GCC inline assembler, mixing register sizes (x86)

Question

Does anyone know how I can get rid of the following assembler warning?

Code is x86, 32 bit:

int test (int x)
{
  int y;
  // do a bit-rotate by 8 on the lower word. leave upper word intact.
  asm ("rorw $8, %0\n\t": "=q"(y) :"0"(x));
  return y;
}

If I compile it I get the following (very valid) warning:

Warning: using `%ax' instead of `%eax' due to `w' suffix

What I'm looking for is a way to tell the compiler/assembler that I want to access the lower 16 bit sub-register of %0. Accessing the byte sub-registers (in this case AL and AH) would be nice to know as well.

I've already chosen the "q" modifier, so the compiler is forced to use EAX, EBX, ECX or EDX. I've made sure the compiler has to pick a register that has sub-registers.

I know that I can force the asm-code to use a specific register (and its sub-registers), but I want to leave the register-allocation job up to the compiler.

Answer 1

You can use %w0 if I remember right. I just tested it, too. :-)

int
test(int x)
{
    int y;
    asm ("rorw $8, %w0" : "=q" (y) : "0" (x));
    return y;
}

Edit: In response to the OP, yes, you can do the following too:

int
test(int x)
{
    int y;
    asm ("xchg %b0, %h0" : "=Q" (y) : "0" (x));
    return y;
}

For x86 it's documented in the x86 Operand Modifiers section of the Extended Asm part of the manual.

For non-x86 instruction sets, you may have to dig through their .md files in the GCC source. For example, gcc/config/i386/i386.md was the only place to find this before it was officially documented.

(Related: In GNU C inline asm, what are the size-override modifiers for xmm/ymm/zmm for a single operand? for vector registers.)

Answer 2

Long ago, but I'll likely need this for my own future reference...

Adding on to Chris's fine answer says, the key is using a modifier between the '%' and the number of the output operand. For example, "MOV %1, %0" might become "MOV %q1, %w0".

I couldn't find anything in constraints.md, but /gcc/config/i386/i386.c had this potentially useful comment in the source for print_reg():

/* Print the name of register X to FILE based on its machine mode and number.
   If CODE is 'w', pretend the mode is HImode.
   If CODE is 'b', pretend the mode is QImode.
   If CODE is 'k', pretend the mode is SImode.
   If CODE is 'q', pretend the mode is DImode.
   If CODE is 'x', pretend the mode is V4SFmode.
   If CODE is 't', pretend the mode is V8SFmode.
   If CODE is 'h', pretend the reg is the 'high' byte register.
   If CODE is 'y', print "st(0)" instead of "st", if the reg is stack op.
   If CODE is 'd', duplicate the operand for AVX instruction.
 */

A comment below for ix86_print_operand() offer an example:

b -- print the QImode name of the register for the indicated operand.

%b0 would print %al if operands[0] is reg 0.

A few more useful options are listed under Output Template of the GCC Internals documentation:

‘%cdigit’ can be used to substitute an operand that is a constant value without the syntax that normally indicates an immediate operand.

‘%ndigit’ is like ‘%cdigit’ except that the value of the constant is negated before printing.

‘%adigit’ can be used to substitute an operand as if it were a memory reference, with the actual operand treated as the address. This may be useful when outputting a “load address” instruction, because often the assembler syntax for such an instruction requires you to write the operand as if it were a memory reference.

‘%ldigit’ is used to substitute a label_ref into a jump instruction.

‘%=’ outputs a number which is unique to each instruction in the entire compilation. This is useful for making local labels to be referred to more than once in a single template that generates multiple assembler instructions.

The '%c2' construct allows one to properly format an LEA instruction using an offset:

#define ASM_LEA_ADD_BYTES(ptr, bytes)                            \
    __asm volatile("lea %c1(%0), %0" :                           \
                   /* reads/writes %0 */  "+r" (ptr) :           \
                   /* reads */ "i" (bytes));

Note the crucial but sparsely documented 'c' in '%c1'. This macro is equivalent to

ptr = (char *)ptr + bytes

but without making use of the usual integer arithmetic execution ports.

Edit to add:

Making direct calls in x64 can be difficult, as it requires yet another undocumented modifier: '%P0' (which seems to be for PIC)

#define ASM_CALL_FUNC(func)                                         \
    __asm volatile("call %P0") :                                    \
              /* no writes */ :                                     \
              /* reads %0 */ "i" (func))                           

A lower case 'p' modifier also seems to function the same in GCC, although only the capital 'P' is recognized by ICC. More details are probably available at /gcc/config/i386/i386.c. Search for "'p'".

Answer 3

While I'm thinking about it ... you should replace the "q" constraint with a capital "Q" constraint in Chris's second solution:

int
test(int x)
{
    int y;
    asm ("xchg %b0, %h0" : "=Q" (y) : "0" (x));
    return y;
}

"q" and "Q" are slightly different in 64-bit mode, where you can get the lowest byte for all of the integer registers (ax, bx, cx, dx, si, di, sp, bp, r8-r15). But you can only get the second-lowest byte (e.g. ah) for the four original 386 registers (ax, bx, cx, dx).

Answer 4

So apparently there are tricks to do this... but it may not be so efficient. 32-bit x86 processors are generally slow at manipulating 16-bit data in general purpose registers. You ought to benchmark it if performance is important.

Unless this is (a) performance critical and (b) proves to be much faster, I would save myself some maintenance hassle and just do it in C:

uint32_t y, hi=(x&~0xffff), lo=(x&0xffff);
y = hi + (((lo >> 8) + (lo << 8))&0xffff);

With GCC 4.2 and -O2 this gets optimized down to six instructions...