Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Mon, 7 Jan 2002 15:36:21 -0700

John, 

The original code was looking in the symbol table for the nivxi library
and calling its initialization only when the library was present.

I don't know what version you are using, but I seem to recall that
Andrew Johnson made source code changes to ifdef out the auto
initialization of the nivxi library in recent versions of drvEpvxi. I
think that Andrew now has it so that, if you would like to turn the
nivxi library auto initialization back on, you must specify -DNICPU030
on the command line when building the driver. See attached. 

Alternatively, you could call the nivxi library's initialization from
the vxWorks startup script.

Jeff

> -----Original Message-----
> From: [email protected] [mailto:[email protected]]
> Sent: Monday, August 27, 2001 4:24 PM
> To: Jeff Hill
> Cc: Andrew Johnson; Marty Kraimer
> Subject: Re: drvEpvix.c
> 
> Hi Jeff,
> 
> Jeff Hill wrote:
> >
> > Is there any reason why you shouldn't return the code to its
original
> > state with respect to this particular issue?
> 
> I thought that the CPU030 stuff might have been responsible for some
of
> the problems I was seeing under the PowerPC and so I took it out, but
that
> was before I found the bus timeout issue that really solved it.  I'm
still
> not convinced that we should load it unless it's needed, even if it is
> benign - it's still taking up room that some of our IOCs can make use
of.
> 
> Looking at the changes I made, all I really changed was to comment out
the
> NICPU030 macro definition and make the call to nicpu030_init() truly
> conditional on that macro.  The ability to exclude the cpu030 code was
> already there, I just tweaked it so it worked, and changed the default
> setting.  I can see that changing the default setting was what is
causing
> the problem, but my argument is that the default was wrong in the
first
> place.
> 
> All that the sites needing those routines have to do is define the
macro
> by appending -DNICPU030 to ARCH_DEP_CPPFLAGS in a
> config/CONFIG_SITE.Vx.<arch> file.  Maybe the macro should be renamed
to
> something like NIVXILIB to make it clearer what it is switching, in
which
> case the routines should also be renamed to match.  Unless we put in
some
> inverted-logic 'exclude' switch it's harder for those who don't want
this
> code to exclude it (originally by editing their copy of drvEpvxi.c,
but we
> don't want sites to have to do this kind of thing) than for those who
do
> to include it with an entry somewhere in config.
> 
> In a properly unbundled version the build system could switch this
based
> on the presence of a variable in the config/RELEASE file pointing to
the
> location of the NIVXI library, which presumably has to be loaded
anyway if
> it's needed.
> 
> > Who is the manufacture of the PPC based VXI slot zero controller
> > that you are using, or is APS using a pure VME to MXI to VXI
> > solution as in the past.
> 
> We're just using a VME to MXI-2 to VXI solution as before.
> 
> - Andrew
> --
> The world is such a cheerful place when viewed from upside-down
> It makes a rise of every fall, a smile of every frown

> -----Original Message-----
> From: John Faucett [mailto:[email protected]]
> Sent: Monday, January 07, 2002 2:49 PM
> To: [email protected]
> Subject: VXI driver support
> 
> I'm writing driver support for the KineticSystems V345 output
> register and using the V215 driver that comes with EPICS 3.13.4 as a
> pattern.  This is my first try at a VXI module.  The IOC is a KSC
> V150 and builds with hkbaja60 routines.
> 
> The drvVxi routine does not recognize that the V345 module is
> present.  The vxiMemProbe call in open_vxi_device returns a status
> value of -1 for all logical addresses (the V345 is set to logical
> address 1).  Therefore epvxiResourceMangerOK never becomes TRUE
> causing the device support to never fully initialize.
> 
> Is some sort of initialization for the V150 required before iocInit
runs?

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System