[Arm-netbook] bunnie about riscv

[Luke Kenneth Casson Leighton]

On Mon, Jun 12, 2017 at 1:02 AM, Neil Jansen <njansen1 at gmail.com> wrote:

> results.  bunnie was playing devils advocate by saying even if you did
> this, there are still things that can be present but in an obfuscated
> manner, that could be malicious or careless.

 there was a research paper a few years back which outlined that it
would only take about 3,000 gates to compromise a processor (easily
enough to implement a full RISC CPU).

 that's about a million times less than what is in the current intel processors.

 the point of the exercise was to illustrate how pointless it is do
perform reverse-engineering of modern CPUs, given that the review
process would be insanely complex and would almost certainly miss such
obfuscated / hidden backdoors.

 this is why both the chinese and the russian governments now design
and make their own CPUs.  in the case of china that's FROM SCRATCH.
and using only trusted foundries.

l.