How does float store in C language?

float how memory is stored

Type	Number of storage bits			Total number of digits	Offset value (offset)
	Digit (S)	Order Code (E)	Mantissa (M)
Short real number (float)	1	8	23	32	127
Long real number (double)	1	11	52	64	1023

N (10) = 123.456,

Translates into binary representations:

N (2) = 1111011.01110100101111001

= 1. 11101101110100101111001 (...) *2^6

then e–127 = 6; E =127 + 6 = 133 (Ten) = 10000101 (2)

M =111 0110 1110 1001 0111 1001 ( max digits omitted 1, total 23bit)

together it is:

S E M

0 10000101 1110110 1110 1001 0111 1001

4bit One interval:

0100 0010 1111 0110 1110 1001 0111 1001

4 2 F 6 E 9 7 9

The first two days of interviewing a company, did not answer a question, summarized as follows, for reference.

Q: How much memory is consumed by instance variables of this struct type:
struct STRUCT1
{
int i;
SHORTJ;
Charc;
};
I ask: What language is the machine what the compiler environment?
He answered, VC6.0.
Me: The memory is aligned and then takes up 8byte.
He went on to ask: how to make it only 7byte?
My first reaction is to use a bit segment, is preparing to answer, and feel wrong, the bit segment can not let it not aligned AH. Thought for a few seconds, or not, had to return to his said not to play ...


This afternoon went to the library to turn over "code optimization: Efficient Use of memory", found that there are two methods mentioned:
Method1:
#pragma pack (push)
#pragma pack (1)
struct STRUCT1
{
int i;
SHORTJ;
Charc;
};
#pragma pack (POP)
The original #pragma pack had several forms, and I was exposed to the #pragma pack (n), that is, the variable is aligned in N-byte.

Method2:
Modify the parameters of the compilation directive to prevent memory alignment:
VC + +:/zn1 (VS2005, right-click Project-Properties-Configuration Properties-c/c++-code generation-Structure Member alignment-select "1 bytes (/ZN1):" (that is, prevent memory alignment), default is to use the default value, that is, the structure of the largest space-occupying members to align. To align the size of the.
Borland C + +:/-A1


Method2is to disable memory alignment for the entire project,and the Law1memory alignment can be disabled for a specific structure,it provides greater flexibility.



In addition, the book also mentions:
char is not aligned;
Short along the even address alignment;
The int/float is aligned along the address of a multiple of the value 4.
Double the address aligned along a multiple of the value 8.


VS2005, the default is to use the default value, that is, according to the structure of the largest space-occupying members to align, we can test the following structure:
struct STRUCT2
{
Chari
SHORTD;
Doublec;
SHORTJ;
};

Leave the default value or modify the compilation parameter/zn?, guess how much the result of sizeof? :)
printf ("struct2:%d\n", sizeof (STRUCT2));
Struct2 St;
printf ("%p\n", &st.i);
printf ("%p\n", &ST.D);
printf ("%p\n", &st.c);
printf ("%p\n", &ST.J);

Supplement: manual alignment of the data:

Char *p;
int temp = align_power-1;
p= (char*) malloc (need_size + temp;
p= (char*) malloc (((int) p+temp) &temp);//modified P, so remember to return to the position before releasing P
Note: Align_power is required for Zishing, char* (which can also be int*) is a pointer type. Also, remember to point to the first address of the requested memory before releasing p.

How does float store in C language?