[Libre-soc-dev] daily kan-ban update 04may2021

Tobias Platen libre-soc at platen-software.de
Tue May 4 19:58:03 BST 2021


On Tue, 2021-05-04 at 19:27 +0100, Luke Kenneth Casson Leighton wrote:
> ---
> crowd-funded eco-conscious hardware: 
> https://www.crowdsupply.com/eoma68
> 
> On Tue, May 4, 2021 at 7:18 PM Tobias Platen
> <libre-soc at platen-software.de> wrote:
> 
> > Yes I have read all emails I have got. First I had a look to
> > mmu/fsm.py
> > line 351 to 366. spr1_o.data is set to spr and a_i is ignored. They
> > both need to be passed to the SPR regfile. spr1_i.data is still
> > unused.
> 
> should be exactly the same as spr main_stage.py
> https://git.libre-soc.org/?p=soc.git;a=blob;f=src/soc/fu/spr/main_stage.py;hb=HEAD
I saw, there are fast SPRs first and then slow SPRs TODO. spr1_o and
spr1_i are used for the slow ones.
> 
> except, of course, also placing a *copy* of the data into the local
> dsisr, dar, etc. etc.
> 
> you'll need to think through all read/write scenarios, to make sure
> the cached copy is up-to-date with the value that ends up in the SPR
> regfile.
> 
> for example:
> 
> * LD/ST exception occurs, DAR/DSISR is overwritten *in the cached
> copies only* (LoadStore1)
> * MFSPR instruction is issued, value is read *from the SPR regfile*
> not the cached copy in LoadStore1.
> 
> this would be a severe problem.
> 
> l.
> 
> _______________________________________________
> Libre-soc-dev mailing list
> Libre-soc-dev at lists.libre-soc.org
> http://lists.libre-soc.org/mailman/listinfo/libre-soc-dev




More information about the Libre-soc-dev mailing list