On Thursday 12 March 2009 16:02:50 Bill Hart wrote: > Arghh!! Where on earth is this coming from: > > p->MUL_KaRaTSUBa_THRESHOLD = vec.MUL_KaRaTSUBa_THRESHOLD; \ > p->MUL_TOOM3_THRESHOLD = vec.MUL_TOOM3_THRESHOLD; \ > p->SQR_KaRaTSUBa_THRESHOLD = vec.SQR_KaRaTSUBa_THRESHOLD; \ > p->SQR_TOOM3_THRESHOLD = vec.SQR_TOOM3_THRESHOLD; > > I've changed about 6 instances of A-Z a-z to '[A-Z]' '[a-z]' in > configure.in and whilst earlier in fat.c things get put in lower case, > here they do not!! > > I have commited the new configure.in. Jason can you run autoconf and > recommit and I'll check out the entire thing again. Maybe autoconf is > caching something old on my machine. >
Done > Bill. > > 2009/3/12 Bill Hart <goodwillh...@googlemail.com>: > > Great page, but not nearly enough on tr. > > > > 2009/3/12 <ja...@njkfrudils.plus.com>: > >> here a page on portable shell programming > >> > >> http://www.gnu.org/software/autoconf/manual/autoconf.html#Limitations-of > >>-Usual-Tools > >> > >> On Thursday 12 March 2009 14:53:52 Bill Hart wrote: > >>> It looks like BSD systems will break if the brackets are inserted. > >>> > >>> Bill. > >>> > >>> 2009/3/12 <ja...@njkfrudils.plus.com>: > >>> > On Thursday 12 March 2009 14:43:00 Bill Hart wrote: > >>> >> This seems to work: > >>> >> > >>> >> echo AHELLO | tr [A-Z] [a-z] > >>> > > >>> > http://www.opengroup.org/onlinepubs/009695399/utilities/tr.html> > >>> > > >>> > recomends also quoting the expressions > >>> > > >>> >> both on linux and Sun. > >>> >> > >>> >> Bill. > >>> >> > >>> >> 2009/3/12 <ja...@njkfrudils.plus.com>: > >>> >> > There was a few other translates in there as well > >>> >> > > >>> >> > On Thursday 12 March 2009 14:37:08 Bill Hart wrote: > >>> >> >> Yes, I do believe this is the issue. It is only replacing capital > >>> >> >> A with little a and ignoring the other letters. Here is the > >>> >> >> result after making the change you suggest: > >>> >> >> > >>> >> >> /bin/bash ../libtool --tag=CC --mode=compile gcc -std=gnu99 > >>> >> >> -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I.. > >>> >> >> -DOPERATION_`echo fat_sqr_basecase | sed 's/_$//'` -O2 -m64 -c > >>> >> >> -o fat_sqr_basecase.lo fat_sqr_basecase.c > >>> >> >> gcc -std=gnu99 -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP > >>> >> >> -I.. -DOPERATION_fat_sqr_basecase -O2 -m64 -c fat_sqr_basecase.c > >>> >> >> -fPIC -DPIC -o .libs/fat_sqr_basecase.o > >>> >> >> In file included from .././mpn/x86_64/fat/sqr_basecase.c:23, > >>> >> >> from fat_sqr_basecase.c:9: > >>> >> >> ../mpn/generic/sqr_basecase.c: In function > >>> >> >> '__gmpn_sqr_basecase_fat': ../mpn/generic/sqr_basecase.c:273: > >>> >> >> error: 'struct cpuvec_t' has no member named > >>> >> >> 'SQR_KARATSUBA_THRESHOLD' > >>> >> >> > >>> >> >> Now nothing changed case. > >>> >> >> > >>> >> >> I'll look into getting Sun tr working on that machine. > >>> >> >> > >>> >> >> It's right of course. The field is all in lower case in > >>> >> >> gmp-impl.h. > >>> >> >> > >>> >> >> Bill. > >>> >> >> > >>> >> >> 2009/3/11 <ja...@njkfrudils.plus.com>: > >>> >> >> > On Wednesday 11 March 2009 15:26:13 Bill Hart wrote: > >>> >> >> >> The issue does not occur on sage.math. Thus we can infer: > >>> >> >> >> > >>> >> >> >> * It isn't a core2 issue > >>> >> >> >> * It isn't a gcc issue > >>> >> >> >> * It isn't a Sun CC issue > >>> >> >> >> > >>> >> >> >> The only thing which seems different about that machine is > >>> >> >> >> that it is a Sun. I'm betting it is some kind of > >>> >> >> >> preprocessing/scripting issue due to different tools on that > >>> >> >> >> machine. > >>> >> >> >> > >>> >> >> >> Bill. > >>> >> >> >> > >>> >> >> >> 2009/3/11 Bill Hart <goodwillh...@googlemail.com>: > >>> >> >> >> > I managed to tease out what kind of machine it is, and it is > >>> >> >> >> > definitely a Core2. No idea why it only has an amd directory > >>> >> >> >> > for libraries. > >>> >> >> >> > > >>> >> >> >> > Anyhow, here is the failure with gcc -4.3.3 on fulvia: > >>> >> >> >> > > >>> >> >> >> > /bin/bash ../libtool --tag=CC --mode=compile gcc -std=gnu99 > >>> >> >> >> > -DHAVE_CONFIG_H -I. -I. -I.. -D__GMP_WITHIN_GMP -I.. > >>> >> >> >> > -DOPERATION_`echo fat_sqr_basecase | sed 's/_$//'` -O2 > >>> >> >> >> > -m64 -c -o fat_sqr_basecase.lo fat_sqr_basecase.c > >>> >> >> >> > gcc -std=gnu99 -DHAVE_CONFIG_H -I. -I. -I.. > >>> >> >> >> > -D__GMP_WITHIN_GMP -I.. -DOPERATION_fat_sqr_basecase -O2 > >>> >> >> >> > -m64 -c fat_sqr_basecase.c -fPIC -DPIC -o > >>> >> >> >> > .libs/fat_sqr_basecase.o > >>> >> >> >> > In file included from .././mpn/x86_64/fat/sqr_basecase.c:23, > >>> >> >> >> > from fat_sqr_basecase.c:9: > >>> >> >> >> > ../mpn/generic/sqr_basecase.c: In function > >>> >> >> >> > '__gmpn_sqr_basecase_fat': > >>> >> >> >> > ../mpn/generic/sqr_basecase.c:273: error: 'struct cpuvec_t' > >>> >> >> >> > has no member named > >>> >> >> >> > 'SQR_KaRaTSUBa_THRESHOLD' > >>> >> >> > > >>> >> >> > Why do we have lower case "a" here ? > >>> >> >> > > >>> >> >> > at line 2596 in configure.in , I bet sun uses a different "tr" > >>> >> >> > perhaps try > >>> >> >> > tr [[:upper:]] [[:lower:]] > >>> >> >> > > >>> >> >> >> > make[2]: *** [fat_sqr_basecase.lo] Error 1 > >>> >> >> >> > make[2]: Leaving directory `/home/wbhart/mpir-fulvia/mpn' > >>> >> >> >> > make[1]: *** [all-recursive] Error 1 > >>> >> >> >> > > >>> >> >> >> > Bill. > >>> >> >> >> > > >>> >> >> >> > 2009/3/11 Bill Hart <goodwillh...@googlemail.com>: > >>> >> >> >> >> Config.guess identifies fulvia as a core2, yet the only gcc > >>> >> >> >> >> library files that seem to be present for 64 bit are > >>> >> >> >> >> labelled amd64. That is probably something to do with the > >>> >> >> >> >> failures. > >>> >> >> >> >> > >>> >> >> >> >> I'm unable to get cat /proc/cpuinfo being a Sun, but here > >>> >> >> >> >> is the result of uname -a > >>> >> >> >> >> > >>> >> >> >> >> SunOS fulvia 5.10 Generic_127128-11 i86pc i386 i86pc > >>> >> >> >> >> > >>> >> >> >> >> Not very helpful sorry. > >>> >> >> >> >> > >>> >> >> >> >> Bill. > >>> >> >> >> >> > >>> >> >> >> >> 2009/3/11 Bill Hart <goodwillh...@googlemail.com>: > >>> >> >> >> >>> I did make clean. I just tried it again and it definitely > >>> >> >> >> >>> passes on varro. > >>> >> >> >> >>> > >>> >> >> >> >>> I don't think I'd be too worried about it. The issues on > >>> >> >> >> >>> fulvia are certainly worth worrying about. I'll paste the > >>> >> >> >> >>> failures here in a few hours (I'm tied up with something > >>> >> >> >> >>> else right now). > >>> >> >> >> >>> > >>> >> >> >> >>> Bill. > >>> >> >> >> >>> > >>> >> >> >> >>> 2009/3/11 <ja...@njkfrudils.plus.com>: > >>> >> >> >> >>>> On Wednesday 11 March 2009 03:36:15 Bill Hart wrote: > >>> >> >> >> >>>>> Here are the results of tests so far on SkyNet: > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> eno - 4.3.0 n f / 4.3.3 n f > >>> >> >> >> >>>>> cicero 4.3.0 n f > >>> >> >> >> >>>>> cato 4.1.2 n > >>> >> >> >> >>>>> varro A 4.0.1 n f / 4.3.3 n f > >>> >> >> >> >>>>> fulvia 4.3.3 n * *** / S n ** *** ; > >>> >> >> >> >>>>> menas 4.2.1 n f / 4.3.3 n f > >>> >> >> >> >>>>> iras 4.1.2 n / 4.3.3 n > >>> >> >> >> >>>>> cleo 4.1.2 n / 4.3.3 n > >>> >> >> >> >>>>> rm mark 4.3.3 n / S > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> * 30/36 yasm tests fail!! > >>> >> >> >> >>>>> ** make check failed for yasm > >>> >> >> >> >>>>> *** make failed for --enable-fat > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> where: > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> A = Apple CC > >>> >> >> >> >>>>> S = Sun CC > >>> >> >> >> >>>>> 4.x.y = gcc version > >>> >> >> >> >>>>> n = normal build passed > >>> >> >> >> >>>>> f = fat build passed > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> I haven't been able to figure out a combination of PATH > >>> >> >> >> >>>>> and $LD_LIBRARY_PATH to get the Sun CC and linker > >>> >> >> >> >>>>> working on Mark. I've done it before by copying all the > >>> >> >> >> >>>>> relevant binaries to directories in my home dir, but > >>> >> >> >> >>>>> surely there is a way to get it to work. Anyone know? > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> I don't think I had directories wrong for gcc 4.3.3 on > >>> >> >> >> >>>>> Fulvia, so I think this is a real build failure. I may > >>> >> >> >> >>>>> have gotten the Sun directories wrong however. But > >>> >> >> >> >>>>> likely if it doesn't build with gcc it won't with Sun > >>> >> >> >> >>>>> CC. > >>> >> >> >> >>>>> > >>> >> >> >> >>>>> I'm pretty surprised that fat binary worked on varro, as > >>> >> >> >> >>>>> it is a PPC, not an x86. Perhaps fat binary just does > >>> >> >> >> >>>>> nothing on unsupported archs. > >>> >> >> >> >>>> > >>> >> >> >> >>>> Looking at configure , fat build should fail on non x86 , > >>> >> >> >> >>>> it doesn't set the paths right. Did you make clean > >>> >> >> >> >>>> between builds ? > >>> >> >> >> >>>> > >>> >> >> >> >>>>> Bill. > > --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en -~----------~----~----~----~------~----~------~--~---