - to the point that I had to re-order the arithmetic.
> Bad arithmetic isn't unheard of.
> 
> Aloha,
> Maren

-- 
Eric Norum
eric@norum.ca