There is a Blackfin anonomoly that requires the RTS instruction to be a 4 cycles from the start of a subroutine?
Is there a similar restriction on exception handlers?
I am trying to write fast exception handlers
e.g. recent post on why link / unlink used in exception handler that does not call a subroutine
e.g. queried why ssync( ) becomes stop interrupts, ssync, restart interrupts inside exception handler
The sort of situation I could imagine causing a problem is an exception handler where the RETX is around "a pipeline" distance from the start of the exception handler
The minimum exception length (except for RETX) would be save P0, R0, set P0, set R0 value, save R0 value, recover R0, P0, RETX which would be close to being shorter than "the one pipeline" distance. If you thropw a ssync in there, then I think it would be way past that limit
Are there any problems on minimum length?