cosmopolitan/third_party/chibicc/codegen.c

2460 lines
69 KiB
C

#include "third_party/chibicc/chibicc.h"
#define GP_MAX 6
#define FP_MAX 8
int depth;
static char argreg8[][5] = {"%dil", "%sil", "%dl", "%cl", "%r8b", "%r9b"};
static char argreg16[][5] = {"%di", "%si", "%dx", "%cx", "%r8w", "%r9w"};
static char argreg32[][5] = {"%edi", "%esi", "%edx", "%ecx", "%r8d", "%r9d"};
static char argreg64[][5] = {"%rdi", "%rsi", "%rdx", "%rcx", "%r8", "%r9"};
static Obj *current_fn;
static char *lastline;
FILE *output_stream;
void flushln(void) {
if (!lastline) return;
fputs(lastline, output_stream);
fputc('\n', output_stream);
free(lastline);
lastline = NULL;
}
static void processln(char *nextline) {
if (lastline) {
// unsophisticated optimization pass to reduce asm noise a little bit
if ((!strcmp(lastline, "\txor\t%eax,%eax") &&
!strcmp(nextline, "\tcltq")) ||
(!strcmp(lastline, "\tmov\t$0x1,%eax") &&
!strcmp(nextline, "\tcltq")) ||
(!strcmp(lastline, "\tmovslq\t(%rax),%rax") &&
!strcmp(nextline, "\tcltq"))) {
free(nextline);
} else if (!strcmp(lastline, "\tmov\t(%rax),%rax") &&
!strcmp(nextline, "\tpush\t%rax")) {
free(lastline);
free(nextline);
lastline = strdup("\tpush\t(%rax)");
} else if (!strcmp(lastline, "\tmov\t$0x1,%eax") &&
!strcmp(nextline, "\tpush\t%rax")) {
free(lastline);
free(nextline);
lastline = strdup("\tpush\t$1");
} else if (!strcmp(lastline, "\tpush\t(%rax)") &&
!strcmp(nextline, "\tpop\t%rdi")) {
free(lastline);
free(nextline);
lastline = strdup("\tmov\t(%rax),%rdi");
} else if (!strcmp(lastline, "\tpush\t%rax") &&
!strcmp(nextline, "\tpop\t%rdi")) {
free(lastline);
free(nextline);
lastline = strdup("\tmov\t%rax,%rdi");
} else {
flushln();
lastline = nextline;
}
} else {
lastline = nextline;
}
}
static void emitlin(char *nextline) {
processln(strdup(nextline));
}
void println(char *fmt, ...) {
va_list ap;
char *nextline;
va_start(ap, fmt);
emitlin(xvasprintf(fmt, ap));
va_end(ap);
}
int count(void) {
static int i = 1;
return i++;
}
void push(void) {
emitlin("\tpush\t%rax");
depth++;
}
void pop(char *arg) {
println("\tpop\t%s", arg);
depth--;
fflush(output_stream);
DCHECK_GE(depth, 0);
}
void push2(void) {
emitlin("\tpush\t%rdx");
emitlin("\tpush\t%rax");
depth++;
depth++;
}
void pop2(char *a, char *b) {
println("\tpop\t%s", a);
println("\tpop\t%s", b);
depth--;
depth--;
DCHECK_GE(depth, 0);
}
void pushreg(char *arg) {
println("\tpush\t%%%s", arg);
depth++;
}
void popreg(char *arg) {
println("\tpop\t%%%s", arg);
depth--;
DCHECK_GE(depth, 0);
}
static const char *nameof(Obj *obj) {
if (obj->asmname) {
return obj->asmname;
} else {
return obj->name;
}
}
static void pushx(void) {
emitlin("\tsub\t$16,%rsp");
emitlin("\tmovdqu\t%xmm0,(%rsp)");
depth += 2;
}
static void popx(int reg) {
println("\tmovdqu\t(%%rsp),%%xmm%d", reg);
emitlin("\tadd\t$16,%rsp");
depth -= 2;
DCHECK_GE(depth, 0);
}
static void pushf(Type *ty) {
if (ty->vector_size == 16) {
pushx();
} else {
emitlin("\tpush\t%rax");
emitlin("\tmovsd\t%xmm0,(%rsp)");
depth++;
}
}
static void popf(Type *ty, int reg) {
if (ty->vector_size == 16) {
popx(reg);
} else {
println("\tmovsd\t(%%rsp),%%xmm%d", reg);
emitlin("\tadd\t$8,%rsp");
depth--;
}
}
static void print_profiling_nop(void) {
emitlin("\t.byte\t0x0f,0x1f,0x44,0x00,0x00");
}
static void print_visibility(Obj *obj) {
if (obj->visibility) {
if (!strcmp(obj->visibility, "hidden")) {
println("\t.hidden\t%s", nameof(obj));
} else if (!strcmp(obj->visibility, "protected")) {
println("\t.protected %s", nameof(obj));
} else {
println("\t.globl\t%s", nameof(obj));
}
} else if (obj->is_static) {
println("\t.local\t%s", nameof(obj));
} else {
println("\t.globl\t%s", nameof(obj));
}
if (obj->is_weak) {
println("\t.weak\t%s", nameof(obj));
}
}
static void print_align(int align) {
if (align > 1) println("\t.align\t%d", align);
}
void print_loc(int64_t file, int64_t line) {
static int64_t lastfile = -1;
static int64_t lastline = -1;
char *locbuf, *p;
if (file != lastfile || line != lastline) {
locbuf = malloc(2 + 4 + 1 + 20 + 1 + 20 + 1);
p = stpcpy(locbuf, "\t.loc\t");
p += int64toarray_radix10(file, p);
*p++ = ' ';
int64toarray_radix10(line, p);
emitlin(locbuf);
lastfile = file;
lastline = line;
}
}
static void print_mov_imm(int64_t val, const char *reg64, const char *reg32) {
if (val) {
if (val != -1) {
if (0 < val && val <= INT32_MAX) {
println("\tmov\t$%#lx,%s", val, reg32);
} else if (INT32_MIN <= val && val < 0) {
println("\tmov\t$%#lx,%s", (int64_t)(int32_t)val, reg64);
} else {
println("\tmov\t$%#lx,%s", val, reg64);
}
} else {
println("\tor\t$-1,%s", reg64);
}
} else {
println("\txor\t%s,%s", reg32, reg32);
}
}
static bool is_quotable_string(const char *s, size_t n) {
size_t i;
if (!n) return false;
if (s[n - 1]) return false;
for (i = 0; i < n - 1; ++i) {
if (s[i] == '"') return false;
if (s[i] == '\\') return false;
if (s[i] < 0x20 || s[i] >= 0x7f) {
return false;
}
}
return true;
}
static char *reg_dx(int sz) {
switch (sz) {
case 1:
return "%dl";
case 2:
return "%dx";
case 4:
return "%edx";
case 8:
return "%rdx";
}
UNREACHABLE();
}
static char *reg_ax(int sz) {
switch (sz) {
case 1:
return "%al";
case 2:
return "%ax";
case 4:
return "%eax";
case 8:
return "%rax";
}
UNREACHABLE();
}
// Compute the absolute address of a given node.
// It's an error if a given node does not reside in memory.
// asm() wants this to not clobber flags or regs other than rax.
void gen_addr(Node *node) {
switch (node->kind) {
case ND_VAR:
// Variable-length array, which is always local.
if (node->var->ty->kind == TY_VLA) {
println("\tmov\t%d(%%rbp),%%rax", node->var->offset);
return;
}
// Local variable
if (node->var->is_local) {
println("\tlea\t%d(%%rbp),%%rax", node->var->offset);
return;
}
if (opt_pic) {
if (node->var->is_tls) {
// Dynamic thread-local variable
println("\tmov\t%%fs:0,%%rax");
println("\tmov\t%s@dtpoff(%%rax),%%rax", nameof(node->var));
return;
}
// Function or global variable
println("\tmov\t%s@gotpcrel(%%rip),%%rax", nameof(node->var));
return;
}
// Static thread-local variable
if (node->var->is_tls) {
println("\tmov\t%%fs:%s@tpoff,%%rax", nameof(node->var));
return;
}
// Here, we generate an absolute address of a function or a global
// variable. Even though they exist at a certain address at runtime,
// their addresses are not known at link-time for the following
// two reasons.
//
// - Address randomization: Executables are loaded to memory as a
// whole but it is not known what address they are loaded to.
// Therefore, at link-time, relative address in the same
// exectuable (i.e. the distance between two functions in the
// same executable) is known, but the absolute address is not
// known.
//
// - Dynamic linking: Dynamic shared objects (DSOs) or .so files
// are loaded to memory alongside an executable at runtime and
// linked by the runtime loader in memory. We know nothing
// about addresses of global stuff that may be defined by DSOs
// until the runtime relocation is complete.
//
// In order to deal with the former case, we use RIP-relative
// addressing, denoted by `(%rip)`. For the latter, we obtain an
// address of a stuff that may be in a shared object file from the
// Global Offset Table using `@GOTPCREL(%rip)` notation.
// Function
if (node->ty->kind == TY_FUNC) {
if (!opt_pic) {
println("\tmov\t$%s,%%eax", nameof(node->var));
} else if (node->var->is_definition) {
println("\tlea\t%s(%%rip),%%rax", nameof(node->var));
} else {
println("\tmov\t%s@gotpcrel(%%rip),%%rax", nameof(node->var));
}
return;
}
// Global variable
if (opt_pic) {
println("\tlea\t%s(%%rip),%%rax", nameof(node->var));
} else {
println("\tmov\t$%s,%%eax", nameof(node->var));
}
return;
case ND_CAST:
gen_expr(node->lhs);
gen_cast(node->lhs->ty, node->ty);
return;
case ND_DEREF:
gen_expr(node->lhs);
return;
case ND_COMMA:
gen_expr(node->lhs);
gen_addr(node->rhs);
return;
case ND_ASSIGN:
case ND_COND:
if (node->ty->kind == TY_STRUCT || node->ty->kind == TY_UNION) {
gen_expr(node);
return;
}
break;
case ND_MEMBER:
gen_addr(node->lhs);
if (node->member->offset) {
println("\tlea\t%d(%%rax),%%rax", node->member->offset);
}
return;
case ND_FUNCALL:
if (node->ret_buffer) {
gen_expr(node);
return;
}
break;
case ND_VLA_PTR:
println("\tlea\t%d(%%rbp),%%rax", node->var->offset);
return;
default:
error_tok(node->tok, "not an lvalue %d", node->kind);
}
}
// Load a value from where %rax is pointing to.
static void load(Type *ty) {
switch (ty->kind) {
case TY_ARRAY:
case TY_STRUCT:
case TY_UNION:
case TY_FUNC:
case TY_VLA:
// If it is an array, do not attempt to load a value to the
// register because in general we can't load an entire array to a
// register. As a result, the result of an evaluation of an array
// becomes not the array itself but the address of the array.
// This is where "array is automatically converted to a pointer to
// the first element of the array in C" occurs.
return;
case TY_FLOAT:
if (ty->vector_size == 16) {
if (ty->align >= 16) {
emitlin("\tmovaps\t(%rax),%xmm0");
} else {
emitlin("\tmovdqu\t(%rax),%xmm0");
}
} else {
emitlin("\tmovss\t(%rax),%xmm0");
return;
}
case TY_DOUBLE:
if (ty->vector_size == 16) {
if (ty->align >= 16) {
emitlin("\tmovapd\t(%rax),%xmm0");
} else {
emitlin("\tmovdqu\t(%rax),%xmm0");
}
} else {
emitlin("\tmovsd\t(%rax),%xmm0");
}
return;
case TY_LDOUBLE:
emitlin("\tfldt\t(%rax)");
return;
case TY_INT128:
emitlin("\tmov\t8(%rax),%rdx");
emitlin("\tmov\t(%rax),%rax");
return;
default:
break;
}
char *insn = ty->is_unsigned ? "movz" : "movs";
// When we load a char or a short value to a register, we always
// extend them to the size of int, so we can assume the lower half of
// a register always contains a valid value. The upper half of a
// register for char, short and int may contain garbage. When we load
// a long value to a register, it simply occupies the entire register.
if (ty->size == 1) {
println("\t%sbl\t(%%rax),%%eax", insn);
} else if (ty->size == 2) {
println("\t%swl\t(%%rax),%%eax", insn);
} else if (ty->size == 4) {
emitlin("\tmovslq\t(%rax),%rax");
} else {
emitlin("\tmov\t(%rax),%rax");
}
}
static void gen_memcpy(size_t size) {
switch (size) {
case 0:
break;
case 1:
emitlin("\tmov\t(%rax),%r8b");
emitlin("\tmov\t%r8b,(%rdi)");
break;
case 2:
emitlin("\tmov\t(%rax),%r8w");
emitlin("\tmov\t%r8w,(%rdi)");
break;
case 3:
emitlin("\tmov\t(%rax),%r8w");
emitlin("\tmov\t%r8w,(%rdi)");
emitlin("\tmov\t2(%rax),%r8b");
emitlin("\tmov\t%r8b,2(%rdi)");
break;
case 4:
emitlin("\tmov\t(%rax),%r8d");
emitlin("\tmov\t%r8d,(%rdi)");
break;
case 5 ... 7:
emitlin("\tmov\t(%rax),%r8d");
println("\tmov\t%d(%%rax),%%r9d", size - 4);
emitlin("\tmov\t%r8d,(%rdi)");
println("\tmov\t%%r9d,%d(%%rdi)", size - 4);
break;
case 8:
emitlin("\tmov\t(%rax),%r8");
emitlin("\tmov\t%r8,(%rdi)");
break;
case 9 ... 15:
emitlin("\tmov\t(%rax),%r8");
println("\tmov\t%d(%%rax),%%r9", size - 8);
emitlin("\tmov\t%r8,(%rdi)");
println("\tmov\t%%r9,%d(%%rdi)", size - 8);
break;
case 16:
emitlin("\tmovdqu\t(%rax),%xmm2");
emitlin("\tmovdqu\t%xmm2,(%rdi)");
break;
case 17 ... 32:
emitlin("\tmovdqu\t(%rax),%xmm2");
println("\tmovdqu\t%d(%%rax),%%xmm3", size - 16);
emitlin("\tmovdqu\t%xmm2,(%rdi)");
println("\tmovdqu\t%%xmm3,%d(%%rdi)", size - 16);
break;
default:
emitlin("\tmov\t%rax,%rsi");
print_mov_imm(size, "%rcx", "%ecx");
emitlin("\trep movsb");
break;
}
}
// Store %rax to an address that the stack top is pointing to.
static void store(Type *ty) {
pop("%rdi");
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
gen_memcpy(ty->size);
return;
case TY_FLOAT:
if (ty->vector_size == 16) {
if (ty->align >= 16) {
emitlin("\tmovaps\t%xmm0,(%rdi)");
} else {
emitlin("\tmovdqu\t%xmm0,(%rdi)");
}
} else {
emitlin("\tmovss\t%xmm0,(%rdi)");
}
return;
case TY_DOUBLE:
if (ty->vector_size == 16) {
if (ty->align >= 16) {
emitlin("\tmovapd\t%xmm0,(%rdi)");
} else {
emitlin("\tmovdqu\t%xmm0,(%rdi)");
}
} else {
emitlin("\tmovsd\t%xmm0,(%rdi)");
}
return;
case TY_LDOUBLE:
emitlin("\tfstpt\t(%rdi)");
return;
case TY_INT128:
emitlin("\tmov\t%rax,(%rdi)");
emitlin("\tmov\t%rdx,8(%rdi)");
return;
}
if (ty->vector_size == 16) {
if (ty->align >= 16) {
emitlin("\tmovdqa\t%xmm0,(%rdi)");
} else {
emitlin("\tmovdqu\t%xmm0,(%rdi)");
}
}
if (ty->size == 1) {
emitlin("\tmov\t%al,(%rdi)");
} else if (ty->size == 2) {
emitlin("\tmov\t%ax,(%rdi)");
} else if (ty->size == 4) {
emitlin("\tmov\t%eax,(%rdi)");
} else {
emitlin("\tmov\t%rax,(%rdi)");
}
}
void cmp_zero(Type *ty) {
switch (ty->kind) {
case TY_FLOAT:
emitlin("\txorps\t%xmm1,%xmm1");
emitlin("\tucomiss\t%xmm1,%xmm0");
return;
case TY_DOUBLE:
emitlin("\txorpd\t%xmm1,%xmm1");
emitlin("\tucomisd\t%xmm1,%xmm0");
return;
case TY_LDOUBLE:
emitlin("\tfldz");
emitlin("\tfucomip");
emitlin("\tfstp\t%st");
return;
case TY_INT128:
emitlin("\tmov\t%rax,%r11");
emitlin("\tor\t%rdx,%r11");
return;
}
if (is_integer(ty) && ty->size <= 4) {
emitlin("\ttest\t%eax,%eax");
} else {
emitlin("\ttest\t%rax,%rax");
}
}
// Structs or unions equal or smaller than 16 bytes are passed
// using up to two registers.
//
// If the first 8 bytes contains only floating-point type members,
// they are passed in an XMM register. Otherwise, they are passed
// in a general-purpose register.
//
// If a struct/union is larger than 8 bytes, the same rule is
// applied to the the next 8 byte chunk.
//
// This function returns true if `ty` has only floating-point
// members in its byte range [lo, hi).
static bool has_flonum(Type *ty, int lo, int hi, int offset) {
if (ty->kind == TY_STRUCT || ty->kind == TY_UNION) {
for (Member *mem = ty->members; mem; mem = mem->next) {
if (!has_flonum(mem->ty, lo, hi, offset + mem->offset)) {
return false;
}
}
return true;
}
if (ty->kind == TY_ARRAY) {
for (int i = 0; i < ty->array_len; i++) {
if (!has_flonum(ty->base, lo, hi, offset + ty->base->size * i)) {
return false;
}
}
return true;
}
return offset < lo || hi <= offset || ty->kind == TY_FLOAT ||
ty->kind == TY_DOUBLE;
}
static bool has_flonum1(Type *ty) {
return has_flonum(ty, 0, 8, 0);
}
static bool has_flonum2(Type *ty) {
return has_flonum(ty, 8, 16, 0);
}
static void push_struct(Type *ty) {
int sz = ROUNDUP(ty->size, 8);
println("\tsub\t$%d,%%rsp", sz);
depth += sz / 8;
for (int i = 0; i < ty->size; i++) {
println("\tmov\t%d(%%rax),%%r10b", i);
println("\tmov\t%%r10b,%d(%%rsp)", i);
}
}
static void push_args2(Node *args, bool first_pass) {
if (!args) return;
push_args2(args->next, first_pass);
if ((first_pass && !args->pass_by_stack) ||
(!first_pass && args->pass_by_stack)) {
return;
}
gen_expr(args);
switch (args->ty->kind) {
case TY_STRUCT:
case TY_UNION:
push_struct(args->ty);
break;
case TY_FLOAT:
case TY_DOUBLE:
pushf(args->ty);
break;
case TY_LDOUBLE:
emitlin("\tsub\t$16,%rsp");
emitlin("\tfstpt\t(%rsp)");
depth += 2;
break;
case TY_INT128:
push2();
break;
default:
push();
}
if (args->realign_stack) {
pushreg("rbx");
}
}
// Load function call arguments. Arguments are already evaluated and
// stored to the stack as local variables. What we need to do in this
// function is to load them to registers or push them to the stack as
// specified by the x86-64 psABI. Here is what the spec says:
//
// - Up to 6 arguments of integral type are passed using RDI, RSI,
// RDX, RCX, R8 and R9.
//
// - Up to 8 arguments of floating-point type are passed using XMM0 to
// XMM7.
//
// - If all registers of an appropriate type are already used, push an
// argument to the stack in the right-to-left order.
//
// - Each argument passed on the stack takes 8 bytes, and the end of
// the argument area must be aligned to a 16 byte boundary.
//
// - If a function is variadic, set the number of floating-point type
// arguments to RAX.
static int push_args(Node *node) {
int stack = 0, gp = 0, fp = 0;
// If the return type is a large struct/union, the caller passes
// a pointer to a buffer as if it were the first argument.
if (node->ret_buffer && node->ty->size > 16) gp++;
// Load as many arguments to the registers as possible.
for (Node *arg = node->args; arg; arg = arg->next) {
Type *ty = arg->ty;
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
if (ty->size > 16) {
arg->pass_by_stack = true;
stack += ROUNDUP(ty->size, 8) / 8;
} else {
bool fp1 = has_flonum1(ty);
bool fp2 = has_flonum2(ty);
if (fp + fp1 + fp2 < FP_MAX && gp + !fp1 + !fp2 < GP_MAX) {
fp = fp + fp1 + fp2;
gp = gp + !fp1 + !fp2;
} else {
arg->pass_by_stack = true;
stack += ROUNDUP(ty->size, 8) / 8;
}
}
break;
case TY_FLOAT:
case TY_DOUBLE:
if (fp++ >= FP_MAX) {
if ((stack & 1) && arg->ty->vector_size == 16) {
arg->realign_stack = true;
++stack;
}
arg->pass_by_stack = true;
++stack;
}
break;
case TY_LDOUBLE:
arg->pass_by_stack = true;
stack += 2;
break;
case TY_INT128:
if (gp + 1 >= GP_MAX) {
if (stack & 1) {
arg->realign_stack = true;
++stack;
}
arg->pass_by_stack = true;
stack += 2;
} else {
gp += 2;
}
break;
default:
if (gp++ >= GP_MAX) {
arg->pass_by_stack = true;
stack++;
}
break;
}
}
if ((depth + stack) % 2 == 1) {
emitlin("\tsub\t$8,%rsp");
depth++;
stack++;
}
push_args2(node->args, true);
push_args2(node->args, false);
// If the return type is a large struct/union, the caller passes
// a pointer to a buffer as if it were the first argument.
if (node->ret_buffer && node->ty->size > 16) {
println("\tlea\t%d(%%rbp),%%rax", node->ret_buffer->offset);
push();
}
return stack;
}
static void copy_ret_buffer(Obj *var) {
/* todo(jart): wat */
Type *ty = var->ty;
int gp = 0, fp = 0;
if (has_flonum1(ty)) {
assert(ty->size == 4 || 8 <= ty->size);
if (ty->size == 4) {
println("\tmovss\t%%xmm0,%d(%%rbp)", var->offset);
} else {
println("\tmovsd\t%%xmm0,%d(%%rbp)", var->offset);
}
fp++;
} else {
for (int i = 0; i < MIN(8, ty->size); i++) {
println("\tmov\t%%al,%d(%%rbp)", var->offset + i);
emitlin("\tshr\t$8,%rax");
}
gp++;
}
if (ty->size > 8) {
if (has_flonum2(ty)) {
assert(ty->size == 12 || ty->size == 16);
if (ty->size == 12) {
println("\tmovss\t%%xmm%d,%d(%%rbp)", fp, var->offset + 8);
} else {
println("\tmovsd\t%%xmm%d,%d(%%rbp)", fp, var->offset + 8);
}
} else {
char *reg1 = (gp == 0) ? "%al" : "%dl";
char *reg2 = (gp == 0) ? "%rax" : "%rdx";
for (int i = 8; i < MIN(16, ty->size); i++) {
println("\tmov\t%s,%d(%%rbp)", reg1, var->offset + i);
println("\tshr\t$8,%s", reg2);
}
}
}
}
static void copy_struct_reg(void) {
/* todo(jart): wat */
Type *ty = current_fn->ty->return_ty;
int gp = 0, fp = 0;
emitlin("\tmov\t%rax,%rdi");
if (has_flonum(ty, 0, 8, 0)) {
assert(ty->size == 4 || 8 <= ty->size);
if (ty->size == 4) {
emitlin("\tmovss\t(%rdi),%xmm0");
} else if (ty->size == 8) {
emitlin("\tmovsd\t(%rdi),%xmm0");
} else {
emitlin("\tmovdqu\t(%rdi),%xmm0");
}
fp++;
} else {
emitlin("\txor\t%eax,%eax");
for (int i = MIN(8, ty->size) - 1; i >= 0; i--) {
emitlin("\tshl\t$8,%rax");
println("\tmov\t%d(%%rdi),%%al", i);
}
gp++;
}
if (ty->size > 8) {
if (has_flonum(ty, 8, 16, 0)) {
assert(ty->size == 12 || ty->size == 16);
if (ty->size == 4) {
println("\tmovss\t8(%%rdi),%%xmm%d", fp);
} else {
println("\tmovsd\t8(%%rdi),%%xmm%d", fp);
}
} else {
char *reg1 = (gp == 0) ? "%al" : "%dl";
char *reg2 = (gp == 0) ? "%rax" : "%rdx";
char *reg3 = (gp == 0) ? "%eax" : "%edx";
println("\txor\t%s,%s", reg3, reg3);
for (int i = MIN(16, ty->size) - 1; i >= 8; i--) {
println("\tshl\t$8,%s", reg2);
println("\tmov\t%d(%%rdi),%s", i, reg1);
}
}
}
}
static void copy_struct_mem(void) {
Type *ty = current_fn->ty->return_ty;
Obj *var = current_fn->params;
println("\tmov\t%d(%%rbp),%%rdi", var->offset);
for (int i = 0; i < ty->size; i++) {
println("\tmov\t%d(%%rax),%%dl", i);
println("\tmov\t%%dl,%d(%%rdi)", i);
}
}
static void builtin_alloca(void) {
// Align size to 16 bytes.
emitlin("\tadd\t$15,%rdi");
emitlin("\tand\t$-16,%rdi");
// Shift the temporary area by %rdi.
println("\tmov\t%d(%%rbp),%%rcx", current_fn->alloca_bottom->offset);
emitlin("\
\tsub\t%rsp,%rcx\n\
\tmov\t%rsp,%rax\n\
\tsub\t%rdi,%rsp\n\
\tmov\t%rsp,%rdx\n\
1:\n\
\ttest\t%rcx,%rcx\n\
\tje\t2f\n\
\tmov\t(%rax),%r8b\n\
\tmov\t%r8b,(%rdx)\n\
\tinc\t%rdx\n\
\tinc\t%rax\n\
\tdec\t%rcx\n\
\tjmp\t1b\n\
2:");
// Move alloca_bottom pointer.
println("\tmov\t%d(%%rbp),%%rax", current_fn->alloca_bottom->offset);
emitlin("\tsub\t%rdi,%rax");
println("\tmov\t%%rax,%d(%%rbp)", current_fn->alloca_bottom->offset);
}
static void gen_comis(Node *node, const char *op, int a, int b,
const char *pred) {
if (a != b) {
gen_expr(node->args->next);
pushf(node->args->next->ty);
gen_expr(node->args);
popf(node->args->next->ty, 1);
} else {
gen_expr(node->args);
}
emitlin("\txor\t%eax,%eax");
println("\t%s\t%%xmm%d,%%xmm%d", op, a, b);
println("\tset%s\t%%al", pred);
}
static bool gen_builtin_funcall(Node *node, const char *name) {
if (!strcmp(name, "alloca")) {
gen_expr(node->args);
emitlin("\tmov\t%rax,%rdi");
builtin_alloca();
return true;
} else if (!strcmp(name, "trap")) {
emitlin("\tint3");
return true;
} else if (!strcmp(name, "unreachable")) {
emitlin("\tud2");
return true;
} else if (!strcmp(name, "frame_address")) {
if (is_const_expr(node->args) && !eval(node->args)) {
emitlin("\tmov\t%rbp,%rax");
} else {
error_tok(node->args->tok, "must be 0");
}
return true;
} else if (!strcmp(name, "ctz")) {
gen_expr(node->args);
emitlin("\tbsf\t%eax,%eax");
return true;
} else if (!strcmp(name, "ctzl") || !strcmp(name, "ctzll")) {
gen_expr(node->args);
emitlin("\tbsf\t%rax,%rax");
return true;
} else if (!strcmp(name, "clz")) {
gen_expr(node->args);
emitlin("\
\tbsr\t%eax,%eax\n\
\txor\t$31,%eax");
return true;
} else if (!strcmp(name, "clzl") || !strcmp(name, "clzll")) {
gen_expr(node->args);
emitlin("\
\tbsr\t%rax,%rax\n\
\txor\t$63,%eax");
return true;
} else if (!strcmp(name, "ffs") || !strcmp(name, "ffsl") ||
!strcmp(name, "ffsll")) {
char regprefix;
gen_expr(node->args);
emitlin("\tor\t$-1,%edi");
regprefix = endswith(name, "l") ? 'r' : 'e';
println("\tbsf\t%%%cax,%%%cax", regprefix, regprefix);
emitlin("\tcmovz\t%edi,%eax");
emitlin("\tinc\t%eax");
return true;
} else if (!strcmp(name, "bswap16")) {
gen_expr(node->args);
emitlin("\txchg\t%al,%ah");
return true;
} else if (!strcmp(name, "bswap32")) {
gen_expr(node->args);
emitlin("\tbswap\t%eax");
return true;
} else if (!strcmp(name, "bswap64")) {
gen_expr(node->args);
emitlin("\tbswap\t%rax");
return true;
} else if (!strcmp(name, "popcount")) {
gen_expr(node->args);
if (opt_popcnt) {
emitlin("\tpopcnt\t%eax,%eax");
} else {
emitlin("\tmov\t%eax,%edi");
emitlin("\tcall\t__popcountsi2");
}
return true;
} else if (!strcmp(name, "popcountl") || !strcmp(name, "popcountll")) {
gen_expr(node->args);
if (opt_popcnt) {
emitlin("\tpopcnt\t%rax,%rax");
} else {
emitlin("\tmov\t%rax,%rdi");
emitlin("\tcall\t__popcountdi2");
}
return true;
} else if (!strcmp(name, "memcpy")) {
if (is_const_expr(node->args->next->next)) {
gen_expr(node->args);
push();
gen_expr(node->args->next);
emitlin("\tmov\t(%rsp),%rdi");
gen_memcpy(eval(node->args->next->next));
pop("%rax");
return true;
}
} else if (!strcmp(name, "isgreater")) {
gen_comis(node, "comisd", 1, 0, "a");
return true;
} else if (!strcmp(name, "isgreaterequal")) {
gen_comis(node, "comisd", 1, 0, "ae");
return true;
} else if (!strcmp(name, "isless")) {
gen_comis(node, "comisd", 0, 1, "a");
return true;
} else if (!strcmp(name, "islessequal")) {
gen_comis(node, "comisd", 0, 1, "ae");
return true;
} else if (!strcmp(name, "islessgreater")) {
gen_comis(node, "ucomisd", 1, 0, "ne");
return true;
} else if (!strcmp(name, "isunordered")) {
gen_comis(node, "ucomisd", 1, 0, "p");
return true;
} else if (!strcmp(name, "isnan")) {
gen_comis(node, "ucomisd", 0, 0, "p");
return true;
} else if (!strcmp(name, "nanf")) {
emitlin("\
\tmov\t$0x7fc00000,%eax\n\
\tmovd\t%eax,%xmm0");
return true;
} else if (!strcmp(name, "nan")) {
emitlin("\
\tmov\t$0x7fffffffffffffff,%rax\n\
\tmovq\t%rax,%xmm0");
return true;
} else if (!strcmp(name, "nanl")) {
emitlin("\
\tpush\t$0x7fc00000\n\
\tflds\t(%rsp)\n\
\tpop\t%rax");
return true;
} else if (!strcmp(name, "inff")) {
emitlin("\
\tmov\t$0x7f800000,%eax\n\
\tmovd\t%eax,%xmm0");
return true;
} else if (!strcmp(name, "inf")) {
emitlin("\
\tmov\t$0x7ff0000000000000,%rax\n\
\tmovq\t%rax,%xmm0");
return true;
} else if (!strcmp(name, "infl")) {
emitlin("\
\tpush\t$0x7f800000\n\
\tflds\t(%rsp)\n\
\tpop\t%rax");
return true;
} else if (!strcmp(name, "isinf")) {
gen_expr(node->args);
emitlin("\
\tmov\t$0x7fffffffffffffff,%rax\n\
\tmovq\t%rax,%xmm1\n\
\tandps\t%xmm1,%xmm0\n\
\tmov\t$0x7fefffffffffffff,%rax\n\
\tmovq\t%rax,%xmm1\n\
\txor\t%eax,%eax\n\
\tcomisd\t%xmm1,%xmm0\n\
\tseta\t%al");
return true;
} else if (!strcmp(name, "isfinite")) {
gen_expr(node->args);
emitlin("\
\tmov\t$0x7fffffffffffffff,%rax\n\
\tmovq\t%rax,%xmm1\n\
\tandps\t%xmm1,%xmm0\n\
\tmov\t$0x7fefffffffffffff,%rax\n\
\tmovq\t%rax,%xmm1\n\
\txor\t%eax,%eax\n\
\tcomisd\t%xmm0,%xmm1\n\
\tsetnb\t%al");
return true;
} else if (!strcmp(name, "signbitf")) {
gen_expr(node->args);
emitlin("\
\tmovd\t%xmm0,%eax\n\
\tand\t$-2147483648,%eax");
return true;
} else if (!strcmp(name, "signbit")) {
gen_expr(node->args);
emitlin("\
\tmovmskpd\t%xmm0,%eax\n\
\tand\t$1,%eax");
return true;
} else if (!strcmp(name, "signbitl")) {
gen_expr(node->args);
emitlin("\
\tfxam\n\
\tfnstsw\t%ax\n\
\tfstp\t%st\n\
\tand\t$0x200,%eax");
return true;
}
return false;
}
static int GetSseIntSuffix(Type *ty) {
switch (ty->kind) {
case TY_CHAR:
return 'b';
case TY_SHORT:
return 'w';
case TY_INT:
return 'd';
case TY_LONG:
return 'q';
default:
UNREACHABLE();
}
}
static bool IsOverflowArithmetic(Node *node) {
return (node->kind == ND_ADD || node->kind == ND_SUB ||
node->kind == ND_MUL || node->kind == ND_NEG) &&
node->overflow;
}
static void HandleOverflow(const char *ax) {
pop("%rdi");
println("\tmov\t%s,(%%rdi)", ax);
emitlin("\tseto\t%al");
emitlin("\tmovzbl\t%al,%eax");
}
// Generate code for a given node.
void gen_expr(Node *node) {
char fbuf[32];
print_loc(node->tok->file->file_no, node->tok->line_no);
switch (node->kind) {
case ND_NULL_EXPR:
return;
case ND_NUM: {
switch (node->ty->kind) {
case TY_FLOAT: {
union {
float f32;
uint32_t u32;
} u = {node->fval};
g_ffmt_p(fbuf, &u.f32, 7, sizeof(fbuf), 0);
println("\tmov\t$%#x,%%eax\t# float %s", u.u32, fbuf);
emitlin("\tmovq\t%rax,%xmm0");
return;
}
case TY_DOUBLE: {
union {
double f64;
uint64_t u64;
} u = {node->fval};
g_dfmt_p(fbuf, &u.f64, 16, sizeof(fbuf), 0);
println("\tmov\t$%#lx,%%rax\t# double %s", u.u64, fbuf);
emitlin("\tmovq\t%rax,%xmm0");
return;
}
case TY_LDOUBLE: {
union {
long double f80;
uint64_t u64[2];
} u;
memset(&u, 0, sizeof(u));
u.f80 = node->fval;
g_xfmt_p(fbuf, &u.f80, 19, sizeof(fbuf), 0);
println("\tmov\t$%lu,%%rax\t# long double %s", u.u64[0], fbuf);
emitlin("\tmov\t%rax,-16(%rsp)");
println("\tmov\t$%lu,%%rax", u.u64[1]);
emitlin("\tmov\t%rax,-8(%rsp)");
emitlin("\tfldt\t-16(%rsp)");
return;
}
case TY_INT128:
print_mov_imm(node->val, "%rax", "%eax");
if (node->ty->is_unsigned) {
emitlin("\txor\t%edx,%edx");
} else {
emitlin("\tcqto");
}
return;
default:
print_mov_imm(node->val, "%rax", "%eax");
return;
}
}
case ND_NEG:
if (IsOverflowArithmetic(node)) {
gen_expr(node->overflow);
push();
}
gen_expr(node->lhs);
switch (node->ty->kind) {
case TY_FLOAT:
emitlin("\tmov\t$-2147483648,%eax");
emitlin("\tmovd\t%eax,%xmm1");
if (node->ty->vector_size == 16) {
emitlin("\tpshufd\t$0,%xmm1");
}
emitlin("\txorps\t%xmm1,%xmm0");
return;
case TY_DOUBLE:
emitlin("\tmov\t$1,%eax");
emitlin("\tror\t%rax");
emitlin("\tmovq\t%rax,%xmm1");
if (node->ty->vector_size == 16) {
emitlin("\tpshufd\t$0b01000100,%xmm1");
}
emitlin("\txorpd\t%xmm1,%xmm0");
return;
case TY_LDOUBLE:
emitlin("\tfchs");
return;
case TY_INT128:
emitlin("\tneg\t%rax");
emitlin("\tadc\t$0,%rdx");
emitlin("\tneg\t%rdx");
return;
}
char *ax;
if (node->lhs->ty->kind == TY_LONG || node->lhs->ty->base) {
ax = "%rax";
} else {
ax = "%eax";
}
println("\tneg\t%s", ax);
if (IsOverflowArithmetic(node)) {
HandleOverflow(ax);
}
return;
case ND_VAR:
gen_addr(node);
load(node->ty);
return;
case ND_MEMBER: {
gen_addr(node);
load(node->ty);
Member *mem = node->member;
if (mem->is_bitfield) {
println("\tshl\t$%d,%%rax", 64 - mem->bit_width - mem->bit_offset);
if (mem->ty->is_unsigned) {
println("\tshr\t$%d,%%rax", 64 - mem->bit_width);
} else {
println("\tsar\t$%d,%%rax", 64 - mem->bit_width);
}
}
return;
}
case ND_DEREF:
gen_expr(node->lhs);
load(node->ty);
return;
case ND_ADDR:
gen_addr(node->lhs);
return;
case ND_ASSIGN:
gen_addr(node->lhs);
push();
gen_expr(node->rhs);
if (node->lhs->kind == ND_MEMBER && node->lhs->member->is_bitfield) {
// If the lhs is a bitfield, we need to read the current value
// from memory and merge it with a new value.
emitlin("\tmov\t%rax,%r8");
Member *mem = node->lhs->member;
emitlin("\tmov\t%rax,%rdi");
println("\tshl\t$%d,%%rdi", 64 - mem->bit_width);
println("\tshr\t$%d,%%rdi", 64 - mem->bit_width - mem->bit_offset);
emitlin("\tmov\t(%rsp),%rax");
load(mem->ty);
unsigned long mask = ((1ul << mem->bit_width) - 1) << mem->bit_offset;
println("\tmov\t$%#lx,%%r9", ~mask);
emitlin("\tand\t%r9,%rax");
emitlin("\tor\t%rdi,%rax");
store(node->ty);
emitlin("\tmov\t%r8,%rax");
return;
}
store(node->ty);
return;
case ND_STMT_EXPR:
for (Node *n = node->body; n; n = n->next) gen_stmt(n);
return;
case ND_COMMA:
gen_expr(node->lhs);
gen_expr(node->rhs);
return;
case ND_CAST:
gen_expr(node->lhs);
gen_cast(node->lhs->ty, node->ty);
return;
case ND_MEMZERO:
switch (node->var->ty->size) {
case 1:
println("\tmovb\t$0,%d(%%rbp)", node->var->offset);
break;
case 2:
println("\tmovw\t$0,%d(%%rbp)", node->var->offset);
break;
case 4:
println("\tmovl\t$0,%d(%%rbp)", node->var->offset);
break;
case 8:
println("\tmovq\t$0,%d(%%rbp)", node->var->offset);
break;
case 9 ... 16:
emitlin("\txor\t%eax,%eax");
println("\tmov\t%%rax,%d(%%rbp)", node->var->offset);
println("\tmov\t%%rax,%d(%%rbp)",
node->var->offset + 8 - (16 - node->var->ty->size));
break;
default:
// `rep stosb` is equivalent to `memset(%rdi, %al, %rcx)`.
print_mov_imm(node->var->ty->size, "%rcx", "%ecx");
println("\tlea\t%d(%%rbp),%%rdi", node->var->offset);
emitlin("\txor\t%eax,%eax");
emitlin("\trep stosb");
break;
}
return;
case ND_COND: {
int c = count();
gen_expr(node->cond);
cmp_zero(node->cond->ty);
println("\tje\t.L.else.%d", c);
gen_expr(node->then);
println("\tjmp\t.L.end.%d", c);
println(".L.else.%d:", c);
gen_expr(node->els);
println(".L.end.%d:", c);
return;
}
case ND_NOT:
gen_expr(node->lhs);
cmp_zero(node->lhs->ty);
emitlin("\tsete\t%al");
emitlin("\tmovzbl\t%al,%rax");
return;
case ND_BITNOT:
gen_expr(node->lhs);
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\tnot\t%rax");
emitlin("\tnot\t%rdx");
} else if (node->lhs->ty->vector_size == 16) {
emitlin("\tpcmpeqd\t%xmm1,%xmm1");
emitlin("\tpxor\t%xmm1,%xmm0");
} else {
emitlin("\tnot\t%rax");
}
return;
case ND_LOGAND: {
int c = count();
gen_expr(node->lhs);
cmp_zero(node->lhs->ty);
println("\tje\t.L.false.%d", c);
gen_expr(node->rhs);
cmp_zero(node->rhs->ty);
println("\tje\t.L.false.%d", c);
emitlin("\tmov\t$1,%eax");
println("\tjmp\t.L.end.%d", c);
println(".L.false.%d:", c);
emitlin("\txor\t%eax,%eax");
println(".L.end.%d:", c);
return;
}
case ND_LOGOR: {
int c = count();
gen_expr(node->lhs);
cmp_zero(node->lhs->ty);
println("\tjne\t.L.true.%d", c);
gen_expr(node->rhs);
cmp_zero(node->rhs->ty);
println("\tjne\t.L.true.%d", c);
emitlin("\txor\t%eax,%eax");
println("\tjmp\t.L.end.%d", c);
println(".L.true.%d:", c);
emitlin("\tmov\t$1,%eax");
println(".L.end.%d:", c);
return;
}
case ND_FUNCALL: {
const char *funcname = NULL;
if (node->lhs->kind == ND_VAR) {
if (startswith(nameof(node->lhs->var), "__builtin_")) {
funcname = nameof(node->lhs->var) + 10;
if (gen_builtin_funcall(node, funcname)) {
return;
}
} else if (!opt_no_builtin) {
if (gen_builtin_funcall(node, nameof(node->lhs->var))) {
return;
}
}
}
int stack_args = push_args(node);
if (!funcname) {
if (node->lhs->kind == ND_VAR && !node->lhs->var->is_local &&
!node->lhs->var->is_tls && node->lhs->ty->kind == TY_FUNC) {
if (!opt_pic || node->lhs->var->is_definition) {
funcname = nameof(node->lhs->var);
} else {
funcname = xasprintf("%s@gotpcrel(%%rip)", nameof(node->lhs->var));
}
} else {
gen_expr(node->lhs);
funcname = NULL;
}
}
int gp = 0, fp = 0;
// If the return type is a large struct/union, the caller passes
// a pointer to a buffer as if it were the first argument.
if (node->ret_buffer && node->ty->size > 16) {
pop(argreg64[gp++]);
}
for (Node *arg = node->args; arg; arg = arg->next) {
Type *ty = arg->ty;
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
if (ty->size > 16) continue;
bool fp1 = has_flonum1(ty);
bool fp2 = has_flonum2(ty);
if (fp + fp1 + fp2 < FP_MAX && gp + !fp1 + !fp2 < GP_MAX) {
if (fp1) {
popf(ty, fp++);
} else {
pop(argreg64[gp++]);
}
if (ty->size > 8) {
if (fp2) {
popf(ty, fp++);
} else {
pop(argreg64[gp++]);
}
}
}
break;
case TY_FLOAT:
case TY_DOUBLE:
if (fp < FP_MAX) {
popf(ty, fp++);
}
break;
case TY_LDOUBLE:
break;
case TY_INT128:
if (gp + 1 < GP_MAX) {
int a = gp++;
int b = gp++;
pop2(argreg64[a], argreg64[b]);
}
break;
default:
if (gp < GP_MAX) {
pop(argreg64[gp++]);
}
break;
}
}
if (funcname) {
if (node->lhs->ty->is_variadic) {
print_mov_imm(fp, "%rax", "%eax");
}
println("\tcall\t%s", funcname);
} else {
if (!node->lhs->ty->is_variadic) {
emitlin("\tcall\t*%rax");
} else {
emitlin("\tmov\t%rax,%r10");
print_mov_imm(fp, "%rax", "%eax");
emitlin("\tcall\t*%r10");
}
}
if (stack_args) {
println("\tadd\t$%d,%%rsp", stack_args * 8);
}
depth -= stack_args;
// It looks like the most significant 48 or 56 bits in RAX may
// contain garbage if a function return type is short or bool/char,
// respectively. We clear the upper bits here.
switch (node->ty->kind) {
case TY_BOOL:
emitlin("\tmovzbl\t%al,%eax");
return;
case TY_CHAR:
if (node->ty->is_unsigned) {
emitlin("\tmovzbl\t%al,%eax");
} else {
emitlin("\tmovsbl\t%al,%eax");
}
return;
case TY_SHORT:
if (node->ty->is_unsigned) {
emitlin("\tmovzwl\t%ax,%eax");
} else {
emitlin("\tcwtl");
}
return;
}
// If the return type is a small struct, a value is returned
// using up to two registers.
if (node->ret_buffer && node->ty->size <= 16) {
copy_ret_buffer(node->ret_buffer);
println("\tlea\t%d(%%rbp),%%rax", node->ret_buffer->offset);
}
return;
}
case ND_LABEL_VAL:
if (opt_pic) {
println("\tlea\t%s(%%rip),%%rax", node->unique_label);
} else {
println("\tmov\t$%s,%%eax", node->unique_label);
}
return;
case ND_CAS: {
gen_expr(node->cas_addr);
push();
gen_expr(node->cas_new);
push();
gen_expr(node->cas_old);
emitlin("\tmov\t%rax,%r8");
load(node->cas_old->ty->base);
pop("%rdx"); // new
pop("%rdi"); // addr
int sz = node->cas_addr->ty->base->size;
println("\tlock cmpxchg %s,(%%rdi)", reg_dx(sz));
emitlin("\tsete\t%cl");
emitlin("\tje\t1f");
println("\tmov\t%s,(%%r8)", reg_ax(sz));
emitlin("1:");
emitlin("\tmovzbl\t%cl,%eax");
return;
}
case ND_EXCH: {
gen_expr(node->lhs);
push();
gen_expr(node->rhs);
pop("%rdi");
int sz = node->lhs->ty->base->size;
println("\txchg\t%s,(%%rdi)", reg_ax(sz));
return;
}
case ND_FPCLASSIFY:
gen_fpclassify(node->fpc);
return;
}
switch (node->lhs->ty->kind) {
case TY_FLOAT:
case TY_DOUBLE: {
gen_expr(node->rhs);
pushf(node->rhs->ty);
gen_expr(node->lhs);
popf(node->rhs->ty, 1);
char sd = node->lhs->ty->kind == TY_FLOAT ? 's' : 'd';
char ps = node->lhs->ty->vector_size == 16 ? 'p' : 's';
switch (node->kind) {
case ND_ADD:
println("\tadd%c%c\t%%xmm1,%%xmm0", ps, sd);
return;
case ND_SUB:
println("\tsub%c%c\t%%xmm1,%%xmm0", ps, sd);
return;
case ND_MUL:
println("\tmul%c%c\t%%xmm1,%%xmm0", ps, sd);
return;
case ND_DIV:
println("\tdiv%c%c\t%%xmm1,%%xmm0", ps, sd);
return;
case ND_EQ:
case ND_NE:
case ND_LT:
case ND_LE:
println("\tucomis%c\t%%xmm0,%%xmm1", sd);
if (node->kind == ND_EQ) {
emitlin("\
\tsete\t%al\n\
\tsetnp\t%dl\n\
\tand\t%dl,%al");
} else if (node->kind == ND_NE) {
emitlin("\
\tsetne\t%al\n\
\tsetp\t%dl\n\
\tor\t%dl,%al");
} else if (node->kind == ND_LT) {
emitlin("\tseta\t%al");
} else {
emitlin("\tsetae\t%al");
}
emitlin("\tand\t$1,%al");
emitlin("\tmovzbl\t%al,%eax");
return;
}
error_tok(node->tok, "invalid expression");
}
case TY_LDOUBLE: {
gen_expr(node->lhs);
gen_expr(node->rhs);
switch (node->kind) {
case ND_ADD:
emitlin("\tfaddp");
return;
case ND_SUB:
emitlin("\tfsubrp");
return;
case ND_MUL:
emitlin("\tfmulp");
return;
case ND_DIV:
emitlin("\tfdivrp");
return;
case ND_EQ:
case ND_NE:
case ND_LT:
case ND_LE:
emitlin("\tfcomip");
emitlin("\tfstp\t%st");
if (node->kind == ND_EQ)
emitlin("\tsete\t%al");
else if (node->kind == ND_NE)
emitlin("\tsetne\t%al");
else if (node->kind == ND_LT)
emitlin("\tseta\t%al");
else
emitlin("\tsetae\t%al");
emitlin("\tmovzbl\t%al,%eax");
return;
}
error_tok(node->tok, "invalid expression");
}
}
if (IsOverflowArithmetic(node)) {
gen_expr(node->overflow);
push();
}
if (node->lhs->ty->vector_size == 16) {
gen_expr(node->rhs);
pushx();
gen_expr(node->lhs);
popx(1);
} else if (node->rhs->ty->kind == TY_INT128) {
gen_expr(node->rhs);
push2();
gen_expr(node->lhs);
pop2("%rdi", "%rsi");
} else if (!opt_pic && is_const_expr(node->rhs)) {
/* shortcut path for immediates */
char **label = NULL;
uint64_t val = eval2(node->rhs, &label);
gen_expr(node->lhs);
if (label) {
println("\tmov\t$%s%+ld,%%edi", *label, val);
} else {
print_mov_imm(val, "%rdi", "%edi");
}
} else {
gen_expr(node->rhs);
push();
gen_expr(node->lhs);
pop("%rdi");
}
char *ax, *di, *dx;
if (node->lhs->ty->kind == TY_LONG || node->lhs->ty->base) {
ax = "%rax";
di = "%rdi";
dx = "%rdx";
} else {
ax = "%eax";
di = "%edi";
dx = "%edx";
}
switch (node->kind) {
case ND_ADD:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\tadd\t%rdi,%rax");
emitlin("\tadc\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
println("\tpadd%c\t%%xmm1,%%xmm0", GetSseIntSuffix(node->lhs->ty));
} else {
println("\tadd\t%s,%s", di, ax);
}
break;
case ND_SUB:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\tsub\t%rdi,%rax");
emitlin("\tsbb\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
println("\tpsub%c\t%%xmm1,%%xmm0", GetSseIntSuffix(node->lhs->ty));
} else {
println("\tsub\t%s,%s", di, ax);
}
break;
case ND_MUL:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\
\timul\t%rdi,%rdx\n\
\timul\t%rax,%rsi\n\
\tadd\t%rdx,%rsi\n\
\tmul\t%rdi\n\
\tadd\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
switch (node->lhs->ty->kind) {
case TY_CHAR:
emitlin("\
\tmovaps\t%xmm1,%xmm2\n\
\tmovaps\t%xmm0,%xmm3\n\
\tpunpcklbw %xmm0,%xmm3\n\
\tpunpcklbw %xmm1,%xmm2\n\
\tpmullw\t%xmm3,%xmm2\n\
\tpunpckhbw %xmm0,%xmm0\n\
\tpcmpeqd\t%xmm3,%xmm3\n\
\tpunpckhbw %xmm1,%xmm1\n\
\tpmullw\t%xmm0,%xmm1\n\
\tandps\t%xmm3,%xmm2\n\
\tmovaps\t%xmm2,%xmm0\n\
\tandps\t%xmm3,%xmm1\n\
\tpackuswb %xmm1,%xmm0");
break;
case TY_SHORT:
emitlin("\tpmullw\t%xmm1,%xmm0");
break;
case TY_INT:
if (opt_sse4) {
emitlin("\tpmulld\t%xmm1,%xmm0");
} else {
emitlin("\
\tmovaps\t%xmm0,%xmm2\n\
\tpsrlq\t$32,%xmm0\n\
\tpmuludq\t%xmm1,%xmm2\n\
\tpsrlq\t$32,%xmm1\n\
\tpmuludq\t%xmm1,%xmm0\n\
\tpshufd\t$8,%xmm2,%xmm2\n\
\tpshufd\t$8,%xmm0,%xmm1\n\
\tpunpckldq %xmm1,%xmm2\n\
\tmovaps\t%xmm2,%xmm0");
}
case TY_LONG:
emitlin("\
\tmovaps\t%xmm1,%xmm2\n\
\tmovaps\t%xmm0,%xmm3\n\
\tpmuludq\t%xmm1,%xmm3\n\
\tmovaps\t%xmm2,%xmm4\n\
\tmovaps\t%xmm0,%xmm1\n\
\tpsrlq\t$32,%xmm4\n\
\tpsrlq\t$32,%xmm1\n\
\tpmuludq\t%xmm4,%xmm0\n\
\tpmuludq\t%xmm2,%xmm1\n\
\tpaddq\t%xmm0,%xmm1\n\
\tpsllq\t$32,%xmm1\n\
\tpaddq\t%xmm1,%xmm3\n\
\tmovaps\t%xmm3,%xmm0");
break;
default:
UNREACHABLE();
}
} else {
println("\timul\t%s,%s", di, ax);
}
break;
case ND_DIV:
case ND_REM:
if (node->lhs->ty->kind == TY_INT128) {
bool skew;
if ((skew = (depth & 1))) {
emitlin("\tsub\t$8,%rsp");
depth++;
}
emitlin("\
\tmov\t%rsi,%rcx\n\
\tmov\t%rdx,%rsi\n\
\tmov\t%rdi,%rdx\n\
\tmov\t%rax,%rdi");
if (node->kind == ND_DIV) {
if (node->ty->is_unsigned) {
emitlin("\tcall\t__udivti3");
} else {
emitlin("\tcall\t__divti3");
}
} else {
if (node->ty->is_unsigned) {
emitlin("\tcall\t__umodti3");
} else {
emitlin("\tcall\t__modti3");
}
}
if (skew) {
emitlin("\tadd\t$8,%rsp");
depth--;
}
} else if (node->lhs->ty->vector_size == 16) {
error_tok(node->tok, "no div/rem sse instruction");
} else {
if (node->ty->is_unsigned) {
emitlin("\txor\t%edx,%edx");
println("\tdiv\t%s", di);
} else {
if (node->lhs->ty->size == 8) {
emitlin("\tcqto");
} else {
emitlin("\tcltd");
}
println("\tidiv\t%s", di);
}
if (node->kind == ND_REM) {
emitlin("\tmov\t%rdx,%rax");
}
}
break;
case ND_BINAND:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\tand\t%rdi,%rax");
emitlin("\tand\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
emitlin("\tpand\t%xmm1,%xmm0");
} else {
println("\tand\t%s,%s", di, ax);
}
break;
case ND_BINOR:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\tor\t%rdi,%rax");
emitlin("\tor\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
emitlin("\tpor\t%xmm1,%xmm0");
} else {
println("\tor\t%s,%s", di, ax);
}
break;
case ND_BINXOR:
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\txor\t%rdi,%rax");
emitlin("\txor\t%rsi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
emitlin("\tpxor\t%xmm1,%xmm0");
} else {
println("\txor\t%s,%s", di, ax);
}
break;
case ND_EQ:
case ND_NE:
case ND_LT:
case ND_LE:
if (node->lhs->ty->kind == TY_INT128) {
switch (node->kind) {
case ND_EQ:
emitlin("\
\txor\t%rax,%rdi\n\
\txor\t%rdx,%rsi\n\
\tor\t%rsi,%rdi\n\
\tsete\t%al");
break;
case ND_NE:
emitlin("\
\txor\t%rax,%rdi\n\
\txor\t%rdx,%rsi\n\
\tor\t%rsi,%rdi\n\
\tsetne\t%al");
break;
case ND_LT:
if (node->lhs->ty->is_unsigned) {
emitlin("\
\tcmp\t%rdi,%rax\n\
\tmov\t%rdx,%rax\n\
\tsbb\t%rsi,%rax\n\
\tsetc\t%al");
} else {
emitlin("\
\tcmp\t%rdi,%rax\n\
\tmov\t%rdx,%rax\n\
\tsbb\t%rsi,%rax\n\
\tsetl\t%al");
}
break;
case ND_LE:
if (node->lhs->ty->is_unsigned) {
emitlin("\
\tcmp\t%rax,%rdi\n\
\tsbb\t%rdx,%rsi\n\
\tsetnc\t%al");
} else {
emitlin("\
\tcmp\t%rax,%rdi\n\
\tsbb\t%rdx,%rsi\n\
\tsetge\t%al");
}
break;
}
} else if (node->lhs->ty->vector_size == 16) {
switch (node->kind) {
case ND_EQ:
switch (node->lhs->ty->kind) {
case TY_CHAR:
case TY_SHORT:
case TY_INT:
println("\tpcmpeq%c\t%%xmm1,%%xmm0",
GetSseIntSuffix(node->lhs->ty));
break;
default:
error_tok(node->tok, "todo sse eq");
}
break;
case ND_NE:
switch (node->lhs->ty->kind) {
case TY_CHAR:
case TY_SHORT:
case TY_INT:
println("\tpcmpeq%c\t%%xmm1,%%xmm0",
GetSseIntSuffix(node->lhs->ty));
emitlin("\tpcmpeqd\t%xmm1,%xmm1\n"
"\tandnps\t%xmm1,%xmm0");
break;
default:
error_tok(node->tok, "todo sse ne");
}
break;
case ND_LT:
if (node->lhs->ty->is_unsigned) {
switch (node->lhs->ty->kind) {
case TY_CHAR:
case TY_SHORT:
println("\tpsubus%c\t%%xmm0,%%xmm1",
GetSseIntSuffix(node->lhs->ty));
emitlin("\tmovaps\t%xmm1,%xmm0\n"
"\txorps\t%xmm1,%xmm1");
println("\tpcmpeq%c\t%%xmm1,%%xmm0",
GetSseIntSuffix(node->lhs->ty));
emitlin("\tpcmpeqd\t%xmm1,%xmm1\n"
"\tandnps\t%xmm1,%xmm0");
break;
default:
error_tok(node->tok, "todo sse ltu");
}
} else {
switch (node->lhs->ty->kind) {
case TY_CHAR:
case TY_SHORT:
case TY_INT:
println("\tpcmpgt%c\t%%xmm0,%%xmm1",
GetSseIntSuffix(node->lhs->ty));
emitlin("\tmovaps\t%xmm1,%xmm0");
break;
default:
error_tok(node->tok, "todo sse lt");
}
}
break;
case ND_LE:
if (node->lhs->ty->is_unsigned) {
switch (node->lhs->ty->kind) {
case TY_CHAR:
emitlin("\tpminub\t%xmm0,%xmm1\n"
"\tpcmpeqb\t%xmm1,%xmm0");
break;
case TY_SHORT:
emitlin("\tpsubusw\t%xmm1,%xmm0\n"
"\txorps\t%xmm1,%xmm1\n"
"\tpcmpeqw\t%xmm1,%xmm0");
break;
case TY_INT:
emitlin("\tmov\t$-2147483648,%eax\n"
"\tmovd\t%eax,%xmm2\n"
"\tpshufd\t$0,%xmm2");
emitlin("\tpcmpgtd\t%xmm1,%xmm0");
emitlin("\tpcmpeqd\t%xmm1,%xmm1\n"
"\tandnps\t%xmm1,%xmm0");
break;
default:
error_tok(node->tok, "todo sse leu");
}
} else {
switch (node->lhs->ty->kind) {
case TY_SHORT:
emitlin("\tpminsw\t%xmm0,%xmm1\n"
"\tpcmpeqw\t%xmm1,%xmm0");
break;
case TY_CHAR:
case TY_INT:
println("\tpcmpgt%c\t%%xmm1,%%xmm0",
GetSseIntSuffix(node->lhs->ty));
emitlin("\tpcmpeqd\t%xmm1,%xmm1\n"
"\tandnps\t%xmm1,%xmm0");
break;
default:
error_tok(node->tok, "todo sse le");
}
}
break;
}
} else {
println("\tcmp\t%s,%s", di, ax);
if (node->kind == ND_EQ) {
emitlin("\tsete\t%al");
} else if (node->kind == ND_NE) {
emitlin("\tsetne\t%al");
} else if (node->kind == ND_LT) {
if (node->lhs->ty->is_unsigned) {
emitlin("\tsetb\t%al");
} else {
emitlin("\tsetl\t%al");
}
} else if (node->kind == ND_LE) {
if (node->lhs->ty->is_unsigned) {
emitlin("\tsetbe\t%al");
} else {
emitlin("\tsetle\t%al");
}
}
}
emitlin("\tmovzbl\t%al,%eax");
break;
case ND_SHL:
emitlin("\tmov\t%edi,%ecx");
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\
\tshld\t%cl,%rax,%rdx\n\
\tshl\t%cl,%rax\n\
\txor\t%edi,%edi\n\
\tand\t$64,%cl\n\
\tcmovne\t%rax,%rdx\n\
\tcmovne\t%rdi,%rax");
} else if (node->lhs->ty->vector_size == 16) {
error_tok(node->tok, "todo sse shl");
} else {
println("\tshl\t%%cl,%s", ax);
}
break;
case ND_SHR:
emitlin("\tmov\t%edi,%ecx");
if (node->lhs->ty->is_unsigned) {
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\
\tshrd\t%cl,%rdx,%rax\n\
\tshr\t%cl,%rdx\n\
\txor\t%edi,%edi\n\
\tand\t$64,%cl\n\
\tcmovne\t%rdx,%rax\n\
\tcmovne\t%rdi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
error_tok(node->tok, "todo sse shr");
} else {
println("\tshr\t%%cl,%s", ax);
}
} else {
if (node->lhs->ty->kind == TY_INT128) {
emitlin("\
\tshrd\t%cl,%rdx,%rax\n\
\tsar\t%cl,%rdx\n\
\tmov\t%rdx,%rdi\n\
\tsar\t$63,%rdi\n\
\tand\t$64,%cl\n\
\tcmovne\t%rdx,%rax\n\
\tcmovne\t%rdi,%rdx");
} else if (node->lhs->ty->vector_size == 16) {
error_tok(node->tok, "todo sse sar");
} else {
println("\tsar\t%%cl,%s", ax);
}
}
break;
default:
error_tok(node->tok, "invalid expression");
}
if (IsOverflowArithmetic(node)) {
HandleOverflow(ax);
}
}
void gen_stmt(Node *node) {
print_loc(node->tok->file->file_no, node->tok->line_no);
switch (node->kind) {
case ND_IF: {
int c = count();
gen_expr(node->cond);
cmp_zero(node->cond->ty);
println("\tje\t.L.else.%d", c);
gen_stmt(node->then);
println("\tjmp\t.L.end.%d", c);
println(".L.else.%d:", c);
if (node->els) gen_stmt(node->els);
println(".L.end.%d:", c);
return;
}
case ND_FOR: {
int c = count();
if (node->init) gen_stmt(node->init);
println(".L.begin.%d:", c);
if (node->cond) {
gen_expr(node->cond);
cmp_zero(node->cond->ty);
println("\tje\t%s", node->brk_label);
}
gen_stmt(node->then);
println("%s:", node->cont_label);
if (node->inc) gen_expr(node->inc);
println("\tjmp\t.L.begin.%d", c);
println("%s:", node->brk_label);
return;
}
case ND_DO: {
int c = count();
println(".L.begin.%d:", c);
gen_stmt(node->then);
println("%s:", node->cont_label);
gen_expr(node->cond);
cmp_zero(node->cond->ty);
println("\tjne\t.L.begin.%d", c);
println("%s:", node->brk_label);
return;
}
case ND_SWITCH: // TODO(jart): 64-bit?
gen_expr(node->cond);
for (Node *n = node->case_next; n; n = n->case_next) {
char *ax = (node->cond->ty->size == 8) ? "%rax" : "%eax";
char *di = (node->cond->ty->size == 8) ? "%rdi" : "%edi";
if (n->begin == n->end) {
println("\tcmp\t$%ld,%s", n->begin, ax);
println("\tje\t%s", n->label);
continue;
}
// [GNU] Case ranges
println("\tmov\t%s,%s", ax, di);
println("\tsub\t$%ld,%s", n->begin, di);
println("\tcmp\t$%ld,%s", n->end - n->begin, di);
println("\tjbe\t%s", n->label);
}
if (node->default_case) {
println("\tjmp\t%s", node->default_case->label);
}
println("\tjmp\t%s", node->brk_label);
gen_stmt(node->then);
println("%s:", node->brk_label);
return;
case ND_CASE:
println("%s:", node->label);
gen_stmt(node->lhs);
return;
case ND_BLOCK:
for (Node *n = node->body; n; n = n->next) gen_stmt(n);
return;
case ND_GOTO:
println("\tjmp\t%s", node->unique_label);
return;
case ND_GOTO_EXPR:
gen_expr(node->lhs);
emitlin("\tjmp\t*%rax");
return;
case ND_LABEL:
println("%s:", node->unique_label);
gen_stmt(node->lhs);
return;
case ND_RETURN:
if (node->lhs) {
gen_expr(node->lhs);
Type *ty = node->lhs->ty;
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
if (ty->size <= 16) {
copy_struct_reg();
} else {
copy_struct_mem();
}
break;
}
}
println("\tjmp\t.L.return.%s", nameof(current_fn));
return;
case ND_EXPR_STMT:
gen_expr(node->lhs);
return;
case ND_ASM:
gen_asm(node->azm);
return;
}
error_tok(node->tok, "invalid statement");
}
// Assign offsets to local variables.
static void assign_lvar_offsets(Obj *prog) {
for (Obj *fn = prog; fn; fn = fn->next) {
if (!fn->is_function) continue;
// If a function has many parameters, some parameters are
// inevitably passed by stack rather than by register.
// The first passed-by-stack parameter resides at RBP+16.
int top = 16;
int bottom = 0;
int gp = 0, fp = 0;
// Assign offsets to pass-by-stack parameters.
for (Obj *var = fn->params; var; var = var->next) {
Type *ty = var->ty;
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
if (ty->size <= 16) {
bool fp1 = has_flonum(ty, 0, 8, 0);
bool fp2 = has_flonum(ty, 8, 16, 8);
if (fp + fp1 + fp2 < FP_MAX && gp + !fp1 + !fp2 < GP_MAX) {
fp = fp + fp1 + fp2;
gp = gp + !fp1 + !fp2;
continue;
}
}
break;
case TY_FLOAT:
case TY_DOUBLE:
if (fp++ < FP_MAX) continue;
break;
case TY_LDOUBLE:
break;
case TY_INT128:
gp++;
if (gp++ < GP_MAX) continue;
default:
if (gp++ < GP_MAX) continue;
}
top = ROUNDUP(top, 8);
var->offset = top;
top += var->ty->size;
}
// Assign offsets to pass-by-register parameters and local variables.
for (Obj *var = fn->locals; var; var = var->next) {
if (var->offset) continue;
// AMD64 System V ABI has a special alignment rule for an array of
// length at least 16 bytes. We need to align such array to at least
// 16-byte boundaries. See p.14 of
// https://github.com/hjl-tools/x86-psABI/wiki/x86-64-psABI-draft.pdf.
int align = (var->ty->kind == TY_ARRAY && var->ty->size >= 16)
? MAX(16, var->align)
: var->align;
bottom += var->ty->size;
bottom = ROUNDUP(bottom, align);
var->offset = -bottom;
}
fn->stack_size = ROUNDUP(bottom, 16);
}
}
static void emit_data(Obj *prog) {
for (Obj *var = prog; var; var = var->next) {
if (var->is_function || !var->is_definition) continue;
flushln();
fputc('\n', output_stream);
print_visibility(var);
int align = (var->ty->kind == TY_ARRAY && var->ty->size >= 16)
? MAX(16, var->align)
: var->align;
if (opt_common && var->is_tentative) {
println("\t.comm\t%s,%d,%d", nameof(var), var->ty->size, align);
} else {
if (var->section) {
println("\t.section %s,\"aw\",@%s", var->section,
var->init_data ? "progbits" : "nobits");
} else if (var->is_tls) {
println("\t.section .t%s,\"awT\",@%s",
var->init_data ? "progbits" : "nobits",
var->init_data ? "data" : "bss");
#if 0 /* TODO: unflag if assigned to array */
} else if (align <= 1 && var->is_string_literal) {
println("\t.section .rodata.str1.1,\"aSM\",@progbits,1");
#endif
} else if (opt_data_sections) {
println("\tsection .%s.%s", var->init_data ? "data" : "bss",
nameof(var));
} else {
println("\t.%s", var->init_data ? "data" : "bss");
}
print_align(align);
println("\t.type\t%s,@object", nameof(var));
println("\t.size\t%s,%d", nameof(var), var->ty->size);
println("%s:", nameof(var));
if (var->init_data) {
int pos = 0;
Relocation *rel = var->rel;
if (!rel && is_quotable_string(var->init_data, var->ty->size)) {
println("\t.asciz\t\"%s\"", var->init_data);
} else {
while (pos < var->ty->size) {
if (rel && rel->offset == pos) {
assert(pos + 8 <= var->ty->size);
println("\t.quad\t%s%+ld", *rel->label, rel->addend);
rel = rel->next;
pos += 8;
} else {
println("\t.byte\t%d", var->init_data[pos++]);
}
}
}
} else {
println("\t.zero\t%d", var->ty->size);
}
}
}
}
static void store_fp(int r, int offset, int sz) {
switch (sz) {
case 4:
println("\tmovss\t%%xmm%d,%d(%%rbp)", r, offset);
return;
case 8:
println("\tmovsd\t%%xmm%d,%d(%%rbp)", r, offset);
return;
case 16:
println("\tmovaps\t%%xmm%d,%d(%%rbp)", r, offset);
return;
}
UNREACHABLE();
}
static void store_gp(int r, int offset, int sz) {
switch (sz) {
case 1:
println("\tmov\t%s,%d(%%rbp)", argreg8[r], offset);
return;
case 2:
println("\tmov\t%s,%d(%%rbp)", argreg16[r], offset);
return;
case 4:
println("\tmov\t%s,%d(%%rbp)", argreg32[r], offset);
return;
case 8:
println("\tmov\t%s,%d(%%rbp)", argreg64[r], offset);
return;
default:
for (int i = 0; i < sz; i++) {
println("\tmov\t%s,%d(%%rbp)", argreg8[r], offset + i);
println("\tshr\t$8,%s", argreg64[r]);
}
return;
}
}
static void emit_text(Obj *prog) {
for (Obj *fn = prog; fn; fn = fn->next) {
if (!fn->is_function || !fn->is_definition) continue;
// No code is emitted for "static inline" functions
// if no one is referencing them.
if (!fn->is_live) continue;
flushln();
fputc('\n', output_stream);
if (fn->section) {
println("\t.section %s,\"a\",@progbits", fn->section);
} else if (opt_function_sections) {
println("\tsection .text.%s", nameof(fn));
} else {
emitlin("\t.text");
}
print_visibility(fn);
print_align(fn->align);
println("\t.type\t%s,@function", nameof(fn));
println("%s:", nameof(fn));
current_fn = fn;
// Prologue
emitlin("\tpush\t%rbp");
emitlin("\tmov\t%rsp,%rbp");
if (opt_nop_mcount) {
print_profiling_nop();
} else if (opt_fentry) {
emitlin("\tcall\t__fentry__@gotpcrel(%rip)");
} else if (opt_pg) {
emitlin("\tcall\tmcount@gotpcrel(%rip)");
} else {
print_profiling_nop();
}
println("\tsub\t$%d,%%rsp", fn->stack_size);
println("\tmov\t%%rsp,%d(%%rbp)", fn->alloca_bottom->offset);
// Save arg registers if function is variadic
if (fn->va_area) {
int gp = 0, fp = 0;
for (Obj *var = fn->params; var; var = var->next) {
if (is_flonum(var->ty)) {
fp++;
} else {
gp++;
}
}
int off = fn->va_area->offset;
// va_elem
println("\tmovl\t$%d,%d(%%rbp)", gp * 8, off); // gp_offset
println("\tmovl\t$%d,%d(%%rbp)", fp * 8 + 48, off + 4); // fp_offset
println("\tmov\t%%rbp,%d(%%rbp)", off + 8); // overflow_arg_area
println("\taddq\t$16,%d(%%rbp)", off + 8);
println("\tmov\t%%rbp,%d(%%rbp)", off + 16); // reg_save_area
println("\taddq\t$%d,%d(%%rbp)", off + 24, off + 16);
// __reg_save_area__
println("\tmov\t%%rdi,%d(%%rbp)", off + 24);
println("\tmov\t%%rsi,%d(%%rbp)", off + 32);
println("\tmov\t%%rdx,%d(%%rbp)", off + 40);
println("\tmov\t%%rcx,%d(%%rbp)", off + 48);
println("\tmov\t%%r8,%d(%%rbp)", off + 56);
println("\tmov\t%%r9,%d(%%rbp)", off + 64);
println("\tmovsd\t%%xmm0,%d(%%rbp)", off + 72);
println("\tmovsd\t%%xmm1,%d(%%rbp)", off + 80);
println("\tmovsd\t%%xmm2,%d(%%rbp)", off + 88);
println("\tmovsd\t%%xmm3,%d(%%rbp)", off + 96);
println("\tmovsd\t%%xmm4,%d(%%rbp)", off + 104);
println("\tmovsd\t%%xmm5,%d(%%rbp)", off + 112);
println("\tmovsd\t%%xmm6,%d(%%rbp)", off + 120);
println("\tmovsd\t%%xmm7,%d(%%rbp)", off + 128);
}
// Save passed-by-register arguments to the stack
int gp = 0, fp = 0;
for (Obj *var = fn->params; var; var = var->next) {
if (var->offset > 0) continue;
Type *ty = var->ty;
switch (ty->kind) {
case TY_STRUCT:
case TY_UNION:
assert(ty->size <= 16);
if (has_flonum(ty, 0, 8, 0)) {
store_fp(fp++, var->offset, MIN(8, ty->size));
} else {
store_gp(gp++, var->offset, MIN(8, ty->size));
}
if (ty->size > 8) {
if (has_flonum(ty, 8, 16, 0)) {
store_fp(fp++, var->offset + 8, ty->size - 8);
} else {
store_gp(gp++, var->offset + 8, ty->size - 8);
}
}
break;
case TY_FLOAT:
case TY_DOUBLE:
store_fp(fp++, var->offset, ty->size);
break;
case TY_INT128:
store_gp(gp++, var->offset + 0, 8);
store_gp(gp++, var->offset + 8, 8);
break;
default:
store_gp(gp++, var->offset, ty->size);
}
}
// Emit code
gen_stmt(fn->body);
assert(!depth);
// [https://www.sigbus.info/n1570#5.1.2.2.3p1] The C spec defines
// a special rule for the main function. Reaching the end of the
// main function is equivalent to returning 0, even though the
// behavior is undefined for the other functions.
if (strcmp(nameof(fn), "main") == 0) {
emitlin("\txor\t%eax,%eax");
}
// Epilogue
println(".L.return.%s:", nameof(fn));
if (fn->is_noreturn) {
emitlin("\tud2");
} else {
emitlin("\tleave");
emitlin("\tret");
}
println("\t.size\t%s,.-%s", nameof(fn), nameof(fn));
if (fn->is_constructor) {
emitlin("\t.section .ctors,\"aw\",@progbits");
emitlin("\t.align\t8");
println("\t.quad\t%s", nameof(fn));
}
if (fn->is_destructor) {
emitlin("\t.section .dtors,\"aw\",@progbits");
emitlin("\t.align\t8");
println("\t.quad\t%s", nameof(fn));
}
}
}
static void emit_staticasms(StaticAsm *a) {
if (!a) return;
emit_staticasms(a->next);
gen_asm(a->body);
}
void codegen(Obj *prog, FILE *out) {
output_stream = out;
File **files = get_input_files();
for (int i = 0; files[i]; i++) {
println("\t.file\t%d %`'s", files[i]->file_no, files[i]->name);
}
assign_lvar_offsets(prog);
emit_staticasms(staticasms);
emit_data(prog);
emit_text(prog);
flushln();
}