bxpfac@umiami.ir.miami.edu (12/26/90)
From: Bimal (devebw9f@miavax.ir.miami.edu)
In one of my applications, most of my integer numbers are small in
magnitude and positive. I am considering using unsigned char in
order to conserve memory space.
Consider the following code fragments:
A. usigned char i;
for (i = 0; i < 100; i++) {
...
B. typedef struct {
unsigned char a;
unsigned char b;
unsigned short c;
} EXAMPLE ;
EXAMPLE p;
Questions:
1. In example A above, is the use of unsigned char any faster compared to
using short or int? Any drawbacks? (Incidentally, I will be using
either the gcc compiler or the turboc compiler for this).
2. How would one go about reading and writing the struct above using
fprintf and fscanf?
To elaborate further:
fprintf (fd, "%? %? %hu\n", p.a, p.b, p.c);
fscanf (fd, "%? %? %hu" , &p.a, &p.b, &p.c);
/\
what should I use here | ?
Thanks for your response.
Bimalbakke@plains.NoDak.edu (Jeffrey P. Bakke) (12/26/90)
In article <1990Dec25.220800.7543@umiami.ir.miami.edu> devebw9f@miavax.ir.miami.edu writes: > From: Bimal (devebw9f@miavax.ir.miami.edu) > > In one of my applications, most of my integer numbers are small in > magnitude and positive. I am considering using unsigned char in > order to conserve memory space. > > Consider the following code fragments: >[... deleted to save space ...] > > 1. In example A above, is the use of unsigned char any faster compared to > using short or int? Any drawbacks? (Incidentally, I will be using > either the gcc compiler or the turboc compiler for this). Ok, under GCC, I wouldn't know... it would dependent on the system you are compiling for. Under turboC in can depend on a number of things. If, under turbo C you are compiling and plan to be running on a 8088 based machine (not 8086 notice) and you turn the option that allows word alignment for speedup on, there should still be no improvement in speed. If you use word alignment on any machine with a 16-bit bus, it will speed up some operations; but then everything gets aligned on word boundaries that wastes space if you're real critical on memory usage. In either case, it seems to be just as fast to use integers over chars... Under Turbo C, the only time that char instead of int seems to make much sense is when you are working with actual character data. As far as using a char over an int for a counter loop or something, there will be no improvement since the 80x86 cpu has 16-bit (word size) registers anyway for storing and manipulating data. (I know, I know, but I'm assuming non-protected mode programming). Under GCC, you might have similar situations but it would be dependent on your host machine architecture. In my opinion using char over int is only really helpful when you have a large amount of data to store. It really shouldn't make any difference in processing time though. The only time that chars might be useful in a loop over integers is that if the loop carries too far, the character counter will rollover instead of 'overflow' past your boundaries of characters unlike type 'int' -- Jeffrey P. Bakke | There are a finite number of INTERNET: bakke@plains.NoDak.edu | jokes in the world... UUCP : ...!uunet!plains!bakke | The overflow began BITNET : bakke@plains.bitnet | decades ago. "I am not a number, I am a free man!" - The Prisoner