JS WAM mixed-mode lowering for the peerhailer argparser (GP-PERF) - #4209
Draft
s243a wants to merge 6 commits into
Draft
JS WAM mixed-mode lowering for the peerhailer argparser (GP-PERF)#4209s243a wants to merge 6 commits into
s243a wants to merge 6 commits into
Conversation
Emit T4/ITE/det functions for eligible cli_args predicates, intern ground facts after first success, and keep shapes the emitter cannot honour (multi_clause_1, Execute of a non-self callee) on the interpreter. ITE else restores a full machine snapshot; nested Runtime.run of an interpreted callee is isolated so it cannot steal the caller's CPs. Co-authored-by: johns243a <johns243a@gmail.com>
switch_on_structure re-emits the same ground clause on a try/retry/trust chain; treat identical ground copies as one deterministic body so the ground-memo intern still fires (GP-PERF item 2). Co-authored-by: johns243a <johns243a@gmail.com>
Lowered helpers were still allocating I.* objects and snapshotting the whole machine on every T4/ITE step, so mixed mode was no faster than the interpreter. Call Runtime.op_* instead, restore A-registers on builtin-only ITE, intern ground facts without copy_term, and allow Execute of JS WAM builtins (substring_from/3 and friends). Co-authored-by: johns243a <johns243a@gmail.com>
Those paths overflowed string_member/2 (cyclic GetList write) and broke the 17-case corpus. Mixed-mode T4+ITE via Runtime.step remains the correct fast path. Co-authored-by: johns243a <johns243a@gmail.com>
Lowered helpers still allocated I.* objects and snapshot_machine on every ==/2 step. Call the shared op_* implementations (same semantics as Runtime.step, including [|] fid default) and restore only A-registers for pure comparison conditions. Keep copy_term ground intern and keep Execute of non-self callees (including builtins) on the interpreter. Co-authored-by: johns243a <johns243a@gmail.com>
The shared structure_ite/2 split_commit took the first cut inside a nested negation, so T4+ITE rejected the argparser loops. A depth-aware fold in the JS emitter lowers those loops and parse_args/3. Execute of a non-self callee (including builtins) stays interpreted. Regenerated wamjs/js/; document GP-PERF timings (4.639s → 3.144s, 1.48×). Co-authored-by: johns243a <johns243a@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
GP-PERF item 1: mixed-mode JS WAM lowering + ground-fact intern for the peerhailer argparser.
Speedup: 1.48× on the 5067-line differential (A2 interpreter 4.639s → mixed 3.144s). Stretch ≥5× was not met. Corpus 17/17, differential 0 divergences / 0 message mismatches.
What changed
emit_mode(mixed)forexamples/cli_args/wamjs/— lower every eligible predicate, interpret the rest.// wamjs lower fallback:comments.\+does not abort T4 (lenient_loop/5,strict_loop/8,parse_args/3now lower).Runtime.op_*(no per-instructionI.*allocation). InterpreterRuntime.stepshares the same helpers; GetList defaults a missing fid to[|].==/2,=:=/2, …). Unifying conditions stillsnapshot_machine.program.ground_memo) viacopy_term. Live-term intern is unsound.Still interpret (and why)
parse_args/2,first_equals_index/2,looks_like_legacy_flag/1,lenient_result/2,merge_flags/3,is_long_flag/1,js_object_prototype_key/1— Execute of non-self (nestedRuntime.runwithcp=0steals CP).substring_from/3,substring_range/4— Execute ofsub_string/5.drop_brackets/2—multi_clause_1; T4+ITE did not match.Testing
node --test examples/cli_args/wamjs/cliArgs.wamjs.test.mjs→ 17/17run_differential_wamjs.sh→ 5067 lines, 0/0; oracle 0.054s, wamjs 3.144stests/test_wam_javascript_lowered.pl, builtins, fact-sources,CONFORMANCE_TARGETS=javascript— greenUW_PROFILE=1200-line:string_member/2*still 1804 callsConstraints honoured
examples/cli_args/outsidewamjs/are untouched.