Skip to content

improve text encoder encode performance #5448

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 7 commits into
base: main
Choose a base branch
from
+210 −26
Open

improve text encoder encode performance #5448

Changes from 1 commit
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
80befbe
experiment with value view and simdutf
anonrig Oct 31, 2025
a118760
address pr reviews
anonrig Oct 31, 2025
f1bbfe6
address pr reviews
anonrig Oct 31, 2025
6e3972e
get rid of multiple valueviews
anonrig Nov 3, 2025
60c8847
apply optimization to improve invalid utf16
anonrig Nov 3, 2025
7437830
add missing simdutf dependency
anonrig Nov 3, 2025
cbe01b6
apply review recommendations
anonrig Nov 4, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Next Next commit
experiment with value view and simdutf
  • Loading branch information
@anonrig
anonrig committed Nov 3, 2025
commit 80befbec63f7d3cedd334065372f521f53d46a86
63 changes: 46 additions & 17 deletions src/workerd/api/encoding.c++
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,6 +4,7 @@

#include "encoding.h"

#include "simdutf.h"
#include "util.h"

#include <workerd/jsg/jsg.h>
Expand Down Expand Up @@ -464,25 +465,53 @@ jsg::Ref<TextEncoder> TextEncoder::constructor(jsg::Lock& js) {
return js.alloc<TextEncoder>();
}

namespace {
TextEncoder::EncodeIntoResult encodeIntoImpl(
jsg::Lock& js, jsg::JsString input, jsg::BufferSource& buffer) {
auto result = input.writeInto(
js, buffer.asArrayPtr().asChars(), jsg::JsString::WriteFlags::REPLACE_INVALID_UTF8);
return TextEncoder::EncodeIntoResult{
.read = static_cast<int>(result.read),
.written = static_cast<int>(result.written),
};
}
} // namespace

jsg::BufferSource TextEncoder::encode(jsg::Lock& js, jsg::Optional<jsg::JsString> input) {
auto str = input.orDefault(js.str());
auto view = JSG_REQUIRE_NONNULL(jsg::BufferSource::tryAlloc(js, str.utf8Length(js)), RangeError,
"Cannot allocate space for TextEncoder.encode");
[[maybe_unused]] auto result = encodeIntoImpl(js, str, view);
KJ_DASSERT(result.written == view.size());
return kj::mv(view);

// Do the conversion while ValueView is alive, but to a C++ heap buffer (not V8 heap)
kj::Array<kj::byte> output_data;

{
v8::String::ValueView value_view(js.v8Isolate, str);
size_t length = static_cast<size_t>(value_view.length());

if (value_view.is_one_byte()) {
auto data = reinterpret_cast<const char*>(value_view.data8());
size_t utf8_length = simdutf::utf8_length_from_latin1(data, length);
output_data = kj::heapArray<kj::byte>(utf8_length);
[[maybe_unused]] auto written =
simdutf::convert_latin1_to_utf8(data, length, output_data.asChars().begin());
KJ_DASSERT(written == output_data.size());
} else {
auto data = reinterpret_cast<const char16_t*>(value_view.data16());

// Check if UTF-16LE is valid
auto validation_result = simdutf::validate_utf16le(data, length);

if (validation_result) {
// Valid UTF-16LE, convert directly
size_t utf8_length = simdutf::utf8_length_from_utf16le(data, length);
output_data = kj::heapArray<kj::byte>(utf8_length);
[[maybe_unused]] auto written =
simdutf::convert_utf16le_to_utf8(data, length, output_data.asChars().begin());
KJ_DASSERT(written == output_data.size());
} else {
// Invalid UTF-16LE (unpaired surrogates), fix it first
auto well_formed = kj::heapArray<char16_t>(length);
simdutf::to_well_formed_utf16le(data, length, well_formed.begin());

// Now convert the well-formed UTF-16LE to UTF-8
size_t utf8_length = simdutf::utf8_length_from_utf16le(well_formed.begin(), length);
output_data = kj::heapArray<kj::byte>(utf8_length);
[[maybe_unused]] auto written = simdutf::convert_utf16le_to_utf8(
well_formed.begin(), length, output_data.asChars().begin());
KJ_DASSERT(written == output_data.size());
}
}
} // ValueView destroyed here, releasing the heap lock

// Now create BufferSource from the output data (this allocates V8 objects, which is now safe)
return jsg::BufferSource(js, jsg::BackingStore::from(js, kj::mv(output_data)));
}

TextEncoder::EncodeIntoResult TextEncoder::encodeInto(
Expand Down