Trigger runtime benchmarks for released tags

[zhouguangyuan0718]

Depends on xgo-dev/benchmarks#23 landing first.

After the tag-only release job publishes artifacts successfully, this adds a separate notification job that dispatches llgo-tag-released to xgo-dev/benchmarks with:

• the source/repository identity;
• the exact released tag;
• the exact commit;
• the source workflow URL.

It reuses the existing BENCHMARKS_REPOSITORY variable and BENCHMARKS_DISPATCH_TOKEN secret contract. Ordinary main pushes continue to use the existing binary-size notification and do not trigger runtime benchmarks.

Validation: workflow YAML parsed successfully and git diff --check passed.

[github-actions]

LLGo baseline benchmarks

dc57dd01cbf9 | workflow run | long-term charts

Program measurements

Platform	Workload	File size	vs base	Build	vs base	Run	vs base
Linux	cprintf	19288 B	0 B / +0.0%	337.867 ms	-1.967 ms / -0.6% (better)	1.288 ms	-4.609 us / -0.4% (better)
Linux	fmtprintf	1859320 B	0 B / +0.0%	2.670 s	-51.03 ms / -1.9% (better)	3.236 ms	-119.7 us / -3.6% (better)
Linux	println	68776 B	0 B / +0.0%	335.143 ms	-8.449 ms / -2.5% (better)	1.598 ms	-12.92 us / -0.8% (better)
macOS	cprintf	84672 B	0 B / +0.0%	354.930 ms	-12.44 ms / -3.4% (better)	2.729 ms	-94.67 us / -3.4% (better)
macOS	fmtprintf	1892448 B	0 B / +0.0%	2.286 s	-318.4 ms / -12.2% (better)	10.998 ms	-2.038 ms / -15.6% (better)
macOS	println	121360 B	0 B / +0.0%	339.865 ms	-4.13 ms / -1.2% (better)	3.489 ms	+193.6 us / +5.9% (worse)

Core language and compiler benchmarks

Platform	Benchmark	ns/op	vs base
Linux	BenchmarkLookupPCRandom	13.350 ns/op	+0.08 ns/op / +0.6% (worse)
Linux	BenchmarkMergeCompilerFlags	150.400 ns/op	-0.9 ns/op / -0.6% (better)
Linux	BenchmarkMergeLinkerFlags	94.250 ns/op	-0.58 ns/op / -0.6% (better)
Linux	BenchmarkChannelBuffered	34 ns/op	+0.01 ns/op / +0.02942% (worse)
Linux	BenchmarkChannelHandoff	29503 ns/op	+756 ns/op / +2.6% (worse)
Linux	BenchmarkDefer	47.300 ns/op	-0.4 ns/op / -0.8% (better)
Linux	BenchmarkDirectCall	1.566 ns/op	+0.007 ns/op / +0.4% (worse)
Linux	BenchmarkGlobalRead	1.557 ns/op	0 ns/op / +0.0%
Linux	BenchmarkGlobalWrite	2.481 ns/op	0 ns/op / +0.0%
Linux	BenchmarkGoroutine	44768 ns/op	+653 ns/op / +1.5% (worse)
Linux	BenchmarkInterfaceCall	8.144 ns/op	+0.039 ns/op / +0.5% (worse)
Linux	BenchmarkRuntimeGetG	2.182 ns/op	0 ns/op / +0.0%
macOS	BenchmarkLookupPCRandom	11.520 ns/op	-0.46 ns/op / -3.8% (better)
macOS	BenchmarkMergeCompilerFlags	108 ns/op	-13.5 ns/op / -11.1% (better)
macOS	BenchmarkMergeLinkerFlags	68.070 ns/op	-3.24 ns/op / -4.5% (better)
macOS	BenchmarkChannelBuffered	22.160 ns/op	-3.42 ns/op / -13.4% (better)
macOS	BenchmarkChannelHandoff	7040 ns/op	+791 ns/op / +12.7% (worse)
macOS	BenchmarkDefer	29.180 ns/op	-6.43 ns/op / -18.1% (better)
macOS	BenchmarkDirectCall	1.021 ns/op	+0.001 ns/op / +0.1% (worse)
macOS	BenchmarkGlobalRead	1.033 ns/op	-0.113 ns/op / -9.9% (better)
macOS	BenchmarkGlobalWrite	1.017 ns/op	-0.042 ns/op / -4.0% (better)
macOS	BenchmarkGoroutine	31100 ns/op	+3790 ns/op / +13.9% (worse)
macOS	BenchmarkInterfaceCall	4.446 ns/op	-0.436 ns/op / -8.9% (better)
macOS	BenchmarkRuntimeGetG	2.387 ns/op	-0.197 ns/op / -7.6% (better)

Compared with cf91afb98fca measured in the same runner job.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[zhouguangyuan0718]

Superseded by #2366, which sends the same llgo-tag-released event from the shared benchmark notification workflow so both runtime performance and open-source compatibility runs are covered.