mpir/doc/devel/Configure.txt

Copyright 2008 William Hart

This file is part of the MPIR Library.

The MPIR library is free software; you can redistribute it and/or modify
it under the terms of the GNU Lesser General Public License as published by
the Free Software Foundation; either version 2.1 of the License, or (at your
option) any later version.

The MPIR Library is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY
or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU Lesser General Public
License for more details.

You should have received a copy of the GNU Lesser General Public License
along with the MPIR Library; see the file COPYING.LIB.  If not, write to
the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA
02110-1301, USA.

-define MPIR_COPYRIGHT to GPL text

-Initialise autotools

-Various command line options to configure, including:

--enable-assert enables asserts
--enable_alloca use alloca for memory allocation
--enable-cxx want c++ headers
--enable-fft want FFT for large integer multiplication
--enable-mpbsd enable berkeley mp functions
--enable-nails enable nails
--enable-profiling enable profiling (prof, gprof or instrumentation)

set fomit_frame_pointer to -fomit-frame-pointer unless profiling is gprof

--with-readline (mainly for readline support in calc demo)
--enable-fat build a fat binary

-set tmp_host to $host_cpu with .'s replaced by _'s

-#undef all the HAVE_HOST_CPU possibilities (that will be tested)

-set cclist to "gcc cc"
-set gcc-cflags and gcc_64_cflags to -O3
-set cc-cflags and cc_64_cflags to -O
-set cyclecounter_size 2

-substitute HAVE_HOST_CPU_FAMILY_power and HAVE_HOST_CPU_FAMILY_powerpc with 0

-go through each architecture and set things like:

X86_PATTERN | athlon64-*-* | core2-*-* | x86_64-*-*)
-abilist 32
-cclist gcc icc cc
-gcc_cflags -O2 $fomit_frame_pointer
-gcc_32_cflags_maybe -m32
icc_cflags -no-gcc
icc_cflags_optlist opt
icc_cflags_opt -O3 -O2 -O1
any_32_testlist sizeof-long-4
CALLING_CONVENTIONS_OBJS x86call.lo c86check$U.lo
SPEED_CYCLECOUNTER_OBJ pentium.lo

-set gcc_cflags_cpu and gcc_cflags_arch for each of $host_cpu in i386*, i486*, i586, pentiummmx, i686, pentiumpro, pentium2, pentium3, k6, k62, k63, athlon, i787, pentium4, viac32, viac3*, athlon64, core2, x86_64 (default to i486 options)

-set path to correct mpn directory according to type processor

-override the settings above if a fat binary was selected (path becomes x86/fat)

- for $host_cpu in athlon64, core2, pentium4, x86_64 (i.e. 64 bit machines) set 64 bit only options
cclist_64 gcc
gcc_64_cflags -O2 -m64
gcc_64_cflags_optlist cpu
CALLING_CONVENTIONS_OBJS_6 amd64call.lo amd64check$U.lo
SPEED_CYCLECOUNTER_OBJ_64 x86_64.lo
cyclecounter_size_64 2
abilist 64 32
path x86_64 / x86_64/pentium4 / x86_64/core2
on pentium4 only gcc is available for cclist_64

-set options for none-*-* for generic C only

-if $host in *-*-mingw set gcc_cflags_optlist gcc_cflags_nocygwin appropriately

-write a file(?) AC_FD_CC with User options ABI, CC, CFLAGS, CPPFLAGS, MPN_PATH and GMP: abilist, cclist

-test that the CFLAGS work

-set abi_last to the last thing in abilist

-set ABI to the ABI the user specified if it is a valid option

-set abi1, abi2 to the different possibilities

-set cclist_chosen to the right cclist for this abi and default to what the user specified if they specified one

-set up the CC and CFLAGS and check it works

-check if longlong works for a limb if we are supposed to be using it

-do various test on the CC (not relevant to x86)

-exclude various bad compilers and operating systems that don't save XMM registers properly

-#define HAVE_ABI_32 or 64 according to the ABI

-set extra_functions according to ABI

-set SPEED_CYCLECOUNTER_OBJ and cyclecounter_size according to ABI

-if SPEED_CYCLECOUNTER_OBJ is defined #define HAVE_SPEED_CYCLECOUNTER to cyclecounter_size

-set CALLING_CONVENTIONS_OBJS according to ABI

-if CALLING_CONVENTIONS_OBJS is defined #define #HAVE_CALLING_CONVENTIONS 1

-choose MPN_PATH according to ABI or use the user defined path

-if limbchosen is longlong then DEFN_LONG_LONG_LIMB is set to #define _LONG_LONG_LIMB 1 else to #undef _LONG_LONG_LIMB

-run the standard autoconf things for CC, etc

-GMP_H_ANSI puts C compiler into ANSI mode if possible

-various GMP macros which check whether CC, CPP work for build, PROG_EXTEXT works for build (??), whether ANSI C works, whether LIBM is  available for build

-the assembler is invoked via CC or we use yasm for .as files

-check CPP compiler works and set WANT_CXX if C++ is wanted

-some path setups for the Cray

-if asked for a NAILS build add the nails directory to path

-if there is any sse2 or mmx in the path, remove it if the assembler doesn't support it

-update the file AC_FD_CC with: ABI, CC, CFLAGS, CPPFLAGS, GMP_LDFLAGS, CXX, CXXFLAGS, path

-configuration for windows dlls

-enable CXX in libtool only if we want it, by setting with_tags appropriately

-error if trying to build static and shared libraries at the same time when $libname.a is in $library_names_spec

-check for a whole load of system specific header files, like float.h, sys/time.h, etc

-check various C functions are available

-check various types are available

-AC_C_STRINGIZE 
-AC_C_VOLATILE 
-AC_C_RESTRICT 

-GMP_C_STDARG
-GMP_C_ATTRIBUTE_CONST
-GMP_C_ATTRIBUTE_MALLOC
-GMP_C_ATTRIBUTE_MODE
-GMP_C_ATTRIBUTE_NORETURN
-GMP_H_EXTERN_INLINE

-These various macros above set various things that end up being substituted into files, e.g. all instances of _extern_inline_ get replaced with something like inline where they occur in header files, etc.

-AC_CHECK_LIBM
-AC_SUBST(LIBM)

-GMP_FUNC_ALLOCA
-GMP_OPTION_ALLOCA

-GMP_H_HAVE_FILE

-AC_C_BIGENDIAN sets HAVE_LIMB_BIG_ENDIAN or HAVE_LIMB_LITTLE_ENDIAN

-GMP_C_DOUBLE_FORMAT

-check for a pile of functions known to be missing in certain configurations

-enable C++ tests when required

-create symlinks to the required assembly files in path and link them to mpn/

-take care of functions that can be provided by multifunction files, e.g. aorsmul_1

-define possible mpn suffixes to S s c as

-GMP_FILE_TO_FUNCTION associates the various mpn functions to actual files

-pile of setups for fat binaries

-for each mpn suffix, set tmp_ext to no

-for each gmp_mpn_function do:
GMP_MULFUNC_CHOICES
find the appropriate file
check if it supports the nails specified
add the appropriate object file name to mpn_objects and mpn_objs_in_libgmp
AC_CONFIG_LINKS
add the filename to gmp_srclinks
define the right HAVE_NATIVE_fn's
deal with sqr_basecase if found

-set up autoheader stuff to undef all the HAVE_NATIVE_*

-if assembler is going to be needed, check for M4, or use yasm

-the following macros are for ASM syntax checking and for various macros defined for .asm assembly files
 the macros are defined in acinclude.m4:
  GMP_ASM_TEXT
  GMP_ASM_DATA
  GMP_ASM_LABEL_SUFFIX
  GMP_ASM_GLOBL - global symbol prefix
  GMP_ASM_GLOBL_ATTR
  GMP_ASM_UNDERSCORE
  GMP_ASM_RODATA
  GMP_ASM_TYPE
  GMP_ASM_SIZE
  GMP_ASM_LSYM_PREFIX - local symbol prefix
  GMP_ASM_W32
  GMP_ASM_ALIGN_LOG - whether alignment is logarithmic

-include various m4 files and do various assembly defines per host cpu

-create a link to the appropriate gmp-mparam.h file and set gmp_mparam_source and gmp_srclinks appropriately

-copy SQR_KARATSUBA_THRESHOLD into config.m4

-check size of some types, e.g. unsigned short, and appropriately define BITS_PER_MP_LIMB and SIZEOF_UNSIGNED

-check compiler limb size matches gmp-mparam.h

-define GMP_LIMB_BITS, GMP_NAIL_BITS and GMP_NUMB_BITS

-exclude the mpn random functions from mpbsd

-make the file tune/sqr_basecase.c if needed according to whether sqr_basecase is an assembly or C file

-various configurations for the demo programs

-GMP_FINISH creates config.m4

-create makefiles from the Makefile.in's

This document, configure.txt can be found in /doc/devel of the MPIR source tree.