Rodrigo Branco

How about a take on TSX?

TSX has more promising usages than MPX, but to me is another failure due to "whoever engineers it do not understand the actual SW stack and is an arrogant prick trying to get career advantages"

MPX has harmed a lot because at first it created perception that the problem is solved and then it created perception that no such HW solution can be adopted. Combined, this probably delayed memory tagging by a decade.

I agree. MPX promised too much, delivered too little and damaged the real path forward which is memory tagging. It also had ridiculous ideas, like be a NOP in uarchs that dont support it to avoid recompilation, but in that case, one would have no security checks?

It was fun to use MPX for academic gimmicks but if the performance cost for the intended use is higher than a software only approach then something is wrong

If the intended use is fuzzing, then ASan is indeed somewhat better. But if you try to get the same security guarantees as MPX provided, you would likely have worse performance.