Vol. 1 7-19
PROGRAMMING WITH GENERAL-PURPOSE INSTRUCTIONS
this instruction are LODSB (load byte string), LODSW (load word string), and LODSD (load doubleword string). This
instruction is usually used in a loop, where other instructions process each element of the string after they are
loaded into the target register.
The STOS instruction stores the source string element from the EAX (doubleword string), AX (word string), or AL
(byte string) register into the memory location identified with the EDI register. The “short forms” for this instruc-
tion are STOSB (store byte string), STOSW (store word string), and STOSD (store doubleword string). This instruc-
tion is also normally used in a loop. Here a string is commonly loaded into the register with a LODS instruction,
operated on by other instructions, and then stored again in memory with a STOS instruction.
The I/O instructions (see Section 7.3.10, “I/O Instructions”) also perform operations on strings in memory.
7.3.9.2
Repeated String Operations
Each of the string instructions described in Section 7.3.9.1 perform one iteration of a string operation. To operate
on strings longer than a doubleword, the string instructions can be combined with a repeat prefix (REP) to create a
repeating instruction or be placed in a loop.
When used in string instructions, the ESI and EDI registers are automatically incremented or decremented after
each iteration of an instruction to point to the next element (byte, word, or doubleword) in the string. String oper-
ations can thus begin at higher addresses and work toward lower ones, or they can begin at lower addresses and
work toward higher ones. The DF flag in the EFLAGS register controls whether the registers are incremented (DF =
0) or decremented (DF = 1). The STD and CLD instructions set and clear this flag, respectively.
The following repeat prefixes can be used in conjunction with a count in the ECX register to cause a string instruc-
tion to repeat:
•
REP — Repeat while the ECX register not zero.
•
REPE/REPZ — Repeat while the ECX register not zero and the ZF flag is set.
•
REPNE/REPNZ — Repeat while the ECX register not zero and the ZF flag is clear.
When a string instruction has a repeat prefix, the operation executes until one of the termination conditions spec-
ified by the prefix is satisfied. The REPE/REPZ and REPNE/REPNZ prefixes are used only with the CMPS and SCAS
instructions. Also, note that a REP STOS instruction is the fastest way to initialize a large block of memory.
7.3.9.3
Fast-String Operation
To improve performance, more recent processors support modifications to the processor’s operation during the
string store operations initiated with the MOVS, MOVSB, STOS, and STOSB instructions. This optimized operation,
called fast-string operation, is used when the execution of one of those instructions meets certain initial condi-
tions (see below). Instructions using fast-string operation effectively operate on the string in groups that may
include multiple elements of the native data size (byte, word, doubleword, or quadword). With fast-string opera-
tion, the processor recognizes interrupts and data breakpoints only on boundaries between these groups. Fast-
string operation is used only if the source and destination addresses both use either the WB or WC memory types.
The initial conditions for fast-string operation are implementation-specific and may vary with the native string size.
Examples of parameters that may impact the use of fast-string operation include the following:
•
the alignment indicated in the EDI and ESI alignment registers;
•
the address order of the string operation;
•
the value of the initial operation counter (ECX); and
•
the difference between the source and destination addresses.
NOTE
Initial conditions for fast-string operation in future Intel 64 or IA-32 processor families may differ
from above. The Intel® 64 and IA-32 Architectures Optimization Reference Manual may contain
model-specific information.
Software can disable fast-string operation by clearing the fast-string-enable bit (bit 0) of IA32_MISC_ENABLE
MSR. However, Intel recommends that system software always enable fast-string operation.