[WASM R4] Audit official profiles and standard-library completeness by cpunion · Pull Request #235 · cpunion/llgo

cpunion · 2026-09-05T13:18:17Z

Scope

Consolidated R4 development PR for xgo-dev#2152. R4 raises WebAssembly completeness in two high-priority areas:

llgo run, build, and test toolchain completeness.
Compatibility with the official Go js/wasm and wasip1/wasm profiles, standard-library behavior, and executable test//GOROOT acceptance.

The required LLGo profiles are EC32, EC64, WC32, GJS, and GWASI. Official-Go GJS/GWASI rows are used as reference executions. Multi-worker/parallel goroutines, WasmGC, exception handling, and JSPI/stack switching remain later work and are not claimed by this PR.

Design references: parent proposal xgo-dev/llgo#2152, typed reflection bridge sub-proposal #2556, and Emscripten host sub-proposal #2557.

Current R4 head: 05e962ddf, rebased on upstream main 309f75c5e (LLVM 22).

What this adds

Toolchain and profile completeness

Raw GOOS=js GOARCH=wasm and single-worker GOOS=wasip1 GOARCH=wasm builds select LLGo's linear-memory collector by default. Unsupported raw WASI threads are rejected instead of silently selecting nogc.
GJS reuses the selected GOROOT's public syscall/js implementation; LLGo supplies the host import/event bridge and follows the same wasm_exec.js contract.
Official-Go GWASI references use GOROOT's go_wasip1_wasm_exec. LLGo GWASI/WC32 run through Wasmtime with matching filesystem, working-directory, and feature configuration.
Source-patch inputs, Wasm ABI, native toolchain identity, test-package identity, synthetic debug locations, and goroutine stack sizing participate correctly in package/cache invalidation.
run, build, and test are exercised through real profile acceptance, not compile-only probes.

Runtime and standard-library completeness

Raw GJS/GWASI support finalizers and weak pointers with the linear collector.
GJS covers nested/interleaved callbacks, channel and timer waits, panic recovery, typed byte copies, JS errors, reference accounting, and exactly-once side effects under both LLGo and official Go.
WASI polling covers readiness, timeout, deadline changes, cancellation, and close wakeups, with explicit C32-compatible stat/poll layouts.
Wasm panic/caller handling preserves logical Go frames across defer, recovery, and the Wasm shadow stack, including the previously failing issue14646, issue5856, and issue22662 cases.
Standard-library DNS behavior returns the appropriate net.DNSError for optional localhost records rather than skipping the supported test.
RSA/TLS fixtures reuse generated keys where key generation itself is not under test, while retaining coverage of real key-generation paths.

Linear GC correctness

Runtime continuation storage can be kept live without conservatively scanning its entire capacity. This avoids retaining arbitrary objects through stale words in multi-megabyte Fiber/Asyncify buffers.
Every execution context publishes its compiler root chain and the used logical stack range. The active range is bounded by the current SP; suspended contexts use their saved SP.
Emscripten Fiber layout is checked at C compile time, including the saved stack-pointer offset.
The main-stack regression suspends with the only live reference encoded as a uintptr, lets another goroutine perform two collections, then verifies the object survives. This specifically covers WASI's heap-allocated logical main stack.
Static-data root discovery uses a one-link marker before mutable data, so immutable packed .rodata is not scanned. The marker adds only one pointer-sized boundary object and preserves LTO/GlobalDCE.
Native TLS package blocks are published as one exact context root while their owning Fiber is suspended. The collector does not conservatively scan the complete Emscripten TLS or Fiber allocation; five-profile stress tests force two collections before reusing a mixed scalar/pointer initializer.
The collector remains a single-mutator design. Embedded targets reuse the unchanged default allocation behavior; separate ESP32/ESP32-C3 gates guard against Wasm-only regressions.

Acceptance policy

Five startup sentinels gate the full matrices; browser execution is independently required.
Compatible test/ cases run in two shards per profile. Runnable GOROOT directives run in four shards per profile, with retained logs and JSON reports.
Exact reviewed notapplicable classifications are distinct from expected failures, failures, incomplete execution, source exclusion, and cases that never ran. Top-level test skips fail acceptance; only intentional skipped subtests inside a passing test are accepted.
test/cgo remains explicitly outside GOARCH=wasm because Go's cgo frontend rejects that architecture, even for LLGo's named C ABI targets. In contrast, every LLGo Wasm profile now executes the compatible runtime/cgo.Handle tests; only official-Go reference profiles exclude them.
The pre-existing Go 1.27 genmeth1.go SSA assertion is tracked separately in go/ssa: Go 1.27 genmeth1.go panics while instantiating generic method expressions xgo-dev/llgo#2526 with one exact xfail entry.
Missing required tools such as llvm-nm are errors, not silent skips.

Current validation

Final-head independent GC: run 34444621899 — passed Linux x86-64/ARM64, macOS, Windows, ESP32/Xtensa, and ESP32-C3/RISC-V. The collector remains explicitly single-mutator; unsupported concurrent entry is rejected.
Rebased-head full Wasm acceptance: run 34460966471 — browser acceptance and all five profile sentinels passed; the complete compatible test/ and GOROOT matrices are running with no observed failure.
Rebased-head standard-library/reference matrix: run 34460966559 — all five rows passed.
Rebased-head/latest-main size and build benchmark: run 34460966769 — passed. The size gate remains open: versus 309f75c5e, cprintf adds 35,526 bytes on EC32 and 37,096 bytes on EC64; fmtprintf shrinks by 136–175 KiB on the C-ABI profiles, while GJS/GWASI include the larger Go-compatible runtime and host surface.

Local post-rebase checks pass:

go test ./test/goroot -count=1
go test ./internal/build -run 'TestGC(Independent|Finalizer)Arena|TestWasmStaticRootMarker|TestWasmFuncInfoNoScanLinking' -count=1
(cd runtime && go test ./internal/gcroot ./internal/clite/emscripten ./internal/wasmcontext ./internal/runtime/tinygogc -count=1)
go build ./cmd/llgo
node dev/test_wasm_fs.mjs
git diff --check

Remaining merge gates

Complete all five final-head Wasm acceptance and GOROOT matrices, including the two previously borderline issue30116* timeout cases.
Review final-head stdlib/reference reports and browser acceptance.
Compare final-head cprintf, println, and fmtprintf sizes against the rebased main baseline; investigate any material increase.
Run the general coverage workflow on the final head and require patch/project coverage to pass. Acceptance-only code without CI execution is not considered complete.
After the focused Wasm gates are green, run the remaining general repository workflows and resolve review comments.

cpunion · 2026-09-08T18:18:45Z

4685f7d validation update: all five profile sentinels, browser acceptance, and the standard-library reference workflow passed. Full test/ and GOROOT shards are still running. The two completed reference failures are test/go/memprofile: the assertion incorrectly applied LLGo’s pointer-width-scaled block size to official Go wasm, whose tiny allocator still uses 16-byte blocks. Local e305bad restricts that implementation-specific assertion to runtime.Compiler == "llgo"; official Go GJS/GWASI, native Go, and the LLGo GJS regression all pass. Holding this small test-only fix for the next batch while the current matrix collects other evidence. The new Wasm size benchmark completed: versus 03abd35, cprintf is approximately +0.6 KiB and fmtprintf +2 KiB; the total R4 size target is not yet met. Compile-phase diagnostics for the two known resource failures are isolated in fork run34260202320 (one job, unchanged per-case limits), currently waiting for a runner. SetGCPercent/GOGC collector integration is being validated separately; no claim of full R4 completion or full-matrix success.

cpunion · 2026-09-08T18:34:15Z

Moving the consolidated R4 contribution and further CI to xgo-dev/llgo, as requested. The active fork acceptance run has been cancelled; completed evidence and discussion remain here. The fork branch is retained, and the replacement upstream PR will be linked shortly. This is a migration, not a claim of completed R4 acceptance.

cpunion · 2026-09-08T18:38:48Z

Continuing R4 CI in cpunion/llgo per the updated runner-capacity decision. No replacement R4 PR was created upstream. The branch now includes e305bad, which fixes the official-Go wasm tiny-allocation assertion; focused official-Go GJS/GWASI, native Go, and LLGo GJS checks pass. The previous full acceptance run was cancelled during the attempted migration and is not claimed as completed. A fresh fork CI run will validate this head; pending GC-pacing and caller-stack experiments are not included yet.

js.FuncOf queued every callback and ran it later on a new goroutine. syscall.fsCall waits on a buffered channel after fs.write, so that path parked, fiber-swapped, and aborted with Asyncify unreachable. Keep host events such as setTimeout queued, but run the callback on the current goroutine when Go is still inside Value.Call/Invoke.

Copy targets/wasm_fs.js next to emcc glue for .html, .js, and .mjs executables, and insert the script tag into generated HTML so syscall/fs_js.go has a browser fs/process/path host. Map Emscripten errno 13 to ECONNABORTED. Skip a same-path copy and close the source before replace so Windows rebuilds do not fail with Access is denied. Do not os.Remove a directory after a failed rename; that replaced occupied/ with a regular file and broke TestWindowsLinkObjFilesExactOutput/copy_error. Skip the 0555 CreateTemp probe on Windows, where directory mode bits do not block new files.

github-actions · 2026-09-10T10:03:49Z

LLGo baseline benchmarks

05e962ddf030 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs base	Text size	vs base	Build	vs base	Run	vs base
Linux	cprintf	20264 B	+416 B / +2.1% (worse)	387 B	0 B / +0.0%	441.532 ms	+28.24 ms / +6.8% (worse)	1.294 ms	-21.03 us / -1.6% (better)
Linux	cprintf-lto	20016 B	+416 B / +2.1% (worse)	368 B	0 B / +0.0%	428.812 ms	+18.11 ms / +4.4% (worse)	1.415 ms	+126.6 us / +9.8% (worse)
Linux	fmtprintf	1645312 B	+28928 B / +1.8% (worse)	510635 B	+18328 B / +3.7% (worse)	3.013 s	+168.6 ms / +5.9% (worse)	3.026 ms	+111.2 us / +3.8% (worse)
Linux	fmtprintf-lto	1496600 B	+29408 B / +2.0% (worse)	452298 B	+18186 B / +4.2% (worse)	8.815 s	+201.6 ms / +2.3% (worse)	2.860 ms	+6.501 us / +0.2% (worse)
Linux	println	62912 B	+160 B / +0.3% (worse)	14806 B	-233 B / -1.5% (better)	419.115 ms	+13.32 ms / +3.3% (worse)	1.631 ms	-82.44 us / -4.8% (better)
Linux	println-lto	54488 B	+176 B / +0.3% (worse)	12211 B	-220 B / -1.8% (better)	603.763 ms	+26.54 ms / +4.6% (worse)	1.608 ms	-52.83 us / -3.2% (better)
macOS	cprintf	84480 B	0 B / +0.0%	17533 B	+416 B / +2.4% (worse)	606.097 ms	-489.5 ms / -44.7% (better)	2.509 ms	-4.806 ms / -65.7% (better)
macOS	cprintf-lto	84288 B	0 B / +0.0%	13297 B	+416 B / +3.2% (worse)	623.849 ms	-215.4 ms / -25.7% (better)	2.626 ms	-1.498 ms / -36.3% (better)
macOS	fmtprintf	1490672 B	+17184 B / +1.2% (worse)	889816 B	+23188 B / +2.7% (worse)	2.992 s	-174.1 ms / -5.5% (better)	6.740 ms	+622 us / +10.2% (worse)
macOS	fmtprintf-lto	1175888 B	+16448 B / +1.4% (worse)	861852 B	+21784 B / +2.6% (worse)	7.052 s	-1.482 s / -17.4% (better)	5.314 ms	+454 us / +9.3% (worse)
macOS	println	114864 B	0 B / +0.0%	35373 B	+272 B / +0.8% (worse)	638.278 ms	+18.66 ms / +3.0% (worse)	3.930 ms	+458 us / +13.2% (worse)
macOS	println-lto	118736 B	0 B / +0.0%	33009 B	+280 B / +0.9% (worse)	673.411 ms	-230.7 ms / -25.5% (better)	3.651 ms	-292.9 us / -7.4% (better)
Windows MinGW	cprintf	19456 B	0 B / +0.0%	4550 B	0 B / +0.0%	915.798 ms	-37.64 ms / -3.9% (better)	2.816 ms	+22.2 us / +0.8% (worse)
Windows MinGW	cprintf-lto	17920 B	0 B / +0.0%	4486 B	0 B / +0.0%	948.457 ms	-22.33 ms / -2.3% (better)	2.699 ms	-139.7 us / -4.9% (better)
Windows MinGW	fmtprintf	1910784 B	+24576 B / +1.3% (worse)	608694 B	+15664 B / +2.6% (worse)	3.375 s	+33.03 ms / +1.0% (worse)	7.304 ms	+59.5 us / +0.8% (worse)
Windows MinGW	fmtprintf-lto	1958912 B	+31232 B / +1.6% (worse)	556358 B	+14992 B / +2.8% (worse)	8.518 s	+190.5 ms / +2.3% (worse)	6.856 ms	+263 us / +4.0% (worse)
Windows MinGW	println	71680 B	0 B / +0.0%	23638 B	-336 B / -1.4% (better)	915.135 ms	-10.07 ms / -1.1% (better)	5.501 ms	-88.8 us / -1.6% (better)
Windows MinGW	println-lto	65024 B	-512 B / -0.8% (better)	20454 B	-272 B / -1.3% (better)	1.094 s	+4.377 ms / +0.4% (worse)	5.521 ms	-297.5 us / -5.1% (better)
Windows MinGW 386	cprintf	43008 B	+5120 B / +13.5% (worse)	5326 B	0 B / +0.0%	1.182 s	+60.53 ms / +5.4% (worse)	5.195 ms	+211.4 us / +4.2% (worse)
Windows MinGW 386	cprintf-lto	20992 B	0 B / +0.0%	5094 B	0 B / +0.0%	1.159 s	+26.24 ms / +2.3% (worse)	5.139 ms	+92 us / +1.8% (worse)
Windows MinGW 386	fmtprintf	1880576 B	+37376 B / +2.0% (worse)	485454 B	+18208 B / +3.9% (worse)	4.345 s	+264.9 ms / +6.5% (worse)	11.854 ms	+328.2 us / +2.8% (worse)
Windows MinGW 386	fmtprintf-lto	2217472 B	+53760 B / +2.5% (worse)	460242 B	+13240 B / +3.0% (worse)	10.314 s	+540.8 ms / +5.5% (worse)	11.142 ms	-1.026 ms / -8.4% (better)
Windows MinGW 386	println	92160 B	+5632 B / +6.5% (worse)	20182 B	-192 B / -0.9% (better)	1.112 s	+18.46 ms / +1.7% (worse)	8.680 ms	-232 us / -2.6% (better)
Windows MinGW 386	println-lto	70656 B	0 B / +0.0%	18050 B	-212 B / -1.2% (better)	1.310 s	-12.85 ms / -1.0% (better)	8.592 ms	+24.3 us / +0.3% (worse)
Windows MinGW ARM64	cprintf	18944 B	0 B / +0.0%	4436 B	0 B / +0.0%	1.450 s	-52.71 ms / -3.5% (better)	6.106 ms	-1.128 ms / -15.6% (better)
Windows MinGW ARM64	cprintf-lto	17920 B	0 B / +0.0%	4368 B	0 B / +0.0%	1.467 s	-29.65 ms / -2.0% (better)	6.432 ms	-230.9 us / -3.5% (better)
Windows MinGW ARM64	fmtprintf	1803776 B	+29696 B / +1.7% (worse)	526184 B	+20272 B / +4.0% (worse)	4.240 s	+28.14 ms / +0.7% (worse)	13.354 ms	+569.6 us / +4.5% (worse)
Windows MinGW ARM64	fmtprintf-lto	1890304 B	+36352 B / +2.0% (worse)	492308 B	+19284 B / +4.1% (worse)	10.029 s	+299.4 ms / +3.1% (worse)	12.405 ms	-686.8 us / -5.2% (better)
Windows MinGW ARM64	println	68608 B	0 B / +0.0%	22712 B	-168 B / -0.7% (better)	1.414 s	-36.87 ms / -2.5% (better)	11.256 ms	-99.7 us / -0.9% (better)
Windows MinGW ARM64	println-lto	65024 B	0 B / +0.0%	20000 B	-160 B / -0.8% (better)	1.616 s	-34.31 ms / -2.1% (better)	10.586 ms	-310.9 us / -2.9% (better)
Windows MSVC	cprintf	120320 B	0 B / +0.0%	65782 B	0 B / +0.0%	928.965 ms	-170.6 ms / -15.5% (better)	3.382 ms	-570.6 us / -14.4% (better)
Windows MSVC	cprintf-lto	119808 B	0 B / +0.0%	65718 B	0 B / +0.0%	1.128 s	+192.9 ms / +20.6% (worse)	4.031 ms	+663.2 us / +19.7% (worse)
Windows MSVC	fmtprintf	1640448 B	+24576 B / +1.5% (worse)	704198 B	+15664 B / +2.3% (worse)	3.711 s	+31.82 ms / +0.9% (worse)	10.078 ms	+1.154 ms / +12.9% (worse)
Windows MSVC	fmtprintf-lto	1647616 B	+32256 B / +2.0% (worse)	659126 B	+15040 B / +2.3% (worse)	8.889 s	-100.7 ms / -1.1% (better)	10.180 ms	+865.7 us / +9.3% (worse)
Windows MSVC	println	192512 B	-512 B / -0.3% (better)	119046 B	-336 B / -0.3% (better)	922.909 ms	-1.322 ms / -0.1% (better)	6.953 ms	-123.8 us / -1.7% (better)
Windows MSVC	println-lto	189952 B	0 B / +0.0%	116438 B	-240 B / -0.2% (better)	1.108 s	+1.387 ms / +0.1% (worse)	7.045 ms	-212.2 us / -2.9% (better)
Windows MSVC 386	cprintf	9728 B	0 B / +0.0%	3931 B	0 B / +0.0%	963.145 ms	-51.73 ms / -5.1% (better)	6.487 ms	+170.1 us / +2.7% (worse)
Windows MSVC 386	cprintf-lto	9216 B	0 B / +0.0%	3853 B	0 B / +0.0%	988.649 ms	-212 ms / -17.7% (better)	5.425 ms	-840 us / -13.4% (better)
Windows MSVC 386	fmtprintf	1207296 B	+21504 B / +1.8% (worse)	468789 B	+18192 B / +4.0% (worse)	4.357 s	+225.4 ms / +5.5% (worse)	12.014 ms	-1.129 ms / -8.6% (better)
Windows MSVC 386	fmtprintf-lto	1251840 B	+24576 B / +2.0% (worse)	439481 B	+13008 B / +3.1% (worse)	9.498 s	+785.1 ms / +9.0% (worse)	12.556 ms	+903.7 us / +7.8% (worse)
Windows MSVC 386	println	34816 B	0 B / +0.0%	19025 B	-208 B / -1.1% (better)	954.640 ms	-31.86 ms / -3.2% (better)	9.446 ms	-1.072 ms / -10.2% (better)
Windows MSVC 386	println-lto	32768 B	0 B / +0.0%	17191 B	-160 B / -0.9% (better)	1.147 s	-33.82 ms / -2.9% (better)	9.754 ms	-1.196 ms / -10.9% (better)
Windows MSVC ARM64	cprintf	11264 B	0 B / +0.0%	3976 B	0 B / +0.0%	2.062 s	+42.47 ms / +2.1% (worse)	7.060 ms	+569.2 us / +8.8% (worse)
Windows MSVC ARM64	cprintf-lto	10752 B	0 B / +0.0%	3868 B	0 B / +0.0%	2.103 s	+73.34 ms / +3.6% (worse)	7.586 ms	+793.1 us / +11.7% (worse)
Windows MSVC ARM64	fmtprintf	1392128 B	+29184 B / +2.1% (worse)	525380 B	+19920 B / +3.9% (worse)	6.571 s	+208.7 ms / +3.3% (worse)	14.136 ms	+386 us / +2.8% (worse)
Windows MSVC ARM64	fmtprintf-lto	1421824 B	+35328 B / +2.5% (worse)	493028 B	+19296 B / +4.1% (worse)	15.799 s	+268.3 ms / +1.7% (worse)	13.765 ms	-745.2 us / -5.1% (better)
Windows MSVC ARM64	println	41984 B	0 B / +0.0%	22072 B	-144 B / -0.6% (better)	2.011 s	+17.35 ms / +0.9% (worse)	12.093 ms	+409.1 us / +3.5% (worse)
Windows MSVC ARM64	println-lto	39936 B	-512 B / -1.3% (better)	19884 B	-160 B / -0.8% (better)	2.340 s	-6.579 ms / -0.3% (better)	11.794 ms	-545.6 us / -4.4% (better)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs base
Linux	BenchmarkLookupPCRandom	13.370 ns/op	0 ns/op / +0.0%
Linux	BenchmarkMergeCompilerFlags	141.200 ns/op	+1.3 ns/op / +0.9% (worse)
Linux	BenchmarkMergeLinkerFlags	89.460 ns/op	+0.43 ns/op / +0.5% (worse)
Linux	BenchmarkChannelBuffered	52.950 ns/op	+13.97 ns/op / +35.8% (worse)
Linux	BenchmarkChannelHandoff	26300 ns/op	-1187 ns/op / -4.3% (better)
Linux	BenchmarkDefer	50.730 ns/op	+1.31 ns/op / +2.7% (worse)
Linux	BenchmarkDirectCall	1.557 ns/op	-0.005 ns/op / -0.3% (better)
Linux	BenchmarkGlobalRead	1.866 ns/op	-0.002 ns/op / -0.1% (better)
Linux	BenchmarkGlobalWrite	2.480 ns/op	0 ns/op / +0.0%
Linux	BenchmarkGoroutine	31861 ns/op	-568 ns/op / -1.8% (better)
Linux	BenchmarkInterfaceCall	8.404 ns/op	+0.3 ns/op / +3.7% (worse)
Linux	BenchmarkRuntimeGetG	1.867 ns/op	-0.655 ns/op / -26.0% (better)
macOS	BenchmarkLookupPCRandom	16.560 ns/op	+4.13 ns/op / +33.2% (worse)
macOS	BenchmarkMergeCompilerFlags	129.800 ns/op	+6.8 ns/op / +5.5% (worse)
macOS	BenchmarkMergeLinkerFlags	104.100 ns/op	+29.93 ns/op / +40.4% (worse)
macOS	BenchmarkChannelBuffered	57.130 ns/op	+32 ns/op / +127.3% (worse)
macOS	BenchmarkChannelHandoff	8560 ns/op	+58 ns/op / +0.7% (worse)
macOS	BenchmarkDefer	64.650 ns/op	+31.84 ns/op / +97.0% (worse)
macOS	BenchmarkDirectCall	1.398 ns/op	+0.331 ns/op / +31.0% (worse)
macOS	BenchmarkGlobalRead	1.442 ns/op	+0.284 ns/op / +24.5% (worse)
macOS	BenchmarkGlobalWrite	1.777 ns/op	+0.695 ns/op / +64.2% (worse)
macOS	BenchmarkGoroutine	62622 ns/op	+30650 ns/op / +95.9% (worse)
macOS	BenchmarkInterfaceCall	6.733 ns/op	+2.079 ns/op / +44.7% (worse)
macOS	BenchmarkRuntimeGetG	3.297 ns/op	+1.062 ns/op / +47.5% (worse)
Windows MinGW	BenchmarkLookupPCRandom	9.576 ns/op	+0.032 ns/op / +0.3% (worse)
Windows MinGW	BenchmarkMergeCompilerFlags	385.700 ns/op	-2.9 ns/op / -0.7% (better)
Windows MinGW	BenchmarkMergeLinkerFlags	349.700 ns/op	+16.5 ns/op / +5.0% (worse)
Windows MinGW	BenchmarkChannelBuffered	36.840 ns/op	+6.45 ns/op / +21.2% (worse)
Windows MinGW	BenchmarkChannelHandoff	1136 ns/op	+19 ns/op / +1.7% (worse)
Windows MinGW	BenchmarkDefer	46.300 ns/op	+2.18 ns/op / +4.9% (worse)
Windows MinGW	BenchmarkDirectCall	1.357 ns/op	0 ns/op / +0.0%
Windows MinGW	BenchmarkGlobalRead	1.358 ns/op	0 ns/op / +0.0%
Windows MinGW	BenchmarkGlobalWrite	2.165 ns/op	-0.001 ns/op / -0.04617% (better)
Windows MinGW	BenchmarkGoroutine	56820 ns/op	-1116 ns/op / -1.9% (better)
Windows MinGW	BenchmarkInterfaceCall	7.988 ns/op	-0.01 ns/op / -0.1% (better)
Windows MinGW	BenchmarkRuntimeGetG	1.694 ns/op	+0.285 ns/op / +20.2% (worse)
Windows MinGW 386	BenchmarkLookupPCRandom	26.620 ns/op	-0.03 ns/op / -0.1% (better)
Windows MinGW 386	BenchmarkMergeCompilerFlags	774.600 ns/op	+12.4 ns/op / +1.6% (worse)
Windows MinGW 386	BenchmarkMergeLinkerFlags	702.100 ns/op	+15.3 ns/op / +2.2% (worse)
Windows MinGW 386	BenchmarkChannelBuffered	45.530 ns/op	+1.43 ns/op / +3.2% (worse)
Windows MinGW 386	BenchmarkChannelHandoff	965.900 ns/op	-52.1 ns/op / -5.1% (better)
Windows MinGW 386	BenchmarkDefer	43.460 ns/op	-1.04 ns/op / -2.3% (better)
Windows MinGW 386	BenchmarkDirectCall	1.549 ns/op	+0.002 ns/op / +0.1% (worse)
Windows MinGW 386	BenchmarkGlobalRead	1.548 ns/op	-0.002 ns/op / -0.1% (better)
Windows MinGW 386	BenchmarkGlobalWrite	7.779 ns/op	-0.003 ns/op / -0.03855% (better)
Windows MinGW 386	BenchmarkGoroutine	88288 ns/op	+1153 ns/op / +1.3% (worse)
Windows MinGW 386	BenchmarkInterfaceCall	9.630 ns/op	+0.02 ns/op / +0.2% (worse)
Windows MinGW 386	BenchmarkRuntimeGetG	2.486 ns/op	+0.314 ns/op / +14.5% (worse)
Windows MinGW ARM64	BenchmarkLookupPCRandom	12.050 ns/op	-0.05 ns/op / -0.4% (better)
Windows MinGW ARM64	BenchmarkMergeCompilerFlags	569.700 ns/op	-3.6 ns/op / -0.6% (better)
Windows MinGW ARM64	BenchmarkMergeLinkerFlags	532.300 ns/op	-11.9 ns/op / -2.2% (better)
Windows MinGW ARM64	BenchmarkChannelBuffered	44.790 ns/op	-0.99 ns/op / -2.2% (better)
Windows MinGW ARM64	BenchmarkChannelHandoff	2321 ns/op	+164 ns/op / +7.6% (worse)
Windows MinGW ARM64	BenchmarkDefer	54.910 ns/op	+1.23 ns/op / +2.3% (worse)
Windows MinGW ARM64	BenchmarkDirectCall	0.589 ns/op	0 ns/op / +0.0%
Windows MinGW ARM64	BenchmarkGlobalRead	0.884 ns/op	+0.2204 ns/op / +33.2% (worse)
Windows MinGW ARM64	BenchmarkGlobalWrite	0.590 ns/op	+0.0001 ns/op / +0.01696% (worse)
Windows MinGW ARM64	BenchmarkGoroutine	58978 ns/op	+917 ns/op / +1.6% (worse)
Windows MinGW ARM64	BenchmarkInterfaceCall	4.715 ns/op	-0.003 ns/op / -0.1% (better)
Windows MinGW ARM64	BenchmarkRuntimeGetG	1.800 ns/op	-0.004 ns/op / -0.2% (better)
Windows MSVC	BenchmarkLookupPCRandom	13.170 ns/op	+0.26 ns/op / +2.0% (worse)
Windows MSVC	BenchmarkMergeCompilerFlags	633.200 ns/op	+14.7 ns/op / +2.4% (worse)
Windows MSVC	BenchmarkMergeLinkerFlags	563.400 ns/op	+34.7 ns/op / +6.6% (worse)
Windows MSVC	BenchmarkChannelBuffered	47.240 ns/op	+10.98 ns/op / +30.3% (worse)
Windows MSVC	BenchmarkChannelHandoff	1064 ns/op	-55 ns/op / -4.9% (better)
Windows MSVC	BenchmarkDefer	53.580 ns/op	-1.52 ns/op / -2.8% (better)
Windows MSVC	BenchmarkDirectCall	1.548 ns/op	+0.001 ns/op / +0.1% (worse)
Windows MSVC	BenchmarkGlobalRead	1.546 ns/op	-0.315 ns/op / -16.9% (better)
Windows MSVC	BenchmarkGlobalWrite	2.470 ns/op	+0.021 ns/op / +0.9% (worse)
Windows MSVC	BenchmarkGoroutine	83905 ns/op	+543 ns/op / +0.7% (worse)
Windows MSVC	BenchmarkInterfaceCall	9.289 ns/op	-0.318 ns/op / -3.3% (better)
Windows MSVC	BenchmarkRuntimeGetG	2.167 ns/op	-0.005 ns/op / -0.2% (better)
Windows MSVC 386	BenchmarkLookupPCRandom	26.590 ns/op	+0.09 ns/op / +0.3% (worse)
Windows MSVC 386	BenchmarkMergeCompilerFlags	775.800 ns/op	+67 ns/op / +9.5% (worse)
Windows MSVC 386	BenchmarkMergeLinkerFlags	715.900 ns/op	+36.4 ns/op / +5.4% (worse)
Windows MSVC 386	BenchmarkChannelBuffered	44.540 ns/op	+1.79 ns/op / +4.2% (worse)
Windows MSVC 386	BenchmarkChannelHandoff	1004 ns/op	-51 ns/op / -4.8% (better)
Windows MSVC 386	BenchmarkDefer	46.040 ns/op	-2.01 ns/op / -4.2% (better)
Windows MSVC 386	BenchmarkDirectCall	1.547 ns/op	-0.003 ns/op / -0.2% (better)
Windows MSVC 386	BenchmarkGlobalRead	1.550 ns/op	-0.312 ns/op / -16.8% (better)
Windows MSVC 386	BenchmarkGlobalWrite	7.790 ns/op	+0.009 ns/op / +0.1% (worse)
Windows MSVC 386	BenchmarkGoroutine	88210 ns/op	+4744 ns/op / +5.7% (worse)
Windows MSVC 386	BenchmarkInterfaceCall	9.669 ns/op	+0.06 ns/op / +0.6% (worse)
Windows MSVC 386	BenchmarkRuntimeGetG	2.165 ns/op	+0.238 ns/op / +12.4% (worse)
Windows MSVC ARM64	BenchmarkLookupPCRandom	12.060 ns/op	-0.01 ns/op / -0.1% (better)
Windows MSVC ARM64	BenchmarkMergeCompilerFlags	559.300 ns/op	+10.9 ns/op / +2.0% (worse)
Windows MSVC ARM64	BenchmarkMergeLinkerFlags	536 ns/op	+14.4 ns/op / +2.8% (worse)
Windows MSVC ARM64	BenchmarkChannelBuffered	43.660 ns/op	+0.19 ns/op / +0.4% (worse)
Windows MSVC ARM64	BenchmarkChannelHandoff	3120 ns/op	-190 ns/op / -5.7% (better)
Windows MSVC ARM64	BenchmarkDefer	60.550 ns/op	-0.44 ns/op / -0.7% (better)
Windows MSVC ARM64	BenchmarkDirectCall	0.590 ns/op	-0.0001 ns/op / -0.01696% (better)
Windows MSVC ARM64	BenchmarkGlobalRead	0.664 ns/op	+0.0003 ns/op / +0.04521% (worse)
Windows MSVC ARM64	BenchmarkGlobalWrite	3.794 ns/op	+0.031 ns/op / +0.8% (worse)
Windows MSVC ARM64	BenchmarkGoroutine	56768 ns/op	-371 ns/op / -0.6% (better)
Windows MSVC ARM64	BenchmarkInterfaceCall	4.728 ns/op	+0.01 ns/op / +0.2% (worse)
Windows MSVC ARM64	BenchmarkRuntimeGetG	1.769 ns/op	-0.001 ns/op / -0.1% (better)

Timer runtime benchmarks

Platform	Operation and runtime	ns/op	vs base
Linux	AfterFuncZeroDelivery/Go	599.400 ns/op	-6.3 ns/op / -1.0% (better)
Linux	AfterFuncZeroDelivery/LLGo	53579 ns/op	+408 ns/op / +0.8% (worse)
Linux	CreateStop/Go	148.600 ns/op	-1.1 ns/op / -0.7% (better)
Linux	CreateStop/LLGo	1130 ns/op	+261.2 ns/op / +30.1% (worse)
Linux	RearmStopped/Go	54.190 ns/op	-1.12 ns/op / -2.0% (better)
Linux	RearmStopped/LLGo	743.700 ns/op	-343.3 ns/op / -31.6% (better)
Linux	ResetActive/Go	41.540 ns/op	-0.03 ns/op / -0.1% (better)
Linux	ResetActive/LLGo	352.700 ns/op	-25.8 ns/op / -6.8% (better)
Linux	ResetHeap1024/Go	41.410 ns/op	-0.03 ns/op / -0.1% (better)
Linux	ResetHeap1024/LLGo	152.300 ns/op	+7 ns/op / +4.8% (worse)
macOS	AfterFuncZeroDelivery/Go	700.600 ns/op	+184.6 ns/op / +35.8% (worse)
macOS	AfterFuncZeroDelivery/LLGo	106465 ns/op	+22151 ns/op / +26.3% (worse)
macOS	CreateStop/Go	171.800 ns/op	-5.8 ns/op / -3.3% (better)
macOS	CreateStop/LLGo	889.500 ns/op	-43.5 ns/op / -4.7% (better)
macOS	RearmStopped/Go	75.700 ns/op	+7.1 ns/op / +10.3% (worse)
macOS	RearmStopped/LLGo	481.100 ns/op	-124.6 ns/op / -20.6% (better)
macOS	ResetActive/Go	52.910 ns/op	-10.26 ns/op / -16.2% (better)
macOS	ResetActive/LLGo	177.200 ns/op	-0.8 ns/op / -0.4% (better)
macOS	ResetHeap1024/Go	75.040 ns/op	+26.88 ns/op / +55.8% (worse)
macOS	ResetHeap1024/LLGo	139.400 ns/op	+37.3 ns/op / +36.5% (worse)
Windows MinGW	AfterFuncZeroDelivery/Go	372.900 ns/op	-18.3 ns/op / -4.7% (better)
Windows MinGW	AfterFuncZeroDelivery/LLGo	99310 ns/op	-1335 ns/op / -1.3% (better)
Windows MinGW	CreateStop/Go	89.640 ns/op	-1.7 ns/op / -1.9% (better)
Windows MinGW	CreateStop/LLGo	390.900 ns/op	+20.6 ns/op / +5.6% (worse)
Windows MinGW	RearmStopped/Go	24.480 ns/op	+0.02 ns/op / +0.1% (worse)
Windows MinGW	RearmStopped/LLGo	247.700 ns/op	-6.9 ns/op / -2.7% (better)
Windows MinGW	ResetActive/Go	14.770 ns/op	-0.1 ns/op / -0.7% (better)
Windows MinGW	ResetActive/LLGo	126.700 ns/op	-2.3 ns/op / -1.8% (better)
Windows MinGW	ResetHeap1024/Go	14.770 ns/op	-0.12 ns/op / -0.8% (better)
Windows MinGW	ResetHeap1024/LLGo	117.100 ns/op	-0.6 ns/op / -0.5% (better)
Windows MinGW 386	AfterFuncZeroDelivery/Go	960.200 ns/op	-1 ns/op / -0.1% (better)
Windows MinGW 386	AfterFuncZeroDelivery/LLGo	190286 ns/op	+705 ns/op / +0.4% (worse)
Windows MinGW 386	CreateStop/Go	192.200 ns/op	-1 ns/op / -0.5% (better)
Windows MinGW 386	CreateStop/LLGo	2146 ns/op	-30 ns/op / -1.4% (better)
Windows MinGW 386	RearmStopped/Go	63.540 ns/op	+0.13 ns/op / +0.2% (worse)
Windows MinGW 386	RearmStopped/LLGo	382.900 ns/op	+9.3 ns/op / +2.5% (worse)
Windows MinGW 386	ResetActive/Go	39.070 ns/op	+0.01 ns/op / +0.0256% (worse)
Windows MinGW 386	ResetActive/LLGo	946.400 ns/op	-54.6 ns/op / -5.5% (better)
Windows MinGW 386	ResetHeap1024/Go	39.460 ns/op	0 ns/op / +0.0%
Windows MinGW 386	ResetHeap1024/LLGo	195.400 ns/op	-0.4 ns/op / -0.2% (better)
Windows MinGW ARM64	AfterFuncZeroDelivery/Go	663.400 ns/op	-1 ns/op / -0.2% (better)
Windows MinGW ARM64	AfterFuncZeroDelivery/LLGo	126531 ns/op	-8486 ns/op / -6.3% (better)
Windows MinGW ARM64	CreateStop/Go	200.100 ns/op	+0.7 ns/op / +0.4% (worse)
Windows MinGW ARM64	CreateStop/LLGo	412.400 ns/op	+15.2 ns/op / +3.8% (worse)
Windows MinGW ARM64	RearmStopped/Go	70.560 ns/op	0 ns/op / +0.0%
Windows MinGW ARM64	RearmStopped/LLGo	288.600 ns/op	+5.6 ns/op / +2.0% (worse)
Windows MinGW ARM64	ResetActive/Go	31.040 ns/op	-0.04 ns/op / -0.1% (better)
Windows MinGW ARM64	ResetActive/LLGo	129.100 ns/op	-4.6 ns/op / -3.4% (better)
Windows MinGW ARM64	ResetHeap1024/Go	31.150 ns/op	+0.09 ns/op / +0.3% (worse)
Windows MinGW ARM64	ResetHeap1024/LLGo	140.100 ns/op	+1.5 ns/op / +1.1% (worse)
Windows MSVC	AfterFuncZeroDelivery/Go	563.100 ns/op	+6.7 ns/op / +1.2% (worse)
Windows MSVC	AfterFuncZeroDelivery/LLGo	157350 ns/op	+1165 ns/op / +0.7% (worse)
Windows MSVC	CreateStop/Go	117.700 ns/op	+0.9 ns/op / +0.8% (worse)
Windows MSVC	CreateStop/LLGo	460 ns/op	+13.9 ns/op / +3.1% (worse)
Windows MSVC	RearmStopped/Go	31.310 ns/op	-0.58 ns/op / -1.8% (better)
Windows MSVC	RearmStopped/LLGo	299.500 ns/op	-4.3 ns/op / -1.4% (better)
Windows MSVC	ResetActive/Go	20.220 ns/op	+0.06 ns/op / +0.3% (worse)
Windows MSVC	ResetActive/LLGo	169.600 ns/op	+10.4 ns/op / +6.5% (worse)
Windows MSVC	ResetHeap1024/Go	20.460 ns/op	+0.04 ns/op / +0.2% (worse)
Windows MSVC	ResetHeap1024/LLGo	138.900 ns/op	-2.8 ns/op / -2.0% (better)
Windows MSVC 386	AfterFuncZeroDelivery/Go	957.300 ns/op	+9.4 ns/op / +1.0% (worse)
Windows MSVC 386	AfterFuncZeroDelivery/LLGo	180994 ns/op	+9070 ns/op / +5.3% (worse)
Windows MSVC 386	CreateStop/Go	195.400 ns/op	+3.8 ns/op / +2.0% (worse)
Windows MSVC 386	CreateStop/LLGo	1756 ns/op	+137 ns/op / +8.5% (worse)
Windows MSVC 386	RearmStopped/Go	63.610 ns/op	+0.2 ns/op / +0.3% (worse)
Windows MSVC 386	RearmStopped/LLGo	337.300 ns/op	-7.6 ns/op / -2.2% (better)
Windows MSVC 386	ResetActive/Go	38.980 ns/op	-0.08 ns/op / -0.2% (better)
Windows MSVC 386	ResetActive/LLGo	949 ns/op	-21.5 ns/op / -2.2% (better)
Windows MSVC 386	ResetHeap1024/Go	39.490 ns/op	+0.06 ns/op / +0.2% (worse)
Windows MSVC 386	ResetHeap1024/LLGo	179.800 ns/op	+1.2 ns/op / +0.7% (worse)
Windows MSVC ARM64	AfterFuncZeroDelivery/Go	668.400 ns/op	-6.4 ns/op / -0.9% (better)
Windows MSVC ARM64	AfterFuncZeroDelivery/LLGo	145633 ns/op	+3516 ns/op / +2.5% (worse)
Windows MSVC ARM64	CreateStop/Go	199.600 ns/op	-14.6 ns/op / -6.8% (better)
Windows MSVC ARM64	CreateStop/LLGo	407.400 ns/op	-1.7 ns/op / -0.4% (better)
Windows MSVC ARM64	RearmStopped/Go	70.620 ns/op	-0.05 ns/op / -0.1% (better)
Windows MSVC ARM64	RearmStopped/LLGo	295.900 ns/op	+5.7 ns/op / +2.0% (worse)
Windows MSVC ARM64	ResetActive/Go	31 ns/op	-0.11 ns/op / -0.4% (better)
Windows MSVC ARM64	ResetActive/LLGo	136.500 ns/op	-7.1 ns/op / -4.9% (better)
Windows MSVC ARM64	ResetHeap1024/Go	31.080 ns/op	-0.05 ns/op / -0.2% (better)
Windows MSVC ARM64	ResetHeap1024/LLGo	144.500 ns/op	+1 ns/op / +0.7% (worse)

Compared with 309f75c5e5b3 measured in the same runner job.

github-actions · 2026-09-10T10:04:09Z

LLGo WebAssembly build benchmarks

05e962ddf030 | workflow run | long-term charts

WebAssembly output sizes

Profile and compiler	Wasm module	vs base	Generated JS glue	vs base
cprintf/ec32/LLGo	148741 B	+35526 B / +31.4% (worse)	80425 B	+9701 B / +13.7% (worse)
cprintf/ec64/LLGo	155891 B	+37096 B / +31.2% (worse)	84170 B	+10149 B / +13.7% (worse)
cprintf/js/LLGo	146203 B	+79453 B / +119.0% (worse)	80011 B	+11512 B / +16.8% (worse)
cprintf/wasip1/LLGo	149164 B	+75576 B / +102.7% (worse)	0 B	0 B / 0.0%
cprintf/wc32/LLGo	149386 B	+32282 B / +27.6% (worse)	0 B	0 B / 0.0%
ec32/LLGo	148113 B	+35630 B / +31.7% (worse)	80425 B	+9701 B / +13.7% (worse)
ec64/LLGo	155226 B	+37198 B / +31.5% (worse)	84170 B	+10149 B / +13.7% (worse)
fmtprintf/ec32/LLGo	2752823 B	-146485 B / -5.1% (better)	106316 B	+8889 B / +9.1% (worse)
fmtprintf/ec64/LLGo	2912524 B	-174870 B / -5.7% (better)	117510 B	+9074 B / +8.4% (worse)
fmtprintf/js/Go	2526852 B	0 B / +0.0%	0 B	0 B / 0.0%
fmtprintf/js/LLGo	2750155 B	+427717 B / +18.4% (worse)	91271 B	-4091 B / -4.3% (better)
fmtprintf/wasip1/Go	2500019 B	0 B / +0.0%	0 B	0 B / 0.0%
fmtprintf/wasip1/LLGo	2593551 B	+478302 B / +22.6% (worse)	0 B	0 B / 0.0%
fmtprintf/wc32/LLGo	2569011 B	-136082 B / -5.0% (better)	0 B	0 B / 0.0%
js/Go	1895533 B	0 B / +0.0%	0 B	0 B / 0.0%
js/LLGo	145767 B	+79695 B / +120.6% (worse)	80011 B	+11512 B / +16.8% (worse)
wasip1/Go	1909947 B	0 B / +0.0%	0 B	0 B / 0.0%
wasip1/LLGo	148584 B	+76049 B / +104.8% (worse)	0 B	0 B / 0.0%
wc32/LLGo	148803 B	+32375 B / +27.8% (worse)	0 B	0 B / 0.0%

LLGo WebAssembly build measurements

Profile	Build	vs base
cprintf/ec32	5.975 s	+1.148 s / +23.8% (worse)
cprintf/ec64	6.019 s	+1.535 s / +34.2% (worse)
cprintf/js	6.063 s	+2.178 s / +56.1% (worse)
cprintf/wasip1	5.029 s	+2.382 s / +89.9% (worse)
cprintf/wc32	5.035 s	+1.685 s / +50.3% (worse)
ec32	6.153 s	+1.305 s / +26.9% (worse)
ec64	6.235 s	+1.765 s / +39.5% (worse)
fmtprintf/ec32	40.814 s	+3.526 s / +9.5% (worse)
fmtprintf/ec64	36.579 s	-147.7 ms / -0.4% (better)
fmtprintf/js	38.755 s	+8.873 s / +29.7% (worse)
fmtprintf/wasip1	35.496 s	+9.524 s / +36.7% (worse)
fmtprintf/wc32	35.319 s	-3.228 s / -8.4% (better)
js	6.277 s	+2.336 s / +59.3% (worse)
wasip1	5.114 s	+2.368 s / +86.2% (worse)
wc32	4.993 s	+1.643 s / +49.0% (worse)

Compared with 309f75c5e5b3 measured in the same runner job.

cpunion · 2026-09-10T13:46:30Z

R1-R3 are now merged and the profile model has been simplified. This oversized R4 aggregate is superseded by W1 in #246. The branch is retained as a migration source; only re-audited necessary changes will be ported.

cpunion force-pushed the codex/wasm-r4-stdlib-reference-20260905 branch from 9ba1e90 to e79eba5 Compare September 5, 2026 14:47

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from 65b7d19 to 5ae575d Compare September 5, 2026 14:47

cpunion force-pushed the codex/wasm-r4-stdlib-reference-20260905 branch from e79eba5 to ecd2e88 Compare September 5, 2026 15:22

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from 5ae575d to 671e903 Compare September 5, 2026 15:22

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from 671e903 to 622addb Compare September 7, 2026 03:28

cpunion changed the title ~~[WASM R4] Expand stdlib acceptance to formatting, conversion and pipe IO~~ [WASM R4] Consolidate stdlib acceptance and three-workload benchmarks Sep 7, 2026

cpunion changed the base branch from codex/wasm-r4-stdlib-reference-20260905 to codex/wasm-r4-main-base-20260907 September 7, 2026 03:28

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from f3ed85d to 7ccc420 Compare September 7, 2026 06:16

cpunion changed the title ~~[WASM R4] Consolidate stdlib acceptance and three-workload benchmarks~~ [WASM R4] Complete single-worker profiles and full acceptance Sep 7, 2026

cpunion changed the title ~~[WASM R4] Complete single-worker profiles and full acceptance~~ [WASM R4] Audit official profiles and standard-library completeness Sep 7, 2026

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from 60f1d3b to 41bde95 Compare September 7, 2026 14:42

This was referenced Sep 7, 2026

fix(build): fingerprint injected Go source patch bodies #242

Closed

fix(ssa): reserve loop-local stack slots once per call #243

Closed

go/ssa: Go 1.27 genmeth1.go panics while instantiating generic method expressions xgo-dev/llgo#2526

Open

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch 3 times, most recently from 5f659cb to 03abd35 Compare September 8, 2026 15:42

cpunion closed this Sep 8, 2026

cpunion reopened this Sep 8, 2026

cpunion and others added 25 commits September 10, 2026 17:28

fix(wasm): keep continuation storage opaque to GC scans

71618b4

fix(wasm): exclude immutable data from GC roots

dcef0ea

fix(wasm): retain native TLS roots during collection

8626ecd

test(wasm): tolerate conservative finalizer roots

d460c79

fix(wasm): harden the browser host shim

395f3a6

docs(wasm): describe full profile acceptance

2e5ba59

benchmark(wasm): count required host sidecars

fe9620a

fix(wasi): retain the Asyncify allocation base

a8215a1

test(wasm): retire passing GOROOT exclusions

1c3c532

refactor(wasm): remove unreachable libffi backend

0c883b8

test(wasm): execute host-only runtime contracts

b1b1af3

test(cl): cover received and malformed pointer method receivers

91e7715

test(cl): exercise uintptr directives through their generic origin

d57bda3

cl: rename locality directives as internal

28d6b2c

ci(wasm): avoid unrelated PCLN relinks in acceptance

0799953

test(wasm): exercise main Goexit lifecycle from host

c690ef3

test(wasm): execute LLGo runtime/cgo compatibility

61d1c7c

test(goroot): retire stale exclusion assertions

14abb03

fix(wasm): install the host callback bridge once

c909845

test(wasm): cover complex and suspending reflect bridges

cad7836

test(wasm): exclude unsupported runtime/cgo frontend

814cf97

fix(wasm): preserve Emscripten process exit

50a9e08

docs(wasm): specify R4 compatibility boundaries

05e962d

cpunion force-pushed the codex/wasm-r4-stdlib-text-io-20260905 branch from a4afd6c to 05e962d Compare September 10, 2026 09:29

cpunion closed this Sep 10, 2026

Navigation Menu

[WASM R4] Audit official profiles and standard-library completeness - #235

[WASM R4] Audit official profiles and standard-library completeness#235
cpunion wants to merge 217 commits into
mainfrom
codex/wasm-r4-stdlib-text-io-20260905

cpunion commented Sep 5, 2026 •

edited

Loading

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

github-actions Bot commented Sep 10, 2026

Uh oh!

github-actions Bot commented Sep 10, 2026

Uh oh!

cpunion commented Sep 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

cpunion commented Sep 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Scope

What this adds

Toolchain and profile completeness

Runtime and standard-library completeness

Linear GC correctness

Acceptance policy

Current validation

Remaining merge gates

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

cpunion commented Sep 8, 2026

Uh oh!

github-actions Bot commented Sep 10, 2026

LLGo baseline benchmarks

Program measurements

Timer runtime benchmarks

Uh oh!

github-actions Bot commented Sep 10, 2026

LLGo WebAssembly build benchmarks

WebAssembly output sizes

LLGo WebAssembly build measurements

Uh oh!

cpunion commented Sep 10, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

cpunion commented Sep 5, 2026 •

edited

Loading