Re: Any News on EPICS 3.14.11 with vxWorks 6.7

["Ernest L. Williams Jr." <ernesto@slac.stanford.edu>]

Hi Jeff,

Awesome!!!

I am able to see this now.

I have submitted a new case with WindRiver and will raise the priority 
to get things resolved as soon as possible.
They are working on this as we speak. :)

Thanks again for isolating this compiler bug.
I still don't understand why our colleagues that use PPC604 and EPICS 
R3.14.11 with vxWorks 6.6/6.7 don't see this problem.
Well, in any case do you have a work-around that will step around this 
in order to build cas and excas?


Cheers,
Ernest



Jeff Hill wrote:
> Ernest,
>
> The following code reproduces the vxWorks 6.6 gcc compiler bug (as seen with
> EPICS R3.14.11).
>
> [hill@integrity cas]$
> /ade/vxWorks/6/gnu/4.1.2-vxworks-6.6/x86-linux2/bin/ccppc -c
> -DCPU=PPC604  -DvxWorks      -O2   -Wall     -mcpu=604 -mstrict-align
> -fno-implicit-fp    -fno-builtin   -MMD -I. -I../O.Common -I. -I..
> -I../../../../src/cas/generic -I../../../../src/cas/io/bsdSocket
> -I../../../../src/cas/generic/st -I../../../../src/cas/../ca
> -I../../../../include/os/vxWorks -I../../../../include
> -I/ade/vxWorks/6/vxworks-6.6/target/h
> -I/ade/vxWorks/6/vxworks-6.6/target/h/wrn/coreip -v
> Using built-in specs.
> Target: powerpc-wrs-vxworks
> Configured with: /scratch/nathan/vxworks/src/gcc-4.1/configure
> --build=i686-pc-linux-gnu --host=i686-pc-linux-gnu
> --target=powerpc-wrs-vxworks --enable-threads --disable-libmudflap
> --disable-libssp --disable-libgomp --disable-libstdcxx-pch
> --enable-sjlj-exceptions --disable-hosted-libstdcxx
> --enable-version-specific-runtime-libs --with-gnu-as --with-gnu-ld
> --enable-languages=c,c++ --disable-shared --disable-hosted-libstdcxx
> --with-cxxabi=/scratch/nathan/vxworks/src/dinkum-20021215/include
> --with-pkgversion='Wind River VxWorks G++ 4.1-82'
> --with-bugurl=support@windriver.com --disable-nls --prefix=/opt/codesourcery
> --exec-prefix='/opt/codesourcery/x86-linux2'
> --libdir='/opt/codesourcery/lib'
> --program-transform-name='s,^gcc$,cc,;s,$,ppc,'
> --with-gxx-include-dir=''\''/opt/codesourcery/'\''include/c++/4.1'
> --with-license=/scratch/nathan/vxworks/powerpc/obj/host-libs-4.1-82-powerpc-
> wrs-vxworks-i686-pc-linux-gnu/usr --with-csl-license-version=20071015
> --with-csl-license-feature=gcc_Power_VxWorks
> --enable-poison-system-directories
> Thread model: vxworks
> gcc version 4.1.2 (Wind River VxWorks G++ 4.1-82)
>
> [hill@integrity cas]$
> /ade/vxWorks/6/gnu/4.1.2-vxworks-6.6/x86-linux2/bin/ccppc -c
> -DCPU=PPC604  -DvxWorks      -O2   -Wall     -mcpu=604 -mstrict-align
> -fno-implicit-fp    -fno-builtin   -MMD -I. -I../O.Common -I. -I..
> -I../../../../src/cas/generic -I../../../../src/cas/io/bsdSocket
> -I../../../../src/cas/generic/st -I../../../../src/cas/../ca
> -I../../../../include/os/vxWorks -I../../../../include
> -I/ade/vxWorks/6/vxworks-6.6/target/h
> -I/ade/vxWorks/6/vxworks-6.6/target/h/wrn/coreip tmp.cc
>
> tmp.cc: In member function 'DblRet Cat::meow()':
> tmp.cc:49: error: unable to find a register to spill in class 'FLOAT_REGS'
> tmp.cc:49: error: this is the insn:
> (insn:HI 30 29 85 3 (set (reg:DF 121 [ D.1042 ])
>         (mem/s/c:DF (plus:SI (reg/f:SI 113 sfp)
>                 (const_int 8 [0x8])) [9 D.1040+0 S8 A64])) 301
> {*movdf_hardfloat32} (insn_list:REG_DEP_ANTI 29 (nil))
>     (nil))
>
> ------------------------snip-snip------------------
>
> class DblRet {
> public:
>     DblRet ();
> private:
>     double delay;
> };
>
> struct Cow {
>     void milk ();
> private:
>     void moo ();
> };
>
> class Fish
> {
> public:
>     Fish ( int & );
>     ~Fish ();
> };
>
> inline Fish :: Fish ( int & )
> {
> }
>
> inline Fish :: ~Fish ()
> {
> }
>
> inline void Cow :: milk ()
> {
>     extern int & fishy;
>     Fish fish ( fishy );
>     moo ();
> }
>
> class Cat {
> public:
>     DblRet meow ();
> private:
>     Cow * m_pCow;
> };
>
> DblRet Cat :: meow ()
> {
>     m_pCow->milk ();
>     return DblRet ();
> }
>
> ------------------------snip-snip------------------
>
> Jeff
> ______________________________________________________
> Jeffrey O. Hill           Email        johill@lanl.gov
> LANL MS H820              Voice        505 665 1831
> Los Alamos NM 87545 USA   FAX          505 665 5107
>
> Message content: TSPA
>
>
>   
> > -----Original Message-----
> > From: Ernest L. Williams Jr. [mailto:ernesto@slac.stanford.edu]
> > Sent: Wednesday, December 09, 2009 12:46 PM
> > To: Jeff Hill
> > Cc: 'Andrew Johnson'; 'Yves Lussignol'; kazuro.furukawa@kek.jp; 'Jean-
> > Francois Gournay'; timo.korhonen@psi.ch; 'Eric Bjorklund'; Ernest L.
> > Williams Jr.
> > Subject: Re: Any News on EPICS 3.14.11 with vxWorks 6.7
> >
> > Jeff Hill wrote:
> >     
> > > Hi Ernest,
> > >
> > > Executive summary:
> > > 1) Apparently, GCC doesn’t fail with the MVME550 but this is very
> > >       
> > similar
> >     
> > > processor. What -mcpu flag was used with the MVME5500?
> > >
> > >       
> > Not sure, if they have a separate CONFIG file for MVME5500
> > What we have with the distribution of BASE from APS is the following:
> > <>/configure/os/CONFIG.Common.vxWorks-ppc604
> >
> > The flag is :
> > ARCH_DEP_CFLAGS_4 = -mcpu=604 -mstrict-align -fno-implicit-fp
> >
> >
> >     
> > > 2) I will need to try to reproduce it here with our vxWorks 6.6. That
> > >       
> > will
> >     
> > > require installation of EPICS R3.14.11 onto a new host here.
> > >
> > >       
> > Okay, be sure to use the latest GNU Compiler patches from Windriver:
> >
> >     
> > > 3) I wonder if -mcpu=604e or -mcpu=7450 would a closer match to the
> > >       
> > board;
> >     
> > > perhaps that’s worth a try.
> > >
> > >       
> > I tried both options; still does not work. Fails the same way.
> >
> >
> >     
> > > Sorry about the delay responding, Los Alamos was, for all practical
> > > purposes, in stasis yesterday due to snow, wind, and power outages (I
> > >       
> > spent
> >     
> > > most of the day shoveling out three driveways).
> > >
> > > A quick review of the current situation is in order. There was a no-
> > >       
> > altivec
> >     
> > > flag for the C++ compiler that might have helped, but it didn’t, and now
> > >       
> > we
> >     
> > > see that the issue does not occur with the MVME5500 but does occur with
> > >       
> > the
> >     
> > > MVME6600. The MVME5500 has an MPC7455 processor chip and the MVME6100
> > >       
> > has an
> >     
> > > MPC7457 processor chip. We are using -mcpu=604 when the compiler fails,
> > >       
> > and
> >     
> > > I am not sure what -mcpu option was used for the MVME5500 when it did
> > > compile. Perhaps -mcpu=604e or -mcpu=7450 would be worth trying?
> > >
> > >       
> > Again, I tried the different mcpu flags that you recommended. Still no
> > change; build fails.
> >
> > By the way build for PPC603 architecture builds just fine. :)
> > I am only having a problem with PPC604
> >
> >
> >
> >     
> > > http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=MPC7457
> > >
> > > http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=MPC7455
> > >
> > > I don’t claim to have read up on the subtle differences between these
> > >       
> > chips.
> >     
> > > Nevertheless, the compiler should produce object code for the supported
> > >       
> > CPU
> >     
> > > architecture that we select so we have a compiler bug; we are just
> > >       
> > looking
> >     
> > > for a workaround when selecting different architectures.
> > >
> > >       
> > I need some work-around in the cas source code; I definitely need this
> > because of the PV Gateway dependencies.
> >
> >
> >     
> > > I _have_ tried to reproduce this in a small code but this is time
> > >       
> > consuming
> >     
> > > because we have been using Ernest's compiler to test. I will need to try
> > >       
> > to
> >     
> > > reproduce it here.
> > >
> > > Jeff
> > > ______________________________________________________
> > > Jeffrey O. Hill           Email        johill@lanl.gov
> > > LANL MS H820              Voice        505 665 1831
> > > Los Alamos NM 87545 USA   FAX          505 665 5107
> > >
> > > Message content: TSPA
> > >
> > >
> > >
> > >       
> > > > -----Original Message-----
> > > > From: Ernest L. Williams Jr. [mailto:ernesto@slac.stanford.edu]
> > > > Sent: Tuesday, December 08, 2009 10:27 AM
> > > > To: Jeff Hill
> > > > Cc: Andrew Johnson; Yves Lussignol; kazuro.furukawa@kek.jp; Jean-
> > > >         
> > Francois
> >     
> > > > Gournay; timo.korhonen@psi.ch; Ernest L. Williams Jr.
> > > > Subject: Any News on EPICS 3.14.11 with vxWorks 6.7
> > > >
> > > > Hi Jeff,
> > > >
> > > > Just wanted to follow up on the issue with vxWorks GNU compiler:
> > > >
> > > > I don't understand why our colleagues don't see this problem?
> > > >
> > > >
> > > > /afs/slac/package/vxworks/devel/6.7/gnu/4.1.2-vxworks-6.7/x86-
> > > > linux2/bin/ccppc
> > > > -c            -DCPU=PPC604  -DvxWorks -DBSD=44     -g   -Wall
> > > > -mcpu=604 -mstrict-align -fno-implicit-fp -mlongcall    -fno-builtin
> > > > -DBSD=44  -I. -I../O.Common -I. -I.. -I../../../../src/cas/generic
> > > > -I../../../../src/cas/io/bsdSocket -I../../../../src/cas/generic/st
> > > > -I../../../../src/cas/../ca -I../../../../include/os/vxWorks
> > > > -I../../../../include
> > > > -I/afs/slac/package/vxworks/devel/6.7/vxworks-6.7/target/h
> > > > -I/afs/slac/package/vxworks/devel/6.7/vxworks-6.7/target/h/wrn/coreip
> > > > ../../../../src/cas/generic/st/casStreamOS.cc
> > > > In file included from
> > > > /afs/slac/package/vxworks/devel/6.7/vxworks-6.7/target/h/taskLib.h:257,
> > > >                  from
> > > > /afs/slac/package/vxworks/devel/6.7/vxworks-
> > > > 6.7/target/h/private/selectLibP.h:24,
> > > >                  from
> > > > /afs/slac/package/vxworks/devel/6.7/vxworks-
> > > >         
> > 6.7/target/h/selectLib.h:75,
> >     
> > > >                  from ../../../../include/os/vxWorks/osdSock.h:35,
> > > >                  from ../../../../include/osiSock.h:21,
> > > >                  from ../../../../include/fdManager.h:28,
> > > >                  from ../../../../src/cas/generic/st/casStreamOS.cc:18:
> > > > /afs/slac/package/vxworks/devel/6.7/vxworks-
> > > >         
> > 6.7/target/h/vsbConfig.h:42:2:
> >     
> > > > warning: #warning "VxWorks Source Build (VSB) project not specified;
> > > > using default VxWorks UP configuration under
> > > > $WIND_BASE/target/lib/h/config"
> > > > ../../../../src/cas/generic/st/casStreamOS.cc: In member function
> > > > 'virtual epicsTimerNotify::expireStatus casStreamIOWakeup::expire(const
> > > > epicsTime&)':
> > > > ../../../../src/cas/generic/st/casStreamOS.cc:283: error: unable to
> > > >         
> > find
> >     
> > > > a register to spill in class 'FLOAT_REGS'
> > > > ../../../../src/cas/generic/st/casStreamOS.cc:283: error: this is the
> > > > insn:
> > > > (insn 162 161 163 9 ../../../../src/cas/generic/st/casStreamOS.cc:280
> > > > (set (reg:DF 119 [ D.21342 ])
> > > >         (mem/s/c:DF (plus:SI (reg/f:SI 113 sfp)
> > > >                 (const_int 24 [0x18])) [0 D.21330+0 S8 A64])) 301
> > > > {*movdf_hardfloat32} (nil)
> > > >     (nil))
> > > > ../../../../src/cas/generic/st/casStreamOS.cc:283: confused by earlier
> > > > errors, bailing out
> > > > make[4]: *** [casStreamOS.o] Error 1
> > > > make[4]: Leaving directory
> > > > `/afs/slac.stanford.edu/g/lcls/vol8/epics/base/base-R3-14-
> > > > 11/src/cas/build/O.vxWorks-ppc604_long'
> > > > make[3]: *** [install.vxWorks-ppc604_long] Error 2
> > > > make[3]: Leaving directory
> > > > `/afs/slac.stanford.edu/g/lcls/vol8/epics/base/base-R3-14-
> > > > 11/src/cas/build'
> > > > make[2]: *** [build.install] Error 2
> > > > make[2]: Leaving directory
> > > > `/afs/slac.stanford.edu/g/lcls/vol8/epics/base/base-R3-14-11/src/cas'
> > > > make[1]: *** [cas.install] Error 2
> > > > make[1]: Leaving directory
> > > > `/afs/slac.stanford.edu/g/lcls/vol8/epics/base/base-R3-14-11/src'
> > > >
> > > >
> > > > Thanks,
> > > > Ernest
> > > >
> > > >         
> > >       
>