I am moving this on to devel to speed the process up but I see one thing and something I need to think about.
(1) init.c has DOS (Windows CR/LF) and needs to be UNIX style. I ran dos2unix on it. (2) The time seems high. Running on psim (PowerPC simulator) which reports times in instructions, I see this: joel@rtbf64c b-psim]$ grep e_receive log/* log/tm09:rtems_message_queue_receive: available - 150 log/tm10:rtems_message_queue_receive: not available NO_WAIT - 88 log/tm10:rtems_message_queue_receive: not available caller blocks - 801 log/tmoverhd:overhead: rtems_message_queue_receive - 5 [joel@rtbf64c b-psim]$ grep q_receive log/* log/psxtmmq01:mq_receive: available - 164 log/psxtmmqrcvblock01:mq_receive: not available: block - 1156 Notice that tm10's blocking case is 801 instructions vs 1156. So there are 355 more instructions on the posix case. Looking at that more closely, I think I believe it is accurate that posix mq's are more expensive than Classic API message queues. I will post the patch with a couple of clean ups for further review. --joel On Thu, Nov 15, 2018 at 1:54 PM Shashvat Jain <shashvatjain2...@gmail.com> wrote: > > > ---------- Forwarded message --------- > From: Shashvat Jain <shashvatjain2...@gmail.com> > Date: Fri, Nov 16, 2018 at 12:32 AM > Subject: [POSIX Benchmark Test] psxtmmqrcvblock01 made (GCI 2018) > To: <devel@rtems.org> > > > hello , > here is the patch which includes the psxtmmqrcvblock01 test for test case > "mq_receive : not available : block" > please verify the output and the code. > > Thank you > > --Shashvat >
_______________________________________________ devel mailing list devel@rtems.org http://lists.rtems.org/mailman/listinfo/devel
