November 2007 - CLN-list

future GCC 4.3 and modules.h
by Ralf Wildenhues 10 Jan '08

10 Jan '08

Hello CLN people, Consider this code snippet: extern "C" void func () {} __asm__("\t.globl " "_GLOBAL__I_" "func"); static struct S { inline S () { } } S; On x86_64, with `g++-4.2 -fPIC -c a.cc', this results in | 0000000000000032 T _GLOBAL__I_func with mainline GCC (what will eventually be 4.3 or some other number), it has | 0000000000000032 t _GLOBAL__I_a.cc (and of course that will be `T' with the appropriate .globl directive). This affects .libs/cl_DF_globals.o and some of the other globals objects, and causes a link failure of the examples due to the CL_REQUIRE undefined references (global constructor keyed to ...). The hack below makes things work with mainline, but still needs to be redone properly, with a configure-time check and so on to not regress on g++ <= 4.2. If you like I can take a stab at writing so; I assume the test should make use of $lt_compiler_pic_CXX and the header file provide different defines based on whether -DPIC was given? (I'm not all that experienced with this stuff, so pointers are greatly appreciated.) Cheers, Ralf Index: include/cln/modules.h =================================================================== RCS file: /home/cvs/cln/include/cln/modules.h,v retrieving revision 1.19 diff -u -r1.19 modules.h --- include/cln/modules.h 7 May 2006 21:05:24 -0000 1.19 +++ include/cln/modules.h 26 May 2007 16:44:38 -0000 @@ -219,7 +219,7 @@ }; \ static _CL_REQUIRE_CLASSNAME(module,__LINE__) \ _CL_REQUIRE_CLASSNAME(module##_requirer,__LINE__); - #else + #elif 0 // gcc-3.0 -fuse-cxa-atexit doesn't have a single per-module destructor // function anymore. Instead, for each object's static constructor it // executes, it pushes the corresponding object's destructor onto a list. @@ -255,9 +255,42 @@ }; \ static _CL_REQUIRE_CLASSNAME(module,__LINE__) \ _CL_REQUIRE_CLASSNAME(module##_requirer,__LINE__); - #endif #define _CL_REQUIRE_CLASSNAME(module,line) __CL_REQUIRE_CLASSNAME(module,line) #define __CL_REQUIRE_CLASSNAME(module,line) cl_module__##module##__##line + #else + #define CL_PROVIDE(module) \ + extern "C" void cl_module__##module##__ctorend (void); \ + CL_GLOBALIZE_JUMP_LABEL(cl_module__##module##__ctorend) \ + CL_GLOBALIZE_CTORDTOR_LABEL( \ + ASM_UNDERSCORE_PREFIX CL_GLOBAL_CONSTRUCTOR_PREFIX \ + #module ".cc") \ + static int cl_module__##module##__counter; \ + struct cl_module__##module##__controller { \ + inline cl_module__##module##__controller () \ + { if (cl_module__##module##__counter++) \ + { CL_JUMP_TO(cl_module__##module##__ctorend); } \ + } \ + }; \ + static cl_module__##module##__controller cl_module__##module##__ctordummy; + #define CL_PROVIDE_END(module) \ + struct cl_module__##module##__destroyer { \ + inline cl_module__##module##__destroyer () \ + { CL_OUTPUT_LABEL (ASM_UNDERSCORE_PREFIX "cl_module__" #module "__ctorend"); } \ + }; \ + static cl_module__##module##__destroyer cl_module__##module##__dtordummy; + #define CL_REQUIRE(module) \ + extern "C" void cl_module__##module##__ctor (void) \ + __asm__ (ASM_UNDERSCORE_PREFIX CL_GLOBAL_CONSTRUCTOR_PREFIX \ + #module ".cc"); \ + struct _CL_REQUIRE_CLASSNAME(module,__LINE__) { \ + inline _CL_REQUIRE_CLASSNAME(module,__LINE__) () \ + { cl_module__##module##__ctor (); } \ + }; \ + static _CL_REQUIRE_CLASSNAME(module,__LINE__) \ + _CL_REQUIRE_CLASSNAME(module##_requirer,__LINE__); + #define _CL_REQUIRE_CLASSNAME(module,line) __CL_REQUIRE_CLASSNAME(module,line) + #define __CL_REQUIRE_CLASSNAME(module,line) cl_module__##module##__##line +#endif #else #define CL_PROVIDE(module) #define CL_PROVIDE_END(module)

3 15

Re: [cln] sparc64 and OpenBSD
by Paul Irofti 18 Dec '07

18 Dec '07

On Mon, Nov 19, 2007 at 01:07:40PM +0200, Paul Irofti wrote: > On Mon, Nov 19, 2007 at 12:23:58PM +0200, Paul Irofti wrote: > [--snip--] > > My new error is: > > > [--attached--] > /usr/local/bin/libtool --mode=compile c++ -O2 -pipe -Wformat -Wbounded > -I/usr/local/include -I../include -I../include -I./polynomial -I./base > -I./base/hash -I./base/digitseq -I./base/digit -Ibase -c > ./polynomial/elem/cl_UP.cc > c++ -O2 -pipe -Wformat -Wbounded -I/usr/local/include -I../include > -I../include -I./polynomial -I./base -I./base/hash -I./base/digitseq > -I./base/digit -Ibase -c ./polynomial/elem/cl_UP.cc -fPIC -DPIC -o > .libs/cl_UP.o > In file included from polynomial/elem/cl_UP.cc:36: > polynomial/elem/cl_UP_GF2.h: In function `const cln::_cl_UP > cln::gf2_mul(cln::cl_heap_univpoly_ring*, const cln::_cl_UP&, const > cln::_cl_UP&)': > polynomial/elem/cl_UP_GF2.h:846: error: cannot convert `uintD*' to `uint32*' > for argument `3' to `uint32 cln::gf2_mul32(unsigned int, unsigned int, > uint32*)' > polynomial/elem/cl_UP_GF2.h:862: error: cannot convert `uintD*' to `uint32*' > for argument `3' to `uint32 cln::gf2_mul32(unsigned int, unsigned int, > uint32*)' It seems that this is generated by my former fix (the one I was inquiring about). I can't just remap a 64-bit to a 32-bit function so a devel will have to write an additional assembly routine for 64-bit sparc here: .global C(gf2_mul16),C(gf2_mul32),C(gf2_mul64) ^^^^^^^^^^^^ The file is src/polynomial/elem/cl_asm_sparc_GF2.cc.

2 9

Build problems on Solaris 8
by Robert W. Johnstone 16 Dec '07

16 Dec '07

Hello, I've tried building CLN 1.1.13 on a machine using Solaris 8. I'm having difficulty getting the library to pass all its tests. I've tried building CLN using gcc 4.1.1, with the following combinations of flags: "-O", "-O1", and "-O1 -fno-schedule-insns". The library fails many of the integer_floor_tests. I've attached the log from running tests/exam. Additionally, I've tried compiling with gcc 3.4.0 with the flags "-O -fno-schedule_insns". The library then fails sfloat_mul_tests 72 through 75. I'm uncertain what would be causing the errors. Thank-you, ---------------------------------------- Robert Johnstone Simon Fraser University http://www.sfu.ca/~rjohnsto/

3 4

problems building on powerpc64
by Dmitry Khutornoy 05 Dec '07

05 Dec '07

Hello! I got problems with building CLN on PowerPC64. uname -a Linux blade1 2.6.5-7.252-pseries64 #1 SMP Tue Feb 14 11:11:04 UTC 2006 ppc64 ppc64 ppc64 GNU/Linux make ... In file included from base/cl_N.h:6, from base/cl_as_err.cc:7: ../include/cln/number.h: In constructor `cln::cl_number::cl_number()': ../include/cln/number.h:191: error: `cl_FN_tag' undeclared (first use this function) ../include/cln/number.h:191: error: (Each undeclared identifier is reported only once for each function it appears in.) And it seems i know why does it happen: <config.h> ... #ifndef __powerpc64__ #define __powerpc64__ 1 #endif ... <object.h> ... #if defined(__alpha__) || defined(__ia64__) || defined(__mips64__) || defined(__powerpc64__) || defined(__sparc64__) || defined(__x86_64__) #define cl_word_alignment 8 #endif ... // Definition of the tags. #if !defined(CL_WIDE_POINTERS) #if (cl_word_alignment == 2) #define cl_FN_tag 1 #define cl_SF_tag 3 // must satisfy the cl_immediate_p predicate! #endif #if (cl_word_alignment == 4) #define cl_FN_tag 1 #define cl_SF_tag 2 #endif #else // CL_WIDE_POINTERS // Single Floats are immediate as well. #define cl_FN_tag 1 #define cl_SF_tag 2 #define cl_FF_tag 3 #endif There is no branch defining tags for powerpc64 (cl_word_alignment==8) Please, tell me what values should i write in #define or explain what am i doing wrong. Thank you! -- Best regards, Dmitry mailto:khutornoy@gmail.com

3 5

drop i386 asm?
by Alexei Sheplyakov 29 Nov '07

29 Nov '07

Hello, \begin{quote} CLN is speed efficient: * The kernel of CLN has been written in assembly language for some CPUs (`i386', `m68k', `sparc', `mips', `arm'). \end{quote} I can't say anything about m68k, arm, and mips, but on x86 GCC generated code runs a bit faster. I checked this for -march={pentium[34], nocona,athlon,athlon-mp,k8}. Hence I propose to drop x86 asm, or at least to make -DNO_ASM the default. Best regards, Alexei -- All science is either physics or stamp collecting.

2 1

Convert CLN's rationals to GMP
by Florian Pigorsch 27 Nov '07

27 Nov '07

Hi, I need to convert rationals of the CLN library (cln::cl_RA) to the rational format of GMP (mpq_t). Since CLN internally uses GMP this should be simple, but I didn't find a way to do the conversion. Does anyone know how to do it? Thanks in advance, Florian

1 0

Re: [cln] sparc64 and OpenBSD
by Richard B. Kreckel 19 Nov '07

19 Nov '07

Hi Paul, (I'm CCing the mailing list.) Paul Irofti wrote: > I'm currently porting cln on OpenBSD. I'm having some problems with the > sparc64 arch. I fixed some bugs and have others in my to do list. Thanks you very much. Your patches are against CLN 1.1.13. Could you, please, check out CVS HEAD instead (a.k.a. CLN 1.2-pre)? At least the "\n\t" patch is already in there. $ cvs -d :pserver:anoncvs@cvs.ginac.de:/home/cvs/cln login [enter the password `anoncvs'] $ cvs -d :pserver:anoncvs@cvs.ginac.de:/home/cvs/cln -z 9 co cln A while ago Sven Verdoolaege gave sparc64 a try. But he seems to have given up rather quickly: <http://www.ginac.de/pipermail/cln-list/2006-December/000265.html> > Currently I'm stuck on one in polynomial/elem. It has to do with a > multiplication. > > In src/polynomial/elem/cl_UP_GF2.h, why do you define the gf2_mul_table > only for arches that are not sparc? Calls to this table are made further > for arches that have intDsize=64 (that includes sparc64) and thus error > out. > > I'll further let the error messages and the patches talk for them selfs. > > Please get back to me so we can fix these and make cln available for > OpenBSD. FreeBSD already marked sparc64 as broken arch and NetBSD, > apparently, didn't even test it on this platform. I'm no Sparc expert. To me, some of the patches don't talk for themselves. Can you, please, add short Changelog entries? > ------------------------------------------------------------------------ > > /usr/local/bin/libtool --mode=compile c++ -O2 -pipe -Wformat -Wbounded > -I/usr/local/include -I../include -I../include -I./polynomial -I./base > -I./base/hash -I./base/digitseq -I./base/digit -Ibase -c > ./polynomial/elem/cl_UP.cc > c++ -O2 -pipe -Wformat -Wbounded -I/usr/local/include -I../include > -I../include -I./polynomial -I./base -I./base/hash -I./base/digitseq > -I./base/digit -Ibase -c ./polynomial/elem/cl_UP.cc -fPIC -DPIC -o > .libs/cl_UP.o > In file included from ../include/cln/univpoly.h:6, > from polynomial/elem/cl_UP.cc:11: > ../include/cln/object.h:29:1: warning: "cl_word_alignment" redefined > ../include/cln/object.h:26:1: warning: this is the location of the previous > definition > In file included from polynomial/elem/cl_UP.cc:36: > polynomial/elem/cl_UP_GF2.h: In function `uint64 cln::gf2_mul32_(unsigned > int, unsigned int)': > polynomial/elem/cl_UP_GF2.h:766: error: `gf2_mul_table' undeclared (first > use this function) > polynomial/elem/cl_UP_GF2.h:766: error: (Each undeclared identifier is > reported only once for each function it appears in.) > > > > ------------------------------------------------------------------------ > > $OpenBSD$ > --- src/base/cl_low.h.orig Mon Nov 12 17:43:08 2007 > +++ src/base/cl_low.h Mon Nov 12 17:48:07 2007 > @@ -238,7 +238,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 ar > #define mulu32(x,y,hi_zuweisung,lo_zuweisung) \ > ({ var register uint64 _hi; \ > var register uint64 _lo; \ > - __asm__("umul %2,%3,%1\n\trd %y,%0" \ > + __asm__("umul %2,%3,%1\n\trd %%y,%0" \ > : "=r" (_hi), "=r" (_lo) \ > : "r" ((uint32)(x)), "r" ((uint32)(y)) \ > ); \ > @@ -429,7 +429,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 ar > __asm__ __volatile__ ( \ > "wr %%g0,%%g0,%%y\n\t" \ > "udiv %2,%3,%0\n\t" \ > - "umul %0,%3,%1" \ > + "umul %0,%3,%1\n\t" \ > "sub %2,%1,%1" \ > : "=&r" (__q), "=&r" (__r) \ > : "r" (__x), "r" (__y)); \ > @@ -518,7 +518,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 ar > __asm__ __volatile__ ( \ > "wr %%g0,%%g0,%%y\n\t" \ > "udiv %2,%3,%0\n\t" \ > - "umul %0,%3,%1" \ > + "umul %0,%3,%1\n\t" \ > "sub %2,%1,%1" \ > : "=&r" (__q), "=&r" (__r) \ > : "r" (__x), "r" (__y)); \ > @@ -568,7 +568,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 ar > __asm__ __volatile__ ( \ > "wr %%g0,%%g0,%%y\n\t" \ > "udiv %2,%3,%0\n\t" \ > - "umul %0,%3,%1" \ > + "umul %0,%3,%1\n\t" \ > "sub %2,%1,%1" \ > : "=&r" (__q), "=&r" (__r) \ > : "r" (__x), "r" (__y)); \ > @@ -680,7 +680,7 @@ inline uint32 mulu32_unchecked (uint32 arg1, uint32 ar > __asm__ __volatile__ ( \ > "wr %2,%%g0,%%y\n\t" \ > "udiv %3,%4,%0\n\t" \ > - "umul %0,%4,%1" \ > + "umul %0,%4,%1\n\t" \ > "sub %3,%1,%1" \ > : "=&r" (__q), "=&r" (__r) \ > : "r" (__xhi), "r" (__xlo), "r" (__y)); \ > > > ------------------------------------------------------------------------ > > $OpenBSD$ > --- src/base/digitseq/cl_asm_sparc64_.cc.orig Fri Jan 28 13:37:28 2000 > +++ src/base/digitseq/cl_asm_sparc64_.cc Wed Nov 14 16:42:47 2007 > @@ -42,6 +42,8 @@ > > .seg "text" > > + .register %g2,#scratch > + > .global C(mulu16_),C(mulu32_),C(mulu32_unchecked),C(mulu64_) > .global C(divu_6432_3232_),C(divu_3216_1616_) > .global C(copy_loop_up),C(copy_loop_down),C(fill_loop_up),C(fill_loop_down) Your other patch: > $OpenBSD$ > --- include/cln/object.h.orig Thu Nov 15 11:58:06 2007 > +++ include/cln/object.h Thu Nov 15 11:58:13 2007 > @@ -22,7 +22,7 @@ namespace cln { > #if defined(__m68k__) > #define cl_word_alignment 2 > #endif > -#if defined(__i386__) || defined(__mips__) || defined(__mipsel__) || defined(__sparc__) || defined(__hppa__) || defined(__arm__) || defined(__rs6000__) || defined(__m88k__) || defined(__convex__) || defined(__s390__) > +#if defined(__i386__) || defined(__mips__) || defined(__mipsel__) || (defined(__sparc__) && !defined(__sparc64__))|| defined(__hppa__) || defined(__arm__) || defined(__rs6000__) || defined(__m88k__) || defined(__convex__) || defined(__s390__) > #define cl_word_alignment 4 > #endif > #if defined(__alpha__) || defined(__ia64__) || defined(__mips64__) || defined(__powerpc64__) || defined(__sparc64__) || defined(__x86_64__) > > > > $OpenBSD$ > --- src/polynomial/elem/cl_UP_GF2.h.orig Thu Nov 15 11:51:38 2007 > +++ src/polynomial/elem/cl_UP_GF2.h Thu Nov 15 11:56:45 2007 > @@ -746,10 +746,10 @@ uint32 gf2_mul32 (uint32 x, uint32 y, uint32* plo) > > // Multiply two GF2-polynomials of degree < intDsize. > // Stores the low part, returns the high part. > -#if (intDsize==32) > +#if (intDsize==32) || defined(__sparc64__) > #define gf2_mul_uintD gf2_mul32 > #endif > -#if (intDsize==64) > +#if (intDsize==64) && !defined(__sparc64__) > inline uint64 gf2_mul32_ (uint32 x, uint32 y) > { > var uint16 *ptr; // uint16 ptr[0x10]; Aha, so __sparc64__ *is* defined. Good. The earlier patch by Sven Verdoolaege implied that it wasn't. So, you got it to compile in the end? Did you run make check? What did it report? Looking forward to see CLN fixed on sparc64 -richy. -- Richard B. Kreckel <http://www.ginac.de/~kreckel/>

2 2

Bug in F_2[X] comparisson
by Remco Bloemen 04 Nov '07

04 Nov '07

Dear all, Lately i have been implementing a few algorithms from "Modern Computer Algebra". In the process of implementing Z[x] factorization i noticed the following strange behavior: cl_modint_ring basering = find_modint_ring(2); cl_univpoly_modint_ring ring = find_univpoly_ring(basering); cl_UP_MI x= ring->zero(); (x != ring->zero()) returns false (correct) (x == ring->zero()) returns false (incorrect) If i first compare x == x the behavior changes: (x == x) returns true (correct) (x != ring->zero()) returns true (incorrect) (x == ring->zero()) returns false (incorrect) Attached is a demonstrational program. I'm using cln version 1.1.13 compiled with gcc-4.2.2 on x86 (prescott). Could someone please tell me if this is fixed in cvs or being worked on?

3 5