Can SSE4.1 Instructions Provide a Vectorized Solution for Faster IPv4 Address Extraction?

Optimal Solution for Extracting IPv4 Address from String

Introduction
The provided code retrieves an IPv4 address from a string. While optimized for certain constraints, there may be faster or alternative methods to consider.

Vectorized Solution
For maximal throughput, a vectorized solution using SSE4.1 instructions is recommended.

Here's the code:

__m128i shuffleTable[65536];    //can be reduced 256x times

UINT32 MyGetIP(const char *str) {
    __m128i input = _mm_lddqu_si128((const __m128i*)str);   //"192.167.1.3"
    ...   // Code omitted for brevity
    return _mm_extract_epi32(prod, 0);
}

Explanation
This solution relies on a precomputed lookup table, shuffleTable, which efficiently rearranges bytes into four 4-byte blocks. Each block represents a part of the IP address. The solution is highly optimized for throughput and achieves impressive speeds of over 300 million addresses processed per second.

Initialization of shuffleTable
The shuffleTable lookup table is generated dynamically. Its purpose is to provide a permutation for rearrangement.

void MyInit() {
    ...   // Code omitted for brevity
}

Testing and Comparison
Testing shows that this vectorized solution is significantly faster than the original code:

Time = 0.406   (1556701184)
Time = 3.133   (1556701184)

Conclusion
This vectorized solution provides a substantial speed improvement compared to the original code. It leverages vectorized instructions and a precomputed lookup table to optimize IPv4 address extraction, resulting in a throughput of over 300 million addresses processed per second.

The above is the detailed content of Can SSE4.1 Instructions Provide a Vectorized Solution for Faster IPv4 Address Extraction?. For more information, please follow other related articles on the PHP Chinese website!