feat(amd64): floating-point immediates through a synthesised pool

Assisted-by: GLM 5.3 Flash

asm/assemble.go

@@ -5,6 +5,7 @@ package asm
 
 import (
 	"fmt"
+	"strconv"
 	"strings"
 
 	"sourcedock.dev/petrbalvin/gasm-devkit/ast"
@@ -28,7 +29,7 @@ import (
 // emitted: the bytes match go tool asm only for NOSPLIT functions or
 // zero-frame leaves, where the toolchain emits no guard either.
 func Assemble(t *ast.Text) ([]byte, map[string]int, error) {
-	code, _, labels, _, _, err := assemble(t, nil)
+	code, _, labels, _, _, _, err := assemble(t, nil)
 	return code, labels, err
 }
 
@@ -66,9 +67,9 @@ type spadjStep struct {
 // assemble encodes a TEXT body, returning the machine code, the static-symbol
 // patch sites (for the file-level layout to resolve), the label table and the
 // stack-adjustment boundaries.
-func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, []spadjStep, []LineEntry, error) {
+func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, []spadjStep, []LineEntry, []floatPoolEntry, error) {
 	if err := checkAdjspBalance(t); err != nil {
-		return nil, nil, nil, nil, nil, err
+		return nil, nil, nil, nil, nil, nil, err
 	}
 	fi := computeFrame(t)
 	chain := jumpChain(t)
@@ -88,6 +89,8 @@ func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, [
 	offsets := map[string]int{}
 	pcs := make([]int, len(t.Body))
 	var guardJBlong, guardJBElong, moreJMPlong bool
+	poolSeen := map[string]bool{}
+	var poolList []floatPoolEntry
 	for {
 		guard := fi.guardLen(guardJBlong, guardJBElong)
 		pos := guard + len(fi.prologue)
@@ -98,7 +101,7 @@ func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, [
 			case *ast.Instr:
 				sz, err := instrSize(s, fi, long[i], link)
 				if err != nil {
-					return nil, nil, nil, nil, nil, fmt.Errorf("%s: %w", s.Mnemonic.Text, err)
+					return nil, nil, nil, nil, nil, nil, fmt.Errorf("%s: %w", s.Mnemonic.Text, err)
 				}
 				sizes[i] = sz
 				pcs[i] = pos
@@ -229,12 +232,18 @@ func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, [
 				spadjStep{pos + epi, 0},
 			)
 		}
-		code, ps, err := encodeInstr(s, pos, offsets, fi, long[i], resolve, link)
+		code, ps, pool, err := encodeInstr(s, pos, offsets, fi, long[i], resolve, link)
 		if err != nil {
-			return nil, nil, nil, nil, nil, fmt.Errorf("%s: %w", s.Mnemonic.Text, err)
+			return nil, nil, nil, nil, nil, nil, fmt.Errorf("%s: %w", s.Mnemonic.Text, err)
+		}
+		for _, entry := range pool {
+			if !poolSeen[entry.name] {
+				poolSeen[entry.name] = true
+				poolList = append(poolList, entry)
+			}
 		}
 		if len(code) != sizes[i] {
-			return nil, nil, nil, nil, nil, fmt.Errorf("%s: size mismatch (%d vs %d)", s.Mnemonic.Text, len(code), sizes[i])
+			return nil, nil, nil, nil, nil, nil, fmt.Errorf("%s: size mismatch (%d vs %d)", s.Mnemonic.Text, len(code), sizes[i])
 		}
 		if strings.ToUpper(s.Mnemonic.Text) == "CALL" {
 			for k := range ps {
@@ -272,7 +281,7 @@ func assemble(t *ast.Text, link *linkInfo) ([]byte, []sbPatch, map[string]int, [
 		pos += len(suffix)
 	}
 	_ = pos
-	return out, patches, offsets, steps, lines, nil
+	return out, patches, offsets, steps, lines, poolList, nil
 }
 
 // jumpChain precomputes jump-to-jump folding: a label whose first instruction
@@ -593,7 +602,7 @@ func instrSize(s *ast.Instr, fi frameInfo, long bool, link *linkInfo) (int, erro
 		}
 		return jumpSize(mnem, long), nil
 	}
-	code, _, err := encodeInstr(s, 0, nil, fi, false, nil, link)
+	code, _, _, err := encodeInstr(s, 0, nil, fi, false, nil, link)
 	if err != nil {
 		return 0, err
 	}
@@ -628,7 +637,7 @@ func jumpSize(mnem string, long bool) int {
 // (relative to pc, the instruction's own offset).  A RET in a frame-pointer
 // function is prefixed with the epilogue.  resolve, when non-nil, redirects a
 // jump label through the jump-to-jump chain before the offset lookup.
-func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, long bool, resolve func(string) string, link *linkInfo) ([]byte, []sbPatch, error) {
+func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, long bool, resolve func(string) string, link *linkInfo) ([]byte, []sbPatch, []floatPoolEntry, error) {
 	mnem := strings.ToUpper(s.Mnemonic.Text)
 
 	var prefix []byte
@@ -638,6 +647,7 @@ func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, lon
 
 	var code []byte
 	var ps []sbPatch
+	var pool []floatPoolEntry
 	var err error
 	if isJumpMnemonic(mnem) {
 		if (mnem == "CALL" || mnem == "JMP") && isSBCall(s) {
@@ -646,7 +656,7 @@ func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, lon
 			// or the linker.
 			code, ps, err = encodeSBCall(s, link)
 			if err != nil {
-				return nil, nil, err
+				return nil, nil, nil, err
 			}
 			for i := range ps {
 				ps[i].kind = RelCall
@@ -656,23 +666,23 @@ func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, lon
 				ps[i].off += body
 				ps[i].after = body + len(code)
 			}
-			return append(prefix, code...), ps, nil
+			return append(prefix, code...), ps, nil, nil
 		}
 		if (mnem == "CALL" || mnem == "JMP") && indirectJumpTarget(s) {
 			// JMP/CALL through a register or memory: no relocation and no
 			// label to resolve, the operand fully determines the bytes.
 			code, err = encodeIndirectJump(s, mnem)
 			if err != nil {
-				return nil, nil, err
+				return nil, nil, nil, err
 			}
-			return append(prefix, code...), nil, nil
+			return append(prefix, code...), nil, nil, nil
 		}
 		code, err = encodeJump(s, mnem, pc+len(prefix), offsets, long, resolve)
 	} else {
-		code, ps, err = encodeNormal(s, fi, link)
+		code, ps, pool, err = encodeNormal(s, fi, link)
 	}
 	if err != nil {
-		return nil, nil, err
+		return nil, nil, nil, err
 	}
 	// Anchor the patch fields at function-relative positions: off indexes the
 	// disp32 field, after is the address just past the instruction.
@@ -681,31 +691,65 @@ func encodeInstr(s *ast.Instr, pc int, offsets map[string]int, fi frameInfo, lon
 		ps[i].off += body
 		ps[i].after = body + len(code)
 	}
-	return append(prefix, code...), ps, nil
+	return append(prefix, code...), ps, pool, nil
 }
 
-func encodeNormal(s *ast.Instr, fi frameInfo, link *linkInfo) ([]byte, []sbPatch, error) {
-	_, size := splitSize(strings.ToUpper(s.Mnemonic.Text))
+func encodeNormal(s *ast.Instr, fi frameInfo, link *linkInfo) ([]byte, []sbPatch, []floatPoolEntry, error) {
+	mnemUpper := strings.ToUpper(s.Mnemonic.Text)
+	if mnemUpper == "FUNCDATA" || mnemUpper == "PCDATA" {
+		code, err := encodeBookkeeping(mnemUpper, s)
+		if err != nil {
+			return nil, nil, nil, err
+		}
+		return code, nil, nil, nil
+	}
+	_, size := splitSize(mnemUpper)
 	if size == 0 {
 		size = 8
 	}
 	ops := make([]Operand, len(s.Operands))
 	for i, op := range s.Operands {
-		o, err := operandFromAST(op, size, fi, link)
+		o, err := operandFromAST(mnemUpper, op, size, fi, link)
 		if err != nil {
-			return nil, nil, err
+			return nil, nil, nil, err
 		}
 		ops[i] = o
 	}
 	e := &enc{}
 	if err := e.encode(s.Mnemonic.Text, ops); err != nil {
-		return nil, nil, err
+		return nil, nil, nil, err
 	}
 	ps := make([]sbPatch, len(e.patches))
 	for i, p := range e.patches {
 		ps[i] = sbPatch{off: p.off, name: p.name, addend: p.addend}
 	}
-	return e.out, ps, nil
+	return e.out, ps, e.floatPoolList(), nil
+}
+
+// encodeBookkeeping accepts-and-ignores FUNCDATA and PCDATA at the statement
+// level, before operand conversion: the toolchain's shapes are FUNCDATA
+// $n, sym(SB) and PCDATA $n, $m, and neither contributes a byte to the
+// function body.  The symbol reference must not run through the SB-operand
+// path, which demands file-level resolution the statement never needs.
+func encodeBookkeeping(upper string, s *ast.Instr) ([]byte, error) {
+	if len(s.Operands) != 2 {
+		return nil, fmt.Errorf("%s expects 2 operands, got %d", upper, len(s.Operands))
+	}
+	a, b := s.Operands[0], s.Operands[1]
+	if a.Kind != ast.OpImmediate || !a.Imm.HasVal {
+		return nil, fmt.Errorf("%s: first operand must be an integer immediate", upper)
+	}
+	switch upper {
+	case "FUNCDATA":
+		if b.Kind != ast.OpAddr || b.Addr.Sym == nil || b.Addr.Sym.Pseudo != "SB" {
+			return nil, fmt.Errorf("FUNCDATA: second operand must be a symbol reference")
+		}
+	case "PCDATA":
+		if b.Kind != ast.OpImmediate || !b.Imm.HasVal {
+			return nil, fmt.Errorf("PCDATA: second operand must be an integer immediate")
+		}
+	}
+	return nil, nil
 }
 
 // encodeJump encodes a JMP/CALL/Jcc with a relative offset resolved from the
@@ -756,7 +800,7 @@ func isSBCall(s *ast.Instr) bool {
 
 // encodeSBCall encodes CALL sym(SB) as E8 rel32 with a patch site.
 func encodeSBCall(s *ast.Instr, link *linkInfo) ([]byte, []sbPatch, error) {
-	o, err := operandFromAST(s.Operands[0], 8, frameInfo{}, link)
+	o, err := operandFromAST(strings.ToUpper(s.Mnemonic.Text), s.Operands[0], 8, frameInfo{}, link)
 	if err != nil {
 		return nil, nil, err
 	}
@@ -813,7 +857,7 @@ func indirectJumpTarget(s *ast.Instr) bool {
 func encodeIndirectJump(s *ast.Instr, mnem string) ([]byte, error) {
 	ops := make([]Operand, len(s.Operands))
 	for i, op := range s.Operands {
-		o, err := operandFromAST(op, 8, frameInfo{}, nil)
+		o, err := operandFromAST(mnem, op, 8, frameInfo{}, nil)
 		if err != nil {
 			return nil, err
 		}
@@ -830,8 +874,11 @@ func encodeIndirectJump(s *ast.Instr, mnem string) ([]byte, error) {
 var spReg = Reg{idx: 4, size: 8}
 
 // operandFromAST converts a parsed operand into an encoder Operand, applying
-// the frame translation to FP/SP pseudo-register operands.
-func operandFromAST(op *ast.Operand, size int, fi frameInfo, link *linkInfo) (Operand, error) {
+// the frame translation to FP/SP pseudo-register operands.  mnemUpper is the
+// instruction's upper-case mnemonic, which the floating-point immediate gate
+// needs: only the SSE mnemonics whose encoding takes an XMM/memory source
+// accept one.
+func operandFromAST(mnemUpper string, op *ast.Operand, size int, fi frameInfo, link *linkInfo) (Operand, error) {
 	switch op.Kind {
 	case ast.OpImmediate:
 		if op.Imm.HasVal {
@@ -841,17 +888,43 @@ func operandFromAST(op *ast.Operand, size int, fi frameInfo, link *linkInfo) (Op
 			}
 			return Imm(v), nil
 		}
+		// A floating-point immediate: $1.5, $-1.0 or the parenthesised
+		// $(-1.0) spelling (the constant-expression folder only folds
+		// integers, so that shape arrives with an empty Immediate and only
+		// the raw spelling carries the value).  The toolchain rewrites it
+		// into a pooled-constant read on the SSE scalar paths and rejects
+		// it everywhere else.
+		if text, neg, ok := floatImmText(op); ok {
+			if !sseFloatImm[mnemUpper] {
+				return nil, fmt.Errorf("%s does not take a floating-point immediate", mnemUpper)
+			}
+			return FloatImm{Text: text, Neg: neg}, nil
+		}
 		return nil, fmt.Errorf("non-integer immediate not supported")
 
 	case ast.OpAddr:
 		a := op.Addr
 
 		// A bracketed register range, [Z0-Z3]: the four-register source of
-		// the 4FMAPS/4VNNIW families.  The EVEX quad-register emit path
-		// needs an encoder operand of its own, so the shape stays a named
-		// gap rather than an encoding.
+		// the 4FMAPS/4VNNIW families.  The range must span four consecutive
+		// same-width vector registers, exactly what the toolchain's parser
+		// takes; the EVEX quad-register emit path reads the low end.
 		if a.Range != nil {
-			return nil, fmt.Errorf("register range %q needs quad-register encoder support", op.Raw)
+			lo, ok := ParseReg(a.Range.Lo)
+			if !ok {
+				return nil, fmt.Errorf("unknown register %q in range", a.Range.Lo)
+			}
+			hi, ok := ParseReg(a.Range.Hi)
+			if !ok {
+				return nil, fmt.Errorf("unknown register %q in range", a.Range.Hi)
+			}
+			if !lo.isVec() || lo.size != hi.size {
+				return nil, fmt.Errorf("register range %q must span four same-width vector registers", op.Raw)
+			}
+			if hi.idx != lo.idx+3 {
+				return nil, fmt.Errorf("register range %q must span four consecutive registers", op.Raw)
+			}
+			return RegList{Lo: lo, Hi: hi}, nil
 		}
 
 		// FP-relative: x+N(FP) → (N + fpAdjust)(SP).  The offset N lives in the
@@ -921,3 +994,37 @@ func operandFromAST(op *ast.Operand, size int, fi frameInfo, link *linkInfo) (Op
 	}
 	return nil, fmt.Errorf("unsupported operand")
 }
+
+// floatImmText recovers a floating-point immediate's magnitude and sign from
+// the parsed operand.  The ordinary spellings arrive in Imm.Float; the
+// parenthesised $(-1.0) leaves the Immediate empty, because the integer
+// folder cannot read it, and only the verbatim operand text still carries
+// the value.  Anything that is not a number a float parser accepts reports
+// not-ok, so every other shape keeps its existing diagnostic.
+func floatImmText(op *ast.Operand) (text string, neg bool, ok bool) {
+	if op.Imm.Float != "" {
+		return op.Imm.Float, op.Imm.Neg, true
+	}
+	if op.Imm.HasVal || op.Imm.Str != "" || op.Imm.Sym != nil {
+		return "", false, false
+	}
+	// joinRaw spaced the token texts; the compact spelling is what matters.
+	compact := strings.ReplaceAll(op.Raw, " ", "")
+	inner, ok := strings.CutPrefix(compact, "$(")
+	if !ok || !strings.HasSuffix(inner, ")") {
+		return "", false, false
+	}
+	inner = strings.TrimSuffix(inner, ")")
+	inner = strings.TrimPrefix(inner, "+")
+	if s, ok := strings.CutPrefix(inner, "-"); ok {
+		neg = true
+		inner = s
+	}
+	if inner == "" || !strings.ContainsAny(inner, "0123456789") {
+		return "", false, false
+	}
+	if _, err := strconv.ParseFloat(inner, 64); err != nil {
+		return "", false, false
+	}
+	return inner, neg, true
+}