ludic/selfhost/emit_expr.ludic
Orkuncakilkaya feb3a71e56 Phase 7i: byte indexing p[i] / p[i] = v (retire peek8/poke8)
Raw byte access is now indexing, not C-style peek/poke:

  peek8(src, i)          -> src[i]        (reads a byte, widened to int)
  poke8(out, j, r)       -> out[j] = r    (narrows the int to a byte)

E_INDEX on a non-slice pointer/string lowers to a `getelementptr i8` + load/zext
(read) or trunc/store (write) — byte-identical to the old peek8/poke8, so the
migration reproduces the compiler exactly. A "byte" element type (llty i8) drives
the widen/narrow. The compiler's own byte work now reads naturally, e.g.
`is_slice_ty` is `t[0] == 91 and t[1] == 93`.

Subtlety fixed on the way: g_addr_ty (the out-param carrying the indexed element
type) must be set AFTER evaluating the index expression, since a member/index in
the index would otherwise clobber it — doing it early made a byte read load a
full pointer from a byte address and crash the self-compile.

Two reseeds: add byte-index support keeping peek8/poke8, then migrate 148 call
sites and delete the intrinsics (+ the now-dead emit_gep_i8). Vocabulary drops
peek8/poke8. Reseeded (22604 lines); C-free fixpoint holds; goldens identical;
18/18; vocab clean.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-08-28 02:01:59 +03:00

237 lines
9.6 KiB
Text

# emit_expr.ludic — lower an expression to IR, returning its register and type.
fn emit_load_at(addr: ptr, ty: ptr) -> Val {
let r = emit_bind(`load {llty(ty)}, ptr {addr}`)
return val(r, ty)
}
# short-circuit `and`/`or`: seed a slot with (left!=0), branch to decide whether
# to overwrite with (right!=0).
fn emit_logic(e: Node) -> Val {
let slot = emit_alloca("i32")
let la = emit_expr(e.a)
let lc = emit_bind(`icmp ne i32 {la.code}, 0`)
let lz = emit_bind(`zext i1 {lc} to i32`)
emit(" store i32 "); emit(lz); emit(", ptr "); emit(slot); emit("\n")
let ev = lbl("sc"); let done = lbl("scend")
if (e.s == "and") { emit(" br i1 "); emit(lc); emit(", label %"); emit(ev); emit(", label %"); emit(done); emit("\n") }
else { emit(" br i1 "); emit(lc); emit(", label %"); emit(done); emit(", label %"); emit(ev); emit("\n") }
emit(ev); emit(":\n")
let rb = emit_expr(e.b)
let rc = emit_bind(`icmp ne i32 {rb.code}, 0`)
let rz = emit_bind(`zext i1 {rc} to i32`)
emit(" store i32 "); emit(rz); emit(", ptr "); emit(slot); emit("\n")
emit(" br label %"); emit(done); emit("\n")
emit(done); emit(":\n")
return val(emit_bind(`load i32, ptr {slot}`), "bool")
}
fn cmp_code(op: ptr) -> ptr {
if (op == ("<")) { return "slt" }
if (op == ("<=")) { return "sle" }
if (op == (">")) { return "sgt" }
if (op == (">=")) { return "sge" }
if (op == ("==")) { return "eq" }
return "ne"
}
fn is_cmp(op: ptr) -> bool {
return (op == ("<")) or (op == ("<=")) or (op == (">")) or (op == (">=")) or (op == ("==")) or (op == ("!="))
}
fn arith_code(op: ptr) -> ptr {
if (op == ("+")) { return "add" }
if (op == ("-")) { return "sub" }
if (op == ("*")) { return "mul" }
if (op == ("/")) { return "sdiv" }
if (op == ("&")) { return "and" }
if (op == ("|")) { return "or" }
if (op == ("^")) { return "xor" }
if (op == ("<<")) { return "shl" }
if (op == (">>")) { return "lshr" } # logical shift (unsigned)
return "srem"
}
# widen an int value to Q16.16 by shifting left 16 (a fixed value passes through)
fn to_fixed(v: Val) -> ptr {
if (v.ty == "fixed") { return v.code }
return emit_bind(`shl i32 {v.code}, 16`)
}
# string operators: `a + b` concatenates, `a == b` / `a != b` compare by content.
# Both call the @fn_str_* prelude (emitted once per program that uses them).
fn emit_str_op(op: ptr, a: Val, b: Val) -> Val {
g_uses_str = true
if (op == ("+")) {
return val(emit_bind(`call ptr @fn_str_concat(ptr {a.code}, ptr {b.code})`), "str")
}
let r = emit_bind(`call i32 @fn_str_eq(ptr {a.code}, ptr {b.code})`)
if (op == ("!=")) {
let c = emit_bind(`icmp eq i32 {r}, 0`)
return val(emit_bind(`zext i1 {c} to i32`), "bool")
}
return val(r, "bool")
}
fn emit_bin(e: Node) -> Val {
if (e.s == "and") or (e.s == "or") { return emit_logic(e) }
let a = emit_expr(e.a)
let b = emit_expr(e.b)
# strings are pointer-typed, so any `+` with a pointer operand is concatenation,
# and `==`/`!=` between pointers is content comparison — except `x == null`,
# which is a pointer-identity test and falls through to the icmp below.
let ptrish = (llty(a.ty) == "ptr") or (llty(b.ty) == "ptr")
let isnull = e.a.kind == E_NULL or e.b.kind == E_NULL
if ptrish {
if (e.s == ("+")) { return emit_str_op("+", a, b) }
if ((e.s == ("==")) or (e.s == ("!="))) and not isnull { return emit_str_op(e.s, a, b) }
}
let fx = (a.ty == "fixed") or (b.ty == "fixed")
if is_cmp(e.s) {
var ac = a.code; var bc = b.code
var ct = "i32"
if fx { ac = to_fixed(a); bc = to_fixed(b) }
else { if (llty(a.ty) == "ptr") or (llty(b.ty) == "ptr") { ct = "ptr" } } # `p == null`, str/record identity
let c = emit_bind(`icmp {cmp_code(e.s)} {ct} {ac}, {bc}`)
return val(emit_bind(`zext i1 {c} to i32`), "bool")
}
if fx {
let af = to_fixed(a); let bf = to_fixed(b)
if (e.s == ("*")) {
let a64 = emit_bind(`sext i32 {af} to i64`)
let b64 = emit_bind(`sext i32 {bf} to i64`)
let m = emit_bind(`mul i64 {a64}, {b64}`)
let sh = emit_bind(`ashr i64 {m}, 16`)
return val(emit_bind(`trunc i64 {sh} to i32`), "fixed")
}
if (e.s == ("/")) {
let a64 = emit_bind(`sext i32 {af} to i64`)
let ash = emit_bind(`shl i64 {a64}, 16`)
let b64 = emit_bind(`sext i32 {bf} to i64`)
let dv = emit_bind(`sdiv i64 {ash}, {b64}`)
return val(emit_bind(`trunc i64 {dv} to i32`), "fixed")
}
let r = emit_bind(`{arith_code(e.s)} i32 {af}, {bf}`)
return val(r, "fixed")
}
let r = emit_bind(`{arith_code(e.s)} i32 {a.code}, {b.code}`)
return val(r, "int")
}
fn emit_call(e: Node) -> Val {
let name = e.a.s
if (name == "self") { if nself == 0 { return val("0", "entity") }; return val(emit_bind(`load i32, ptr {self_stk[nself - 1]}`), "entity") }
if (name == "key") { return val(emit_bind("load i32, ptr @L_key"), "int") }
if (name == "save") { emit(" call void @L_save()\n"); return val("0", "void") }
if (name == "ui_build") { emit(" call void @ui_build()\n"); return val("0", "void") }
if (name == "load") { return val(emit_bind("call i32 @L_load()"), "bool") }
if (name == "quit") { emit(" store i32 0, ptr @L_running\n"); return val("0", "void") }
if (name == "len") { return emit_len(e) }
if (name == "push") { return emit_push(e) }
if (name == "str") { # str(x): int/bool/fixed -> text, a string passes through
let a = emit_expr(e.kids[0])
if (llty(a.ty) == "ptr") { return a }
g_uses_intstr = true
return val(emit_bind(`call ptr @fn_int_str(i32 {a.code})`), "str")
}
if (name == "print") { # print(x): a value + newline (int or string)
let a = emit_expr(e.kids[0])
if (llty(a.ty) == "ptr") { emit(" call i32 (ptr, ...) @printf(ptr @.fmt_line, ptr " + `{a.code})\n`) }
else { emit(" call i32 (ptr, ...) @printf(ptr @.fmt_int, i32 " + `{a.code})\n`) }
return val("0", "void")
}
if (name == "fx") { let a = emit_expr(e.kids[0]); return val(emit_bind(`shl i32 {a.code}, 16`), "fixed") }
if (name == "flr") { let a = emit_expr(e.kids[0]); return val(emit_bind(`ashr i32 {a.code}, 16`), "int") }
if is_intrinsic(name) { return emit_intrinsic(name, e) }
if is_intrinsic2(name) { return emit_intrinsic2(name, e) }
if is_math_builtin(name) { return emit_math_builtin(name, e) }
var fn2 = find_fn(name)
var cname = name
if (fn2 == null) {
# a builtin like clear()/reg() is satisfied by its rt_ function
let rtname = `rt_{name}`
fn2 = find_fn(rtname)
if (fn2 == null) { perr(`unknown function {name}`) }
cname = rtname
}
# evaluate args first (their IR is emitted before the call instruction)
let args = new []ptr
let atys = new []ptr
var i = 0
while i < len(e.kids) { let v = emit_expr(e.kids[i]); push(args, v.code); push(atys, v.ty); i = i + 1 }
let rl = llty(fn2.ty)
emit(" ")
var rreg = "0"
if not (rl == "void") { rreg = nreg(); emit(rreg); emit(" = ") }
emit("call "); emit(rl); emit(" @fn_"); emit(cname); emit("(")
i = 0
while i < len(args) {
if i > 0 { emit(", ") }
emit(llty(atys[i])); emit(" "); emit(args[i])
i = i + 1
}
emit(")\n")
return val(rreg, fn2.ty)
}
fn emit_expr(e: Node) -> Val {
if (e == null) { return val("0", "int") }
if e.kind == E_INT { return val(itoa(e.ival), "int") }
if e.kind == E_FLOAT { return val(itoa(e.ival), "fixed") }
if e.kind == E_BOOL { return val(itoa(e.ival), "bool") }
if e.kind == E_NULL { return val("null", "ptr") }
if e.kind == E_SLICE { # s[a..b] -> a fresh substring
let base = emit_expr(e.a)
let lo = emit_expr(e.b)
let hi = emit_expr(e.c)
g_uses_strslice = true
return val(emit_bind(`call ptr @fn_str_slice(ptr {base.code}, i32 {lo.code}, i32 {hi.code})`), "str")
}
if e.kind == E_STR { return val(emit_str_const(e.s), "str") }
if e.kind == E_NEW {
if is_slice_ty(e.s) { return emit_new_slice(e.s) }
return emit_new_struct(e.s)
}
if e.kind == E_ID {
let li = loc_find(e.s)
if li >= 0 { return emit_load_at(loc_reg[li], loc_ty[li]) }
let g = find_global(e.s)
if (g != null) {
if g.kind == N_CONST { return val(itoa(g.a.ival), "int") }
let r = emit_bind(`load {llty(g.ty)}, ptr @g_{e.s}`)
return val(r, g.ty)
}
# a UI_<name> that is not a const/var resolves to its widget index
if is_ui_ident(e.s) { return val(itoa(ui_index_of(e.s)), "int") }
perr(`unknown identifier {e.s}`)
}
if e.kind == E_MEMBER {
if e.a.kind == E_ID { # `Enum.Variant` -> its ordinal, a compile-time int
let ord = enum_ordinal(e.a.s, e.s)
if ord >= 0 { return val(itoa(ord), "int") }
}
let bt = static_type(e.a) # `x.field` where field is @Computed -> inline it
if (bt != null) {
let cx = computed_expr(bt, e.s)
if (cx != null) { return emit_expr(qualify_fields(cx, e.a)) }
}
let a = emit_member_addr(e); return emit_load_at(a, g_addr_ty)
}
if e.kind == E_INDEX {
let a = emit_index_addr(e)
if (g_addr_ty == "byte") { # a byte read, widened to int
let b = emit_bind(`load i8, ptr {a}`)
return val(emit_bind(`zext i8 {b} to i32`), "int")
}
return emit_load_at(a, g_addr_ty)
}
if e.kind == E_CALL { return emit_call(e) }
if e.kind == E_BIN { return emit_bin(e) }
if e.kind == E_UN {
let a = emit_expr(e.a)
if (e.s == ("-")) { return val(emit_bind(`sub i32 0, {a.code}`), "int") }
if (e.s == "~") { return val(emit_bind(`xor i32 {a.code}, -1`), "int") }
let c = emit_bind(`icmp eq i32 {a.code}, 0`)
return val(emit_bind(`zext i1 {c} to i32`), "bool")
}
perr("cannot emit expression")
return val("0", "int")
}