By: Brendan (btrotter.delete@this.gmail.com), August 11, 2019 7:50 pm
Room: Moderated Discussions
Groo (charlie.delete@this.semiaccurate.com) on August 11, 2019 6:47 pm wrote:
> Linus Torvalds (torvalds.delete@this.linux-foundation.org) on August 11, 2019 1:13 pm wrote:
> > So an actually architected rdrand instruction is very much the right thing to do. But you'd wish
> > that the CPU vendors had verified it better. Intel apparently did do a good job. Nobody has shown
> > any patterns in their rdrand implementation, and it's been available for a longish while.
> >
> Ummm....
>
> https://en.wikipedia.org/wiki/RdRand
>
> Too tired to search through all the literature an posts on this one but it
> does appear RDRAND on Intel was not exactly lily white, pun intended.
It's hard to tell what you're talking about; but the only thing that seemed relevant on that wiki page was the note ("In some Ivy Bridge versions, due to a bug, the RdRand instruction causes an Illegal Instruction exception"); and quite frankly anyone silly enough to expect CPUID to provide usable raw results (without a huge amount of "errata filtering") deserves to be shown the error of their ways.
> Linus Torvalds (torvalds.delete@this.linux-foundation.org) on August 11, 2019 1:13 pm wrote:
> > So an actually architected rdrand instruction is very much the right thing to do. But you'd wish
> > that the CPU vendors had verified it better. Intel apparently did do a good job. Nobody has shown
> > any patterns in their rdrand implementation, and it's been available for a longish while.
> >
> Ummm....
>
> https://en.wikipedia.org/wiki/RdRand
>
> Too tired to search through all the literature an posts on this one but it
> does appear RDRAND on Intel was not exactly lily white, pun intended.
It's hard to tell what you're talking about; but the only thing that seemed relevant on that wiki page was the note ("In some Ivy Bridge versions, due to a bug, the RdRand instruction causes an Illegal Instruction exception"); and quite frankly anyone silly enough to expect CPUID to provide usable raw results (without a huge amount of "errata filtering") deserves to be shown the error of their ways.