Bit Operations & Math

[RagingGrim]

I was just curious. I thought i'd go ahead and learn some assembly. During that time I saw a trick to figure out a characters case ( the third bit of a sequence indicates the case of a character ) so I thought I'd implement it in C before trying it in Nasm.

#include <stdio.h>
#include <stdlib.h>
int IsUpper(char AChar); //Will return 1 for any uppercase char.
int IsUpper(char AChar)
{
    return (0b00100000 & AChar) | 0b1;
}
int main()
{
    printf("%i",26 >> 1); //shl 2 is equal to * by 4
    return 0;
}

The IsUpper function works fine Is it at it's fastest though? Is it correct? I mean the | 0b1 is the same as +1 right? Only the program doesn't have to do the conversion because it's already in binary?

As for the 26 >> 1 , I just remembered that from some delphi challenge ages ago. That's the same as dividing an even integer with two correct? I get an error if i try %f and 26.0 anyway. I was just curious if 26 shr 1 is faster tha 26 / 2 .

Just for the sake of it ^^

[Salem]

> The IsUpper function works fine Is it at it's fastest though? Is it correct?
I dunno. Why don't you write
for ( int ch = 0 ; ch < 256 ; ch++ )
and test every possible char?

FWIW, these bit-tricks only work on ASCII character sets. It won't work for example on EBCDIC.

> I mean the | 0b1 is the same as +1 right?
Try it with an odd number.

> I was just curious if 26 shr 1 is faster tha 26 / 2 .
Nowadays, only the most brain-dead of compilers is incapable of optimising /2 as a shift instruction.
Do yourself a favour and stop searching for micro-optimisations.

A modern compiler has the distilled wisdom of probably 1000's of years of expertise within it. Just once, you might get a very localised 'win' against it. The other 99.99% of the time, you're just going to waste your time.

For instance, x >> n can give you a big surprise on some machines if x is negative.

[RagingGrim]

>> I mean the | 0b1 is the same as +1 right?
>Try it with an odd number.

Sorry didn't state my question correctly , I realise the result would be different because an odd number in binary would have a 2 ** 0 which would result in a 1 being carried etc etc. Thanks for pointing that out though , It'll probably help me remember that :/

I'm not really searching for micro optimistations more than I am trying to gain understanding I'm also making a point of learning nasm (general assembly ?? ) because I believe I'll require it in the future.

I'll test it in a for loop right after this post . I assumed the output would be fine because the logic behind it seems solid ^^ Then again I didn't really study any of this so your probably right with the testing.

char AChar;
    for(AChar='A';AChar-1<'Z';AChar++)
        printf("Testing %c : %i\n",AChar,(0b00100000 & AChar) | 0b1);

I tested it with all the uppercase characters (I mean the alphabet letters , i never intended to test it with them all XD ) and it works fine I'll just make a comment next to the function about the EBCDIC format.
Not that I think I'd ever use an IBM pc but all knowledge is good right?

As always though I find your replies informing and helpful !
Thanks Salem