Files
scylladb/tests/utf8_test.cc
Yibo Cai (Arm Technology China) 6fadba56cc utils: optimize UTF-8 validation
UTF-8 string is now validated by boost::locale::conv::utf_to_utf, it
actually does string conversions which is more than necessary.  As
observed on Arm server, UTF-8 validation can become bottleneck under
heavy loads.

This patch introduces a brand new SIMD implementation supporting both
NEON and SSE, as well as a naive approach to handle short strings.
The naive approach is 3x faster than boost utf_to_utf, whilst SIMD
method outperforms naive approach 3x ~ 5x on Arm and x86. Details at
https://github.com/cyb70289/utf8/.

UTF-8 unit test is added to check various corner cases.

Signed-off-by: Yibo Cai <yibo.cai@arm.com>
Message-Id: <1543978498-12123-1-git-send-email-yibo.cai@arm.com>
2018-12-05 21:51:01 +02:00

166 lines
5.2 KiB
C++

/*
* Copyright (c) 2018, Arm Limited and affiliates. All rights reserved.
*/
/*
* This file is part of Scylla.
*
* Scylla is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* Scylla is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with Scylla. If not, see <http://www.gnu.org/licenses/>.
*/
#define BOOST_TEST_MODULE core
#include <cstdint>
#include <vector>
#include <boost/test/unit_test.hpp>
#include "utils/utf8.hh"
struct test_str {
const void *data;
size_t len;
};
// Positive strings
static const std::vector<test_str> positive = {
{"", 0},
{"\x00", 1},
{"\x66", 1},
{"\x7F", 1},
{"\x00\x7F", 2},
{"\x7F\x00", 2},
{"\xC2\x80", 2},
{"\xDF\xBF", 2},
{"\xE0\xA0\x80", 3},
{"\xE0\xA0\xBF", 3},
{"\xED\x9F\x80", 3},
{"\xEF\x80\xBF", 3},
{"\xF0\x90\xBF\x80", 4},
{"\xF2\x81\xBE\x99", 4},
{"\xF4\x8F\x88\xAA", 4},
};
// Negative strings
static const std::vector<test_str> negative = {
{"\x80", 1},
{"\xBF", 1},
{"\xC0\x80", 2},
{"\xC1\x00", 2},
{"\xC2\x7F", 2},
{"\xDF\xC0", 2},
{"\xE0\x9F\x80", 3},
{"\xE0\xC2\x80", 3},
{"\xED\xA0\x80", 3},
{"\xED\x7F\x80", 3},
{"\xEF\x80\x00", 3},
{"\xF0\x8F\x80\x80", 4},
{"\xF0\xEE\x80\x80", 4},
{"\xF2\x90\x91\x7F", 4},
{"\xF4\x90\x88\xAA", 4},
{"\xF4\x00\xBF\xBF", 4},
{"\x00\x00\x00\x00\x00\xC2\x80\x00\x00\x00\xE1\x80\x80\x00\x00\xC2" \
"\xC2\x80\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00", 32},
{"\x00\x00\x00\x00\x00\xC2\xC2\x80\x00\x00\xE1\x80\x80\x00\x00\x00", 16},
{"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" \
"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xF1\x80", 32},
{"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" \
"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xF1", 32},
{"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" \
"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xF1\x80" \
"\x80", 33},
{"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" \
"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xF1\x80" \
"\xC2\x80", 34},
{"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" \
"\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\xF0" \
"\x80\x80\x80", 35},
};
// Round concatenate positive test strings to 1024 bytes
static void prepare_test_buf(uint8_t *buf, size_t pos_idx) {
int buf_idx = 0;
while (buf_idx < 1024) {
size_t buf_len = 1024 - buf_idx;
if (buf_len >= positive[pos_idx].len) {
memcpy(buf+buf_idx, positive[pos_idx].data, positive[pos_idx].len);
buf_idx += positive[pos_idx].len;
} else {
// Fill remaining buffer with 0
memset(buf+buf_idx, 0, buf_len);
buf_idx += buf_len;
}
if (++pos_idx == positive.size()) {
pos_idx = 0;
}
}
}
BOOST_AUTO_TEST_CASE(test_utf8_positive) {
// Test single positive string
for (auto &test : positive) {
BOOST_CHECK(utils::utf8::validate((const uint8_t*)test.data, test.len));
}
const int max_size = 1024 + 32;
uint64_t buf64[max_size/8 + 2];
// Unalign buffer address: offset 8 bytes boundary by 1 byte
uint8_t *buf = (reinterpret_cast<uint8_t*>(buf64)) + 1;
// Test concatenated and shifted positive strings to cover 1k length
for (size_t i = 0; i < positive.size(); ++i) {
// Round concatenate strings staring from i-th positive string
size_t buf_len = 1024;
prepare_test_buf(buf, i);
// Shift 16 bytes, validate each shift
for (int j = 0; j < 16; ++j) {
BOOST_CHECK(utils::utf8::validate(buf, buf_len));
for (int k = buf_len; k >= 1; --k)
buf[k] = buf[k-1];
buf[0] = '\x55';
++buf_len;
}
}
}
BOOST_AUTO_TEST_CASE(test_utf8_negative) {
// Test single negative string
for (auto &test : negative) {
BOOST_CHECK(!utils::utf8::validate((const uint8_t*)test.data, test.len));
}
// Must be larger than 1024 + 16 + max(negative string length)
uint8_t buf[1024*2];
for (size_t i = 0; i < negative.size(); ++i) {
prepare_test_buf(buf, i % positive.size());
// Append one error string
memcpy(buf+1024, negative[i].data, negative[i].len);
size_t buf_len = 1024 + negative[i].len;
// Shift 16 bytes, validate each shift
for (int j = 0; j < 16; ++j) {
BOOST_CHECK(!utils::utf8::validate(buf, buf_len));
for (int k = buf_len; k >= 1; --k)
buf[k] = buf[k-1];
buf[0] = '\x66';
++buf_len;
}
}
}