A recursive Forth word almost always opens with a guard that returns early,
so every leaf of the recursion costs a call whose whole body is that test.
`Call(self)` now compiles as `<guard> IF <what the guard returns> ELSE
Call(self) THEN`, which is what the callee would have done on entry anyway.
Half of fib's nodes are leaves: Fibonacci(25) 356 -> 237 us, 1.24x sf64 ->
0.83x, so all five benchmarks now beat it.
The guard runs twice along the recursive path, hence the bounds: at most six
effect-free operations, at most four call sites, never a tail call. WS-018.
Promotion was all-or-nothing per word, so one `.` or one host call put the
whole body -- hot loops included -- on the memory data stack, where a
loop-carried add costs 2.2 ns/iteration instead of 0.31. The stack simulator
now runs over each promotable stretch of a word; BEGIN/UNTIL, BEGIN/AGAIN and
BEGIN/WHILE/REPEAT join DO/LOOP as promotable when the construct is provably
stack-neutral; and the inliner no longer moves a loop-bearing callee into a
caller that can never be promoted.
Fixes a bug the BEGIN work uncovered, present since promotion was introduced
and shipped in 0.2.6: the loop fixup and the IF join copied locals one slot at
a time in index order, so a body that permutes the stack lost a value --
`: C 3 4 2 0 DO SWAP LOOP . . ;` printed `4 4` where gforth prints `4 3`.
Four of five benchmarks now beat sf64: Factorial 0.29x, Collatz 0.30x,
NestedLoops 0.27x, GCD 0.67x. Only Fibonacci is behind, at 1.24x. Also scale
GCD, Factorial and NestedLoops, which ran in 14-51 us where scatter and fixed
costs dominated -- that is what exposed GCD as a loss and pointed at BEGIN.
WS-014, WS-015, WS-016, WS-019.
Such a word now compiles to a fast entry (i32 x p) -> (i32 x q) carrying
its stack items as WASM values, plus the usual ( -- ) wrapper that keeps
the table slot, so EXECUTE / interpreter / host words / CATCH see the
unchanged memory ABI. Fib(25) 1035 -> 366 us, 4.3x slower than sf64 ->
1.2x; the default guards-on config 1740 -> 361 us. WS-006.
Two reporting bugs found from the browser shell.
An uncaught ABORT printed 'ABORT (throw -1)'. The standard defines ABORT
as 'empty the data stack and perform the function of QUIT', and QUIT
displays no message; gforth and SwiftForth are both silent. It now takes
the same silent path QUIT got in 0.2.5. CATCH still reports -1 and still
restores the stack depth, and ABORT" still prints its text -- different
word, different code (-2).
Compile-only constructs used in interpretation state claimed to be an
'unknown word', which is misleading for a word the system obviously
knows: ABORT", IF, THEN, LOOP, LITERAL, RECURSE and friends. They now
report 'interpreting a compile-only word: <name> (throw -14)', the
standard condition both reference engines give. The check reuses the
existing INTERPRETER_TOKENS table at the point where interpretation has
already failed, so a genuine typo still reports 'unknown word'.
Ships as v0.2.6.
The CORE word was missing. QUIT empties the return stack, enters
interpretation state, restores SOURCE-ID to the user input device and
returns to the interpreter without a message, leaving the data stack
untouched -- that last part is the whole difference to ABORT, which the
standard defines as 'empty the data stack, then QUIT'.
Implemented on the throw plumbing with the standard code -56, so nested
EVALUATE / INCLUDE frames unwind and are abandoned on the way out. Two
places treat -56 specially: CATCH lets it through (QUIT is a return to
the prompt, not an exception) and evaluate() turns it into a silent Ok
after the compile-state wipe it already performs.
Semantics checked against gforth 0.7.3 and SwiftForth sf64, which agree:
the data stack survives, nothing is printed, the rest of the input is
abandoned, and '1 2 ' QUIT CATCH .' prints nothing while leaving 1 2.
Six tests in outer.rs pin it. Deliberately NOT added to the cross-engine
corpus: what QUIT abandons is the input source, and the three engines are
fed differently there, so a comparison would measure the harness.
The gap survived because the Forth 2012 suite skips QUIT by its own
admission, and HELP's coverage lint compares dictionary against docs --
a word missing from both looks complete. docs/wafer-anki.txt had been
documenting QUIT as if it existed.
ABORT itself was already correct: executed while a definition is open it
clears both stacks and returns to interpretation state.
Ships as v0.2.5.
The doc comments still claimed a leading + binds as a sign. It does not:
sf64 converts +7 as the double 7 with DPL 1, and the code follows that.
Only a leading - is a sign.
Punctuation (`,` `.` `+` `/` `:` and an embedded `-`) after the leftmost
digit now forces double-cell conversion, so `12.34`, `1,234`, `12:30:45`
and `2026-08-06` convert as doubles. Only a trailing `.` worked before,
and `1.5` was an "unknown word" error.
The punctuation is a double-cell marker, not a fractional point, so the
scale has to travel separately: DPL carries the digit count right of the
rightmost punctuation character (negative when there was none), which is
what lets `<# #>` place the point back on output. NH carries the high
cell a single-cell conversion drops, so a token that overflows a cell is
still recoverable as a double.
parse_number and parse_double_number duplicated the prefix and sign
handling and could not share a DPL counter, so they collapse into one
parse_numeric_literal that reports which kind it converted.
Verified token-for-token against sf64. One deliberate divergence: WAFER
keeps accepting a sign before a base prefix (`-$FF`), which sf64 rejects.
- Dictionary::find no longer falls back to the newest entry across all
wordlists when the search order has no match (Forth 2012 16.3.3;
gforth and SwiftForth agree). Cross-engine corpus program guards it.
- ~40 argument-taking host words (RND-SEED, ACCEPT, RESIZE, ALLOCATE,
SEARCH, SUBSTITUTE, ROLL, M*, UM/MOD, SF@/SF!/DF@/DF!, F./FE./FS./F~,
2R@, ...) popped or read stack cells with no underflow check; on an
empty stack the pointer silently drifted past its base. New host_need/
host_fneed/host_fpop checked helpers; class-wide regression test
drives every word on an empty stack.
WS-012 -- INCLUDE/INCLUDED:
- Injected source loader (core stays IO-free: CLI installs a
filesystem reader, web leaves it unset -> defined error). Recursive
include_file feeds files line-by-line through evaluate, so compile
state and SEE capture span lines for free. Cycle detection, depth
cap 16, paths relative to the including file, SOURCE-ID per nesting
level, parent input restored on success/error/BYE.
- CLI file mode now runs through the include machinery: `wafer x.fth`
gets file:line error context and a base dir for nested INCLUDEs.
- Unlocks the REMEMBER+INCLUDE reload loop.
WS-008 -- error reporting remainder:
- Errors inside included files carry `file.fth:12:` context
(anyhow context chain; CLI prints {e:#}).
- describe_uncaught now returns typed WaferError::UncaughtThrow
{ code, message } -- display text unchanged, THROW code reachable
via downcast for CLI/web consumers.
- compile_word emits a WASM name section; wasmtime trap backtraces
name the faulting word and runtime_native prefixes "in <WORD>:".
Batch/consolidated modules stay unnamed (no name plumbing there;
boot primitives rarely trap).
WS-003 -- SwiftForth correctness lane:
- compare_all_programs_sf64 runs the program corpus with sf64 as
oracle; whitespace-token comparison (sf64 prints numbers
space-prefixed and echoes piped lines). 34/35 parity; dot-quote
skipped (interpret-mode ." is a SwiftForth no-op). #[ignore]d like
the gforth lane; `just compare-correctness` runs both.
WS-011 leftovers:
- WORDS ALL: grouped full view -- one section per wordlist (search
order first), then internal words, each with counts. Backed by
Dictionary::visible_entries (name, wid, internal); visible_words
now derives from it.
- .RS / RDEPTH: return-stack introspection in boot.fth over a new
RP@ primitive (IrOp::RpFetch); BEGIN/WHILE walk so the walk never
touches the stack it prints. SPACES clamped per 6.1.2230.
549 unit + 11 compliance + 9(+2) comparison + 5 crypto + 1 bench
green; fmt/clippy clean; --no-default-features and wasm32 web checks
pass.
Implements plans/01-see-introspection.md, all phases.
- see.rs: feature-free IR pretty-printer (format_ir/format_ir_with),
exhaustive over IrOp -- a new variant fails the build, not the output.
- SEE-IR <name>: post-optimization IR view with resolved callee names,
immediate/does> annotations; host-word and interpreter-token stubs.
- SEE <name>: verbatim source capture for colon words (multi-line,
comments preserved, EVALUATE-nesting safe, error-path wiped, MARKER/
REMEMBER/EMPTY roll word sources back too). Data definers (VARIABLE/
CONSTANT/CREATE/BUFFER:/2*/F*/SYNONYM) record synthesized one-liners
at definition time; VALUE/2VALUE/FVALUE/DEFER synthesize at SEE time
so current values and IS targets show. Fallback chain ends at IR dump
or host-word stub -- SEE never dead-ends on a defined word.
- HELP [<name>]: wordhelp.rs doc table with stack effect + one-line
description for EVERY word in a fresh VM (300+ dictionary words plus
all outer-interpreter tokens); a coverage test fails the build if a
word is ever added undocumented. User words echo their leading
( ... -- ... ) comment. SEE/SEE-IR prepend the HELP line as a
\ comment. Bare HELP prints usage.
- boot.fth colon definitions get real sources for free (they flow
through evaluate); INTERPRETER_TOKENS gained the missing ?DO.
524 unit + 11 compliance + 9 comparison + 5 crypto + 1 bench green;
fmt/clippy clean; core still builds --no-default-features; web
wasm-pack build unchanged.
Compiled code could silently move dsp/rsp/fsp out of their stack
regions (e.g. DROP on an empty stack), corrupting later pushes with
no diagnostic -- the addresses stay inside valid linear memory, so
nothing could trap. Host-side checks cannot catch it.
- Guards are emitted at the sp-adjustment choke points (dsp_inc/
dsp_dec, fsp_inc/fsp_dec, rpush/rpop/rpeek, peek, TwoDup/TwoDrop,
promoted prologue/epilogue -- DROP never loads its value, so
guarding pop() alone is not enough). On fault: write the code to
SYSVAR_FAULT_CODE, call _STACK_FAULT_, which THROWs it -- so
guards are CATCHable and print standard messages (-3/-4/-5/-6/
-44/-45).
- The batch/consolidated compile path (all boot primitives) and the
export path are wired too; a thread-local carries the fault index
into the shared emission helpers.
- Config: codegen.stack_guards, default ON. `wafer build` output
defaults OFF (production artifact); WAFER_STACK_GUARDS=0|1
overrides either. Perf comparison lanes run unguarded.
- Measured overhead in release loops: within noise (never-taken
branches).
- toolstest.fth baseline 37 -> 38: line 368's bare interpreted `R>`
used to underflow silently and count as passing; the guard now
correctly reports -6.
- REMEMBER <name>: SwiftForth-style re-runnable marker -- restores
to just AFTER its own definition and survives execution. The
edit-reload-test loop: REMEMBER fresh ... fresh ... fresh.
- EMPTY rolls back to the boot dictionary; GILD re-baselines EMPTY
to the current state. Baseline captured at VM construction.
- MarkerState now also snapshots search order, wid allocation,
compilation wordlist, REPLACES table, and ABORT" texts; restore
discards marker entries newer than the snapshot (was: newer than
the executing marker id only).
- Shared snapshot_marker_state/apply_marker_state used by MARKER,
REMEMBER, EMPTY, and GILD.
Known ambiguity (standard-conformant): a DEFER defined before a
marker but retargeted at a word defined after it dangles after
rollback; stale function-table slots are name-unreachable and get
overwritten by later definitions.
Core:
- Uncaught THROW prints its standard message ("Stack underflow
(throw -4)"; unknown codes as "Catch = <n>") instead of the
"forth-throw" sentinel. ABORT" text is carried as a structured
payload and shown only when the -2 throw goes uncaught -- CATCH
stays silent and the payload cannot go stale. ABORT throws -1
through the same path.
- WORDS: optional same-line substring filter (WORDS FDEPTH), skips
internal words (new INTERNAL header flag, set at create for
underscore-prefixed names), wraps at 78 columns, prints a count.
ORDER names wids (FORTH / wid#N) instead of Rust debug output.
- .S honors BASE. New: F.S (float stack), DUMP (hex+ASCII,
bounds-checked, 4K cap), ? (fetch-and-print, boot.fth). BYE is a
real word now: sets a VM flag the driver honors (exits REPL,
stops rest of line/file).
CLI:
- Persistent history (~/.local/state/wafer/history, 0600 perms,
$WAFER_HISTORY override), Tab completion over the live dictionary
(snapshot refreshed after each line), Up/Down do prefix history
search, Ctrl-C clears the line instead of exiting.
Web:
- History survives reloads (localStorage, cap 200, dedup, init-code
runs excluded), User Words palette populated via new words()
export, stack bar annotates non-decimal BASE, base() reads the
real BASE sysvar instead of returning a hardcoded 10.
sf64 discovery + stdin runner (no -e flag; input lines truncate at
~256 chars, so one statement per line), ucounter-based µs timing —
same wrapper shape as gforth utime. New sf64 + WAFER/sf columns,
informational only (no regression limit). Justfile: bench-compare
target; CARGO_PROFILE_RELEASE_STRIP=none for Darwin 27 dlopen bug.
Rust 1.95 promoted collapsible_match and map_unwrap_or; CI runs
-D warnings so they break the build. Collapse nested `if`s into
match guards across codegen/optimizer/export, and swap
map().unwrap_or(..) for map_or / is_ok_and.
Implement `(LOCAL)` as a host primitive that defers its effect to the
outer-interpreter compile state via two new `PendingAction` variants:
- `DeclareLocal(name)` — a non-sentinel `(LOCAL)` call with `u > 0`
appends the name to `compiling_locals` as an int local.
- `DeclareLocalEnd` — the `0 0 (LOCAL)` sentinel emits reverse-order
`ForthLocalSet` IR for the batch declared since the last sentinel,
reusing the same IR shape as the `{: ... :}` locals flow.
`local_batch_base` tracks where the current batch started; it is
saved/restored across nested compile frames and cleared on
`finish_colon_def`. Int-only, per spec — float locals remain `{F: :}`.
Also fix `\` per §6.2.2535: parse-and-discard must stop at the next
`\n`, not at `#TIB`. Under line-wrapped `evaluate` calls (common in
test files) the old behaviour consumed the trailing `;` of a multi-line
`:` definition, silently leaving state in compile mode.
Tighten `compliance.rs`: `load_file` now returns a line-failure count,
every prerequisite is asserted against `expected_load_failures(path)`,
and a new `load_file_whole` handles multi-line definitions (`DOES>`
split across lines in `errorreport.fth`) that the per-line loader
cannot stitch. Baselines document known gaps for `core.fr` (nested
`:`, SOURCE/>IN via EVALUATE), `coreexttest.fth` (SAVE-INPUT, `.(`
inside `[...]`), `exceptiontest.fth` (one garbled parse after
CATCH/THROW source stacking), and `toolstest.fth` (37 `\?`-guarded
lines where `SOURCE >IN ! DROP` fails to skip under per-line
`evaluate`). Each entry is a tech-debt ledger item, not an allowlist.
Regression tests: LT32 (the localstest case that silently skipped
before `(LOCAL)` existed), the `0 0 (LOCAL)` sentinel-only no-op, a
multi-line `:` followed by `VARIABLE` after a `\` comment, and a
direct `\` stops-at-newline case.
Incidental: clear two `implicit_clone` clippy lints in the RANDOM
determinism test (`.to_vec()` → `.clone()`).
Fix rustfmt drift and two clippy lints (`doc_markdown` missing
backticks around `NativeRuntime`) that surfaced after the Rust 1.94
toolchain update. No functional change.
compile_token matched hardcoded tokens (S, ." etc) before
checking compiling_locals. Local named `s` got hijacked by
the `S` string shortcut. Forth 2012 §13.3.3.2 — locals
supersede dict names in scope. Move locals check to top of
compile_token for uniform precedence.
Tests: S-hijack repro, get+set round-trip, int-uninit pipe
syntax coverage (`{: | name :}`).
`{: F: x F: y :}` now declares float-typed locals that live on the float
stack. `x x F* y y F* F+ FSQRT` writes real float code without manual
FSTACK juggling — previously WAFER had a 100%-compliant float wordset
but no way to name intermediate float values.
New IR ops `ForthFLocalGet(n)` / `ForthFLocalSet(n)` alongside the
existing int-local ops. Each kind has its own index namespace so mixed
declarations like `{: n F: f :}` compose cleanly. Codegen allocates f64
WASM locals after the existing f64 scratch pair; the fsp-bridge logic
mirrors the existing FDup/FSwap path.
Outer interpreter tracks a parallel `compiling_local_kinds` alongside
`compiling_locals` (keeps the 18 existing touch-points unchanged) and
extends `{:` to recognize `F:` as a per-next-name type marker. `TO` and
name resolution branch on kind to pick Int vs Float get/set ops.
Four tests: classic hypot, TO round-trip, mixed int/float args, and
uninitialized float via `|`. Inline-inhibit for the new ops added to
optimizer and is_promotable so they don't sneak into contexts that
would collide with the caller's WASM locals.
State-smart anonymous xt builder. Interpret mode leaves the xt on the
data stack; compile mode emits a literal push into the enclosing word,
so `: APPLY EXECUTE ; [: 1 2 + ;] APPLY` prints 3.
Supported nested inside colon definitions via a new compile-frame stack
(`Vec<CompileFrame>`). Each frame snapshots `compiling_name`,
`compiling_word_id`, `compiling_word_addr`, `compiling_ir`,
`control_stack`, `saw_create_in_def`, `compiling_locals`, and `state`.
The inner [: ... ;] compiles its body as an anonymous word; on ;] the
outer frame pops back and the xt is either pushed to the data stack
(interpret mode) or compiled as a literal (compile mode).
Also fixes a latent bug: `finish_colon_def` used to reveal `latest`,
which breaks when intermediate dict entries (now including quotations)
move `latest`. Each definition now tracks its own `compiling_word_addr`
and uses `reveal_at`, matching the existing DOES> pattern.
Five tests cover interpret, compile, inside-a-colon-def, two-level
nesting, and the control-stack-travels-with-frame regression (outer
IF/ELSE/THEN must still match around an inner [: ;]).
BEGIN-STRUCTURE, END-STRUCTURE, +FIELD, FIELD:, CFIELD:, FFIELD:,
SFFIELD:, DFFIELD: — the Forth 2012 structure-definition family plus
the float-typed variants for symmetry with WAFER's float wordset.
Each defining word carries its own inline CREATE .. DOES> — factoring
through a shared +FIELD helper doesn't work in WAFER, because DOES>-
defining words only dispatch at the outer interpreter, not from compiled
IR. So FIELD: can't call +FIELD and have the DOES> action fire; each
FIELD:/CFIELD:/... repeats the pattern directly.
Three tests cover size computation, field offsets, and mixed cell + char
fields with alignment.
Non-standard but ubiquitous in gforth/SwiftForth/VFX. Adds a shared
rng_state on ForthVM, seeded from nanosecond wall-clock at boot.
`RANDOM ( -- u )` returns a 32-bit pseudo-random cell; `RND-SEED ( u -- )`
reseeds, with 0 forced to a nonzero constant to avoid xorshift's fixed
point.
Three tests cover determinism after seeding, distinct-value spread
across 1000 pulls, and the zero-seed safeguard.
`S name` in interpret mode used to leave (c-addr u) pointing into the
input buffer, so the next REFILL clobbered the bytes. Typing `s test`
then `type` on a fresh line printed "pest" because the new input
overwrote the first chars of the old TIB content.
Move `S` from boot.fth to the Rust outer interpreter alongside `S"` /
`C"`: both interpret and compile modes now copy the token to HERE-space
(stable across REFILL). Compile-mode output is still bit-identical to
writing `S" name"` inline.
Adds `test_s_interpret_survives_refill` regression.
`S name` is the string analogue of `[CHAR] x` and `['] name`: parses the
next whitespace-delimited token, state-smart.
Interpret: leaves ( c-addr u ) pointing into the input buffer.
Compile: appends run-time push of the copied bytes (identical code
to writing S" name" inline).
One line in boot.fth, leverages the existing PARSE-NAME + SLITERAL.
Zero runtime overhead inside : definitions.
Introduces a `crypto` feature (on by default) that wires the RustCrypto
sha1/sha2 crates into a small `HashAlgo` registry. `register_primitives`
iterates `crypto::ALGOS` and installs one Forth host word per algorithm,
each with the stack effect
( c-addr u -- c-addr2 u2 )
reading `u` bytes from `c-addr` and writing the digest into a shared
`HASH_SCRATCH` region in linear memory (carved out between the float
stack and the dictionary).
Adding a new hash is a one-line entry in `ALGOS`. `register_host_primitive`
is now `pub` so downstream crates can extend the VM with their own I/O
host words without forking WAFER — kelvar (a deterministic password
manager on WAFER) is the first consumer.
- 4 unit tests (lib-level sha1/256/512 + registry sanity)
- 5 integration tests (in-VM `SHA1`/`SHA256`/`SHA512` against RFC-3174,
FIPS-180, and the first-round S/KEY seed used by `hel`)
- All 437 existing lib tests still pass; `wafer-web` still builds for
`wasm32-unknown-unknown` with the feature enabled
Implement PAGE (Facility word set) as IR primitive emitting form feed.
Web REPL clears output div on form feed, CLI REPL sends ANSI clear.
Fix init code panel: use default textarea content instead of placeholder
so init code actually executes on first visit. Update wasm-pack 0.10→0.14
and refresh Cargo.lock to latest compatible versions.
Decouple ForthVM from wasmtime via a Runtime trait so the same outer
interpreter, compiler, and 200+ word definitions work on both native
(wasmtime) and browser (js-sys WebAssembly API) backends.
Runtime trait (runtime.rs):
- HostAccess trait for memory/global ops inside host function closures
- HostFn type: Box<dyn Fn(&mut dyn HostAccess) -> Result<()>>
- Runtime trait: memory, globals, table, instantiate, call, register
NativeRuntime (runtime_native.rs):
- Wraps wasmtime Engine/Store/Memory/Table/Global/Func
- CallerHostAccess bridges HostAccess to wasmtime Caller API
- Feature-gated behind "native" (default)
outer.rs refactor:
- ForthVM<R: Runtime> — generic over execution backend
- All 87 host functions converted from Func::new closures to HostFn
- All memory access via rt.mem_read/write_*, global access via rt.get/set_*
- Zero logic changes — pure API conversion
wafer-core feature gates:
- default = ["native"] includes wasmtime + all native modules
- Without "native": pure Rust only (outer, codegen, optimizer, dictionary)
Browser REPL (crates/web):
- WebRuntime: js-sys WebAssembly.Memory/Table/Global/Module/Instance
- WaferRepl: wasm-bindgen entry point (evaluate, data_stack, reset)
- WebAssembly.Function with Safari fallback (wrapper module)
- Frontend: dark terminal UI, word panel, init code editor, history
- Build: wasm-pack build --target web
All 452 tests pass (431 unit + 1 benchmark + 9 comparison + 11 compliance).
wasmtime 31→43, wasm-encoder/wasmparser 0.228→0.246, rustyline 15→18.
API migrations: F64Const now takes Ieee64 wrapper, wasmtime has own
Error type (wasmtime::bail! in host closures), cache_config_load_default
removed. Add performance regression limits to benchmark tests.
Three compile-time words for unstructured control flow:
- AHEAD: unconditional forward branch (code to THEN skipped)
- CS-PICK: duplicate control-flow stack entries (enables multi-exit loops)
- CS-ROLL: rotate control-flow stack entries (reorder IF/THEN resolution)
Also adds POSTPONE support for compile-time keywords (IF, UNTIL, etc.)
via a __CTRL__ host function and unified pending_actions queue.
Key design:
- LoopRestartIfFalse IR op desugars into nested If nodes for CS-PICK'd
BEGIN+UNTIL patterns (multiple backward branches in one loop)
- Flat Block/BranchIfFalse/EndBlock IR ops for CS-ROLL'd IF/THEN
patterns where structured If nesting would consume wrong flags
- First-iteration flag local for AHEAD-into-BEGIN patterns (PT8)
Enables 12th compliance test (compliance_tools): all 11+1 now pass.
Three bugs fixed to safely enable promotion for control flow:
1. compute_stack_needs now recurses into IF/DoLoop/Begin bodies,
correctly calculating preload counts for promoted words with
nested control flow (was flat, causing stack underflow).
2. BeginDoubleWhileRepeat rejected from promotion (boot.fth's
-TRAILING uses this pattern, handler had structural bugs).
3. IF/ELSE branches must have same net stack effect for promotion
(BITSSET? has asymmetric branches: 2 items vs 1).
Performance with promotion enabled:
- Factorial: 0.50x (2x faster than gforth)
- Collatz: 0.38x (2.6x faster than gforth)
- All 427 unit tests, 10/11 compliance, 35/35 behavioral pass
Extends the promoted codegen path (StackSim) with handlers for DoLoop,
BeginWhileRepeat, BeginUntil, BeginAgain, If/Else/Then, RFetch, LoopJ,
and Exit. Includes loop-iteration fixup to copy modified locals back to
loop-top positions, and IF branch state merging.
The promotion is currently gated off for control flow (is_promotable
rejects all loops/IF) pending fix for edge cases in the Forth 2012 test
suite. The infrastructure is ready to enable incrementally.
When briefly enabled for testing, showed dramatic results:
- Factorial: 0.49x (2x faster than gforth)
- Collatz: 0.17x (6x faster than gforth)
Two-path DO/LOOP codegen based on static analysis of the loop body:
- Fast path (no calls, no >R/R> in body): index and limit live purely
in WASM locals with zero return stack traffic per iteration. RFetch (I)
and LoopJ (J) resolve to local.get instead of memory access.
- Slow path (body has calls or explicit RS ops): locals still used for
loop control, but synced to return stack for LEAVE/UNLOOP compatibility.
Also converts J from a host function (WASM→Rust roundtrip per call) to
an IR primitive (IrOp::LoopJ) that compiles to local.get of the outer
loop's index local.
Performance impact (vs gforth, all opts enabled):
- Factorial: 1.02x → 0.94x (now faster than gforth)
- NestedLoops: 717x → 543x (24% faster, still bottlenecked by data stack)
- Fibonacci, GCD, Collatz: unchanged (don't use DO/LOOP)
35 behavioral tests across 8 categories verify identical output between
WAFER and gforth. Performance benchmarks compare execution speed for
Fibonacci, Factorial, GCD, NestedLoops, and Collatz workloads.
WAFER-only correctness tests run in CI without gforth; cross-engine
comparison and performance report are opt-in via --ignored.
Major compliance push bringing WAFER from 3 to 10 passing Forth 2012
compliance test suites (Core, Core Extensions, Core Plus, Double,
Exception, Facility, Locals, Memory, Search Order, String).
Compiler/runtime fixes:
- DEFER: host function via pending_define, works inside colon defs
- COMPILE,: handle_pending_compile in execute_word for [...] sequences
- MARKER: full save/restore with pending_marker_restore mechanism
- IMMEDIATE: changed from XOR toggle to OR set per Forth 2012 spec
- ABORT": throw -2 via THROW, no message display when caught
- M*/: symmetric division to match WAFER's / behavior
- pending_define: single i32 flag → Vec<i32> queue for multi-action words
- Optimizer: prevent inlining words containing EXIT or ForthLocal ops
- +LOOP: corrected boundary check formula with AND step comparison
- REPEAT: accept bare BEGIN (unstructured IF...BEGIN...REPEAT)
- Auto-close unclosed IFs at ; for unstructured control flow
- _create_part_: use reserve_fn_index to preserve dictionary.latest()
Memory layout:
- Separate PICT_BUF and WORD_BUF regions to prevent PAD overlap
- Updated DEPTH hardcoded DATA_STACK_TOP in boot.fth
New word sets:
- [IF]/[ELSE]/[THEN]/[DEFINED]/[UNDEFINED]: conditional compilation
- UNESCAPE/SUBSTITUTE/REPLACES: string substitution (host functions)
- Locals {: syntax: parser, ForthLocalGet/Set IR ops, WASM local codegen
- ENVIRONMENT? support for #LOCALS (returns 16)
- N>R/NR>/SYNONYM: programming-tools extensions
- Search Order: ONLY, ALSO, PREVIOUS, DEFINITIONS, FORTH,
FORTH-WORDLIST, GET-ORDER, SET-ORDER, GET-CURRENT, SET-CURRENT,
WORDLIST, SEARCH-WORDLIST with full multi-wordlist dictionary support
via Arc<Mutex> shared state for immediate effect from compiled code
Remaining: 1 cascade error in Programming-Tools from CS-PICK/CS-ROLL
(unstructured control-flow stack manipulation, requires flat IR).
- SOURCE-ID now returns -1 during EVALUATE (saves/restores SYSVAR_SOURCE_ID)
- BUFFER: aligns HERE to cell boundary before allocating
- S\" returns Vec<u8> instead of String to preserve raw escape bytes
Core_ext: 14→6 errors. Total: 46→44.
parse_s_escape returned String via from_utf8_lossy which replaces
non-UTF-8 bytes (like \xAB = 171) with the 3-byte U+FFFD replacement
character, corrupting both string length and content.
Changed to return Vec<u8> and write raw bytes directly to WASM memory.
Also registered ( as immediate word for FIND, added 'x' char literals.
Core_ext: 14→8 errors.
- Register ( in dictionary as immediate so FIND can discover it
(fixes search-order FIND test: 4→3 errors)
- Add character literal parsing: 'z' → 122 (Forth 2012 number prefix)
- Fix ALLOCATE/RESIZE -1 size validation (memory suite now passes)
Oleksandr Kozachuk
ok2