The change is straightforward — it removes the slow fallback branches guarded by the ENABLE_FAST_TEXTENCODER autogate in both encode() and encodeInto(), making the fast simdutf-based implementation the only code path. The #include <workerd/util/autogate.h> is also correctly removed since it's no longer needed.

One issue I notice: after this PR, the ENABLE_FAST_TEXTENCODER autogate key definition in autogate.h and autogate.c++ becomes dead code — it has no remaining consumers. This should be cleaned up in this PR or a follow-up.

The PR description says "Blocked until fast-text-encoder autogate is rolled out", which is the correct approach — the autogate must be fully rolled out before removing the fallback.

LGTM

github run

Merging this PR will improve performance by ×8.7

⚡ 16 improved benchmarks
✅ 54 untouched benchmarks
⏩ 129 skipped benchmarks¹

Performance Changes

_{Comparing yagiz/remove-slow-text-encoder (82f0c46) with main (755795d)}

@anonrig ... what's the status on this one?

@anonrig ... what's the status on this one?

Autogate not yet released.

Are you planning on releasing the autogate soon? Is there a reason to keep waiting?