Library:
bfix.hpp
A Generic API for Bit Manipulation in C/C++
- insert(bfi()) or extract(bfx()) bit fields from an unsigned char array of arbitrary length
- can handle 32 or 64 bit machines
- big endian, little endian or run time checking
- a C++ header file only library
- simple library, two routines bfi()(insert bit field) and bfx()(extract bit field)
- no library dependencies(except standard header files)
- offset bit numbers are from 1(start of array) to unlimited
- verification tests that show code usage examples, see bfix_test.cpp
- see cut/paste Examples section and API(bfi(), bfx()) section below
- a bfix-compile.tar file is provided with a non-header compiled(clang, clang++, gcc, g++) version
- compiles with clang++/g++
- see article in Embedded Systems Programming, Jul. 1999,
"A Generic API for Bit Manipulation in C"(included)
- this file is the entire documentation needed to use the library
- Author: Richard Hogaboom, richard.hogaboom@gmail.com
Usage:
see Examples: below
Files:
bfix-compile.tar - non-header compiled(clang, clang++, gcc, g++) version
bfix.doc - documentation for library
bfix.hpp - header only C++ implementation
bfix.mk - build script, compiles/executes test code
bfix_test.cpp - test code
Examples:
/* 16 bytes unsigned char array plus 7 bytes pad(3 needed for 32 bit systems and 7 needed for 64 bit systems) */
unsigned char c[16+7];
bfi(c, 1, 1, 1, 0);
bfi(c, 5, 2, 3, 0);
bfi(c, 17, 4, 15, 0);
bfi(c, 97, 32, 0xffffffff, 0);
bfi(c, 49, 8, 0xee, 0);
bfi(c, 8, 8, 0xff, 0);
bitfield1 = bfx(c, 1, 1, 0);
bitfield2 = bfx(c, 5, 2, 0);
bitfield3 = bfx(c, 17, 4, 0);
bitfield4 = bfx(c, 113, 16, 0);
bitfield5 = bfx(c, 49, 8, 0);
API:
/*
* File: bfix.hpp
*
* Based on article in Embedded Systems Programming, Jul. 1999,
* "A Generic API for Bit Manipulation in C" by Richard Hogaboom
*
* Description:
*
* bfi()/bfx() are used to insert and extract bit fields from an arbitrary length array
* of unsigned chars pointed to by an unsigned char* pointer.
*
* Functions:
*
* int
* bfi
* (
* unsigned char *cptr,
* unsigned long bit_offset,
* unsigned long bit_len,
* long value,
* unsigned int endian
* )
* e.g. int return = bfi(cptr, bit_offset, bit_len, value, endian);
*
*
* long
* bfx
* (
* const unsigned char *cptr,
* unsigned long bit_offset,
* unsigned long bit_len,
* unsigned int endian
* )
* e.g. long value = bfx(cptr, bit_offset, bit_len, endian);
*
* Error Handling:
*
* 1. Exceptions - None
* 2. Debugging - const int DEBUG = true for debugging output
* 3. Returns
*
* bfi():
* bit_offset < 1 - error, return -1
* bit_len < 1 - error, return -2
* bit_len > too long - error, return -3
* endian not 0-2 - error, return -4
* return 0 - success
*
* bfx():
* bit_offset < 1 - error, return -1
* bit_len < 1 - error, return -2
* bit_len > too long - error, return -3
* endian not 0-2 - error, return -4
* return value - success
*
* Notes:
* 1. in the following notes any annotation of the form n/m means n for 32 bit systems and
* m for 64 bit systems. operation on 32 or 64 bit systems should be transparent.
*
* 2. bit_len should be <=32/64 to 25/57. it depends on the value of bit_offset. the method
* always uses a memmove() of 4/8 bytes to a long temporary storage in which logical
* operations can be performed. this means that the bit_len can be at most 4/8 bytes,
* but in a case in which the start bit is not at the beginning of a byte, then the
* bit_len can only extend until the end of the 4/8'th byte. if the start bit is the
* last bit of a byte this will limit bit_len to 25/57 bits - the last bit of the first
* byte plus the next 3/7 bytes.
*
* 3. 4(32 bit machines)/8(64 bit machines) bytes are always read from the unsigned char
* array, modified and then written back. this means that if you set the last bit of
* the array, then the next 3/7 bytes will be read and written back, thus seemingly
* overrunning the array. if the 4/8 bytes does not overrun the array then no bits
* beyond the end of the array will be changed. if the 4/8 bytes does overrun the
* array some provision must be made to deal with this possibility. the array could
* be padded by 3/7 extra bytes.
*
* 4. bit_offset+bit_len should not overrun the array.
*
* 5. value should not be too long to fit into the bit field. if it is, the high order bits
* in front of the low order bit_len bits will be truncated.
*
* 6. all bit_len bits will be set and no bit outside the bit field will be changed.
*
* 7. value may be negative and prefix 2's complement sign bits are truncated to fit into
* bit_len bits.
*
* 8. use the lscpu cmd to determine 32/64 bit and endianness:
* $ lscpu
* Architecture: x86_64
* CPU op-mode(s): 32-bit, 64-bit
* Byte Order: Little Endian
*
* 9. use lscpu to determine 32/64 bit machine and big or little endian
*
* 10. since bit_offset is unsigned long it cannot have negative values - however no
* check for too large is done
*
* Author: Richard Hogaboom
* richard.hogaboom@gmail.com
*
*/
/*
* Name:
* bfi()
*
* Purpose:
* extract bit field from an array of chars pointed to by an unsigned char* pointer
*
* Usage:
* unsigned char *cptr
* unsigned long bit_offset
* unsigned long bit_len
* long value
* unsigned int endian
*
* int return = bfi(cptr, bit_offset, bit_len, value, endian);
*
* Returns:
* bit_offset < 1 - error, return -1
* bit_len < 1 - error, return -2
* bit_len > too long - error, return -3
* endian not 0-2 - error, return -4
* return 0 - success
*
* Parameters:
* unsigned char *cptr - pointer to unsigned char array
* unsigned long bit_offset - bit offset(starting from 1) from the start of the char array
* unsigned long bit_len - bit length of field to insert
* long value - value to insert
* unsigned int endian - endian enum, 0(run time checking), 1(big endian), 2(little endian)
*
* Comments:
* 4(32 bit machines)/8(64 bit machines) bytes are always read from the unsigned char
* array, modified and then written back. this means that if you set the last bit of
* the array, then the next 3/7 bytes will be read and written back, thus seemingly
* overrunning the array. if the 4/8 bytes does not overrun the array then no bits
* beyond the end of the array will be changed. if the 4/8 bytes does overrun the
* array some provision must be made to deal with this possibility. the array could
* be padded by 3/7 extra bytes.
*/
inline int
bfi
(
unsigned char *cptr,
unsigned long bit_offset,
unsigned long bit_len,
long value,
unsigned int endian
)
/*
* Name:
* bfx()
*
* Purpose:
* extract bit field from an array of chars pointed to by an unsigned char* pointer
*
* Usage:
* const unsigned char *cptr
* unsigned long bit_offset
* unsigned long bit_len
* unsigned int endian
*
* long value = bfx(cptr, bit_offset, bit_len, endian);
*
* Returns:
* bit_offset < 1 - error, return -1
* bit_len < 1 - error, return -2
* bit_len > too long - error, return -3
* endian not 0-2 - error, return -4
* return value - success
*
* Parameters:
* const unsigned char *cptr - const pointer to unsigned char array
* unsigned long bit_offset - bit offset(starting from 1) from the start of the char array
* unsigned long bit_len - bit length of field to extract
* unsigned int endian - endian enum, 0(run time checking), 1(big endian), 2(little endian)
*
* Comments:
*/
inline long
bfx
(
const unsigned char *cptr,
unsigned long bit_offset,
unsigned long bit_len,
unsigned int endian
)
-
Notifications
You must be signed in to change notification settings - Fork 3
A Generic API for Bit Manipulation in C - C/C++ - a C++ header file only library
License
rahogaboom/bfix
Folders and files
Name | Name | Last commit message | Last commit date | |
---|---|---|---|---|
Repository files navigation
About
A Generic API for Bit Manipulation in C - C/C++ - a C++ header file only library
Topics
Resources
License
Stars
Watchers
Forks
Packages 0
No packages published