5.40 Comparisons of an unpacked struct, a __packed struct, and a struct with individually __packed fields, and of a __packed struct and a #pragma packed struct

These comparisons illustrate the differences between the methods of packing structures.

Comparison of an unpacked struct, a __packed struct, and a struct with individually __packed fields

The differences between not packing a struct, packing an entire struct, and packing individual fields of a struct are illustrated by the three implementations of a struct shown in the following table.

Table 5-10 C code for an unpacked struct, a packed struct, and a struct with individually packed fields

Unpacked struct __packed struct __packed fields
struct foo
{
    char one;
    short two;
    char three;
    int four;
} c;
__packed struct foo
{
    char one;
    short two;
    char three;
    int four;
} c;
struct foo
{
    char one;
    __packed short two;
    char three;
    int four;
} c;

In the first implementation, the struct is not packed. In the second implementation, the entire structure is qualified as __packed. In the third implementation, the __packed attribute is removed from the structure and the individual field that is not naturally aligned is declared as __packed.

The following table shows the corresponding disassembly of the machine code produced by the compiler for each of the sample implementations of the preceding table, where the C code for each implementation has been compiled using the option -O2.

Table 5-11 Disassembly for an unpacked struct, a packed struct, and a struct with individually packed fields

Unpacked struct __packed struct __packed fields
; r0 contains address of c
; char one
LDRB    r1, [r0, #0]
; short two
LDRSH   r2, [r0, #2]
; char three
LDRB    r3, [r0, #4]
; int four
LDR     r12, [r0, #8]
; r0 contains address of c
; char one
LDRB  r1, [r0, #0]
; short two
LDRB  r2, [r0, #1]
LDRSB r12, [r0, #2]
ORR   r2, r12, r2, LSL #8
; char three
LDRB  r3, [r0, #3]
; int four
ADD   r0, r0, #4
BL    __aeabi_uread4
; r0 contains address of c
; char one
LDRB  r1, [r0, #0]
; short two
LDRB  r2, [r0, #1]
LDRSB r12, [r0, #2]
ORR   r2, r12, r2, LSL #8
; char three
LDRB  r3, [r0, #3]
; int four
LDR   r12, [r0, #4]

Note:

The -Ospace and -Otime compiler options control whether accesses to unaligned elements are made inline or through a function call. Using -Otime results in inline unaligned accesses. Using -Ospace results in unaligned accesses made through function calls.

In the disassembly of the unpacked struct example above, the compiler always accesses data on aligned word or halfword addresses. The compiler is able to do this because the struct is padded so that every member of the struct lies on its natural size boundary.

In the disassembly of the __packed struct example above, fields one and three are aligned on their natural size boundaries by default, so the compiler makes aligned accesses. The compiler always carries out aligned word or halfword accesses for fields it can identify as being aligned. For the unaligned field two, the compiler uses multiple aligned memory accesses (LDR/STR/LDM/STM), combined with fixed shifting and masking, to access the correct bytes in memory. The compiler calls the ARM Embedded Application Binary Interface (AEABI) runtime routine __aeabi_uread4 for reading an unsigned word at an unknown alignment to access field four because it is not able to determine that the field lies on its natural size boundary.

In the disassembly of the struct with individually packed fields example above, fields one, two, and three are accessed in the same way as in the case where the entire struct is qualified as __packed. In contrast to the situation where the entire struct is packed, however, the compiler makes a word-aligned access to the field four. This is because the presence of the __packed short within the structure helps the compiler to determine that the field four lies on its natural size boundary.

Comparison of a __packed struct and a #pragma packed struct

The differences between a __packed struct and a #pragma packed struct are illustrated by the two implementations of a struct shown in the following table.

Table 5-12 C code for a packed struct and a pragma packed struct

__packed struct #pragma packed struct
__packed struct foobar
{
    char x;
    short y[10];
};
short get_y0(struct foobar *s)
{
    // Unaligned-capable load
    return *s->y;
}
short *get_y(struct foobar *s)
{
    return s->y;    // Compile error
}
#pragma push
#pragma pack(1)
struct foobar
{
    char x;
    short y[10];
};
#pragma pop
short get_y0(struct foobar *s)
{
    // Unaligned-capable load
    return *s->y;
}
short *get_y(struct foobar *s)
{
    return s->y;    // No error
    // Potentially illegal unaligned load,
    // depending on use of result
}

In the first implementation, taking the address of a field in a __packed struct or a __packed field in a struct yields a __packed pointer, and the compiler generates a type error if you try to implicitly cast this to a non-__packed pointer. In the second implementation, in contrast, taking the address of a field in a #pragma packed struct does not yield a __packed-qualified pointer. However, the field might not be properly aligned for its type, and dereferencing such an unaligned pointer results in Undefined behavior.

Non-ConfidentialPDF file icon PDF versionARM DUI0472M
Copyright © 2010-2016 ARM Limited or its affiliates. All rights reserved.