simon/jolt
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

feat: success checker on by default in direct-link builds, free (jolt audit)

Browse source 
Checking inherently needs an inference pass (~2.6x compile as a standalone
pass). But direct-link builds ALREADY run one inference pass for
specialization (run-passes' infer-top), so checking can ride along: set a
check-mode flag, turn checking? on during that existing pass, and collect
the diagnostics after — ~2% overhead measured on the ray tracer, vs 2.6x
for the separate pass.

So the checker now defaults to `warn` in direct-link builds (where it's
nearly free) and stays OFF in plain REPL/dev builds (no inference to ride,
no forced cost — opt in with JOLT_TYPE_CHECK there). JOLT_TYPE_CHECK still
overrides in both directions (off to disable, error to escalate).

It checks the POST-optimization IR, which matches what the optimized
program actually evaluates — scalar-replace only drops provably-pure code,
an accepted opt-mode divergence, so no real error is hidden. The loaders
enable position tracking whenever checking will run (env-selected or
direct-link). type-check! (the standalone pass) stays for plain builds;
both paths share report-diags!.

cli-test pins: plain build silent, direct-link warns by default,
JOLT_TYPE_CHECK=off disables. Gate green, suite 4718, runtime bench even.
This commit is contained in:
Yogthos 2026-06-13 13:46:16 -04:00
parent 69af83da89
commit 088232b778
6 changed files with 111 additions and 35 deletions
Download patch file Download diff file Expand all files Collapse all files

27
jolt-core/jolt/passes.clj
View file

@ -1311,6 +1311,21 @@
(get fnode :arities))))
def-node)))
;; Piggyback checking (jolt audit). In direct-link mode infer-top already runs
;; one inference pass for specialization; turning checking? on during it makes
;; the success checker nearly free there (no extra traversal — just the
;; per-call error-domain predicates). The back end sets the mode before
;; run-passes and reads take-diags! after. It checks the POST-optimization IR,
;; which matches what the optimized program actually evaluates (scalar-replace
;; only drops provably-pure code, an accepted opt-mode divergence).
(def ^:private check-mode-box (atom {:on false :strict false}))
(defn set-check-mode!
"Enable/disable checking during the next run-passes inference (direct-link)."
[on strict?] (reset! check-mode-box {:on (if on true false) :strict (if strict? true false)}))
(defn take-diags!
"Diagnostics accumulated by the last checking run-passes; clears the buffer."
[] (let [d (vec @diag-box)] (reset! diag-box []) d))
(defn run-passes
"All passes, in order. The back end applies this to every analyzed form. When
inlining is enabled for the unit (user code under direct-linking, jolt-87f),
@ -1327,5 +1342,15 @@
(if (and @dirty (< i 8))
(recur (inc i) n2)
n2)))]
(infer-top opt))
;; specialization inference, optionally also emitting success diagnostics
(if (get @check-mode-box :on)
(do (reset! diag-box [])
(reset! checking-box #{})
(reset! strict-box (get @check-mode-box :strict))
(reset! checking? true)
(let [r (infer-top opt)]
(reset! checking? false)
(reset! strict-box false)
r))
(infer-top opt)))
(const-fold node)))