Claire Xen  🏳️‍⚧️ 🏳️‍🌈 🧙🏻‍♀️ BLM  🏴 🚩

I'm still not over this. In addition to UB the Verilog $random C code also relies on some IDB. 1/https://twitter.com/oe1cxw/status/870972176421330945 …

Namely it assumes that "long" is 32 bits wide and that "float" is using the IEEE binary32 format.

The latter is ridiculous considering that this is a function that has only integers as in- and outputs! (old seed, new seed, rand val)

But internally it first creates a uniformly distributed float random value in range (INT_MIN INT_MAX) and then converts that one to an int.

But (of course!) that random float value is in turn generated from the "true" rng value, which is the new 32 bit seed.

I completely fail to understand how anyone could ever under any circumstances have thought that any of that is a good idea.