A Quick Guide to Go's Assembler

This document is a quick outline of the unusual form of assembly language used by the gc suite of Go compilers (6g, 8g, etc.). The document is not comprehensive.

The assembler is based on the input to the Plan 9 assemblers, which is documented in detail on the Plan 9 site. If you plan to write assembly language, you should read that document although much of it is Plan 9-specific. This document provides a summary of the syntax and describes the peculiarities that apply when writing assembly code to interact with Go.

The most important thing to know about Go's assembler is that it is not a direct representation of the underlying machine. Some of the details map precisely to the machine, but some do not. This is because the compiler suite (see this description) needs no assembler pass in the usual pipeline. Instead, the compiler emits a kind of incompletely defined instruction set, in binary form, which the linker then completes. In particular, the linker does instruction selection, so when you see an instruction like MOV what the linker actually generates for that operation might not be a move instruction at all, perhaps a clear or load. Or it might correspond exactly to the machine instruction with that name. In general, machine-specific operations tend to appear as themselves, while more general concepts like memory move and subroutine call and return are more abstract. The details vary with architecture, and we apologize for the imprecision; the situation is not well-defined.

The assembler program is a way to generate that intermediate, incompletely defined instruction sequence as input for the linker. If you want to see what the instructions look like in assembly for a given architecture, say amd64, there are many examples in the sources of the standard library, in packages such as runtime and math/big. You can also examine what the compiler emits as assembly code:

$ cat x.go
package main

func main() {
	println(3)
}
$ go tool 6g -S x.go        # or: go build -gcflags -S x.go

--- prog list "main" ---
0000 (x.go:3) TEXT    main+0(SB),$8-0
0001 (x.go:3) FUNCDATA $0,gcargs·0+0(SB)
0002 (x.go:3) FUNCDATA $1,gclocals·0+0(SB)
0003 (x.go:4) MOVQ    $3,(SP)
0004 (x.go:4) PCDATA  $0,$8
0005 (x.go:4) CALL    ,runtime.printint+0(SB)
0006 (x.go:4) PCDATA  $0,$-1
0007 (x.go:4) PCDATA  $0,$0
0008 (x.go:4) CALL    ,runtime.printnl+0(SB)
0009 (x.go:4) PCDATA  $0,$-1
0010 (x.go:5) RET     ,
...

The FUNCDATA and PCDATA directives contain information for use by the garbage collector; they are introduced by the compiler.

Symbols

Some symbols, such as PC, R0 and SP, are predeclared and refer to registers. There are two other predeclared symbols, SB (static base) and FP (frame pointer). All user-defined symbols other than jump labels are written as offsets to these pseudo-registers.

The SB pseudo-register can be thought of as the origin of memory, so the symbol foo(SB) is the name foo as an address in memory.

The FP pseudo-register is a virtual frame pointer used to refer to function arguments. The compilers maintain a virtual frame pointer and refer to the arguments on the stack as offsets from that pseudo-register. Thus 0(FP) is the first argument to the function, 8(FP) is the second (on a 64-bit machine), and so on. When referring to a function argument this way, it is conventional to place the name at the beginning, as in first_arg+0(FP) and second_arg+8(FP). Some of the assemblers enforce this convention, rejecting plain 0(FP) and 8(FP). For assembly functions with Go prototypes, go vet will check that the argument names and offsets match.

The SP pseudo-register is a virtual stack pointer used to refer to frame-local variables and the arguments being prepared for function calls. It points to the top of the local stack frame, so references should use negative offsets in the range [−framesize, 0): x-8(SP), y-4(SP), and so on. On architectures with a real register named SP, the name prefix distinguishes references to the virtual stack pointer from references to the architectural SP register. That is, x-8(SP) and -8(SP) are different memory locations: the first refers to the virtual stack pointer pseudo-register, while the second refers to the hardware's SP register.

Instructions, registers, and assembler directives are always in UPPER CASE to remind you that assembly programming is a fraught endeavor. (Exception: the g register renaming on ARM.)

In Go object files and binaries, the full name of a symbol is the package path followed by a period and the symbol name: fmt.Printf or math/rand.Int. Because the assembler's parser treats period and slash as punctuation, those strings cannot be used directly as identifier names. Instead, the assembler allows the middle dot character U+00B7 and the division slash U+2215 in identifiers and rewrites them to plain period and slash. Within an assembler source file, the symbols above are written as fmt·Printf and math∕rand·Int. The assembly listings generated by the compilers when using the -S flag show the period and slash directly instead of the Unicode replacements required by the assemblers.

Most hand-written assembly files do not include the full package path in symbol names, because the linker inserts the package path of the current object file at the beginning of any name starting with a period: in an assembly source file within the math/rand package implementation, the package's Int function can be referred to as ·Int. This convention avoids the need to hard-code a package's import path in its own source code, making it easier to move the code from one location to another.

Directives

The assembler uses various directives to bind text and data to symbol names. For example, here is a simple complete function definition. The TEXT directive declares the symbol runtime·profileloop and the instructions that follow form the body of the function. The last instruction in a TEXT block must be some sort of jump, usually a RET (pseudo-)instruction. (If it's not, the linker will append a jump-to-itself instruction; there is no fallthrough in TEXTs.) After the symbol, the arguments are flags (see below) and the frame size, a constant (but see below):

TEXT runtime·profileloop(SB),NOSPLIT,$8
	MOVQ	$runtime·profileloop1(SB), CX
	MOVQ	CX, 0(SP)
	CALL	runtime·externalthreadhandler(SB)
	RET

In the general case, the frame size is followed by an argument size, separated by a minus sign. (It's not a subtraction, just idiosyncratic syntax.) The frame size $24-8 states that the function has a 24-byte frame and is called with 8 bytes of argument, which live on the caller's frame. If NOSPLIT is not specified for the TEXT, the argument size must be provided.

Note that the symbol name uses a middle dot to separate the components and is specified as an offset from the static base pseudo-register SB. This function would be called from Go source for package runtime using the simple name profileloop.

For DATA directives, the symbol is followed by a slash and the number of bytes the memory associated with the symbol occupies. The arguments are optional flags and the data itself. For instance,

DATA  runtime·isplan9(SB)/4, $1

declares the local symbol runtime·isplan9 of size 4 and value 1. Again the symbol has the middle dot and is offset from SB.

The GLOBL directive declares a symbol to be global. The arguments are optional flags and the size of the data being declared as a global, which will have initial value all zeros unless a DATA directive has initialized it. The GLOBL directive must follow any corresponding DATA directives. This example

GLOBL runtime·tlsoffset(SB),$4

declares runtime·tlsoffset to have size 4.

There may be one or two arguments to the directives. If there are two, the first is a bit mask of flags, which can be written as numeric expressions, added or or-ed together, or can be set symbolically for easier absorption by a human. Their values, defined in the file src/cmd/ld/textflag.h, are:

Architecture-specific details

It is impractical to list all the instructions and other details for each machine. To see what instructions are defined for a given machine, say 32-bit Intel x86, look in the top-level header file for the corresponding linker, in this case 8l. That is, the file $GOROOT/src/cmd/8l/8.out.h contains a C enumeration, called as, of the instructions and their spellings as known to the assembler and linker for that architecture. In that file you'll find a declaration that begins

enum	as
{
	AXXX,
	AAAA,
	AAAD,
	AAAM,
	AAAS,
	AADCB,
	...

Each instruction begins with a initial capital A in this list, so AADCB represents the ADCB (add carry byte) instruction. The enumeration is in alphabetical order, plus some late additions (AXXX occupies the zero slot as an invalid instruction). The sequence has nothing to do with the actual encoding of the machine instructions. Again, the linker takes care of that detail.

One detail evident in the examples from the previous sections is that data in the instructions flows from left to right: MOVQ $0, CX clears CX. This convention applies even on architectures where the usual mode is the opposite direction.

Here follows some descriptions of key Go-specific details for the supported architectures.

32-bit Intel 386

The runtime pointer to the g structure is maintained through the value of an otherwise unused (as far as Go is concerned) register in the MMU. A OS-dependent macro get_tls is defined for the assembler if the source includes an architecture-dependent header file, like this:

#include "zasm_GOOS_GOARCH.h"

Within the runtime, the get_tls macro loads its argument register with a pointer to the g pointer, and the g struct contains the m pointer. The sequence to load g and m using CX looks like this:

get_tls(CX)
MOVL	g(CX), AX     // Move g into AX.
MOVL	g_m(AX), BX   // Move g->m into BX.

64-bit Intel 386 (a.k.a. amd64)

The assembly code to access the m and g pointers is the same as on the 386, except it uses MOVQ rather than MOVL:

get_tls(CX)
MOVQ	g(CX), AX     // Move g into AX.
MOVQ	g_m(AX), BX   // Move g->m into BX.

ARM

The registers R10 and R11 are reserved by the compiler and linker.

R10 points to the g (goroutine) structure. Within assembler source code, this pointer must be referred to as g; the name R10 is not recognized.

To make it easier for people and compilers to write assembly, the ARM linker allows general addressing forms and pseudo-operations like DIV or MOD that may not be expressible using a single hardware instruction. It implements these forms as multiple instructions, often using the R11 register to hold temporary values. Hand-written assembly can use R11, but doing so requires being sure that the linker is not also using it to implement any of the other instructions in the function.

When defining a TEXT, specifying frame size $-4 tells the linker that this is a leaf function that does not need to save LR on entry.

The name SP always refers to the virtual stack pointer described earlier. For the hardware register, use R13.

Unsupported opcodes

The assemblers are designed to support the compiler so not all hardware instructions are defined for all architectures: if the compiler doesn't generate it, it might not be there. If you need to use a missing instruction, there are two ways to proceed. One is to update the assembler to support that instruction, which is straightforward but only worthwhile if it's likely the instruction will be used again. Instead, for simple one-off cases, it's possible to use the BYTE and WORD directives to lay down explicit data into the instruction stream within a TEXT. Here's how the 386 runtime defines the 64-bit atomic load function.

// uint64 atomicload64(uint64 volatile* addr);
// so actually
// void atomicload64(uint64 *res, uint64 volatile *addr);
TEXT runtime·atomicload64(SB), NOSPLIT, $0-8
	MOVL	4(SP), BX
	MOVL	8(SP), AX
	// MOVQ (%EAX), %MM0
	BYTE $0x0f; BYTE $0x6f; BYTE $0x00
	// MOVQ %MM0, 0(%EBX)
	BYTE $0x0f; BYTE $0x7f; BYTE $0x03
	// EMMS
	BYTE $0x0F; BYTE $0x77
	RET