\C{lang} The NASM Language \H{syntax} Layout of a NASM Source Line Like most assemblers, each NASM source line contains (unless it is a macro, a preprocessor directive or an assembler directive: see \k{preproc} and \k{directive}) some combination of the four fields \c label: instruction operands ; comment As usual, most of these fields are optional; the presence or absence of any combination of a label, an instruction and a \i{comment} is allowed. Of course, the operand field is either required or forbidden by the presence and nature of the instruction field. NASM uses backslash (\\) as the line continuation character; if a line ends with backslash, the next line is considered to be a part of the backslash-ended line. NASM places no restrictions on white space within a line: labels may have white space before them, or instructions may have no space before them, or anything. The \i{colon} after a label is also optional. (Note that this means that if you intend to code \c{lodsb} alone on a line, and type \c{lodab} by accident, then that's still a valid source line which does nothing but define a label. Running NASM with the command-line option \I{label-orphan}\c{-w+orphan-labels} will cause it to warn you if you define a label alone on a line without a \i{trailing colon}.) \i{Valid characters} in labels are letters, numbers, \c{_}, \c{$}, \c{#}, \c{@}, \c{~}, \c{.}, and \c{?}. The only characters which may be used as the \e{first} character of an identifier are letters, \c{.} (with special meaning: see \k{locallab}), \c{_} and \c{?}. An identifier may also be prefixed with a \I{$, prefix}\c{$} to indicate that it is intended to be read as an identifier and not a reserved word; thus, if some other module you are linking with defines a symbol called \c{eax}, you can refer to \c{$eax} in NASM code to distinguish the symbol from the register. Maximum length of an identifier is 4095 characters. The instruction field may contain any machine instruction: Pentium and P6 instructions, FPU instructions, MMX instructions and even undocumented instructions are all supported. The instruction may be prefixed by \c{LOCK}, \c{REP}, \c{REPE}/\c{REPZ}, \c{REPNE}/\c{REPNZ}, \c{XACQUIRE}/\c{XRELEASE} or \c{BND}/\c{NOBND}, in the usual way. Explicit \I{address-size prefixes}address-size and \i{operand-size prefixes} \i\c{A16}, \i\c{A32}, \i\c{A64}, \i\c{O16} and \i\c{O32}, \i\c{O64} are provided - one example of their use is given in \k{mixsize}. You can also use the name of a \I{segment override}segment register as an instruction prefix: coding \c{es mov [bx],ax} is equivalent to coding \c{mov [es:bx],ax}. We recommend the latter syntax, since it is consistent with other syntactic features of the language, but for instructions such as \c{LODSB}, which has no operands and yet can require a segment override, there is no clean syntactic way to proceed apart from \c{es lodsb}. An instruction is not required to use a prefix: prefixes such as \c{CS}, \c{A32}, \c{LOCK} or \c{REPE} can appear on a line by themselves, and NASM will just generate the prefix bytes. In addition to actual machine instructions, NASM also supports a number of pseudo-instructions, described in \k{pseudop}. Instruction \i{operands} may take a number of forms: they can be registers, described simply by the register name (e.g. \c{ax}, \c{bp}, \c{ebx}, \c{cr0}: NASM does not use the \c{gas}-style syntax in which register names must be prefixed by a \c{%} sign), or they can be \i{effective addresses} (see \k{effaddr}), constants (\k{const}) or expressions (\k{expr}). For x87 \i{floating-point} instructions, NASM accepts a wide range of syntaxes: you can use two-operand forms like MASM supports, or you can use NASM's native single-operand forms in most cases. \# Details of \# all forms of each supported instruction are given in \# \k{iref}. For example, you can code: \c fadd st1 ; this sets st0 := st0 + st1 \c fadd st0,st1 ; so does this \c \c fadd st1,st0 ; this sets st1 := st1 + st0 \c fadd to st1 ; so does this Almost any x87 floating-point instruction that references memory must use one of the prefixes \i\c{DWORD}, \i\c{QWORD} or \i\c{TWORD} to indicate what size of \i{memory operand} it refers to. \H{pseudop} \i{Pseudo-Instructions} Pseudo-instructions are things which, though not real x86 machine instructions, are used in the instruction field anyway because that's the most convenient place to put them. The current pseudo-instructions are \i\c{DB}, \i\c{DW}, \i\c{DD}, \i\c{DQ}, \i\c{DT}, \i\c{DO}, \i\c{DY} and \i\c\{DZ}; their \I{storage, uninitialized}\i{uninitialized} counterparts \i\c{RESB}, \i\c{RESW}, \i\c{RESD}, \i\c{RESQ}, \i\c{REST}, \i\c{RESO}, \i\c{RESY} and \i\c\{RESZ}; the \i\c{INCBIN} command, the \i\c{EQU} command, and the \i\c{TIMES} prefix. In this documentation, the notation "\c{D}\e{x}" and "\c{RES}\e{x}" is used to indicate all the \c{DB} and \c{RESB} type directives, respectively. \S{db} \c{D}\e{x}: Declaring Initialized Data \i\c{DB}, \i\c{DW}, \i\c{DD}, \i\c{DQ}, \i\c{DT}, \i\c{DO}, \i\c{DY} and \i\c{DZ} (collectively "\c{D}\e{x}" in this documentation) are used, much as in MASM, to declare initialized data in the output file. They can be invoked in a wide range of ways: \I{floating-point constant}\I{character constant}\I{string constant} \c db 0x55 ; just the byte 0x55 \c db 0x55,0x56,0x57 ; three bytes in succession \c db 'a',0x55 ; character constants are OK \c db 'hello',13,10,'$' ; so are string constants \c dw 0x1234 ; 0x34 0x12 \c dw 'a' ; 0x61 0x00 (it's just a number) \c dw 'ab' ; 0x61 0x62 (character constant) \c dw 'abc' ; 0x61 0x62 0x63 0x00 (string) \c dd 0x12345678 ; 0x78 0x56 0x34 0x12 \c dd 1.234567e20 ; floating-point constant \c dq 0x123456789abcdef0 ; eight byte constant \c dq 1.234567e20 ; double-precision float \c dt 1.234567e20 ; extended-precision float \c{DT}, \c{DO}, \c{DY} and \c{DZ} do not accept integer \i{numeric constants} as operands. \I{masmdb} Starting in NASM 2.15, a the following \i{MASM}-like features have been implemented: \b A \I{?db}\c{?} argument to declare \i{uninitialized storage}: \c db ? ; uninitialized \b A superset of the \i\c{DUP} syntax. The NASM version of this has the following syntax specification; capital letters indicate literal keywords: \c dx := DB | DW | DD | DQ | DT | DO | DY | DZ \c type := BYTE | WORD | DWORD | QWORD | TWORD | OWORD | YWORD | ZWORD \c atom := expression | string | float | '?' \c parlist := '(' value [',' value ...] ')' \c duplist := expression DUP [type] ['%'] parlist \c list := duplist | '%' parlist | type ['%'] parlist \c value := [type] atom | list \c \c stmt := dx value [',' value ...] \> Note that a \e{list} needs to be prefixed with a \I{%db}\c{%} sign unless prefixed by either \c{DUP} or a \e{type} in order to avoid confusing it with a parenthesis starting an expression. The following expressions are all valid: \c db 33 \c db (44) ; Integer expression \c ; db (44,55) ; Invalid - error \c db %(44,55) \c db %('XX','YY') \c db ('AA') ; Integer expression - outputs single byte \c db %('BB') ; List, containing a string \c db ? \c db 6 dup (33) \c db 6 dup (33, 34) \c db 6 dup (33, 34), 35 \c db 7 dup (99) \c db 7 dup dword (?, word ?, ?) \c dw byte (?,44) \c dw 3 dup (0xcc, 4 dup byte ('PQR'), ?), 0xabcd \c dd 16 dup (0xaaaa, ?, 0xbbbbbb) \c dd 64 dup (?) \I{baddb} The use of \c{$} (current address) in a \c{D}\e{x} statement is undefined in the current version of NASM, \e{except in the following cases}: \b For the first expression in the statement, either a \c{DUP} or a data item. \b An expression of the form "\e{value}\c{ - $}", which is converted to a self-relative relocation. Future versions of NASM is likely to produce a different result or issue an error this case. There is no such restriction on using \c{$$} or section-relative symbols. \S{resb} \c{RESB} and Friends: Declaring \i{Uninitialized} Data \i\c{RESB}, \i\c{RESW}, \i\c{RESD}, \i\c{RESQ}, \i\c{REST}, \i\c{RESO}, \i\c{RESY} and \i\c\{RESZ} are designed to be used in the BSS section of a module: they declare \e{uninitialized} storage space. Each takes a single operand, which is the number of bytes, words, doublewords or whatever to reserve. The operand to a \c{RESB}-type pseudo-instruction \e{would} be a \i\e{critical expression} (see \k{crit}), except that for legacy compatibility reasons forward references are permitted, however \e{the code will be extremely fragile and this should be considered a severe programming error.} A warning will be issued; code generating this warning should be remedied as quickly as possible (see the \c{forward} class in \k{warnings}.) For example: \c buffer: resb 64 ; reserve 64 bytes \c wordvar: resw 1 ; reserve a word \c realarray resq 10 ; array of ten reals \c ymmval: resy 1 ; one YMM register \c zmmvals: resz 32 ; 32 ZMM registers \I{masmdb} Since NASM 2.15, the MASM syntax of using \I{?db}\c{?} and \i\c{DUP} in the \c{D}\e{x} directives is also supported. Thus, the above example could also be written: \c buffer: db 64 dup (?) ; reserve 64 bytes \c wordvar: dw ? ; reserve a word \c realarray dq 10 dup (?) ; array of ten reals \c ymmval: dy ? ; one YMM register \c zmmvals: dz 32 dup (?) ; 32 ZMM registers \S{incbin} \i\c{INCBIN}: Including External \i{Binary Files} \c{INCBIN} includes binary file data verbatim into the output file. This can be handy for (for example) including \i{graphics} and \i{sound} data directly into a game executable file. It can be called in one of these three ways: \c incbin "file.dat" ; include the whole file \c incbin "file.dat",1024 ; skip the first 1024 bytes \c incbin "file.dat",1024,512 ; skip the first 1024, and \c ; actually include at most 512 \c{INCBIN} is both a directive and a standard macro; the standard macro version searches for the file in the include file search path and adds the file to the dependency lists. This macro can be overridden if desired. \S{equ} \i\c{EQU}: Defining Constants \c{EQU} defines a symbol to a given constant value: when \c{EQU} is used, the source line must contain a label. The action of \c{EQU} is to define the given label name to the value of its (only) operand. This definition is absolute, and cannot change later. So, for example, \c message db 'hello, world' \c msglen equ $-message defines \c{msglen} to be the constant 12. \c{msglen} may not then be redefined later. This is not a \i{preprocessor} definition either: the value of \c{msglen} is evaluated \e{once}, using the value of \c{$} (see \k{expr} for an explanation of \c{$}) at the point of definition, rather than being evaluated wherever it is referenced and using the value of \c{$} at the point of reference. \S{times} \i\c{TIMES}: \i{Repeating} Instructions or Data The \c{TIMES} prefix causes the instruction to be assembled multiple times. This is partly present as NASM's equivalent of the \i\c{DUP} syntax supported by \i{MASM}-compatible assemblers, in that you can code \c zerobuf: times 64 db 0 or similar things; but \c{TIMES} is more versatile than that. The argument to \c{TIMES} is not just a numeric constant, but a numeric \e{expression}, so you can do things like \c buffer: db 'hello, world' \c times 64-$+buffer db ' ' which will store exactly enough spaces to make the total length of \c{buffer} up to 64. Finally, \c{TIMES} can be applied to ordinary instructions, so you can code trivial \i{unrolled loops} in it: \c times 100 movsb Note that there is no effective difference between \c{times 100 resb 1} and \c{resb 100}, except that the latter will be assembled about 100 times faster due to the internal structure of the assembler. The operand to \c{TIMES} is a critical expression (\k{crit}). Note also that \c{TIMES} can't be applied to \i{macros}: the reason for this is that \c{TIMES} is processed after the macro phase, which allows the argument to \c{TIMES} to contain expressions such as \c{64-$+buffer} as above. To repeat more than one line of code, or a complex macro, use the preprocessor \i\c{%rep} directive. \H{effaddr} Effective Addresses An \i{effective address} is any operand to an instruction which \I{memory reference}references memory. Effective addresses, in NASM, have a very simple syntax: they consist of an expression evaluating to the desired address, enclosed in \i{square brackets}. For example: \c wordvar dw 123 \c mov ax,[wordvar] \c mov ax,[wordvar+1] \c mov ax,[es:wordvar+bx] Anything not conforming to this simple system is not a valid memory reference in NASM, for example \c{es:wordvar[bx]}. More complicated effective addresses, such as those involving more than one register, work in exactly the same way: \c mov eax,[ebx*2+ecx+offset] \c mov ax,[bp+di+8] NASM is capable of doing \i{algebra} on these effective addresses, so that things which don't necessarily \e{look} legal are perfectly all right: \c mov eax,[ebx*5] ; assembles as [ebx*4+ebx] \c mov eax,[label1*2-label2] ; ie [label1+(label1-label2)] Some forms of effective address have more than one assembled form; in most such cases NASM will generate the smallest form it can. For example, there are distinct assembled forms for the 32-bit effective addresses \c{[eax*2+0]} and \c{[eax+eax]}, and NASM will generally generate the latter on the grounds that the former requires four bytes to store a zero offset. NASM has a hinting mechanism which will cause \c{[eax+ebx]} and \c{[ebx+eax]} to generate different opcodes; this is occasionally useful because \c{[esi+ebp]} and \c{[ebp+esi]} have different default segment registers. However, you can force NASM to generate an effective address in a particular form by the use of the keywords \c{BYTE}, \c{WORD}, \c{DWORD} and \c{NOSPLIT}. If you need \c{[eax+3]} to be assembled using a double-word offset field instead of the one byte NASM will normally generate, you can code \c{[dword eax+3]}. Similarly, you can force NASM to use a byte offset for a small value which it hasn't seen on the first pass (see \k{crit} for an example of such a code fragment) by using \c{[byte eax+offset]}. As special cases, \c{[byte eax]} will code \c{[eax+0]} with a byte offset of zero, and \c{[dword eax]} will code it with a double-word offset of zero. The normal form, \c{[eax]}, will be coded with no offset field. The form described in the previous paragraph is also useful if you are trying to access data in a 32-bit segment from within 16 bit code. For more information on this see the section on mixed-size addressing (\k{mixaddr}). In particular, if you need to access data with a known offset that is larger than will fit in a 16-bit value, if you don't specify that it is a dword offset, nasm will cause the high word of the offset to be lost. Similarly, NASM will split \c{[eax*2]} into \c{[eax+eax]} because that allows the offset field to be absent and space to be saved; in fact, it will also split \c{[eax*2+offset]} into \c{[eax+eax+offset]}. You can combat this behaviour by the use of the \c{NOSPLIT} keyword: \c{[nosplit eax*2]} will force \c{[eax*2+0]} to be generated literally. \c{[nosplit eax*1]} also has the same effect. In another way, a split EA form \c{[0, eax*2]} can be used, too. However, \c{NOSPLIT} in \c{[nosplit eax+eax]} will be ignored because user's intention here is considered as \c{[eax+eax]}. In 64-bit mode, NASM will by default generate absolute addresses. The \i\c{REL} keyword makes it produce \c{RIP}-relative addresses. Since this is frequently the normally desired behaviour, see the \c{DEFAULT} directive (\k{default}). The keyword \i\c{ABS} overrides \i\c{REL}. A new form of split effective address syntax is also supported. This is mainly intended for mib operands as used by MPX instructions, but can be used for any memory reference. The basic concept of this form is splitting base and index. \c mov eax,[ebx+8,ecx*4] ; ebx=base, ecx=index, 4=scale, 8=disp For mib operands, there are several ways of writing effective address depending on the tools. NASM supports all currently possible ways of mib syntax: \c ; bndstx \c ; next 5 lines are parsed same \c ; base=rax, index=rbx, scale=1, displacement=3 \c bndstx [rax+0x3,rbx], bnd0 ; NASM - split EA \c bndstx [rbx*1+rax+0x3], bnd0 ; GAS - '*1' indecates an index reg \c bndstx [rax+rbx+3], bnd0 ; GAS - without hints \c bndstx [rax+0x3], bnd0, rbx ; ICC-1 \c bndstx [rax+0x3], rbx, bnd0 ; ICC-2 When broadcasting decorator is used, the opsize keyword should match the size of each element. \c VDIVPS zmm4, zmm5, dword [rbx]{1to16} ; single-precision float \c VDIVPS zmm4, zmm5, zword [rbx] ; packed 512 bit memory \H{const} \i{Constants} NASM understands four different types of constant: numeric, character, string and floating-point. \S{numconst} \i{Numeric Constants} A numeric constant is simply a number. NASM allows you to specify numbers in a variety of number bases, in a variety of ways: you can suffix \c{H} or \c{X}, \c{D} or \c{T}, \c{Q} or \c{O}, and \c{B} or \c{Y} for \i{hexadecimal}, \i{decimal}, \i{octal} and \i{binary} respectively, or you can prefix \c{0h} or \c{0x}, \c{0d} or \c{0t}, \c{0q} or \c{0o}, and \c{0b} or \c{0y) in the style of C. Please note that unlike C, a \c{0} prefix by itself does \e{not} imply an octal constant (this is deprecated in C23.) Previous versions of NASM allowed prefixing \c{$} for hexadecimal in the style of Borland Pascal or Motorola Assemblers. Unfortunately though, the \I{$, prefix}\c{$} prefix does double duty as a prefix on identifiers (see \k{syntax}), so a hex number prefixed with a \c{$} sign would have to have a digit after the \c{$} rather than a letter, which is \e{not} what users would typically expect. This syntax is strongly deprecated, and can be disabled entirely with the \c{[DOLLARHEX]} directive, see \k{dollarhex}. Numeric constants can have underscores (\c{_}) interspersed to break up long strings. Some examples (all producing exactly the same code): \c mov ax,200 ; decimal \c mov ax,0200 ; still decimal \c mov ax,0200d ; explicitly decimal \c mov ax,0d200 ; also decimal \c mov ax,0c8h ; hex \c mov ax,0xc8 ; hex yet again \c mov ax,0hc8 ; still hex \c mov ax,310q ; octal \c mov ax,310o ; octal again \c mov ax,0o310 ; octal yet again \c mov ax,0q310 ; octal yet again \c mov ax,11001000b ; binary \c mov ax,1100_1000b ; same binary constant \c mov ax,1100_1000y ; same binary constant once more \c mov ax,0b1100_1000 ; same binary constant yet again \c mov ax,0y1100_1000 ; same binary constant yet again \c \c ; Deprecated syntax: \c mov ax,$0c8 ; hex again: the 0 is required \S{strings} \I{string}\I{string constants}\i{Character Strings} A character string consists of up to eight characters enclosed in either single quotes (\c{'...'}), double quotes (\c{"..."}) or backquotes (\c{`...`}). Single or double quotes are equivalent to NASM (except of course that surrounding the constant with single quotes allows double quotes to appear within it and vice versa); the contents of those are represented verbatim. Strings enclosed in backquotes support C-style \c{\\}-escapes for special characters. The following \i{escape sequences} are recognized by backquoted strings: \c \' single quote (') \c \" double quote (") \c \` backquote (`) \c \\\ backslash (\) \c \? question mark (?) \c \a BEL (ASCII 7) \c \b BS (ASCII 8) \c \t TAB (ASCII 9) \c \n LF (ASCII 10) \c \v VT (ASCII 11) \c \f FF (ASCII 12) \c \r CR (ASCII 13) \c \e ESC (ASCII 27) \c \377 Up to 3 octal digits - literal byte \c \xFF Up to 2 hexadecimal digits - literal byte \c \u1234 4 hexadecimal digits - Unicode character \c \U12345678 8 hexadecimal digits - Unicode character All other escape sequences are reserved. Note that \c{\\0}, meaning a \c{NUL} character (ASCII 0), is a special case of the octal escape sequence. \i{Unicode} characters specified with \c{\\u} or \c{\\U} are converted to \i{UTF-8}. For example, the following lines are all equivalent: \c db `\u263a` ; UTF-8 smiley face \c db `\xe2\x98\xba` ; UTF-8 smiley face \c db 0E2h, 098h, 0BAh ; UTF-8 smiley face \S{chrconst} \i{Character Constants} A character constant consists of a string up to eight bytes long, used in an expression context. It is treated as if it was an integer. A character constant with more than one byte will be arranged with \i{little-endian} order in mind: if you code \c mov eax,'abcd' then the constant generated is not \c{0x61626364}, but \c{0x64636261}, so that if you were then to store the value into memory, it would read \c{abcd} rather than \c{dcba}. This is also the sense of character constants understood by the Pentium's \i\c{CPUID} instruction. \S{strconst} \i{String Constants} String constants are character strings used in the context of some pseudo-instructions, namely the \I\c{DW}\I\c{DD}\I\c{DQ}\I\c{DT}\I\c{DO}\I\c{DY}\i\c{DB} family and \i\c{INCBIN} (where it represents a filename.) They are also used in certain preprocessor directives. A string constant looks like a character constant, only longer. It is treated as a concatenation of maximum-size character constants for the conditions. So the following are equivalent: \c db 'hello' ; string constant \c db 'h','e','l','l','o' ; equivalent character constants And the following are also equivalent: \c dd 'ninechars' ; doubleword string constant \c dd 'nine','char','s' ; becomes three doublewords \c db 'ninechars',0,0,0 ; and really looks like this Note that when used in a string-supporting context, quoted strings are treated as a string constants even if they are short enough to be a character constant, because otherwise \c{db 'ab'} would have the same effect as \c{db 'a'}, which would be silly. Similarly, three-character or four-character constants are treated as strings when they are operands to \c{DW}, and so forth. \S{unicode} \I{UTF-8}I{UTF-16}\I{UTF-32}\i{Unicode} Strings The special operators \i\c{__?utf16?__}, \i\c{__?utf16le?__}, \i\c{__?utf16be?__}, \i\c{__?utf32?__}, \i\c{__?utf32le?__} and \i\c{__?utf32be?__} allows definition of Unicode strings. They take a string in UTF-8 format and converts it to UTF-16 or UTF-32, respectively. Unless the \c{be} forms are specified, the output is littleendian. For example: \c %define u(x) __?utf16?__(x) \c %define w(x) __?utf32?__(x) \c \c dw u('C:\WINDOWS'), 0 ; Pathname in UTF-16 \c dd w(`A + B = \u206a`), 0 ; String in UTF-32 The UTF operators can be applied either to strings passed to the \c{DB} family instructions, or to character constants in an expression context. \S{fltconst} \i{Floating-Point Constants} \i{Floating-point} constants are acceptable only as arguments to \i\c{DB}, \i\c{DW}, \i\c{DD}, \i\c{DQ}, \i\c{DT}, and \i\c{DO}, or as arguments to the special operators \i\c{__?float8?__}, \i\c{__?float16?__}, \i\c{__?bfloat16?__}, \i\c{__?float32?__}, \i\c{__?float64?__}, \i\c{__?float80m?__}, \i\c{__?float80e?__}, \i\c{__?float128l?__}, and \i\c{__?float128h?__}. See also \k{pkg_fp}. Floating-point constants are expressed in the traditional form: digits, then a period, then optionally more digits, then optionally an \c{E} followed by an exponent. The period is mandatory, so that NASM can distinguish between \c{dd 1}, which declares an integer constant, and \c{dd 1.0} which declares a floating-point constant. NASM also support C99-style hexadecimal floating-point: \c{0x}, hexadecimal digits, period, optionally more hexadeximal digits, then optionally a \c{P} followed by a \e{binary} (not hexadecimal) exponent in decimal notation. As an extension, NASM additionally supports the \c{0h} and \c{$} prefixes for hexadecimal, as well binary and octal floating-point, using the \c{0b} or \c{0y} and \c{0o} or \c{0q} prefixes, respectively. As with integers, the \c{$} prefix for hexadecimal is deprecated. Underscores to break up groups of digits are permitted in floating-point constants as well. Some examples: \c db -0.2 ; "Quarter precision" \c dw -0.5 ; IEEE 754r/SSE5 half precision \c dd 1.2 ; an easy one \c dd 1.222_222_222 ; underscores are permitted \c dd 0x1p+2 ; 1.0x2^2 = 4.0 \c dq 0x1p+32 ; 1.0x2^32 = 4 294 967 296.0 \c dq 1.e10 ; 10 000 000 000.0 \c dq 1.e+10 ; synonymous with 1.e10 \c dq 1.e-10 ; 0.000 000 000 1 \c dt 3.141592653589793238462 ; pi \c do 1.e+4000 ; IEEE 754r quad precision The 8-bit "quarter-precision" floating-point format is sign:exponent:mantissa = 1:4:3 with an exponent bias of 7. This appears to be the most frequently used 8-bit floating-point format, although it is not covered by any formal standard. This is sometimes called a "\i{minifloat}." The \i\c{bfloat16} format is effectively a compressed version of the 32-bit single precision format, with a reduced mantissa. It is effectively the same as truncating the 32-bit format to the upper 16 bits, except for rounding. There is no \c{D}\e{x} directive that corresponds to \c{bfloat16} as it obviously has the same size as the IEEE standard 16-bit half precision format, see however \k{pkg_fp}. The special operators are used to produce floating-point numbers in other contexts. They produce the binary representation of a specific floating-point number as an integer, and can use anywhere integer constants are used in an expression. \c{__?float80m?__} and \c{__?float80e?__} produce the 64-bit mantissa and 16-bit exponent of an 80-bit floating-point number, and \c{__?float128l?__} and \c{__?float128h?__} produce the lower and upper 64-bit halves of a 128-bit floating-point number, respectively. For example: \c mov rax,__?float64?__(3.141592653589793238462) ... would assign the binary representation of pi as a 64-bit floating point number into \c{RAX}. This is exactly equivalent to: \c mov rax,0x400921fb54442d18 NASM cannot do compile-time arithmetic on floating-point constants. This is because NASM is designed to be portable - although it always generates code to run on x86 processors, the assembler itself can run on any system with an ANSI C compiler. Therefore, the assembler cannot guarantee the presence of a floating-point unit capable of handling the \i{Intel number formats}, and so for NASM to be able to do floating arithmetic it would have to include its own complete set of floating-point routines, which would significantly increase the size of the assembler for very little benefit. The special tokens \i\c{__?Infinity?__}, \i\c{__?QNaN?__} (or \i\c{__?NaN?__}) and \i\c{__?SNaN?__} can be used to generate \I{infinity}infinities, quiet \i{NaN}s, and signalling NaNs, respectively. These are normally used as macros: \c %define Inf __?Infinity?__ \c %define NaN __?QNaN?__ \c \c dq +1.5, -Inf, NaN ; Double-precision constants The \c{%use fp} standard macro package contains a set of convenience macros. See \k{pkg_fp}. \S{bcdconst} \I{floating-point, packed BCD constants}Packed BCD Constants x87-style packed BCD constants can be used in the same contexts as 80-bit floating-point numbers. They are suffixed with \c{p} or prefixed with \c{0p}, and can include up to 18 decimal digits. As with other numeric constants, underscores can be used to separate digits. For example: \c dt 12_345_678_901_245_678p \c dt -12_345_678_901_245_678p \c dt +0p33 \c dt 33p \H{expr} \i{Expressions} Expressions in NASM are similar in syntax to those in C. Expressions are evaluated as 64-bit integers which are then adjusted to the appropriate size. NASM supports two special tokens in expressions, allowing calculations to involve the current assembly position: the \I{$, here}\c{$} and \i\c{$$} tokens. \c{$} evaluates to the assembly position at the beginning of the line containing the expression; so you can code an \i{infinite loop} using \c{JMP $}. \c{$$} evaluates to the beginning of the current section; so you can tell how far into the section you are by using \c{($-$$)}. The arithmetic \i{operators} provided by NASM are listed here, in increasing order of \i{precedence}. A \e{boolean} value is true if nonzero and false if zero. The operators which return a boolean value always return 1 for true and 0 for false. \S{exptri} \I{?op}\c{?} ... \c{:}: Conditional Operator The syntax of this operator, similar to the C conditional operator, is: \e{boolean} \c{?} \e{trueval} \c{:} \e{falseval} This operator evaluates to \e{trueval} if \e{boolean} is true, otherwise to \e{falseval}. Note that NASM allows \c{?} characters in symbol names. Therefore, it is highly advisable to always put spaces around the \c{?} and \c{:} characters. \S{expbor}: \i\c{||}: \i{Boolean OR} Operator The \c{||} operator gives a boolean OR: it evaluates to 1 if both sides of the expression are nonzero, otherwise 0. \S{expbxor}: \i\c{^^}: \i{Boolean XOR} Operator The \c{^^} operator gives a boolean XOR: it evaluates to 1 if any one side of the expression is nonzero, otherwise 0. \S{expband}: \i\c{&&}: \i{Boolean AND} Operator The \c{&&} operator gives a boolean AND: it evaluates to 1 if both sides of the expression is nonzero, otherwise 0. \S{exprel}: \i{Comparison Operators} NASM supports the following comparison operators: \b \i\c{=} or \i\c{==} compare for equality. \b \i\c{!=} or \i\c{<>} compare for inequality. \b \i\c{<} compares signed less than. \b \i\c{<=} compares signed less than or equal. \b \i\c{>} compares signed greater than. \b \i\c{>=} compares signed greater than or equal. These operators evaluate to 0 for false or 1 for true. \b \i{<=>} does a signed comparison, and evaluates to -1 for less than, 0 for equal, and 1 for greater than. At this time, NASM does not provide unsigned comparison operators. \S{expor} \i\c{|}: \i{Bitwise OR} Operator The \c{|} operator gives a bitwise OR, exactly as performed by the \c{OR} machine instruction. \S{expxor} \i\c{^}: \i{Bitwise XOR} Operator \c{^} provides the bitwise XOR operation. \S{expand} \i\c{&}: \i{Bitwise AND} Operator \c{&} provides the bitwise AND operation. \S{expshift} \i{Bit Shift} Operators \i\c{<<} gives a bit-shift to the left, just as it does in C. So \c{5<<3} evaluates to 5 times 8, or 40. \i\c{>>} gives an \I{unsigned, bit shift}\e{unsigned} (logical) bit-shift to the right; the bits shifted in from the left are set to zero. \i\c{<<<} gives a bit-shift to the left, exactly equivalent to the \c{<<} operator; it is included for completeness. \i\c{>>>} gives an \I{signed, bit shift}\e{signed} (arithmetic) bit-shift to the right; the bits shifted in from the left are filled with copies of the most significant (sign) bit. \S{expplmi} \I{+ opaddition}\c{+} and \I{- opsubtraction}\c{-}: \i{Addition} and \i{Subtraction} Operators The \c{+} and \c{-} operators do perfectly ordinary addition and subtraction. \S{expmul} \i{Multiplication}, \i{Division} and \i{Modulo} \i\c{*} is the multiplication operator. \i\c{/} and \i\c{//} are both division operators: \c{/} is \I{division, unsigned}\I{unsigned, division}unsigned division and \c{//} is \I{division, signed}\I{signed, division}signed division. Similarly, \i\c{%} and \i\c{%%} provide \I{modulo, unsigned}\I{unsigned, modulo}unsigned and \I{modulo, signed}\I{signed, modulo}signed modulo operators respectively. Since the \c{%} character is used extensively by the macro \i{preprocessor}, you should ensure that both the signed and unsigned modulo operators are followed by white space wherever they appear. NASM, like ANSI C, provides no guarantees about the sensible operation of the signed modulo operator. On most systems it will match the signed division operator, such that: \c b * (a // b) + (a %% b) = a (b != 0) \S{expmul} \I{operators, unary}\i{Unary Operators} The highest-priority operators in NASM's expression grammar are those which only apply to one argument. These are: \b \I{- opunary}\c{-} \I{arithmetic negation}negates (\i{2's complement}) its operand. \b \I{+ opunary}\c{+} does nothing; it's provided for symmetry with \c{-}. \b \I{~ opunary}\c{~} computes the \I{negation, bitwise}\i{bitwise negation} (\i{1's complement}) of its operand. \b \I{! opunary}\c{!} is the \I{negation, boolean}\i{boolean negation} operator. It evaluates to 1 if the argument is 0, otherwise 0. \b \c{SEG} provides the \i{segment address} of its operand (explained in more detail in \k{segwrt}). \b A set of additional operators with leading and trailing double underscores are used to implement the \c{integer functions} of the \c{ifunc} macro package, see \k{pkg_ifunc}. \H{segwrt} \i\c{SEG} and \i\c{WRT} When writing large 16-bit programs, which must be split into multiple \i{segments}, it is often necessary to be able to refer to the \I{segment address}segment part of the address of a symbol. NASM supports the \c{SEG} operator to perform this function. The \c{SEG} operator evaluates to the \i\e{preferred} segment base of a symbol, defined as the segment base relative to which the offset of the symbol makes sense. So the code \c mov ax,seg symbol \c mov es,ax \c mov bx,symbol will load \c{ES:BX} with a valid pointer to the symbol \c{symbol}. Things can be more complex than this: since 16-bit segments and \i{groups} may \I{overlapping segments}overlap, you might occasionally want to refer to some symbol using a different segment base from the preferred one. NASM lets you do this, by the use of the \c{WRT} (With Reference To) keyword. So you can do things like \c mov ax,weird_seg ; weird_seg is a segment base \c mov es,ax \c mov bx,symbol wrt weird_seg to load \c{ES:BX} with a different, but functionally equivalent, pointer to the symbol \c{symbol}. NASM supports far (inter-segment) calls and jumps by means of the syntax \c{call segment:offset}, where \c{segment} and \c{offset} both represent immediate values. So to call a far procedure, you could code either of \c call (seg procedure):procedure \c call weird_seg:(procedure wrt weird_seg) (The parentheses are included for clarity, to show the intended parsing of the above instructions. They are not necessary in practice.) NASM supports the syntax \I\c{CALL FAR}\c{call far procedure} as a synonym for the first of the above usages. \c{JMP} works identically to \c{CALL} in these examples. To declare a \i{far pointer} to a data item in a data segment, you must code \c dw symbol, seg symbol NASM supports no convenient synonym for this, though you can always invent one using the macro processor. \H{strict} \i\c{STRICT}: Inhibiting Optimization When assembling with the optimizer set to level 2 or higher (see \k{opt-O}), NASM will use size specifiers (\c{BYTE}, \c{WORD}, \c{DWORD}, \c{QWORD}, \c{TWORD}, \c{OWORD}, \c{YWORD} or \c{ZWORD}), but will give them the smallest possible size. The keyword \c{STRICT} can be used to inhibit optimization and force a particular operand to be emitted in the specified size. For example, with the optimizer on, and in \c{BITS 16} mode, \c push dword 33 is encoded in three bytes \c{66 6A 21}, whereas \c push strict dword 33 is encoded in six bytes, with a full dword immediate operand \c{66 68 21 00 00 00}. With the optimizer off, the same code (six bytes) is generated whether the \c{STRICT} keyword was used or not. \H{crit} \i{Critical Expressions} Although NASM has an optional multi-pass optimizer, there are some expressions which must be resolvable on the first pass. These are called \e{Critical Expressions}. The first pass is used to determine the size of all the assembled code and data, so that the second pass, when generating all the code, knows all the symbol addresses the code refers to. So one thing NASM can't handle is code whose size depends on the value of a symbol declared after the code in question. For example, \c times (label-$) db 0 \c label: db 'Where am I?' The argument to \i\c{TIMES} in this case could equally legally evaluate to anything at all; NASM will reject this example because it cannot tell the size of the \c{TIMES} line when it first sees it. It will just as firmly reject the slightly \I{paradox}paradoxical code \c times (label-$+1) db 0 \c label: db 'NOW where am I?' in which \e{any} value for the \c{TIMES} argument is by definition wrong! NASM rejects these examples by means of a concept called a \e{critical expression}, which is defined to be an expression whose value is required to be computable in the first pass, and which must therefore depend only on symbols defined before it. The argument to the \c{TIMES} prefix is a critical expression. \H{locallab} \i{Local Labels} NASM gives special treatment to symbols beginning with a \i{period}. A label beginning with a single period is treated as a \e{local} label, which means that it is associated with the previous non-local label. So, for example: \c label1 ; some code \c \c .loop \c ; some more code \c \c jne .loop \c ret \c \c label2 ; some code \c \c .loop \c ; some more code \c \c jne .loop \c ret In the above code fragment, each \c{JNE} instruction jumps to the line immediately before it, because the two definitions of \c{.loop} are kept separate by virtue of each being associated with the previous non-local label. This form of local label handling is borrowed from the old Amiga assembler \i{DevPac}; however, NASM goes one step further, in allowing access to local labels from other parts of the code. This is achieved by means of \e{defining} a local label in terms of the previous non-local label: the first definition of \c{.loop} above is really defining a symbol called \c{label1.loop}, and the second defines a symbol called \c{label2.loop}. So, if you really needed to, you could write \c label3 ; some more code \c ; and some more \c \c jmp label1.loop Sometimes it is useful - in a macro, for instance - to be able to define a label which can be referenced from anywhere but which doesn't interfere with the normal local-label mechanism. Such a label can't be non-local because it would interfere with subsequent definitions of, and references to, local labels; and it can't be local because the macro that defined it wouldn't know the label's full name. NASM therefore introduces a third type of label, which is probably only useful in macro definitions: if a label begins with the \I{label prefix}special prefix \i\c{..@}, then it does nothing to the local label mechanism. So you could code \c label1: ; a non-local label \c .local: ; this is really label1.local \c ..@foo: ; this is a special symbol \c label2: ; another non-local label \c .local: ; this is really label2.local \c \c jmp ..@foo ; this will jump three lines up NASM has the capacity to define other special symbols beginning with a double period: for example, \c{..start} is used to specify the entry point in the \c{obj} output format (see \k{dotdotstart}), \c{..imagebase} is used to find out the offset from a base address of the current image in the \c{win64} output format (see \k{win64pic}). So just keep in mind that symbols beginning with a double period are special.