boards: snps: nsim: align ICCM/DCCM configuration #81417
Conversation
|
Wasn't the idea that these nSIM boards matched some of our hardware development boards/emulation environments, so a single binary could be used in both simulation and hardware environment? |
If so, I'm afraid this correspondence was already lost before my times, except maybe for some boards like HS5/HS6 which I haven't touched in this PR. Still, I think I could match all the others back again if I had some singular source of truth about the hardware configurations we want to simulate. Otherwise I believe it's better to have internally consistent configurations, even if unrealistic. |
tagunil
force-pushed
the
arc-nsim-ccm-fixes
branch
from
9575b77 to
d1cfcce
Compare
|
Hi @ruuddw and @evgeniy-paltsev, let's decide how to proceed. I've just reverted VPX5 back to 256K as requested. Do you have any other configurations you have hard data about, or we could have this quick fix and analyse the configurations later on a case-by-case basis? |
tagunil
force-pushed
the
arc-nsim-ccm-fixes
branch
from
d1cfcce to
321821e
Compare
Fix ICCM/DCCM properties in DTS to match the intended configuration, along with some other minor inconsistencies. Signed-off-by: Ilya Tagunov <[email protected]>
As these boards are completely simulated and do not represent any real hardware configurations anymore, provide them with 1M of both ICCM and DCCM to avoid any out-of-memory test failures. Signed-off-by: Ilya Tagunov <[email protected]>
Fix ICCM/DCCM properties both in DTS and in MDB settings to match the original hardware configuration. Signed-off-by: Ilya Tagunov <[email protected]>
tagunil
force-pushed
the
arc-nsim-ccm-fixes
branch
from
321821e to
ce0987f
Compare
Some of the boards are completely simulated and do not represent any real hardware configurations. Others are diverged from the original intentions. Treat them on a case-by-case basis, separately for EM, SEM and HS.
Original description: