NLnet Labs — @[email protected]

Parsing time stamps faster with SIMD instructions In software, it is common to represent time as a time-stamp string. It is usually specified by a time format string. Some standards use the format %Y%m%d%H%M%S meaning that we print the year, the month, the day, the hours, the minutes and the seconds (e.g., 20230701205436). It is convenient because it is short, easy to read and if you sort the strings lexicographically, you also sort them chronologically. You can generate time stamps using any programming language. We are interested in the problem of parsing these strings. In practice, this means that we want to convert them to an integer presenting the number of seconds since the Unix epoch. The way you typically solve this problem is to use something like the C function strptime. Can we do better? Yes. We can go much faster. Find out how at: lemire.me/blog/2023/07/01/pa… cc @NLnetLabs

Parsing IP addresses crazily fast Most of us are familiar with IP addresses: they are strings of the form “ddd.ddd.ddd.ddd” where ddd is a decimal number of up to three digits in the range 0 to 255. For example, 127.0.0.1 or 192.168.0.2. Each of the four number is a byte value, and the address is n IPv4 network address that fits in 32 bits. There is a more recent extension (IPv6) that spans 128 bits, but the conventional IPv4 addresses are common. They can be part of URLs (e.g., http://127.0.0.1/home). For our fast URL parsing library, I also wrote the counterpart, where we go from the string to the 32-bit value. However, I did not optimize it particularly well because there are many other challenges in efficient URL parsing. Recently, Jeroen Koekkoek from @NLnetLabs brought to my attention that my friend @pshufb has a recent article on the parsing of IP addresses. As usual, it is brillant. Koekkoek is working on accelerating DNS record parsing in his simdzone project and we expect to parse a lot of IP addresses. Wojciech’s results are slightly depressing, however. He suggests that you can beat a standard approach by a factor of 2 or 2.5 in speed, which is excellent, but at the cost of relatively large tables. After revisiting his work, I produce an optimized function which six times faster than the standard one routine in my tests. My source code is available. Read more at lemire.me/blog/2023/06/08/pa… cc @yagiznizipli