You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
LMM users can start cog threads, and should be able to do the same in XMM
modes. This is desirable because XMM is supposed to allow LMM programs to
expand beyond 32KB memory.
Original issue reported on code.google.com by [email protected] on 12 Jan 2013 at 5:24
The text was updated successfully, but these errors were encountered:
I think we need to be careful of how we resolve this issue. I think it would be
a bad idea to just grab 2K bytes of hub memory in every XMM program just in
case someone wants to call _start_cog_thread. What I'll probably do is make the
COG image available only if that function is actually used in a program. Then
you only lose 2K of hub memory if you are actually starting other C threads. Of
course, Steve's comment about HUBTEXT will still apply. Any C code that you
start using _start_cog_thread will have to be in hub memory and also any
functions that it calls. There is currently no way to share the external memory
interface between COGs. That could be added as another issue I guess but I
think the cache thrashing we'd see if we enabled that feature might make it
less than useful.
Original issue reported on code.google.com by
[email protected]
on 12 Jan 2013 at 5:24The text was updated successfully, but these errors were encountered: