beginnings of bit fields

This commit is contained in:
Perry Kivolowitz 2022-06-17 21:30:47 -05:00
parent b29469a255
commit dc6ac5a73a
4 changed files with 409 additions and 0 deletions

View file

@ -0,0 +1,111 @@
# Section 2 / Bit Fields
## Overview
Many C and C++ programmers have never seen bit fields.
Bit fields are a
feature of the C and C++ language which completely hide what is often
called "bit bashing".
Bit bashing is the manipulation of individual bits. Bit
bashing goes to the very core of the C language. Remember that C is a
high level assembly language, as we argue in Section 1 of this book.
And C is the (later) language in which Unix was implemented and indeed,
C was
developed specifically to implement Unix.
Since an operating system directly
interfaces with hardware - the C language grew to have features
to aid Unix implementers.
*With that said, consider this WARNING: the ordering of bits in a bit
field is not guaranteed to be the same on different platforms and even
between different compilers on the same platform.*
Bit fields are implemented within a `struct` by appending a colon plus
a number after the declaration of integer types.
For example:
```c
struct BF {
unsigned char a : 1;
unsigned char b : 2;
unsigned char c : 5;
};
```
The above declares a `struct` whose size is 1 byte. Members of the `struct`
are `a`, `b` and `c` which are 1, 2 and 5 bits in size, respectively.
## Bit Fields Aren't Just For Hardware
Consider a data structure for which there will be potentially millions of
instances in RAM. Or, perhaps billions of instances on disc. Suppose you
need 8 boolean members in every instance. The C++ standard does not
define the size of a `bool` instead leaving it to be implementation
dependent. Some implementations equate `bool` to `int`, four bytes in
length. Some implement `bool` with a `char`, or 1 byte in length.
Let's assume the smallest case and equate a `bool` with `char`. Our
`struct`, for which there may be millions or billions of instances
requires 8 `bool` so therefore 8 bytes. Times millions or billions.
Ouch.
Bit fields can come to your aid here by using a single bit per boolean
value. In the best case, 8 bytes collapse to 1 byte. In a worse case,
8 x 4 = 32 bytes collapsed into 1.
## Without Bit Fields
Let's assume we're working with a byte that is comprised of three
fields layed out as in `struct BF` above. That is, a one, two and
five bit field inside one byte.
Without bit fields, we would have to write this code:
```c
/* Note the absence of defensive programming such as checking
to ensure that byte is not null and that bit_number is not
too large.
*/
void ClearA(unsigned char * byte) {
*byte &= ~1;
}
void SetA(unsigned char * byte) {
*byte &= ~1;
*byte |= 1;
}
void ClearB(unsigned char * byte) {
*byte &= ~6;
}
void SetB(unsigned char * byte, unsigned char value) {
value &= 3; // ensures only bits 0 and 1 can be set
*byte &= ~6; // clears bits 1 and 2 in byte
*byte |= (value << 1); // stores bits 0 and 1 into bits 2 and 3
}
void ClearC(unsigned char * byte) {
*byte &= 7; // squashes bits 3 to 7 to 0
}
void SetC(unsigned char * byte, unsigned char value) {
value &= 0x1F; // ensures only bits 0 to 4 can be set
*byte &= ~(0x1F << 3); // squashes correct bits in byte
*byte |= (value << 3); // or's in the bits at the right place
}
```
In naive assembly language, these functions would look like this:
```asm
```

View file

@ -0,0 +1,65 @@
.global SetA
.global SetB
.global SetC
.global ClearA
.global ClearB
.global ClearC
.text
.align 2
ClearA: ldrb w1, [x0]
mov w2, 1
mvn w2, w2
and w1, w1, w2
strb w1, [x0]
ret
ClearB: ldrb w1, [x0]
mov w2, 6
mvn w2, w2
and w1, w1, w2
strb w1, [x0]
ret
ClearC: ldrb w1, [x0]
and w1, w1, 7
strb w1, [x0]
ret
SetA: ldrb w1, [x0]
mov w2, 1
mvn w2, w2
and w1, w1, w2
orr w1, w1, 1
strb w1, [x0]
ret
SetB: ldrb w3, [x0]
and w1, w1, 3 // value &= 3
lsl w1, w1, 1
mov w2, 6
mvn w2, w2
and w3, w3, w2 // B is cleared
orr w3, w3, w1
strb w3, [x0]
ret
SetC: ldrb w3, [x0]
mov w2, 0x1F
and w1, w1, w2
lsl w1, w1, 3
lsl w2, w2, 3
mvn w2, w2
and w3, w3, w2
orr w3, w3, w1
strb w3, [x0]
ret
.end

View file

@ -0,0 +1,69 @@
#include <stdio.h>
struct BF {
unsigned char a : 1;
unsigned char b : 2;
unsigned char c : 5;
};
unsigned char noBF = 0;
#undef C
#ifdef C
/* Note the absence of defensive programming such as checking
to ensure that byte is not null and that bit_number is not
too large.
*/
void ClearA(unsigned char * byte) {
*byte &= ~1;
}
void SetA(unsigned char * byte) {
*byte &= ~1;
*byte |= 1;
}
void ClearB(unsigned char * byte) {
*byte &= ~6;
}
void SetB(unsigned char * byte, unsigned char value) {
value &= 3; // ensures only bits 0 and 1 can be set
*byte &= ~6; // clears bits 1 and 2 in byte
*byte |= (value << 1); // stores bits 0 and 1 into bits 2 and 3
}
void ClearC(unsigned char * byte) {
*byte &= 7; // squashes bits 3 to 7 to 0
}
void SetC(unsigned char * byte, unsigned char value) {
value &= 0x1F; // ensures only bits 0 to 4 can be set
*byte &= ~(0x1F << 3); // squashes correct bits in byte
*byte |= (value << 3); // or's in the bits at the right place
}
#else
extern void SetA(unsigned char *);
extern void SetB(unsigned char *, unsigned char);
extern void SetC(unsigned char *, unsigned char);
extern void ClearA(unsigned char *);
extern void ClearB(unsigned char *);
extern void ClearC(unsigned char *);
#endif
int main() {
volatile struct BF bf;
bf.a = 1;
bf.b = 2;
bf.c = 3;
ClearA(&noBF);
SetB(&noBF, 1);
SetC(&noBF, 7);
printf("noBF should be 0x3A - value: 0x%X\n", (unsigned int) noBF);
printf("bf should be 0x1D - value: 0x%X\n", (unsigned int) *((unsigned char *) &bf));
return 0;
}

164
section_2/bitfields/test.s Normal file
View file

@ -0,0 +1,164 @@
.arch armv8-a
.file "test.c"
.text
.align 2
.p2align 3,,7
.global Foo
.type Foo, %function
Foo:
.LFB23:
.cfi_startproc
and w0, w0, 255
mvn w0, w0
ret
.cfi_endproc
.LFE23:
.size Foo, .-Foo
.align 2
.p2align 3,,7
.global ClearA
.type ClearA, %function
ClearA:
.LFB24:
.cfi_startproc
ldrb w1, [x0]
and w1, w1, -2
strb w1, [x0]
ret
.cfi_endproc
.LFE24:
.size ClearA, .-ClearA
.align 2
.p2align 3,,7
.global SetA
.type SetA, %function
SetA:
.LFB25:
.cfi_startproc
ldrb w1, [x0]
orr w1, w1, 1
strb w1, [x0]
ret
.cfi_endproc
.LFE25:
.size SetA, .-SetA
.align 2
.p2align 3,,7
.global ClearB
.type ClearB, %function
ClearB:
.LFB26:
.cfi_startproc
ldrb w1, [x0]
and w1, w1, -7
strb w1, [x0]
ret
.cfi_endproc
.LFE26:
.size ClearB, .-ClearB
.align 2
.p2align 3,,7
.global SetB
.type SetB, %function
SetB:
.LFB27:
.cfi_startproc
ldrb w2, [x0]
bfi w2, w1, 1, 2
strb w2, [x0]
ret
.cfi_endproc
.LFE27:
.size SetB, .-SetB
.align 2
.p2align 3,,7
.global ClearC
.type ClearC, %function
ClearC:
.LFB28:
.cfi_startproc
ldrb w1, [x0]
and w1, w1, 7
strb w1, [x0]
ret
.cfi_endproc
.LFE28:
.size ClearC, .-ClearC
.align 2
.p2align 3,,7
.global SetC
.type SetC, %function
SetC:
.LFB29:
.cfi_startproc
ldrb w2, [x0]
ubfiz w1, w1, 3, 5
and w2, w2, 7
orr w2, w2, w1
strb w2, [x0]
ret
.cfi_endproc
.LFE29:
.size SetC, .-SetC
.section .rodata.str1.8,"aMS",@progbits,1
.align 3
.LC0:
.string "noBF should be 0x3A - value: 0x%X\n"
.align 3
.LC1:
.string "bf should be 0x1D - value: 0x%X\n"
.section .text.startup,"ax",@progbits
.align 2
.p2align 3,,7
.global main
.type main, %function
main:
.LFB30:
.cfi_startproc
stp x29, x30, [sp, -32]!
.cfi_def_cfa_offset 32
.cfi_offset 29, -32
.cfi_offset 30, -24
mov w5, 2
mov w4, 3
mov x29, sp
ldrb w0, [sp, 24]
adrp x2, .LANCHOR0
adrp x1, .LC0
add x1, x1, :lo12:.LC0
orr w0, w0, 1
strb w0, [sp, 24]
mov w0, 58
strb w0, [x2, #:lo12:.LANCHOR0]
ldrb w3, [sp, 24]
mov w2, 58
mov w0, 1
bfi w3, w5, 1, 2
strb w3, [sp, 24]
ldrb w3, [sp, 24]
bfi w3, w4, 3, 5
strb w3, [sp, 24]
bl __printf_chk
ldrb w2, [sp, 24]
adrp x1, .LC1
mov w0, 1
add x1, x1, :lo12:.LC1
bl __printf_chk
mov w0, 0
ldp x29, x30, [sp], 32
.cfi_restore 30
.cfi_restore 29
.cfi_def_cfa_offset 0
ret
.cfi_endproc
.LFE30:
.size main, .-main
.global noBF
.bss
.set .LANCHOR0,. + 0
.type noBF, %object
.size noBF, 1
noBF:
.zero 1
.ident "GCC: (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0"
.section .note.GNU-stack,"",@progbits