[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Cryptography] Does RISC V solve Spectre ?



Shouldn't all of that stuff be in the compilers, where it can be done
ahead of time and without occupying silicon nanoacres, and without
converting watts of electricity into heat during runtime?

Speculative caching is strictly less efficient than a compiler-generated
instruction that says "precache xxxxx to yyyyy now" getting executed
well before needing the data.

Speculative execution is strictly less efficient than an instruction
stream that explicitly says to do both things before the branch and then
reuses one set of registers/information or the other for some different
purpose after the branch.

Instructions that are already in an efficient order, getting executed in
order is strictly more efficient than out of order execution.

Fixing this won't create a performance penalty in the long run; it will
fix one instead.

People wanted to change the fundamental chip architectures without
changing the instruction set, so that the stuff compiled for earlier
chips would continue to work.  So they spent silicon and security on
building very fast but buggy interpreters of earlier or simplified
versions of their instruction sets, instead of using an updated and
restructured instruction set that would require recompiling everything
to get the goodies with their new chip architectures.

But it's never been the best answer for efficiency or speed; only the
one with the lowest barrier to entry for deployment.

Definitely deploy RISC.  Deploy each new generation of RISC with a
clean-slate design that uses explicit instructions to just say exactly
what to do and compilers that know how to tune for that architecture
using those instructions.  Don't do contortions trying to emulate some
earlier instruction set, and this problem goes away.

					Bear

Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
The cryptography mailing list
cryptography AT metzdowd.com
http://www.metzdowd.com/mailman/listinfo/cryptography