Opened 15 years ago

Closed 15 years ago

Last modified 15 years ago

#22378 closed defect (fixed)

Error when building Atlas

Reported by: handyman@… Owned by: jameskyle@…
Priority: Normal Milestone:
Component: ports Version: 1.8.1
Keywords: Cc: thorsten.raasch@…, anddam (Andrea D'Amore), yasir.suhail@…, alakazam@…, wayn3w@…, wcsantos@…, ryandesign (Ryan Carsten Schmidt), gallimanuel@…, storiginal@…, creecode@…, mjjohnson.geo@…, jpascale@…, dershow, mklein-de (Michael Klein), tobi_m@…, rechner@…, dplepage@…, furio.mf@…, vakuumfluktuation@…, LenoreHorner
Port: atlas

Description (last modified by jmroot (Joshua Root))

Building atlas
Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
Command output:             mu=4, nu=4, using  0.75 percent of L1 Cache
              Performance = 596.94 (21.84 of copy matmul, 27.64 of clock)
STAGE 2-3: TUNING PREC='z' (precision 3 of 4)
   STAGE 2-3-1 : BUILDING BLOCK MATMUL TUNE
make -f Makefile INSTALL_LOG/zMMRES pre=z 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG
      zL1MATMUL: lat=7, nb=52, pf=512, mu=6, nu=1, ku=1, if=6, nf=1;
                 Performance: 808.74 (37.44 percent of of detected clock rate)
make -f Makefile INSTALL_LOG/zNCNB pre=z 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOGmake -f Makefile INSTALL_LOG/zbestNN_36x36x36 pre=z nb=36 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG      NCgemmNN : muladd=1, lat=7, pf=512, nb=36, mu=6, nu=1 ku=4,
                 ForceFetch=0, ifetch=6 nfetch=1
                 Performance = 717.44 (88.71 of copy matmul, 33.21 of clock)
make -f Makefile INSTALL_LOG/zbestNT_36x36x36 pre=z nb=36 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG      NCgemmNT : muladd=1, lat=8, pf=512, nb=36, mu=6, nu=1 ku=36,
                 ForceFetch=0, ifetch=6 nfetch=1
                 Performance = 693.73 (85.78 of copy matmul, 32.12 of clock)
make -f Makefile INSTALL_LOG/zbestTN_36x36x36 pre=z nb=36 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG      NCgemmTN : muladd=1, lat=7, pf=512, nb=36, mu=6, nu=1 ku=36,
                 ForceFetch=0, ifetch=6 nfetch=1
                 Performance = 712.25 (88.07 of copy matmul, 32.97 of clock)
make -f Makefile INSTALL_LOG/zbestTT_36x36x36 pre=z nb=36 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG      NCgemmTT : muladd=1, lat=4, pf=512, nb=36, mu=6, nu=1 ku=36,
                 ForceFetch=0, ifetch=6 nfetch=1
                 Performance = 705.07 (87.18 of copy matmul, 32.64 of clock)
make -f Makefile MMinstall pre=z 2>&1 | ./xatlas_tee INSTALL_LOG/zMMSEARCH.LOG
   STAGE 2-3-2: CacheEdge DETECTION
make -f Makefile INSTALL_LOG/atlas_zdNKB.h pre=z 2>&1 | ./xatlas_tee INSTALL_LOG/zMMCACHEEDGE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2
Error: Problem while installing atlas

Attachments (2)

atlas-PPC-Altivec.diff (1.4 KB) - added by mklein-de (Michael Klein) 15 years ago.
PPC build failure workaround
atlas-PPC-Altivec-2.diff (9.1 KB) - added by aaaskew@… 15 years ago.
PowerPC Mac Mini support

Download all attachments as: .zip

Change History (61)

comment:1 Changed 15 years ago by handyman@…

Cc: handyman@… added

Cc Me!

comment:2 Changed 15 years ago by jmroot (Joshua Root)

Cc: handyman@… removed
Description: modified (diff)
Owner: changed from macports-tickets@… to jameskyle@…
Priority: HighNormal

comment:3 Changed 15 years ago by jameskyle@…

What platform, arch, and OS version?

comment:4 Changed 15 years ago by handyman@…

MacPorts 1.8.1 OS X 10.6.1

comment:5 Changed 15 years ago by handyman@…

And atlas 3.8.3_1

comment:6 Changed 15 years ago by skymoo (Adam Mercer)

32 or 64bit hardware?

comment:7 Changed 15 years ago by handyman@…

32 bit

comment:8 Changed 15 years ago by thorsten.raasch@…

Cc: thorsten.raasch@… added

Cc Me!

comment:9 Changed 15 years ago by thorsten.raasch@…

I have similar problems under OS X 10.5.8 (PPC G4).

comment:10 in reply to:  9 Changed 15 years ago by yardcat@…

Similar atlas build problem for me too. 10.4.1.1 G4 PPC

comment:11 Changed 15 years ago by jameskyle@…

@thorston, @yardcat.

Recent changes that should help some of the issues on ppc arches are being pushed.

@handyman

Hm, the only difference I see between your and my platform is the 32 bit. I'm wondering if you've sync'd since I made the changes that detect arch for atlas. I'll have to run down a 32bit to test on.

comment:12 Changed 15 years ago by yaseppochi (Stephen J. Turnbull)

I still have the same problem as yardcat with a similar configuration (10.4.11 PPC G4), after updating from svn (at r60547). I did see Atlas get updated by svn, presumably that was the changeset you mention.

comment:13 Changed 15 years ago by anddam (Andrea D'Amore)

Cc: and.damore@… added

Cc Me!

comment:14 Changed 15 years ago by anddam (Andrea D'Amore)

I'm still having the same issue on Leopard PPC on G4 processor, my port tree is on svn and updated.

comment:15 Changed 15 years ago by yasir.suhail@…

I have the same problem on Tiger PPC. It's the latest port because I just did selfupdate. Any workaround?

Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
Command output:               Performance = 255.23 (16.98 of copy matmul, 20.43 of clock)


STAGE 2-4: TUNING PREC='c' (precision 4 of 4)


   STAGE 2-4-1 : BUILDING BLOCK MATMUL TUNE
make -f Makefile IRunMMDef pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG
make -f Makefile INSTALL_LOG/cMMRES pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG
      cL1MATMUL: lat=4, nb=80, pf=1, mu=5, nu=4, ku=80, if=8, nf=1;
                 Performance: 1179.30 (94.42 percent of of detected clock rate)
make -f Makefile INSTALL_LOG/cNCNB pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOGmake -f Makefile INSTALL_LOG/cbestNN_40x40x40 pre=c nb=40 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG      NCgemmNN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1531.78 (129.89 of copy matmul, 122.64 of clock)
make -f Makefile INSTALL_LOG/cbestNT_40x40x40 pre=c nb=40 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG      NCgemmNT : muladd=1, lat=5, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1584.19 (134.33 of copy matmul, 126.84 of clock)
make -f Makefile INSTALL_LOG/cbestTN_40x40x40 pre=c nb=40 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG      NCgemmTN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1487.85 (126.16 of copy matmul, 119.12 of clock)
make -f Makefile INSTALL_LOG/cbestTT_40x40x40 pre=c nb=40 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG      NCgemmTT : muladd=1, lat=2, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1487.37 (126.12 of copy matmul, 119.08 of clock)
make -f Makefile MMinstall pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMSEARCH.LOG


   STAGE 2-4-2: CacheEdge DETECTION
make -f Makefile INSTALL_LOG/atlas_csNKB.h pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMCACHEEDGE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2

comment:16 Changed 15 years ago by yasir.suhail@…

Cc: yasir.suhail@… added

Cc Me!

comment:17 Changed 15 years ago by alakazam@…

Cc: alakazam@… added

Cc Me!

comment:18 Changed 15 years ago by wayn3w@…

Cc: wayn3w@… added

Cc Me!

comment:19 Changed 15 years ago by wcsantos@…

I have the same problem on a Leopard Dual 1GHz PPC G4. My port tree is updated by "sudo port selfupdate" I'm not on the SVN port tree. Any idea how I can get around this problem?

Thank You.

comment:20 Changed 15 years ago by wcsantos@…

Cc: wcsantos@… added

Cc Me!

comment:21 in reply to:  19 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)

Cc: ryandesign@… added

Replying to wcsantos@…:

My port tree is updated by "sudo port selfupdate" I'm not on the SVN port tree.

There is only one port tree. It's maintained in Subversion and synced to rsync (which "port sync" and "port selfupdate" use) every half hour.

comment:22 Changed 15 years ago by gallimanuel@…

Cc: gallimanuel@… added

Cc Me!

comment:23 in reply to:  15 Changed 15 years ago by vmstanford

Same problem on PPC G4 1.33 GHz. Leopard 10.5.8 gcc version 4.0.1 (Apple Inc. build 5493) Xcode 3.1.3.

Thanks for any info.

comment:24 Changed 15 years ago by storiginal@…

Cc: storiginal@… added

Cc Me!

comment:25 Changed 15 years ago by creecode@…

Similar problem on my G4 Dual 1.0GHz, Mac OS X 10.5.8. Can provide error output or debug output from port if that would be of use.

Are we closer to a diagnosis of the problem or a solution? Is there anything I can do to help?

comment:26 Changed 15 years ago by creecode@…

Cc: creecode@… added

Cc Me!

comment:27 Changed 15 years ago by mjjohnson.geo@…

I, too, can't get Atlas to build. (I'm running OS X 10.5.8 on a PowerPC G4.) I'm not totally sure if it's the same problem. The error message is a bit different:

--->  Building atlas
Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
Command output:       ger : chose routine 3:ATL_cger1_2x1p.c written by R. Clint Whaley
            mu=2, nu=1, using  0.78 percent of L1 Cache
              Performance = 186.67 (14.51 of copy matmul, 14.00 of clock)


STAGE 2-4: TUNING PREC='c' (precision 4 of 4)


   STAGE 2-4-1 : BUILDING BLOCK MATMUL TUNE
      cL1MATMUL: lat=4, nb=80, pf=1, mu=5, nu=4, ku=80, if=8, nf=1;
                 Performance: 1017.47 (76.33 percent of of detected clock rate)
      NCgemmNN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1351.98 (132.88 of copy matmul, 101.42 of clock)
      NCgemmNT : muladd=1, lat=5, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1341.32 (131.83 of copy matmul, 100.62 of clock)
      NCgemmTN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1281.64 (125.96 of copy matmul, 96.15 of clock)
      NCgemmTT : muladd=1, lat=2, pf=1, nb=40, mu=5, nu=4 ku=40,
                 ForceFetch=0, ifetch=8 nfetch=1
                 Performance = 1261.96 (124.03 of copy matmul, 94.67 of clock)
make -f Makefile MMinstall pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/zL3TUNE.LOG


   STAGE 2-4-2: CacheEdge DETECTION
make -f Makefile INSTALL_LOG/atlas_csNKB.h pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMCACHEEDGE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2

Error: Status 1 encountered during processing.

Is this the same thing, or should I open a new report?

comment:28 Changed 15 years ago by mjjohnson.geo@…

Cc: mjjohnson.geo@… added

Cc Me!

comment:29 Changed 15 years ago by jpascale@…

Cc: jpascale@… added

Cc Me!

comment:30 in reply to:  15 Changed 15 years ago by jpascale@…

Same problem under "STAGE 2-4-2". I'm running Tiger 10.4.11 on a PowerBook G4 12". I just started using MacPorts one week ago and this is the second port (octave and then inkscape) that has failed due to dependency on atlas.

comment:31 Changed 15 years ago by dershow

Cc: dersh@… added

Cc Me!

comment:32 Changed 15 years ago by mklein-de (Michael Klein)

This seems to be caused by some discrepancy between native GCC and Apple GCC when it comes to Altivec syntax.

The workaround below makes it build, but fails later in the destroot phase (#22670):

ld: file not found: libptcblas.a
Error: Target org.macports.destroot returned: shell command "cd /opt/local/var/macports/build/_opt_local_var_macports_sources_svn.macports.org_math_atlas/work/atlas-3.8.3/build/lib &&  ld -arch ppc -dynamic -dylib -single_module -dead_strip -x  -all_load -L. -L/opt/local/lib/gcc43 -ldylib1.o -dylib_install_name /opt/local/lib/libptcblas.dylib libptcblas.a -o  libptcblas.dylib -latlas  -lSystem" returned error 1
Command output: ld: file not found: libptcblas.a

Changed 15 years ago by mklein-de (Michael Klein)

Attachment: atlas-PPC-Altivec.diff added

PPC build failure workaround

comment:33 Changed 15 years ago by mklein-de (Michael Klein)

Cc: michael.klein@… added

Cc Me!

comment:34 Changed 15 years ago by linn@…

I believe this is same bug (or related) I'm using OS X 10.4.11 on a 1GHz PPC. I've cleaned, selfupdated, etc. without success .. this seems to be blocking several installs including gimp and octave. As a MacPorts newb any help will be a boon.

sudo port install atlas
--->  Computing dependencies for atlas
--->  Fetching atlas
--->  Verifying checksum(s) for atlas
--->  Extracting atlas
--->  Configuring atlas
--->  Building atlas
Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
Command output: pre=d, muladd=1, lat=4, pf=512, nb=64, mu=3, nu=3, ku=64, mflop=527.11

pre=d, muladd=1, lat=4, pf=512, nb=64, mu=3, nu=3, ku=1, mflop=587.62
make[3]: *** [res/dMMRES] Error 255
make[2]: *** [/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/tune/blas/gemm/res/dMMRES] Error 2
/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//bin/atlas_install.c:376: failed assertion `fp'


IN STAGE 1 INSTALL:  SYSTEM PROBE/AUX COMPILE


   Level 1 cache size calculated as 32KB
   dFPU: Combined muladd instruction with 4 cycle pipeline.
         Apparent number of registers : 29
         Register-register performance=1590.42MFLOPS
   sFPU: Combined muladd instruction with 4 cycle pipeline.
         Apparent number of registers : 29
         Register-register performance=1588.03MFLOPS


IN STAGE 2 INSTALL:  TYPE-DEPENDENT TUNING


STAGE 2-1: TUNING PREC='d' (precision 1 of 4)


   STAGE 2-1-1 : BUILDING BLOCK MATMUL TUNE
make -f Makefile INSTALL_LOG/dMMRES pre=d 2>&1 | ./xatlas_tee INSTALL_LOG/dMMSEARCH.LOG
make[1]: *** [build] Error 134
make: *** [build] Error 2

Error: Status 1 encountered during processing.

comment:35 Changed 15 years ago by tobi_m@…

Cc: tobi_m@… added

Cc Me!

comment:36 Changed 15 years ago by trog24 (Frank J. R. Hanstick)

Hello, Same problem on a dual 1.73 GHz PowerPC G4 running MacOS 10.5.8. Difference in performance numbers with no indication of which processor was used:

Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
Command output:       ger : chose routine 3:ATL_cger1_2x1p.c written by R. Clint Whaley
           mu=2, nu=1, using  0.78 percent of L1 Cache
             Performance = 130.02 ( 7.25 of copy matmul,  7.50 of clock)


STAGE 2-4: TUNING PREC='c' (precision 4 of 4)


  STAGE 2-4-1 : BUILDING BLOCK MATMUL TUNE
     cL1MATMUL: lat=4, nb=80, pf=1, mu=5, nu=4, ku=80, if=8, nf=1;
                Performance: 1499.96 (86.55 percent of of detected clock rate)
     NCgemmNN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                ForceFetch=0, ifetch=8 nfetch=1
                Performance = 1793.70 (119.58 of copy matmul, 103.50 of clock)
     NCgemmNT : muladd=1, lat=5, pf=1, nb=40, mu=5, nu=4 ku=40,
                ForceFetch=0, ifetch=8 nfetch=1
                Performance = 1724.19 (114.95 of copy matmul, 99.49 of clock)
     NCgemmTN : muladd=1, lat=4, pf=1, nb=40, mu=5, nu=4 ku=40,
                ForceFetch=0, ifetch=8 nfetch=1
                Performance = 1667.69 (111.18 of copy matmul, 96.23 of clock)
     NCgemmTT : muladd=1, lat=2, pf=1, nb=40, mu=5, nu=4 ku=40,
                ForceFetch=0, ifetch=8 nfetch=1
                Performance = 1461.29 (97.42 of copy matmul, 84.32 of clock)
make -f Makefile MMinstall pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/zL3TUNE.LOG


  STAGE 2-4-2: CacheEdge DETECTION
make -f Makefile INSTALL_LOG/atlas_csNKB.h pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMCACHEEDGE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2

Error: The following dependencies failed to build: gimp-app gimp2 py25-gtk py25-cairo py25-numpy atlas fftw-3 py25-nose py25-setuptools python25 py25-gobject gimp-jp2 gimp-lqr-plugin liblqr gtk-nodoka-engine coreutils gutenprint icns-gimp macclipboard-gimp macfile-gimp ufraw exiv2 gtkimageview xsane sane-backends libusb-compat libusb
Error: Status 1 encountered during processing.

comment:37 Changed 15 years ago by rechner@…

Cc: rechner@… added

Cc Me!

comment:38 in reply to:  36 Changed 15 years ago by dplepage@…

I get the same error as trog24, running MacOS 10.6.2 on a MacBook Pro (Intel Core Duo).

comment:39 Changed 15 years ago by dplepage@…

Cc: dplepage@… added

Cc Me!

comment:40 Changed 15 years ago by aaaskew@…

I get the same error on a PPC G4 running OS X 10.4.11. I have messed around a bit with the atlas configure and make files without any luck. There appears to be a number of things wrong. Could other people confirm the following?

  • Atlas starts by compiling some programs to work out what OS/compiler/architecture it is running on. Take a look at the /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build directory. Run xprobe_arch. It does not appear as if it has been able to determine what CPU I have. I think it should say MACHTYPE=4 which is PPCG4. Instead it says MACHTYPE=0 or UNKNOWN.

There are also xarchinfo_freebsd and xarchinfo_linux programs. Both of these output MACHTYPE=0 / PTR BITS=32, SURE=0. I would also assume to have only one of these rather than both so this would be another indication of Atlas failing to work out what OS/CPU it is running on.

Also the xconfig program outputs the architecture as UNKNOWN when it is run as part of the configure stage.

  • What I have done so far is to see if I can force the build to be PPCG4. To do this I did the following:
1. sudo port clean --work atlas
2. sudo port fetch atlas
3. sudo port extract atlas
4. sudo port patch atlas

Then I added a -A 4 argument to the xconfig lines in the Atlas configure file. I also edited the Atlas CONFIG/src/Makefile, changing the IRun sections for arch, freebsd and linux. For example:

IRunArchInfo_freebsd: xarchinfo_freebsd
        - rm -f config0.out
        $(MAKE) $(atlrun) atldir=$(mydir) exe=xarchinfo_freebsd args="$(args)" \
                redir=config0.out
        - cat config0.out
        - echo MACHTYPE=4 > configEXTRA.out
        - cat configEXTRA.out
5. sudo port -d configure atlas
6. sudo port -d build atlas

Having made these changes I think it is now using the PPCG4 CPU profile rather than UNKNOWN. This still fails though. In the Atlas build directory I have a error_PPCG432AltiVec.tgz file that I don't know enough about to understand what is wrong. What I would say is running the compile and build in debug mode shows an error before what is reported above. I see the following:

/opt/local/bin/gcc-mp-4.3 -DL2SIZE=4194304
 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/include
 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include
 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include/contrib
 -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX -DATL_ARCH_PPCG4 -DATL_CPUMHZ=1416 -DATL_AltiVec -DATL_GAS_PPC -DWALL
 -DATL_UCLEANM -DATL_UCLEANN -DATL_UCLEANK -O3 -maltivec -mabi=altivec -mcpu=7400 -mtune=7400 -no-cpp-precomp -faltivec -fPIC -m32
 -c ATL_cupMBmm0_1_0_b0.c
gcc-mp-4.3: unrecognized option '-no-cpp-precomp'
ATL_cupMBmm0_1_0_b0.c:82:1: warning: "ATL_AltiVec" redefined
<command-line>: warning: this is the location of the previous definition
ATL_cupMBmm0_1_0_b0.c: In function 'ATL_mmcu':
ATL_cupMBmm0_1_0_b0.c:131: error: can't convert value to a vector
ATL_cupMBmm0_1_0_b0.c: In function 'ATL_cupMBmm0_1_0_b0':
ATL_cupMBmm0_1_0_b0.c:212: error: can't convert value to a vector
ATL_cupMBmm0_1_0_b0.c:219: warning: use of 'long' in AltiVec types is deprecated; use 'int'
ATL_cupMBmm0_1_0_b0.c:219: error: can't convert between vector values of different size
make[7]: *** [ATL_cupMBmm0_1_0_b0.o] Error 1
make[6]: *** [ccleanuplib] Error 2
make[5]: *** [clib] Error 2
make[4]: *** [cmmlib] Error 2
make[3]: *** [cinstall] Error 2
make[2]: *** [MMinstall] Error 2

There are also lots of 'gcc-mp-4.3: unrecognized option '-no-cpp-precomp' ' around.

My guess is that the Atlas library is not configured properly to compile with GCC 4.3. The Atlas website mentions GCC 4.2 instead.

Maybe if the maintainer of this Atlas port could raise an issue upstream library the compile errors might be understood and fixed or a new version of the library included?

comment:41 Changed 15 years ago by handyman@…

I noticed Atlas was updated 4 days ago in the SVN, so I just did an update and tried it again… this time it was successful.

comment:42 in reply to:  41 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)

Replying to handyman@…:

I noticed Atlas was updated 4 days ago in the SVN, so I just did an update and tried it again… this time it was successful.

That change was r62043 which just updated the livecheck. It would have had no effect on the ability to install the port. There have been no other changes to the port in the last 6 weeks. I'm glad it's working for you now but r62043 would not be the reason.

comment:43 Changed 15 years ago by aaaskew@…

I have created a new patch which should help PowerPC Mac Mini owners. This might help other PowerPC machines if the configure stage (in debug mode) shows UNKNOWN for the processor type. To test this patch:

cd `port dir atlas`
sudo patch -p0 < ~/Desktop/atlas-PPC-Altivec-2.diff
sudo port clean --all atlas
sudo port -d install atlas

If the processor type is still shown as UNKNOWN then this is something that should be forwarded on to the Atlas people at http://math-atlas.sourceforge.net/faq.html#help The Atlas build works out the processor type using:

sysctl hw.model

Try this on your machine and then look to see if it is defined:

cd `port dir atlas`
cd work/atlas-3.8.3/CONFIG/src/backend 
less archinfo_freebsd.c

...
  case AFPPC: /* don't know */
      if (!CmndOneLine(NULL, "sysctl hw.model", res))
      {
         if (strstr(res, "PowerMac"))
         {
            if (strstr(res,"c1,2")||strstr(res,"c3,1")||strstr(res,"c3,2")||
                strstr(res,"c3,3")||strstr(res,"c3,4")||strstr(res,"c3,5")||
                strstr(res,"c3,6")||strstr(res,"c4,2")||strstr(res,"c4,5")||
                strstr(res,"c5,1"))
               mach = PPCG4;
            else if
...
I added
strstr(res,"c10,1") to this list so that my Mac Mini, which returns PowerMac10,1, would work correctly.

Now I have the same problem as michael.klein above.

Changed 15 years ago by aaaskew@…

Attachment: atlas-PPC-Altivec-2.diff added

PowerPC Mac Mini support

comment:44 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)

I have made a note in the upstream bug report that instead of inferring the CPU type after querying "sysctl hw.model", they should query "sysctl hw.cputype" and "sysctl hw.cpusubtype".

comment:45 Changed 15 years ago by furio.mf@…

Cc: furio.mf@… added

Cc Me!

comment:46 Changed 15 years ago by vakuumfluktuation@…

Cc: vakuumfluktuation@… added

Cc Me!

comment:47 Changed 15 years ago by khepler

Looks like I have the same problem, otherwise let me know. My system seems to properly detect 32bit PPC G4, but I noticed these errors a ways back in the output:

[SNIP]

/opt/local/bin/gcc-mp-4.3 -DL2SIZE=4194304 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include/contrib -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX -DATL_ARCH_PPCG4 -DATL_CPUMHZ=1666 -DATL_AltiVec -DATL_GAS_PPC -DWALL -O3 -maltivec -mabi=altivec -mcpu=7400 -mtune=7400 -no-cpp-precomp -faltivec -fPIC -m32 -DATL_BETA=1 -c cmm_b1.c 
gcc-mp-4.3: unrecognized option '-no-cpp-precomp'
cmm_b1.c: In function 'ATL_cJIK80x80x80TN80x80x0_a1_b1':
cmm_b1.c:147: error: can't convert value to a vector
cmm_b1.c:153: warning: use of 'long' in AltiVec types is deprecated; use 'int'
cmm_b1.c:153: error: can't convert between vector values of different size
make[6]: *** [cBuildCobjs] Error 1
make[5]: *** [cmmucase] Error 2
make cmmucase mmrout=CASES/ATL_smm4x4x16_av.c csC=2 casnam=res/cuser308_80x80x80 pre=c muladd=1 lat=0 M=80 N=80 K=80 mb=80 nb=80 kb=80 mu=4 nu=4 ku=16 lda=80 ldb=80 ldc=0 

BEST USER CASE 307, NB=80: -1.00 MFLOP

[SNIP]

/opt/local/bin/gcc-mp-4.3 -DL2SIZE=4194304 -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include -I/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/..//include/contrib -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX -DATL_ARCH_PPCG4 -DATL_CPUMHZ=1666 -DATL_AltiVec -DATL_GAS_PPC -DWALL -DATL_UCLEANM -DATL_UCLEANN -DATL_UCLEANK -O3 -maltivec -mabi=altivec -mcpu=7400 -mtune=7400 -no-cpp-precomp -faltivec -fPIC -m32 -c ATL_cupMBmm0_1_0_b0.c
gcc-mp-4.3: unrecognized option '-no-cpp-precomp'
ATL_cupMBmm0_1_0_b0.c:82:1: warning: "ATL_AltiVec" redefined
<command-line>: warning: this is the location of the previous definition
ATL_cupMBmm0_1_0_b0.c: In function 'ATL_mmcu':
ATL_cupMBmm0_1_0_b0.c:131: error: can't convert value to a vector
ATL_cupMBmm0_1_0_b0.c: In function 'ATL_cupMBmm0_1_0_b0':
ATL_cupMBmm0_1_0_b0.c:212: error: can't convert value to a vector
ATL_cupMBmm0_1_0_b0.c:219: warning: use of 'long' in AltiVec types is deprecated; use 'int'
ATL_cupMBmm0_1_0_b0.c:219: error: can't convert between vector values of different size
make[7]: *** [ATL_cupMBmm0_1_0_b0.o] Error 1
make[6]: *** [ccleanuplib] Error 2
make[5]: *** [clib] Error 2
make[4]: *** [cmmlib] Error 2
make[3]: *** [res/atlas_csNKB.h] Error 2
make[2]: *** [/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build/tune/blas/gemm/res/atlas_csNKB.h] Error 2
ERROR 664 DURING CACHE EDGE DETECTION!!.
cd /opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build ; make error_report
make -f Make.top error_report
uname -a 2>&1 >> bin/INSTALL_LOG/ERROR.LOG
/opt/local/bin/gcc-mp-4.3 -v 2>&1  >> bin/INSTALL_LOG/ERROR.LOG
Using built-in specs.
Target: ppc-apple-darwin9
Configured with: ../gcc-4.3.4/configure --prefix=/opt/local --build=ppc-apple-darwin9 --enable-languages=c,c++,objc,obj-c++,java,fortran --libdir=/opt/local/lib/gcc43 --includedir=/opt/local/include/gcc43 --infodir=/opt/local/share/info --mandir=/opt/local/share/man --with-local-prefix=/opt/local --with-system-zlib --disable-nls --program-suffix=-mp-4.3 --with-gxx-include-dir=/opt/local/include/gcc43/c++/ --with-gmp=/opt/local --with-mpfr=/opt/local --disable-multilib
Thread model: posix
gcc version 4.3.4 (GCC) 
/opt/local/bin/gcc-mp-4.3 -V 2>&1  >> bin/INSTALL_LOG/ERROR.LOG
gcc-mp-4.3: '-V' option must have argument
make[4]: [error_report] Error 1 (ignored)
/opt/local/bin/gcc-mp-4.3 --version 2>&1  >> bin/INSTALL_LOG/ERROR.LOG
tar cf error_PPCG432AltiVec.tar Make.inc bin/INSTALL_LOG/*
tar: bin/INSTALL_LOG/Stage1.log: file changed as we read it
tar: bin/INSTALL_LOG/dMMSEARCH.LOG: file changed as we read it
tar: bin/INSTALL_LOG/zMMSEARCH.LOG: file changed as we read it
gzip --best error_PPCG432AltiVec.tar
mv error_PPCG432AltiVec.tar.gz error_PPCG432AltiVec.tgz
Error report error_<ARCH>.tgz has been created in your top-level ATLAS
directory.  Be sure to include this file in any help request.
cat: ../../CONFIG/error.txt: No such file or directory
cat: ../../CONFIG/error.txt: No such file or directory

[SNIP]

These blocks of errors repeat several times throughout the build process, but the build only halts once it reaches:

   STAGE 2-4-2: CacheEdge DETECTION
make -f Makefile INSTALL_LOG/atlas_csNKB.h pre=c 2>&1 | ./xatlas_tee INSTALL_LOG/cMMCACHEEDGE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2
Error: Target org.macports.build returned: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
DEBUG: Backtrace: shell command " cd "/opt/local/var/macports/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas/work/atlas-3.8.3/build" && /usr/bin/make build " returned error 2
    while executing
"command_exec build"
    (procedure "portbuild::build_main" line 9)
    invoked from within
"$procedure $targetname"
Warning: the following items did not execute (for atlas): org.macports.build
Error: Status 1 encountered during processing.
To report a bug, see <http://guide.macports.org/#project.tickets>

PowerBook G4: 1.67GHz PPC 7450
Mac OSX Leopard 10.5.8
MacPorts 1.8.2
atlas @3.8.3, Revision 1

comment:48 Changed 15 years ago by jameskyle@…

Resolution: fixed
Status: newclosed

I apologize for the delays, I've been in SE Asia for the last month with limited internet availability.

I also thank all the G4 users for their work here as I do not have a system to test this arch on. I've committed the patches and built atlas on the intel systems I do have and all seems well.

I've committed these just now and the changes should be available in 30m.

comment:49 Changed 15 years ago by francois.work@…

on ppc 10.4.11 macports 1.8.2. I have tried to compile atlas for a couple of days now. I have just got the latest version from svn. it got further. it did compile (port -v build is fine) but the port install still does not work

   STAGE 2-1-5: GEMV TUNE
make -f Makefile INSTALL_LOG/dMVRES pre=d 2>&1 | ./xatlas_tee INSTALL_LOG/dMVTUNE.LOG
make[1]: *** [build] Error 255
make: *** [build] Error 2

Error: Status 1 encountered during processing.
Before reporting a bug, first run the command again with the -d flag to get complete output.

I am compiling / installating with the debug flag. it takes hours to compile

comment:50 Changed 15 years ago by francois.work@…

details from dMVTUNE.LOG:

gcc-mp-4.3: unrecognized option '-no-cpp-precomp'
/opt/local/bin/gcc-mp-4.3 -DL2SIZE=4194304 -I/opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/include -I/opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/..//include -I/opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/..//include/contrib -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX -DATL_ARCH_PPCG4 -DATL_CPUMHZ=799 -DATL_AltiVec -DATL_GAS_PPC -DATL_AltiVec -DATL_AVgcc -DWALL -O3 -maltivec -mabi=altivec -mcpu=7400 -mtune=7400 -no-cpp-precomp -faltivec -fPIC -m32 -o xdmv dgemvtune.o \
                   ATL_dgemv.o ATL_dgemvN_b0.o ATL_dgemvN_b1.o ATL_dgemvN_bX.o ATL_dgemvT_b0.o ATL_dgemvT_b1.o ATL_dgemvT_bX.o /opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/lib/libtstatlas.a /opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/lib/libatlas.a
gcc-mp-4.3: unrecognized option '-no-cpp-precomp'
/opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/bin/ATLrun.sh /opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/tune/blas/gemv xdmv -C 5000 -l 100 -A T \
                  -m 1000 -n 1000 -f 0 -o res/dgemvT_5000_100 
      res/dgemvT_5000_100 : 39.957908 MFLOPS
      res/dgemvT_5000_100 : 69.657903 MFLOPS
      res/dgemvT_5000_100 : 56.145995 MFLOPS
   res/dgemvT_5000_100 : 55.25 MFLOPS


res/dgemvT_5000_100 : VARIATION EXCEEDS TOLERENCE, RERUN WITH HIGHER REPS.

make[3]: *** [res/dMVRES] Error 255
make[2]: *** [/opt/local/var/macports/build/_Users_xx_Desktop_svn_atlas/work/atlas-3.8.3/build/tune/blas/gemv/res/dMVRES] Error 2

Since I can't reopen this ticket I will try to open a new one when debug compile is finished

comment:51 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)

Resolution: fixed
Status: closedreopened

I've re-opened it for you. Please remember WikiFormatting in the future.

comment:52 Changed 15 years ago by jameskyle@…

res/dgemvT_5000_100 : VARIATION EXCEEDS TOLERENCE, RERUN WITH HIGHER REPS.

This appears to be unrelated to the original ticket and not a ports bug. You're failing the atlas timing tests.

The atlas site suggests compiling without any other processes putting a load on the cpu to help prevent this from happening.

comment:53 Changed 15 years ago by jameskyle@…

Resolution: fixed
Status: reopenedclosed

I'm resolving this ticket as fixed since the original submitter's issue is taken care of.

@Francois I would check the ATLAS site or their mailing lists for more suggestions on how to approach failed cpu clock tests.

comment:54 Changed 15 years ago by francois.work@…

I do that because I was compiling qt4-mac at the same time (qt4-mac takes 48h) I am surprised because I have downloaded atlas outside macports and I have compiled it straight without any surprise. Keep in touch today.

comment:55 Changed 15 years ago by francois.work@…

You are right I do not have the same message right now. I create a new defect to stop poluting this case.

comment:56 in reply to:  48 Changed 15 years ago by ryandesign (Ryan Carsten Schmidt)

Replying to jameskyle@…:

I've committed the patches

...in r62767.

comment:57 Changed 15 years ago by joseph.vanpelt@…

Should this be re-opened? I am running 10.5.8 on 32 bit PPC G4: I just issued;

  sudo port sync
  sudo port selfupdate
  sudo port upgrade outdated

Then as instructed above;

  sudo port -v build atlas

The tail end of that build was this (looks successful);

STAGE 4: POST-BUILD TUNING
make -f Makefile IPostTune 2>&1 | ./xatlas_tee INSTALL_LOG/POSTTUNE.LOG
   done.

ATLAS install complete.  Examine 
ATLAS/bin/<arch>/INSTALL_LOG/SUMMARY.LOG for details.
/usr/bin/make clean
rm -rf *.o x* config?.out *core*

Now when I try to install the dependant I still get;

--->  Computing dependencies for py25-gtk
--->  Staging atlas into destroot
Error: Target org.macports.destroot returned: shell command "cd /opt/local/var/macports
/build/_opt_local_var_macports_sources_rsync.macports.org_release_ports_math_atlas
/work/atlas-3.8.3/build/lib &&  ld -arch ppc -dynamic -dylib -single_module -dead_strip
 -x  -all_load -L. -L/opt/local/lib/gcc43 -ldylib1.o -dylib_install_name /opt/local
/lib/libptcblas.dylib libptcblas.a -o  libptcblas.dylib -latlas  -lSystem" returned error 1

Command output: ld: file not found: libptcblas.a

comment:58 Changed 15 years ago by joseph.vanpelt@…

Well, it looks like this is already another new ticket #23349.

comment:59 Changed 15 years ago by LenoreHorner

Cc: LenoreHorner@… added

Cc Me!

Note: See TracTickets for help on using tickets.