Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
I added in the same sed command " -e 's,ARCHDEFS =,& -DATL_AVgcc' "
Put an updated info file (also different build of lib) in
experimental/jfmertens/main/finkinfo/sci
I hope this should fix it _ sorry to ask to test again..
If it does, there will remain to fix the compiler flags for G4, and
maybe a couple
of other things.. (If someone with a G4 could send me a copy of
Make.inc it might help
_ at that stage).
Would of course be much better to find the configure error that
causes G4's not to be recognised ...
(error is at the line "Bad MACHTYPE value=0..." in your log).
But w/o a G4 at hand, this looks though.. First point I'd like to
check though is
> # cat -n ATLAS/CONFIG/src/backend/archinfo_freebsd.c|grep -C4 G4
> 65if (!CmndOneLine(NULL, "sysctl hw.model", res))
> 66{
> 67 if (strstr(res, "PowerMac"))
> 68 {
> 69 if (strstr(res, "4")) mach = PPCG4;
> 70 else if (strstr(res, "3,3")) mach = PPCG4;
> 71 else if (strstr(res, "7,2") || strstr(res,
> "7,3")) mach = PPCG5;
> 72 }
> 73}
What does "sysctl hw.model" yield on a 7400 and on a 7450 ? (other
G4's?)
Jean-Francois
PS: Dominique _ any news from upstream ?
Do you have a pointer to your bug-report ?
On 21 Nov 2007, at 04:24, Alexander Hansen wrote:
>> I tried using:
>
> if test `machine` = ppc7450 ; then sed -i.bak 's,^\( *ARCH =\).*,\1
> PPCG432AltiVec,' Make.inc ; fi
>
> as above, and so I now get errors nearer the end of the first stage of
> the build:
>
> ...
> gcc -DL2SIZE=4194304 -I/sw/src/fink.build/atlas-3.8.0-4/bld/include
> - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include
> - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include/contrib
> - -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX
> - -DATL_CPUMHZ=866 -DATL_AltiVec -DATL_GAS_PPC -O -fomit-frame-
> pointer
> - -m32 -DATL_BETA=1 -c dmm.c
> dmm.c: In function 'ATL_dJIK56x56x56TN56x56x0_a1_b1':
> dmm.c:40: error: 'vector' undeclared (first use in this function)
> dmm.c:40: error: (Each undeclared identifier is reported only once
> dmm.c:40: error: for each function it appears in.)
> dmm.c:40: error: expected ')' before 'float'
> dmm.c:41: error: expected ')' before 'float'
> make[5]: *** [dmmcase0] Error 1
> make[4]: *** [mmcase] Error 2
> Error in command: make mmcase pre=d loopO=JIK ta=T tb=N M=56 N=56
> K=56 mb=56 nb=56 kb=56 lda=56 ldb=56 ldc=0 lda2=56 ldb2=56 ldc2=0 mu=4
> nu=4 ku=56 alpha=1 beta=1 muladd=1 lat=3 cleanup=1 ff=0 if=7 nf=1
> pfA=1
> mmnreg = 33
>
> NB's to try: 56
>
>
> pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63
>
> pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63
>
> pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63
>
> pre=d, muladd=1, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1010.50
> make[3]: *** [RunMMDef] Error 255
> make[2]: *** [IRunMMDef] Error 2
> cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dMMRES
> INSTALL_LOG/.
> cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dNCNB
> INSTALL_LOG/.
> cp
> /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/
> dbestNN_52x52x52
> INSTALL_LOG/.
> cp
> /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/
> dbestNT_52x52x52
> INSTALL_LOG/.
> cp
> /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/
> dbestTN_52x52x52
> INSTALL_LOG/.
> cp
> /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/
> dbestTT_52x52x52
> INSTALL_LOG/.
> cd /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm ; make dinstall
> ./xemit_mm -p d -R -2
> pre=d, CU=0, ma=0, ff=0, if=-1, nf=-1, lo=1, ta=112, tb=111, lat=4,
> mu=4, nu=4, ku=1, m=0, n=0, k=0, lda=0, ldb=0, ldc=0, csA=1, csB=1,
> csC=1, alpha=1, beta=1
>
> line 4023 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> line 4025 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> line 4027 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> line 3770 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> line 3772 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> line 3774 of
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c
> cat: CASES/: Is a directory
> /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/
> emit_mm.c:3737:
> failed assertion `system(ln) == 0'
> make[3]: *** [dinstall] Abort trap
> make[2]: *** [MMinstall] Error 2
> cd /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm ; make
> res/atlas_cacheedge.h pre=d
> make dRunFindCE
> cd /sw/src/fink.build/atlas-3.8.0-4/bld/src/blas/gemm ; make dlib
> make auxillib dcleanuplib dusergemm
> cd /sw/src/fink.build/atlas-3.8.0-4/bld/src/auxil ; make lib
> make[7]: Nothing to be done for `lib'.
> cd KERNEL ; make -
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
-BEGIN PGP SIGNED MESSAGE- Hash: SHA1 Jean-François Mertens wrote: > > On 20 Nov 2007, at 04:02, Jean-François Mertens wrote: > >> At worst, there is a way if I remember correctly to force the use of >> those >> or those architectural defaults; but I would have hoped for a quicker >> and cleaner solution from upstream. > > If you want to test this already now, try inserting a line like the > following > right between configure and make : > > if test `uname -m` = G4 ; then sed -i.bak 's,^\( *ARCH =\).*,\1 > PPCG432AltiVec,' Make.inc ; fi > > but check first that the test is correct (`machine` , or `arch` > might be better ? _ I can't try now on a G4..) > > JF I tried using: if test `machine` = ppc7450 ; then sed -i.bak 's,^\( *ARCH =\).*,\1 PPCG432AltiVec,' Make.inc ; fi as above, and so I now get errors nearer the end of the first stage of the build: ... gcc -DL2SIZE=4194304 -I/sw/src/fink.build/atlas-3.8.0-4/bld/include - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include/contrib - -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX - -DATL_CPUMHZ=866 -DATL_AltiVec -DATL_GAS_PPC -O -fomit-frame-pointer - -m32 -DATL_BETA=1 -c dmm.c dmm.c: In function 'ATL_dJIK56x56x56TN56x56x0_a1_b1': dmm.c:40: error: 'vector' undeclared (first use in this function) dmm.c:40: error: (Each undeclared identifier is reported only once dmm.c:40: error: for each function it appears in.) dmm.c:40: error: expected ')' before 'float' dmm.c:41: error: expected ')' before 'float' make[5]: *** [dmmcase0] Error 1 make[4]: *** [mmcase] Error 2 Error in command: make mmcase pre=d loopO=JIK ta=T tb=N M=56 N=56 K=56 mb=56 nb=56 kb=56 lda=56 ldb=56 ldc=0 lda2=56 ldb2=56 ldc2=0 mu=4 nu=4 ku=56 alpha=1 beta=1 muladd=1 lat=3 cleanup=1 ff=0 if=7 nf=1 pfA=1 mmnreg = 33 NB's to try: 56 pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63 pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63 pre=d, muladd=0, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1006.63 pre=d, muladd=1, lat=1, pf=0, nb=40, mu=5, nu=4, ku=40, mflop=1010.50 make[3]: *** [RunMMDef] Error 255 make[2]: *** [IRunMMDef] Error 2 cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dMMRES INSTALL_LOG/. cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dNCNB INSTALL_LOG/. cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dbestNN_52x52x52 INSTALL_LOG/. cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dbestNT_52x52x52 INSTALL_LOG/. cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dbestTN_52x52x52 INSTALL_LOG/. cp /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/dbestTT_52x52x52 INSTALL_LOG/. cd /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm ; make dinstall ./xemit_mm -p d -R -2 pre=d, CU=0, ma=0, ff=0, if=-1, nf=-1, lo=1, ta=112, tb=111, lat=4, mu=4, nu=4, ku=1, m=0, n=0, k=0, lda=0, ldb=0, ldc=0, csA=1, csB=1, csC=1, alpha=1, beta=1 line 4023 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c line 4025 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c line 4027 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c line 3770 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c line 3772 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c line 3774 of /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c cat: CASES/: Is a directory /sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//tune/blas/gemm/emit_mm.c:3737: failed assertion `system(ln) == 0' make[3]: *** [dinstall] Abort trap make[2]: *** [MMinstall] Error 2 cd /sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm ; make res/atlas_cacheedge.h pre=d make dRunFindCE cd /sw/src/fink.build/atlas-3.8.0-4/bld/src/blas/gemm ; make dlib make auxillib dcleanuplib dusergemm cd /sw/src/fink.build/atlas-3.8.0-4/bld/src/auxil ; make lib make[7]: Nothing to be done for `lib'. cd KERNEL ; make -f dMakefile dlib make[7]: dMakefile: No such file or directory make[7]: *** No rule to make target `dMakefile'. Stop. make[6]: *** [dcleanuplib] Error 2 make[5]: *** [dlib] Error 2 make[4]: *** [dmmlib] Error 2 make[3]: *** [res/atlas_cacheedge.h] Error 2 make[2]: *** [/sw/src/fink.build/atlas-3.8.0-4/bld/tune/blas/gemm/res/atlas_cachedge.h] Error 2 ERROR 639 DURING CACHE EDGE DETECTION!!. cd /sw/src/fink.build/atlas-3.8.0-4/bld ; make error_report make -f Make.top error_report uname -a 2>&1 >> bin/INSTALL_LOG/ERROR.LOG gcc -v 2>&1 >> bin/INSTALL_LOG/ERROR.LOG Using built-in specs. Target: powerpc-apple-darwin8 Configured with: ../gcc-4.2.2/configure --prefix=/sw - --prefix=/sw/lib/gcc4.2 --mandir=/sw/share/man - --infodir=/sw/share/info --enable-languages=c,c++,fortran,objc,java - --host=powerpc-apple-darwin8 --with-gmp=/sw --with-libiconv-prefix=/sw - --with-system-zlib --x-includes=/usr/X11R6/include - --x-libraries=/usr/X11R6/lib Thread model: pos
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
Jean-François Mertens wrote: [] > but check first that the test is correct (`machine` , or `arch` > might be better ? _ I can't try now on a G4..) `machine` gives ppc7450 on a G4. -- Martin - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
On 20 Nov 2007, at 04:02, Jean-François Mertens wrote: > At worst, there is a way if I remember correctly to force the use of > those > or those architectural defaults; but I would have hoped for a quicker > and cleaner solution from upstream. If you want to test this already now, try inserting a line like the following right between configure and make : if test `uname -m` = G4 ; then sed -i.bak 's,^\( *ARCH =\).*,\1 PPCG432AltiVec,' Make.inc ; fi but check first that the test is correct (`machine` , or `arch` might be better ? _ I can't try now on a G4..) JF - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
On 20 Nov 2007, at 03:38, Jean-François Mertens wrote: > There were other msgs when the last version of atlas came up > showing that on some G4's at least it didn't work. > > cf cvs log : > > revision 1.8 > date: 2007/11/09 16:47:53; author: jfmertens; state: Exp; lines: > +28 -9 > should fix problems with uniprocessor machines and with 64-bit > machines; > remain problems for some G4s > > This resulted IIRC from tests by Dominique Dhumieres on a G4; those > showed in > particular that there was a problem much higher: atlas had not > recognized the machine > _ else I would expect a -DATL_ARCH_xyz flag in your compilation too. > Hence it is treated as UNKNOWN, and the defaults are not used. > > And Atlas does have defaults for G4 : ATLAS/CONFIG/ARCHS/ > PPCG432AltiVec.tgz > > I think Dominique sent a bug-report upstream concerning this. > > JF Mertens At worst, there is a way if I remember correctly to force the use of those or those architectural defaults; but I would have hoped for a quicker and cleaner solution from upstream. JF - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
Alexander K. Hansen wrote: > CPU: 867 MHz G4 OS:10.4.11 Xcode: 10.5 > --- Gah! I've got Leopard on the brain! That should be Xcode: 2.5 - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
Re: [Fink-devel] build failure in atlas-3.8.0-4 on a PowerPC
There were other msgs when the last version of atlas came up showing that on some G4's at least it didn't work. cf cvs log : revision 1.8 date: 2007/11/09 16:47:53; author: jfmertens; state: Exp; lines: +28 -9 should fix problems with uniprocessor machines and with 64-bit machines; remain problems for some G4s This resulted IIRC from tests by Dominique Dhumieres on a G4; those showed in particular that there was a problem much higher: atlas had not recognized the machine _ else I would expect a -DATL_ARCH_xyz flag in your compilation too. Hence it is treated as UNKNOWN, and the defaults are not used. And Atlas does have defaults for G4 : ATLAS/CONFIG/ARCHS/ PPCG432AltiVec.tgz I think Dominique sent a bug-report upstream concerning this. JF Mertens On 20 Nov 2007, at 03:11, Alexander K. Hansen wrote: > -BEGIN PGP SIGNED MESSAGE- > Hash: SHA1 > > CPU: 867 MHz G4 > OS:10.4.11 > Xcode: 10.5 > - --- > There isn't an obvious error message where the build stops: > > 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 > > I do see a few errors during the build of the form > > gcc -DL2SIZE=4194304 -I/sw/src/fink.build/atlas-3.8.0-4/bld/include > - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include > - -I/sw/src/fink.build/atlas-3.8.0-4/bld/../ATLAS//include/contrib > - -DAdd_ -DF77_INTEGER=int -DStringSunStyle -DATL_OS_OSX > - -DATL_CPUMHZ=866 -DATL_AltiVec -DATL_GAS_PPC -O -fomit-frame- > pointer > - -m32 -DDREAL -c drot.c > drot.c: In function 'ATL_UROT': > drot.c:22: error: 'vector' undeclared (first use in this function) > drot.c:22: error: (Each undeclared identifier is reported only once > drot.c:22: error: for each function it appears in.) > drot.c:22: error: expected ')' before 'float' > drot.c:22: error: expected ')' before 'float' > make[8]: *** [drottest] Error 1 > > The salient issue would seem to be the "'vector' undeclared". > > A full build log is at > http://akhmac.blogdns.net/~hansen/finklogs/fink-build- > log_atlas_3.8.0-4_2007.11.19-16.16.28 > -BEGIN PGP SIGNATURE- > Version: GnuPG v1.4.6 (GNU/Linux) > Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org > > iD8DBQFHQkJRB8UpO3rKjQ8RAs+/AJ9yNdXFv+gahe1saWAnQRfgUsukfgCeLZW8 > R98LYCGMnPTCmW2KP0syq40= > =XpJ1 > -END PGP SIGNATURE- > > > -- > --- > This SF.net email is sponsored by: Microsoft > Defy all challenges. Microsoft(R) Visual Studio 2005. > http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ > ___ > Fink-devel mailing list > Fink-devel@lists.sourceforge.net > http://news.gmane.org/gmane.os.apple.fink.devel - This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse012070mrt/direct/01/ ___ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
