feat(asm): emit the loong64 stack-split guard for small and medium frames

asm/loong64_assemble.go

@@ -27,6 +27,7 @@ import (
 func assembleLOONG64(t *ast.Text) ([]byte, map[string]int, []Reloc, []LineEntry, []SpadjStep, error) {
 	fi := loong64ComputeFrame(t)
 	prologue := loong64Prologue(fi)
+	guardLen := loong64GuardLen(fi)
 	chain := loong64JumpChain(t)
 	resolve := func(name string) string {
 		if r, ok := chain[name]; ok {
@@ -42,12 +43,12 @@ func assembleLOONG64(t *ast.Text) ([]byte, map[string]int, []Reloc, []LineEntry,
 	// delta by autosize; the boundary is reported at the third instruction's
 	// pc, exactly as the toolchain's pctospadj does.
 	if fi.autosize != 0 {
-		spadj = append(spadj, SpadjStep{PC: 8, Value: fi.autosize})
+		spadj = append(spadj, SpadjStep{PC: guardLen + 8, Value: fi.autosize})
 	}
 
 	// Pass 1: label offsets from the instruction sizes.
 	offsets := map[string]int{}
-	pos := len(prologue)
+	pos := guardLen + len(prologue)
 	for _, stmt := range t.Body {
 		switch s := stmt.(type) {
 		case *ast.Label:
@@ -57,9 +58,25 @@ func assembleLOONG64(t *ast.Text) ([]byte, map[string]int, []Reloc, []LineEntry,
 		}
 	}
 
-	// Pass 2: encode.  Relocation offsets are recorded function-relative.
-	out := append([]byte(nil), prologue...)
-	pc := len(prologue)
+	// Pass 2: encode.  The guard prefix precedes the prologue; its branches
+	// target the morestack block at the end of the function, which the first
+	// pass has sized.
+	bodyLen := 0
+	{
+		p := guardLen + len(prologue)
+		for _, stmt := range t.Body {
+			if in, ok := stmt.(*ast.Instr); ok {
+				p += loong64InstrSize(in, fi)
+			}
+		}
+		bodyLen = p - (guardLen + len(prologue))
+	}
+	var out []byte
+	if fi.needSplit {
+		out = append(out, loong64GuardBytes(fi, guardLen+len(prologue)+bodyLen)...)
+	}
+	out = append(out, prologue...)
+	pc := guardLen + len(prologue)
 	preCount := len(relocs)
 	var lines []LineEntry
 	for _, stmt := range t.Body {
@@ -94,6 +111,12 @@ func assembleLOONG64(t *ast.Text) ([]byte, map[string]int, []Reloc, []LineEntry,
 		out = append(out, code...)
 		pc += len(code)
 	}
+	if fi.needSplit {
+		block, blReloc := loong64MoreStackBlock(pc)
+		out = append(out, block...)
+		relocs = append(relocs, blReloc)
+		pc += len(block)
+	}
 	return out, offsets, relocs, lines, spadj, nil
 }