May 19, 2021 Systems meeting.

Updates to the shmem code to allow returning a handle, being able to close shared memory ...

See merge request cds/advligorts!223

 * Updated cps_xmit/recv, cds_pub_sub, local_dc to work with posix shmem or mbuf
 * Add tests that use posix shmem for daqd, cps_recv/xmit, cds_pub_sub, local_dc

This also allows the use of posix shared memory instead of mbuf if the proper name is passed in.

lib/createEpicsMakefile.pm.

Double DAQ_DCU_SIZE, and eliminate time 2 operations on it throughout the code.

See merge request cds/advligorts!222

of subroutines from very large feCodeGen.pl file to their own
perl module files:
- Kernel Makefile generation -> lib/createKernelMakefile.pm
- Part connections file generation -> lib/createDiagsFile.pm

Userspace alternative to gpstime kernel module

See merge request cds/advligorts!218

awgtpman: changed to ignore type of first adc.  Get IOP rate from IOMem mbuf.

See merge request cds/advligorts!221

Got rid of some uses of DAQ_NUM_SWING_BUFFERS since it seemed they weren't used as swing buffers.  It looks like swing buffering may not be used for DAQ data.

Updates to the gpsclock merge request

See merge request christopher.wipf/advligorts!1

Try to keep things to one build of the code, most/all of this can be done at run time.

 * Remove the GPSCLOCK cmake config option
 * Revert edcu.c and sdf_monitor.c.  These are not used anymore and should just be dropped (for another commit).
 * Removed dead symmetricom/gps code from standalone_edcu.cc.
 * Updated the gps logic used by standalone_edcu.cc (gps.hh) to use gpsclock instead of the system time if /dev/gpstime is not available.
 * Minor refactor of the controllerAppUser.c and controllerIopUser.c to reduce the number of ifdefs (for ease of reading)
 * Reworked local_dc.c to use gpsclock to use /dev/gpstime if available or gpsclock if it is not available.

awgtpman: adds model rate as a global value.  Use model rate for determining maximum number of test points allowed.  Also fix a build issue.

Awg fix warnings

See merge request cds/advligorts!217

awgtpman: changed to ignore type of first adc.  Will work on other than 16 bit adc.  Now gets sample rate from IOP via mbuf.  Uses sample rate to meter the gds heartbeat instead of fixed 65536 rate.  awg data buffers now can handle 2^20 samples per second.  IO mem data moved partially to new shared memory definition style, but still needs an offset.  What is in the first section of this mbuf?

to allow IOP running with an ADC internal clock as the timing source.

This code can be used when models run in userspace and without GPS hardware.
It removes the dependency on the special device file /dev/gpstime.
GPS time is obtained in userspace from the system clock.
Enable at compile time by defining USE_GPSCLOCK.
When building a model, add userspacegps=1 to the CDS parameters.

awgtpman: copy to page buffer in mbuf changed to 16 byte copy for gcc, falling back to 8 byte otherwise.  Speed up is almost 16x.  Mbuf structures are now required to be 16 byte aligned.  AWG_DATA mbuf structs changed to be 16 byte aligned.

awgtpman: deletion of unneeded #if directives in awgtpman.

See merge request cds/advligorts!216

awgtpman: deletion of unneeded #if directives in awgtpman.  Those deleted here seem to be trivially unneeded.  There are many more that require some research.

In addition to cleaning up #if 0 and #if 1 directives, the following symbol assumptions were made:

OS_VXWORKS undefined
TARGET = 0
RMEM_LAYOUT = 2
GS_UNIX_TARGET defined
PORTMAP = 1
OS_SOLARIS undefined
_LINUX defined
_CONFIG_DYNAMIC undefined
__GNUC__ defined
linux defined
P__DARWIN undefined
__CYGWIN__ undefined
RTLINUX undefined
__linux__ defined
P__SOLARIS undefined
HAVE_CONFG undefined
SITE undefined
IFO undefined

Simplified awg data mbuf handles excitation data streaming from awg to models

See merge request cds/advligorts!215

awgptan: build working with explicit sized integers.  mbuf analyze improved to show the first working awg slot.

rcg, awgtpman: simplified awg data mbuf works for streaming to IOP.  user models build but aren't tested.

awg data used to be in the mbuf named <model_name> along with test points, epics data and a DAQ info structure.

data is now passed through <model_name>_awg.  The entire mbuf is mapped with the AWG_DATA structure.

There's no longer any need to calculate offsets.  Access code is much simpler.  Expansion or reconfiguring of the mbuf layout will be straightforward.