[Libre-soc-dev] qemu cosimulation

[Luke Kenneth Casson Leighton]

Lauri I have ironically tracked down a discrepancy in memory at
address 0x100804 which i believe is directly after one of the last
samples.  perhaps not.

i added memory checking after each LD/ST by recording the address of
each operation, and reading the entire 8 byte row.  that address came
up.

this is not guaranteed to be "the" bug but at least now i have both
regs and memory compares.

i have a suspicion that memory in qemu is partly junk, although this
needs to be confirmed with a dump.

would it be reasonable to add a data segment to memmap of all zeros
covering 0x100000 to 0x500000?

that would be the quickest way to zero out that area: pygdbmi is
dreadfully slow, i think it does not properly look for the command
prompt, instead relying on a timeout.

l.