[Libre-soc-dev] LD/ST Elwidth Overrides

lkcl luke.leighton at gmail.com
Wed Sep 1 16:49:38 BST 2021



On September 1, 2021 3:06:22 PM UTC, Lauri Kasanen <cand at gmx.com> wrote:
>On Wed, 01 Sep 2021 14:52:05 +0000
>lkcl <luke.leighton at gmail.com> wrote:

>> for STs which would involve overlapping memory WRITE operations it's
>definitely not fine.
>>
>> question: what should the behaviour be?
>
>I'm against silent clamping. 

this would be changing the ld operation to a lw for elwidth=32, or to a lb for elwidth=8

this is actually a lot of work (a lot of gates) in the decoder, actually changing the operation depending on the 24 bit RM Prefix but only when the operation is detected to be a LD/ST in the first place.

the LDST reversing and the redefinition of LD operations to take part of the immediate as a Shift Register (RC) showed exactly how much Hell that is.

> Leaving it undefined or illegal is fine.

illegal ironically is more gates, because explicit action has to be taken, effectively reinterpreting the illegal combinations as "to indicate that exception shall occur".

l.




More information about the Libre-soc-dev mailing list