I appreciate your response, Damien. I did do the bare minimum of correctness testing and it was the post right after Guenther was congratulated on proving incorrectness:
https://marc.info/?l=openbsd-tech&m=165259528425835&w=2 The post includes software to reproduce the results. I wrote a highly dynamically allocated program to test at intervals of intervals to show at various stages to show the degree that the output remains random This is an example of some output: testing arc4random_uniform(5) and arc4random_uniform_small_unlocked(5): 256X 2048X 16384X 131072X 1048576X .................................................................... 0 original 246 2053 16400 131115 1048306 mine 263 2042 16312 131258 1046989 1 original 234 2013 16337 130894 1047810 mine 249 2022 16304 131161 1049511 2 original 236 2061 16367 130802 1047430 mine 257 2117 16597 130718 1046375 3 original 284 2089 16444 131092 1050332 mine 266 2058 16379 131190 1049877 4 original 280 2024 16372 131457 1049002 mine 245 2001 16328 131033 1050128 max-min values: original 50 76 107 655 2902 mine 21 116 293 540 3753 The program is set up to test values 2 through 50. This will create 224KiB of output. I suspected that you'd prefer that I didn't attach it. Some progress output has been relegated to stderr so that you can easily pipe it to a file. I added some getrlimit an setrlimit functions to maximize memory limits if necessary In the case that I created for you, this will not be needed. It uses 1276K RAM in the current configuration. > Just to bring this back to where we came in: even putting thread-safety > aside (which you absolutely can't): it doesn't matter how much faster > it is, your replacement function isn't useful until you do the work to > demonstrate correctness. > > You have done literally zero so far, not even the bare minimum of > testing the output. As a result your first version was demonstrated to > be completely broken by literally the most basic of possible tests, a > mere 10 lines of code. > > That you left this to others to do tells me that you fundamentally don't > understand the environment in which you're trying to operate, and that > you don't respect the time of the people you're trying to convince. > > Please stop wasting our time. > > -d -- -Luke
arc4random_uniform_small.c
Description: Binary data
