Inline AssemblerThe compiler includes a powerful inlineassembler. With it, assembly language instructions can be used directlyin C and C++ source programs without requiring a separate assemblerprogram. Assembly language enables optimizing critical functions,
interfacing to the BIOS, operating system and special hardware, andaccess capabilities of the processor that are not available from C++. It supports both 16-bit and 32-bit code generation in all memory models.
What's in This Chapter
- Basic features of the inline assembler.
- The ASM statement and the ASM block.
- Using ASM registers.
- Calling C and C++ from assembly language.
- Registers and opcodes.
Advantages of writing inline assembly language functionsUse assembly language functions to:
- Create a subroutine that executes as quickly as possible
- Provide an interface to functions compiled with another compiler
- Acess capabilities of the CPU and the native instruction set that are not available from C++
- Interface to specialized hardware
- Write code where specific instruction selection and ordering is critical
The asm StatementThe asm statement invokes the assembler. Usethis statement wherever a C or C++ statement is legal. You can use asmin any of three ways. The first example shows asm followed simply by an assembly instruction:
asm mov AH,2
asm mov DL,7
asm int 21H
The second example shows asm followed by a set of assemblyinstructions enclosed by braces. An empty set of braces may follow thedirective. asm {
mov AH,2
mov DL,7
int 21H
}
Because the asm statement is a statement separator, assembly instructions can appear on the same line: asm mov AH,2 asm mov DL,7 asm int 21H
The three previous examples generate identical code. Butenclosing assembly language in braces, as in the second example, hasthe advantage of setting the assembly language off from the surroundingC++ code and avoids repeating the asm statement. No assembler instruction can continue onto a second line. Use theform of the last example primarily for writing macros, which must beone line long after expansion.
NoteThe Digital Mars C++ asm statement emulates theBorland asm statement. The _asm and __asm statements emulate the
Microsoft _asm and __asm statements. The ASM BlockA series of assembler instructions enclosed bybraces following the asm keyword are called an "ASM block." Unlike C++blocks, ASM blocks do not affect the scope of variables. Restrictions on using C and C++ in an ASM blockAn ASM block can use the following C and C++ language elements:
- Symbols, including labels, variables, and function names.
- Constants, including symbolic constants and enum members.
- Macros and preprocessor directives.
- Comments delimited by /**/ symbols or defined by // symbols.In Microsoft-compatible mode, semicolons (;) can also be used todelimit comments.
- Type names (where a MASM type would be legal).
- Type names, including declarations of structs and pointers.
- C-style type casts.
NoteMicrosoft and Borland inline assemblers do not support type casts. Inline assembly instructions within C or C++ statements can refer to C or C++ variables by name.
MASM-Style hexadecimal constantsSupport for MASM-style hexadecimal constants provides easy conversion of MASM-style source code. The constants take the form: digit {hex_digit} ('H'| 'h')
You cannot use hexadecimal constants if the -A (for ANSI compatibility) option is used. C and C++ operators in an ASM blockAn ASM block cannot use operators specific to C and C++, such as the left-shift (<<) operator. You can use operators common to C, C++, and MASM within an ASM blockbut interpret them as assembly language operators. C and C++ interpretbrackets ([]) as enclosing array subscripts and scale them to the sizeof an array element. But within an asm statement, C and C++ interpretbrackets as the MASM index operator, which adds an unscaled byte offsetto any adjacent operand.
In Microsoft-compatible mode, the semicolon delimits comments, as in MASM.
Assembly language in an asm statementIn common with otherassemblers, the inline assembler accepts any instruction that is legalin MASM. The following are some of the assembly language features ofthe asm statement:
- Expressions. Inline assembly code can use any MASM expression. AMASM expression is any combination of operands and operators thatevaluates to a single value or address.
- Data directives and operators. An asm statement can definedata objects in the code segment with the MASM directives DB, DW, andDQ. The MASM directives DT, DF, STRUC, and RECORD, and the operatorsDUP, THIS, WIDTH, and MASK are not accepted.
- Macros. An asm statement can use C preprocessor macros eventhough the inline assembler is not a macro assembler and does notsupport MASM macro directives.
Other restrictions on C and C++ symbolsAn asm statement canreference any C or C++ variable name, function, or label in scope,provided those names are not symbolic constants. However, you cannotcall a C++ member function from within an asm statement. Prototype the functions referenced in an asm statement beforeusing them in programs. This lets the compiler distinguish them fromnames and labels.
Each assembly language instruction can contain a single C or C++ symbol.
C or C++ symbols within an ASM block must not have the same spelling as an asm reserved word.
The inline assembler allows structure or union tags in asmstatements, but only as qualifiers of references to members of thestructure or union.
Accessing C or C++ data in an asm statementIn general,instructions in an asm statement can reference any symbol in scopewhere the statement appears. The following statement loads the AXregister with the value of var, a C variable in scope: asm mov AX,var
An asm statement can reference a uniquely named member of aclass, structure, or union without specifying the variable name or typebefore the period operator. But if the member name is not unique, youmust specify a variable or type name before the period operator. If twostructure types have a member name in common, as in this example: struct first_type
{
char *mold;
int common_name;
};
struct second__type
{
char *mildew;
long common_name;
int unique_name;
};
then qualify the reference to
common_name with the tag name: asm mov [bx]first_type.common_name,10
You need not qualify a reference to a unique member name. In thefollowing example, unique_name is an anonymous structure member becauseit is a member of first_type. asm mov [bx].unique_name,10
This statement generates the same instruction whether or not aqualifying name or type is present. For more information, see thesection "Making anonymous references to structure members" later inthis chapter. Functions in inline assembly languageBecause ASM blocks donot require separate source file assembly steps, writing a functionusing ASM blocks is easier than using a separate assembler. Inaddition, the compiler generates function prolog and epilog code. The expon2 function is an example of a function written in inline assembly language:
int expon2(int num, int power)
{
asm
{ mov AX,num // get first argument
mov CX,power // get second argument
shl AX,CL // AX = AX * (2 to the power of CL)
}
}
An inline function refers to its arguments by name and may appear in the same source file as the callers of the function. Refer to
Using Assembly Language Functions for a description of the register stacks used by inline assembly instructions.
Making anonymous references to structure membersYou can make anonymous references to members of a given structure, as in the following: struct x {
int i;
int j;
int k;
} foo;
You can refer to these members i, j, and k anonymously, for example, with the assembly instruction: asm
{ mov BX,4
mov AX,foo[BX]; Refers to member j of foo
}
Using register variablesDigital Mars C++ supports registervariables. Register variables are useful with inline assembly. If asmstatements place results in registers, you can use register variablesto access those values. For more information see "Using Register Variables" in Chapter 5, "Using Assembly Language Functions."
Using the __LOCAL_SIZE symbolWhen using the inlineassembler, the special symbol __LOCAL_SIZE expands to the number ofbytes used by all local symbols. __LOCAL_SIZE is useful in combinationwith __declspec(naked), as __LOCAL_SIZE is the amount of space toreserve on the stack. For example:
__declspec(naked) int test()
{
int x, y, z;
_asm
{ push BP
mov BP,SP
sub SP,__LOCAL_SIZE
mov BX,__LOCAL_SIZE[BP]
mov BX,__LOCAL_SIZE+2[BP]
mov AX,__LOCAL_SIZE
mov AX,__LOCAL_SIZE+2
}
_AX = x + y + z;
_asm
{
mov SP,BP
pop BP
ret
}
}
Using ASM RegistersThe asm statement alters the registersoutside the programmer's explicit assembly language instructions.Registers contain whatever values the normal control flow leaves inthem at the point of the asm statement. For 16-bit memory modelsYou do not need to preserve thefollowing registers when writing inline assembly language: AX, BX, CX,DX, SI, DI, ES, and flags (other than DF). C and C++ do not expect these registers to be maintained betweenstatements, but they do preserve the following registers: CS, DS, SS,SP and BP.
NoteThe compiler does not use registers to hold register variables for functions containing inline assembly code. For 32-bit memory modelsFunctions can change the values in the EAX, ECX, EDX, ESI, EDI registers. Functions must preserve the values in the EBX, ESI, EDI, EBP, ESP,SS, CS, DS registers (plus ES and GS for the NT memory model).
Always set the direction flag to forward.
To maximize speed on 32-bit buses, make sure data aligns along 32-Function return values
- For 16-bit models. If the return value for a function is short (achar, int, or near pointer) store it in the AX register, as in theprevious example, expon2. If the return value is long, store the highword in the DX register and the low word in AX. To return a longervalue, store the value in memory and return a pointer to the value.
- For 32-bit models. Return near pointers, ints, unsignedints, chars, shorts, longs and unsigned longs in EAX. 32-bit modelsreturn far pointers in EDX, EAX, where EDX contains the segment and EAXcontains the offset.
- When C linkage is in effect. Floats are returned in EAX anddoubles in EDX, EAX, where EDX contains the most significant 32 bitsand EAX the least significant.
- When C++ linkage is in effect. The compiler creates a temporary copy on the stack and returns a pointer to it.
Function return values
- For 16-bit memory models. If the return value for a function isshort (a char, int, or near pointer) store it in the AX register, as inthe previous example, expon2. If the return value is long, store thehigh word in the DX register and the low word in AX. To return a longervalue, store the value in memory and return a pointer to the value.
- For 32-bit models. Return near pointers, ints, unsignedints, chars, shorts, longs, and unsigned longs in EAX. 32-bit modelsreturn far pointers in EDX, EAX, where EDX contains the segment and EAXcontains the offset.
- When C linkage is in effect. Floats are returned in EAX anddoubles in EDX, EAX, where EDX contains the most significant 32 bitsand EAX the least significant.
- When C++ linkage is in effect. The compiler creates a temporary copy on the stack and returns a pointer to it.
Interfacing to a member functionThe easiest way to interfacean assembly language routine to a class member function is to provide aC wrapper function that can be called from the assembly languageroutine. An alternative is to write the member function in C++ and compile it with normal out-of-line member functions.
Calling C Functions from an ASM BlockC functions, including C library functions, can be called from within the asm block, as in the following example: #include <stdio.h>
char format [] = "% s %s %s \n";
char alas[] = "Alas,";
char poor[] = "poor";
char Yorick[] = "Yorick!";
void main(void)
{
asm
{ mov AX, offset Yorick
push AX
mov AX, offset poor
push AX
mov AX, offset Alas
push AX
mov AX, offset format
push AX
call printf
}
}
Simply push the needed arguments from right to left beforecalling the function, since function arguments are passed on the stack.To print the message, the example pushes pointers to the three strings,formats them, and then calls printf. Calling C++ functionsAn ASM block can call only global C++functions that are not overloaded because the types of the argumentsare unknown. The compiler issues an error if an ASM block calls anoverloaded global C++ function or a C member function. You can also call a function declared with extern "C" linkage froman asm statement within a C++ program, because all the standard headerfiles declare the library functions to have extern "C" linkage.
Defining ASM blocks as C macrosA C++ macro is aconvenient way to insert assembly language into source code. But,because a macro expands into a single, logical line, take care whenwriting them. If the macro expands into multiple instructions, enclose theinstructions in an ASM block. The asm statement must precede eachinstruction. Also separate comments from code with /**/ charactersrather than //. Unless you take these precautions, the compiler can beconfused by C or C++ statements to the left or right of the assemblycode or interpret instructions as comments when the macro becomes asingle line. Without the closing brace, the compiler cannot tell wherethe assembly language ends.
Warning: Do not use double-slash (//) characters within a macro. The compiler terminates the macro when it sees a double-slash.
An ASM block written as a macro can accept arguments but, unlikea C macro, it cannot return values. But some MASM macros can be writtenas macros for C. The following MASM macro sets a video page to thevalue specified in the argument page:
findpage MACRO page
mov AH, 5
MOV AL,page
int 10h
ENDM
The following C macro does the same thing: #define findpage(page) asm \
{ \
asm mov AH,5 \
asm mov AL,page \
asm int 10h \
}
RegistersThe following registers are supported. Register names are in upper or lower case.
AL,
AH,
AX,
EAX BL,
BH,
BX,
EBX CL,
CH,
CX,
ECX DL,
DH,
DX,
EDX BP,
EBP SP,
ESP DI,
EDI SI,
ESI ES,
CS,
SS,
DS,
GS,
FS CR0,
CR2,
CR3,
CR4 DR0,
DR1,
DR2,
DR3,
DR6,
DR7 TR3,
TR4,
TR5,
TR6,
TR7 ST ST(0),
ST(1),
ST(2),
ST(3),
ST(4),
ST(5),
ST(6),
ST(7) MM0,
MM1,
MM2,
MM3,
MM4,
MM5,
MM6,
MM7 XMM0,
XMM1,
XMM2,
XMM3,
XMM4,
XMM5,
XMM6,
XMM7 OpcodesThe following instructions are supported. Opcode names are in upper or lower case.
| aaa | aad | aam | aas | adc |
| add | addpd | addps | addsd | addss |
| and | andnpd | andnps | andpd | andps |
| arpl | bound | bsf | bsr | bswap |
| bt | btc | btr | bts | call |
| cbw | cdq | clc | cld | clflush |
| cli | clts | cmc | cmova | cmovae |
| cmovb | cmovbe | cmovc | cmove | cmovg |
| cmovge | cmovl | cmovle | cmovna | cmovnae |
| cmovnb | cmovnbe | cmovnc | cmovne | cmovng |
| cmovnge | cmovnl | cmovnle | cmovno | cmovnp |
| cmovns | cmovnz | cmovo | cmovp | cmovpe |
| cmovpo | cmovs | cmovz | cmp | cmppd |
| cmpps | cmps | cmpsb | cmpsd | cmpss |
| cmpsw | cmpxch8b | cmpxchg | comisd | comiss |
| cpuid | cvtdq2pd | cvtdq2ps | cvtpd2dq | cvtpd2pi |
| cvtpd2ps | cvtpi2pd | cvtpi2ps | cvtps2dq | cvtps2pd |
| cvtps2pi | cvtsd2si | cvtsd2ss | cvtsi2sd | cvtsi2ss |
| cvtss2sd | cvtss2si | cvttpd2dq | cvttpd2pi | cvttps2dq |
| cvttps2pi | cvttsd2si | cvttss2si | cwd | cwde |
| da | daa | das | db | dd |
| de | dec | df | di | div |
| divpd | divps | divsd | divss | dl |
| dq | ds | dt | dw | emms |
| enter | f2xm1 | fabs | fadd | faddp |
| fbld | fbstp | fchs | fclex | fcmovb |
| fcmovbe | fcmove | fcmovnb | fcmovnbe | fcmovne |
| fcmovnu | fcmovu | fcom | fcomi | fcomip |
| fcomp | fcompp | fcos | fdecstp | fdisi |
| fdiv | fdivp | fdivr | fdivrp | feni |
| ffree | fiadd | ficom | ficomp | fidiv |
| fidivr | fild | fimul | fincstp | finit |
| fist | fistp | fisub | fisubr | fld |
| fld1 | fldcw | fldenv | fldl2e | fldl2t |
| fldlg2 | fldln2 | fldpi | fldz | fmul |
| fmulp | fnclex | fndisi | fneni | fninit |
| fnop | fnsave | fnstcw | fnstenv | fnstsw |
| fpatan | fprem | fprem1 | fptan | frndint |
| frstor | fsave | fscale | fsetpm | fsin |
| fsincos | fsqrt | fst | fstcw | fstenv |
| fstp | fstsw | fsub | fsubp | fsubr |
| fsubrp | ftst | fucom | fucomi | fucomip |
| fucomp | fucompp | fwait | fxam | fxch |
| fxrstor | fxsave | fxtract | fyl2x | fyl2xp1 |
| hlt | idiv | imul | in | inc |
| ins | insb | insd | insw | int |
| into | invd | invlpg | iret | iretd |
| ja | jae | jb | jbe | jc |
| jcxz | je | jecxz | jg | jge |
| jl | jle | jmp | jna | jnae |
| jnb | jnbe | jnc | jne | jng |
| jnge | jnl | jnle | jno | jnp |
| jns | jnz | jo | jp | jpe |
| jpo | js | jz | lahf | lar |
| ldmxcsr | lds | lea | leave | les |
| lfence | lfs | lgdt | lgs | lidt |
| lldt | lmsw | lock | lods | lodsb |
| lodsd | lodsw | loop | loope | loopne |
| loopnz | loopz | lsl | lss | ltr |
| maskmovdqu | maskmovq | maxpd | maxps | maxsd |
| maxss | mfence | minpd | minps | minsd |
| minss | mov | movapd | movaps | movd |
| movdq2q | movdqa | movdqu | movhlps | movhpd |
| movhps | movlhps | movlpd | movlps | movmskpd |
| movmskps | movntdq | movnti | movntpd | movntps |
| movntq | movq | movq2dq | movs | movsb |
| movsd | movss | movsw | movsx | movupd |
| movups | movzx | mul | mulpd | mulps |
| mulsd | mulss | neg | nop | not |
| or | orpd | orps | out | outs |
| outsb | outsd | outsw | packssdw | packsswb |
| packuswb | paddb | paddd | paddq | paddsb |
| paddsw | paddusb | paddusw | paddw | pand |
| pandn | pavgb | pavgw | pcmpeqb | pcmpeqd |
| pcmpeqw | pcmpgtb | pcmpgtd | pcmpgtw | pextrw |
| pinsrw | pmaddwd | pmaxsw | pmaxub | pminsw |
| pminub | pmovmskb | pmulhuw | pmulhw | pmullw |
| pmuludq | pop | popa | popad | popf |
| popfd | por | prefetchnta | prefetcht0 | prefetcht1 |
| prefetcht2 | psadbw | pshufd | pshufhw | pshuflw |
| pshufw | pslld | pslldq | psllq | psllw |
| psrad | psraw | psrld | psrldq | psrlq |
| psrlw | psubb | psubd | psubq | psubsb |
| psubsw | psubusb | psubusw | psubw | punpckhbw |
| punpckhdq | punpckhqdq | punpckhwd | punpcklbw | punpckldq |
| punpcklqdq | punpcklwd | push | pusha | pushad |
| pushf | pushfd | pxor | rcl | rcpps |
| rcpss | rcr | rdmsr | rdpmc | rdtsc |
| rep | repe | repne | repnz | repz |
| ret | retf | rol | ror | rsm |
| rsqrtps | rsqrtss | sahf | sal | sar |
| sbb | scas | scasb | scasd | scasw |
| seta | setae | setb | setbe | setc |
| sete | setg | setge | setl | setle |
| setna | setnae | setnb | setnbe | setnc |
| setne | setng | setnge | setnl | setnle |
| setno | setnp | setns | setnz | seto |
| setp | setpe | setpo | sets | setz |
| sfence | sgdt | shl | shld | shr |
| shrd | shufpd | shufps | sidt | sldt |
| smsw | sqrtpd | sqrtps | sqrtsd | sqrtss |
| stc | std | sti | stmxcsr | stos |
| stosb | stosd | stosw | str | sub |
| subpd | subps | subsd | subss | sysenter |
| sysexit | test | ucomisd | ucomiss | ud2 |
| unpckhpd | unpckhps | unpcklpd | unpcklps | verr |
| verw | wait | wbinvd | wrmsr | xadd |
| xchg | xlat | xlatb | xor | xorpd |
| xorps |
|
|
|
Pentium 4 (Prescott) Opcodes Supported
| addsubpd | addsubps | fisttp | haddpd | haddps |
| hsubpd | hsubps | lddqu | monitor | movddup |
| movshdup | movsldup | mwait |
|
|
AMD Opcodes
| pavgusb | pf2id | pfacc | pfadd | pfcmpeq |
| pfcmpge | pfcmpgt | pfmax | pfmin | pfmul |
| pfnacc | pfpnacc | pfrcp | pfrcpit1 | pfrcpit2 |
| pfrsqit1 | pfrsqrt | pfsub | pfsubr | pi2fd |
| pmulhrw | pswapd |
