[libre-riscv-dev] Paper: Scoreboards can deadlock

[Samuel Falvo II]

On Fri, Jun 7, 2019 at 10:22 PM Luke Kenneth Casson Leighton
<lkcl at lkcl.net> wrote:
> Did that make any sense, Samuel? Thornton notes that although the logic is

I haven't read the paper in detail; I merely thought it would be of
interest to others here.  I did skim it, though, and it looked like
the author was going off of Thornton's original description of the
6600.  As I was skimming it, it looked to me like everything boiled
down to the left-over state in some of the instruction fields in the
FU.  It seemed clear to me that, if you don't re-initialize the valid
bits on issue, of course you run into the possibility of deadlock.
This would be a case of "don't do that."  To me, it felt like the
authors of the paper were taking Thornton's book for a Chilton's-style
technical teardown of the 6600, when it was really much more generic
than that.

Still, I thought there might be some nugget of useful information in
it especially with respect to their use of formal verification.  That
part I haven't studied yet.

-- 
Samuel A. Falvo II