内存型SQLite3数据库能否比C++数据结构性能更优？

我有一个管理结构体数组的场景，该数组类似SQL表，需要基于结构体的各类属性执行类SQL查询。我希望使用内存型SQLite3数据库替代C++结构体数组，借助其内置的查询、索引特性减少代码量并提升性能，无需自行实现遍历与索引逻辑。

但初始测试显示，仅插入操作，内存型SQLite3就比C++数据结构慢12-30倍。我尝试过Stack Overflow上的部分优化方案，但效果有限。

我的问题是：内存型SQLite3数据库能否在性能上超过C++数据结构？

注：由于插入与查询操作实时交错，我无法采用批量插入、先插入后建索引等优化方式，仅希望SQLite的算法优化能在查询性能上超越自行实现的C++数据结构（例如基于prop3和prop5查询时，无需自行实现索引）。

#include <iostream>
#include <map>
#include <cstdlib>
#include <ctime>

struct item_t {
    unsigned int a;
    unsigned int b;
    bool prop1;
    bool prop2;
    bool prop3;
    bool prop4;
    bool prop5;
};

typedef std::map<unsigned int, item_t> items_t;

int main() {
    srand(RAND_SEED);
    items_t data;
    unsigned int delta = 0x100000000 / ITERATIONS, randInt;
    unsigned int count;

    // 创建ITERATIONS个带随机属性的条目
    for (unsigned int i = 0, a = 0; i < ITERATIONS; ++i, a += delta) {
        randInt = rand();
        data[a] = {
            a, a + delta - 1,
            bool(randInt & 16),
            bool(randInt &  8),
            bool(randInt &  4),
            bool(randInt &  2),
            bool(randInt &  1)
        };
    }

    printf("Inserted %d entries\n", data.size());
    return 0;
}

#include <stdio.h>
#include <ctime>
#include <cstdlib>
#include <sqlite3.h>
#include <cassert>
#include <string.h>


static int sqliteExecCallback(void *unused, int count, char **data, char **columns)
{
    for (int idx = 0; idx < count; idx++) {
        printf("%s\n", data[idx]);
    }

    return 0;
}

int main() {
    srand(RAND_SEED);
    unsigned int delta = 0x100000000 / ITERATIONS, randInt;

    sqlite3 *db;
    char *zErrMsg = 0;
    int rc;

    rc = sqlite3_open(":memory:", &db);

    if (rc) {
        fprintf(stderr, "无法打开数据库: %s\n", sqlite3_errmsg(db));
        return(0);
    }
    rc = sqlite3_exec(
        db,
        "PRAGMA synchronous = OFF;"
        "PRAGMA journal_mode = MEMORY",
        NULL, NULL, &zErrMsg
    );
    if (rc) {
        fprintf(stderr, "无法设置PRAGMA参数: %s\n", sqlite3_errmsg(db));
        return(0);
    }

    rc = sqlite3_exec(
        db,
        "CREATE TABLE IF NOT EXISTS DATA("
        "a INTEGER PRIMARY KEY,"
        "b INTEGER,"
        "prop1 INTEGER,"
        "prop2 INTEGER,"
        "prop3 INTEGER,"
        "prop4 INTEGER,"
        "prop5 INTEGER"
        ");"
        "CREATE INDEX DATA_prop1_idx on DATA(prop1);"
        "CREATE INDEX DATA_prop2_idx on DATA(prop2);"
        "CREATE INDEX DATA_prop3_idx on DATA(prop3);"
        "CREATE INDEX DATA_prop4_idx on DATA(prop4);"
        "CREATE INDEX DATA_prop5_idx on DATA(prop5);",
        0, 0, &zErrMsg
    );

    if (rc) {
        fprintf(stderr, "无法执行建表语句: %s\n", sqlite3_errmsg(db));
        return(0);
    }

    char sql[512] = "INSERT INTO DATA (a, b, prop1, prop2, prop3, prop4, prop5)";
    int len = strlen(sql);
    for (unsigned int i = 0, a = 0; i < ITERATIONS; ++i, a += delta) {
        randInt = rand();
        sprintf(
            sql + len,
            " VALUES (%u, %u, %d, %d, %d, %d, %d);",
            a, a + delta - 1,
            bool(randInt & 16),
            bool(randInt &  8),
            bool(randInt &  4),
            bool(randInt &  2),
            bool(randInt &  1)
        );
        rc = sqlite3_exec(db, sql, 0, 0, &zErrMsg);
    }

    rc = sqlite3_exec(db, "SELECT COUNT(*) FROM DATA", sqliteExecCallback, 0, &zErrMsg);

    if (false && rc) {
        fprintf(stderr, "无法执行COUNT查询: %s\n", sqlite3_errmsg(db));
    }
    return 0;
}

g++ -std=c++20 -O3 -march=native -DITERATIONS=2000000 -DRAND_SEED=1 sqlite.cpp -l sqlite3 > comp.log 2>&1
/bin/time -f "sqlite,%S,%U,%e,%M,%X" --append -o metrics.csv ./a.out > run.log 2>&1

g++ -std=c++20 -O3 -march=native -DITERATIONS=2000000 -DRAND_SEED=1 cpp.cpp -l sqlite3 > comp.log 2>&1
/bin/time -f "cpp,%S,%U,%e,%M,%X" --append -o metrics.csv ./a.out > run.log 2>&1

注：由于请求来自外部，我无法批量执行SQL查询，但接受其他优化方案。

更新：使用sqlite3_prepare_v2后，性能提升了8倍，运行时间降至3-4秒。

内容的提问来源于stack exchange，提问作者Krishna