More on real time on PDP11, Assembler vs C

[G.TOMASEVICH]

In my previous article noting that I ran DR11-W's at 8 kHz, that was
the rate of transferring 16-bit words.  It happens that the output
included mu-law speech in one byte (this is telephony work), and
the input was processed speech and other data.
The only assembly code I used was for the C program interfaces to the
three additional system calls, and these were modeled after 'ustat'.
The tight loop that tested for DR11-W ready, then restarted it,
took about 60 microsec on the average.  It used register variables
and included a counter to bail out in case the DR11-W went dead.  I watched
the DR11-W ready bits with a storage 'scope to make sure it worked.  The 'scope
was the most useful software debugging tool.
	George Tomasevich, AT&T Bell Laboratories