Testing aplc compiled APL programs on Epiphany cores.
Posted: Thu Jun 13, 2013 12:22 amSorry for the long gap since the last post -
it turned out we had some gotchas to overcome before we could compile a generic APL program for the Epiphany.
They both arose from the very crude "Epiphany writes to buffer - wait - host reads buffer" replacement for std-out, that I adopted from the example "hello_world.c", and not from the apl to c translation process, which has worked fine.
Firstly, the placing of the buffer was such that when a multi-line aplc program cleaned up its temporary workings in the heap, it erased the buffer.
So the heap was slightly shrunk, and the buffer placed just above it in external dram_1.
Secondly, the above "wait" before the host read the data was not long enough : simple test statements output correctly, but multi-line aplc output was getting chopped off. Just increasing the delay fixed that - thanks particularly to yspair for tracking down that, and for much other help.
See the next postings for examples of simple APL programs running.
it turned out we had some gotchas to overcome before we could compile a generic APL program for the Epiphany.
They both arose from the very crude "Epiphany writes to buffer - wait - host reads buffer" replacement for std-out, that I adopted from the example "hello_world.c", and not from the apl to c translation process, which has worked fine.
Firstly, the placing of the buffer was such that when a multi-line aplc program cleaned up its temporary workings in the heap, it erased the buffer.
So the heap was slightly shrunk, and the buffer placed just above it in external dram_1.
Secondly, the above "wait" before the host read the data was not long enough : simple test statements output correctly, but multi-line aplc output was getting chopped off. Just increasing the delay fixed that - thanks particularly to yspair for tracking down that, and for much other help.
See the next postings for examples of simple APL programs running.