Carry flag

Uses

The carry flag is affected by the result of most arithmetic (and typically several bitwise) instructions and is also used as an input to many of them. Several of these instructions have two forms which either read or ignore the carry. In assembly languages these instructions are represented by mnemonics such as ADD/SUB, ADC/SBC (ADD/SUB including carry), SHL/SHR (bit shifts), ROL/ROR (bit rotates), RCR/RCL (rotate through carry), and so on. The use of the carry flag in this manner enables multi-word add, subtract, shift, and rotate operations.

An example is what happens if one were to add 255 and 255 using 8-bit registers. The result should be 510 which is the 9-bit value 111111110 in binary. The 8 least significant bits always stored in the register would be 11111110 binary (254 decimal) but since there is carry out of bit 7 (the eight bit), the carry is set, indicating that the result needs 9 bits. The valid 9-bit result is the concatenation of the carry flag with the result.

For x86 ALU size of 8 bits, an 8-bit two's complement interpretation, the addition operation 11111111 + 11111111 results in 111111110, Carry_Flag set, Sign_Flag set, and Overflow_Flag clear.

If 11111111 represents two's complement signed integer −1 (ADD al,-1), then the interpretation of the result is -2 because Overflow_Flag is clear, and Carry_Flag is ignored. The sign of the result is negative, because Sign_Flag is set. 11111110 is the two's complement form of signed integer −2.

If 11111111 represents unsigned integer binary number 255 (ADD al,255), then the interpretation of the result would be 254, which is not correct, because the most significant bit of the result went into the Carry_Flag, which therefore cannot be ignored. The Overflow_Flag and the Sign_Flag are ignored.

Another example may be an 8-bit register with the bit pattern 01010101 and the carry flag set; if we execute a rotate left through carry instruction, the result would be 10101011 with the carry flag cleared because the most significant bit (bit 7) was rotated into the carry while the carry was rotated into the least significant bit (bit 0).

The early microprocessors Intel 4004 and Intel 8008 had specific instructions to set as well as reset the carry flag explicitly. However, the later Intel 8080 (and Z80) did not include an explicit reset carry opcode as this could be done equally fast via one of the bitwise AND, OR or XOR instructions (which do not use the carry flag).

The carry flag is also often used following comparison instructions, which are typically implemented by subtractive operations, to allow a decision to be made about which of the two compared values is lower than (or greater or equal to) the other. Branch instructions which examine the carry flag are often represented by mnemonics such as BCC and BCS to branch if the carry is clear, or branch if the carry is set respectively. When used in this way the carry flag provides a mechanism for comparing the values as unsigned integers. This is in contrast to the overflow flag which provides a mechanism for comparing the values as signed integer values.

Vs. borrow flag

While the carry flag is well-defined for addition, there are two ways in common use to use the carry flag for subtraction operations.

The first uses the bit as a borrow flag, setting it if a

The second uses the identity that −x = (not x)+1 directly (i.e. without storing the carry bit inverted) and computes a−b as a+(not b)+1. The carry flag is set according to this addition, and subtract with carry computes a+not(b)+C, while subtract without carry acts as if the carry bit were set. The result is that the carry bit is set if a≥b, and clear if a{{cite journal

The processors listed above, which include the most popular microprocessors of the last few decades, call these operations "subtract with borrow" and "subtract with carry", respectively, but the nomenclature is far from consistent. The VAX, NS320xx, Fairchild Clipper and Atmel AVR architectures use the borrow bit convention, but call their a−b−C operation "subtract with carry" (SBWC, SUBC, SUBWC and SBC). The PA-RISC and PICmicro architectures use the carry bit convention, but call their a+not(b)+C operation "subtract with borrow" (SUBB and SUBWFB). Still others, such as the H8, call theirs "subtract extended" (SUBX). SPARC uses the borrow convention, the SUBX mnemonic, and the "subtract with carry" name.

The Motorola 6809 uses the borrow bit convention and both nomenclatures, calling the operation "subtract with borrow", but assigning it the mnemonic abbreviation SBC.{{cite book |archive-url=https://web.archive.org/web/20250123202438/https://www.maddes.net/m6809pm/appendix_a.htm#page=109 |archive-date=2025-01-23 |url-status=bot: unknown |access-date=2025-01-03

The ST6 8-bit microcontrollers go both ways in a different sense. Although they do not have any sort of "subtract with carry" instruction, they do have a carry bit which is set by a subtract instruction, and the convention depends on the processor model. The ST60 processor uses the "carry" convention, while the ST62 and ST63 processors use the "borrow" convention. To summarize, and support the assertion "most common", here's a chart of processor families:

Convention Name "Borrow" "Carry" Naming unclear Borrow Intel 8051 VAX MC6809 8080/Z80/x86 NS32k ST62/ST63 Motorola 6800 AVR H8 calls it SUBX = subtract extended Motorola 68000 Fairchild Clipper NEC μPD78C1xA STM8 PDP/11 has a borrow bit, but no subtract using it. 78K/RL78 SPARC?

Carry PA-RISC S/360 ST60 PICmicro 6502 ARM COP8 MSP430 POWER/PowerPC 88000 Am29000 i960

The borrow convention gets called "subtract with carry" a lot, but the "borrow" terminology is used by the 600-pound gorilla of x86, and also the historically important Z80 and 68k families. The most popular architecture calling it "carry" is the AVR of Arduino fame. (Although VAX and SPARC are historically important.)

The carry convention is pretty consistently called "carry" (including by Intel and Motorola), with ARM, POWER, and S/360 all being current, and the 6502 being historically important. The most popular architecture calling it "borrow" is PICmicro series.

References

References

1. The well known Z80 and 68000 are examples of such CPUs, among many others.
2. "Intel Architecture Software Developer's Manual, Volume 2: Instruction Set Reference Manual".
3. "IBM System/360 Principles of Operation".
4. (October 2004). "ST6 Family Programming Manual". [[STMicroelectronics]].