Closed TopiLeppanen closed 1 year ago
Can you mention this in CHANGES? Did you run all tests?
Updated CHANGES.
No, didn't run all the tests. Since this fix is so clearly specific to AlmaIF and on the the RTL-side, I only ran the tests from testsuite/systemtest_long/procgen/PlatformIntegrator/AlmaIFIntegrator, which are the only tests using AlmaIF.
There was extra 2 subtracted from on-chip address width when calculating the memory mask to decide whether the memory access is directed to on-chip or to the AXI master. What this caused was that basically only the first quarter of the on-chip memory was actually accessible.
Thankfully this seems to have been fixed by @karihepola in AamuDSP RTL, but didn't seem to get upstreamed back then.