AI_Assistant/pcbjam
Watch
1
0
Fork
You've already forked pcbjam
 0
Code Issues Pull requests Projects Releases 4 Packages Wiki Activity Actions 2

port staging 29c61b8's zero-duration nanosleep guard (allocator must not suspend)

Browse source 
mimalloc's mi_atomic_yield() is sleep(0) on wasm, reached from malloc's slow
path under cross-thread delayed-free contention. Under JSPI a yield there
suspends the activation INSIDE the allocator — any other activation that runs
next can re-enter mimalloc mid-operation. ms==0 now busy-returns like stock
emscripten. (The removelist half of the staging fix is obsolete here — no
asyncify, no removelist. The mimalloc-storm harness + spec arrive with the
staging merge.)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01NDeBaKKhQztd8KiVtHuyXr
This commit is contained in:
Viktor Vaczi 2026-08-13 09:04:43 +02:00
commit 94cac5e4d4
1 changed files with 18 additions and 5 deletions
Download patch file Download diff file Expand all files Collapse all files

23
wasm/shims/nanosleep_yield.c
View file

@ -23,6 +23,16 @@
* emscripten_thread_sleep (the real underlying blocking sleep — workers may block). * emscripten_thread_sleep (the real underlying blocking sleep — workers may block).
* *
* SCOPE: only the main browser thread yields; only it must never block the event loop. * SCOPE: only the main browser thread yields; only it must never block the event loop.
*
* ZERO-DURATION GUARD (ported from staging 29c61b8): a nanosleep of 0 ms
* returns immediately WITHOUT yielding. mimalloc's spin-wait hint
* mi_atomic_yield() is sleep(0) on wasm, reached from malloc's slow path
* under cross-thread delayed-free contention. Under JSPI a yield there
* SUSPENDS the activation inside the allocator — any other activation that
* then runs can re-enter mimalloc mid-operation. A zero-duration sleep never
* promised an event-loop turn — stock emscripten busy-waits and returns
* immediately. mimalloc is the module's only zero-duration sleeper; the
* worker-boot deadlock this shim fixes needs only the ms-scale yields.
*/ */
#include <emscripten/emscripten.h> #include <emscripten/emscripten.h>
#include <emscripten/threading.h> #include <emscripten/threading.h>
@ -69,13 +79,16 @@ int nanosleep( const struct timespec* req, struct timespec* rem )
if( req ) if( req )
{ {
double ms = (double) req->tv_sec * 1000.0 + (double) req->tv_nsec / 1.0e6; double ms = (double) req->tv_sec * 1000.0 + (double) req->tv_nsec / 1.0e6;
if( emscripten_is_main_runtime_thread() ) if( ms > 0.0 )
{ {
if( !pcbjam_context_sleep_ms || !pcbjam_context_sleep_ms( ms ) ) if( emscripten_is_main_runtime_thread() )
__wasm_main_thread_yield_ms( ms ); /* yield -> event loop runs -> Worker boots */ {
if( !pcbjam_context_sleep_ms || !pcbjam_context_sleep_ms( ms ) )
__wasm_main_thread_yield_ms( ms ); /* yield -> event loop runs -> Worker boots */
}
else
emscripten_thread_sleep( ms ); /* worker: real blocking sleep */
} }
else
emscripten_thread_sleep( ms ); /* worker: real blocking sleep */
} }
if( rem ) if( rem )
{ {