I don't even know if building a fat binary with MinGW64 is even possible, even in theory.
My guess is 2.5.0 was simply ignoring the --enable-fat flag. As far as I know, building with the 32 bit version of MinGW supports --enable-fat though. But that is no use to you as you want a 64 bit binary. To build a binary that runs on a wide range of processors, you could try ./configure --build=x86_64-w64-mingw32 This should run on all x86_64 processors. This won't be as fast as a true fat binary though. It just selects lowest common denominator assembly files. Bill. On 26 May 2012 00:22, Pavel Holoborodko <pa...@holoborodko.com> wrote: > 2.5.0 compiles fine with --enable-fat with all tests passing. > > Then how I can compile mpir so that binaries will work on wide range > of CPUs? Generic C build, --march, else? Could you teach me? > > Thanks in advance. > Pavel. > > On May 25, 2012, at 22:06, Bill Hart <goodwillh...@googlemail.com> wrote: > >> It looks like you supplied --enable-fat when configuring MPIR. >> >> This is not supported under MinGW. >> >> Bill. >> >> On 25 May 2012 04:03, Pavel Holoborodko <pa...@holoborodko.com> wrote: >>> Thank you for taking this problem under consideration. >>> >>> To see the full picture I've tried to compile MPIR from SVN\trunk. >>> There are many errors coming from Assembler (as was anticipated by Brian). >>> Cause seems to be simple enough to fix, but I have little experience with >>> Assembler. >>> >>> Hope this report will help somehow. Please see error messages at the end of >>> the mail. >>> >>> *** >>> On a very different topic. >>> >>> I've been using MPIR (through MPFR and MPC) to develop special plugin for >>> MATLAB, which provides numerical methods in arbitrary precision for the >>> system. >>> It covers basic math, linear algebra, optimization, ODE, integration, >>> special functions, etc. >>> >>> Product is commercial, full name is Multiprecision Computing Toolbox for >>> MATLAB, website: http://advanpix.com. >>> MPIR library is acknowledged on the main page of the project, as well as in >>> "Info" section (shipped with plugin): >>> http://www.advanpix.com/documentation/mp-info/ >>> >>> I have no idea if this is appropriate, but can I ask you to reference my >>> toolbox from MPIR page? >>> *** >>> >>> Errors coming from Assembler: >>> >>> libtool: compile: x86_64-w64-mingw32-gcc -std=gnu99 -DHAVE_CONFIG_H -I. >>> -I.. -D__GMP_WITHIN_GMP -I.. -DOPERATION_fat_divrem_1 -O2 -m64 -c >>> fat_divrem_1.c -DDLL >>> _EXPORT -DPIC -o .libs/fat_divrem_1.o >>> C:\Users\Pavel\AppData\Local\Temp\ccHSijVV.s: Assembler messages: >>> C:\Users\Pavel\AppData\Local\Temp\ccHSijVV.s:25: Error: incorrect register >>> `C%:e\sUis'e russ\ePda vweilt\hA p`pq' suDfaftiax\ >>> ocal\Temp\ccpHqXIY.s: Assembler messages: >>> C:\Users\Pavel\AppData\Local\Temp\ccpHqXIY.s:23: Erromake[2]: r:*** >>> [fat_divexact_byfobm1.lo] Error 1 >>> ncorrect register `%ebx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccj23bx2.s: Assembler messages: >>> C:\Users\Pavel\AppData\Local\Temp\ccj23bx2.s:31: Error: register type >>> mismatch for `bsf' >>> C:\Users\Pavel\AppData\Local\Temp\ccj23bx2.s:82: Error: incorrect register >>> `%r9d' used with `q'C :s\uUfsfeirxs\P >>> vel\ACp:p\DUastae\rLso\cPaalv\eTle\mApp\pcDcaXt0aE\aLQo5c.asl:\ >>> ATsesmepm\bclcejr2 3mbexs2s.asg:make[2]: e1s1*** [fat_divexact_by3c.lo] >>> Error 1:0 >>> : >>> CE:\Ursreorrs:\ Pavienlc\oArprpeDcatt ar\eLgoicsatle\rT e`m%pr\9cdc'X >>> 0uEsaeQd5 .wsi:t5h3 :` q' Esurfrfoirx: >>> ncorrect register `%ecx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:67: Error: incorrect register >>> `%r8d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:113: Error: incorrect register >>> `%edx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:120: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:121: Error: operand type >>> mismatch for `make[2]: ad*** [fat_divexact_1.lo] Error 1c' >>> >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:128: Error: incorrect register >>> `%r9d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:131: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:132: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:161: Error: incorrect register >>> `%edx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:168: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:169: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:176: Error: incorrect register >>> `%r9d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:179: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:180: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:195: Error: incorrect register >>> `%ecx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:209: Error: incorrect register >>> `%r8d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:225: Error: incorrect register >>> `%esi' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:232: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:239: Error: incorrect register >>> `%r9d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:243: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:290: Error: incorrect register >>> `%ebp' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:311: Error: incorrect register >>> `%edx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:318: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:319: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:326: Error: incorrect register >>> `%r10d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:329: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:330: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:355: Error: incorrect register >>> `%esi' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:362: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:369: Error: incorrect register >>> `%r10d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:373: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:394: Error: incorrect register >>> `%ecx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:410: Error: incorrect register >>> `%r8d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:432: Error: register type >>> mismatch for `bsf' >>> C:\Users\Pavel\AppData\Local\Temp\ccX0EaQ5.s:451: Error: incorrect register >>> `%ebp'libtool: compile: x86_64-w64-mingw32-gcc -std=gnu99 -DHAVE_CONFIG_H >>> -I. -I.. >>> -D__GMP_WITHIN_GMP -I.. -DOPERATION_fat_divrem_2 -O2 -m64 -c fat_divrem_2.c >>> -DDLL_EXPORT -DPIC -o .libs/fat_divrem_2.o u >>> d with `q' suffix >>> make[2]: *** [fat_divrem_1.lo] Error 1 >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s: Assembler messages: >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:42: Error: operand type >>> mismatch for `sub' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:43: Error: operand type >>> mismatch for `sbb' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:88: Error: incorrect register >>> `%edx' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:95: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:96: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:103: Error: incorrect register >>> `%r14d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:109: Error: operand type >>> mismatch for `add' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:110: Error: operand type >>> mismatch for `adc' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:121: Error: incorrect register >>> `%r9d' used with `q' suffix >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:137: Error: operand type >>> mismatch for `sub' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:148: Error: operand type >>> mismatch for `sub' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:149: Error: operand type >>> mismatch for `sbb' >>> C:\Users\Pavel\AppData\Local\Temp\ccG16MSv.s:185: Error: operand type >>> mismatch for `add' >>> make[2]: *** [fat_divrem_2.lo] Error 1 >>> >>> >>> >>> >>> On Thu, May 24, 2012 at 10:19 PM, Bill Hart <goodwillh...@googlemail.com> >>> wrote: >>>> >>>> This does look like a real live bug. Hopefully someone will fix it >>>> soon. Thanks for the report, and sorry for the delay in replying. >>>> >>>> Bill. >>>> >>>> On May 14, 3:45 am, Pavel Holoborodko <pa...@holoborodko.com> wrote: >>>>> Dear All, >>>>> >>>>> This is my first mail to the group. I want to thank developers for >>>>> creating >>>>> this beautiful and very helpful library. >>>>> I've been using it for several years (relying on MSVC build - special >>>>> thanks to Brian Gladman) in my projects. >>>>> >>>>> Now I'm transferring to mingw64 environment, and have encountered >>>>> problem >>>>> in building shared library of MPIR-2.5.1. >>>>> >>>>> I've tried to compile the latest stable version MPIR.2.5.1 using mingw64 >>>>> with following configuration: >>>>> >>>>> ./configure --prefix=/mingw --host=x86_64-w64-mingw32 --enable-fat >>>>> --disable-static --enable-shared --enable-gmpcompat >>>>> >>>>> After that compilation went fine, but linker choked up with: >>>>> >>>>> libtool: link: x86_64-w64-mingw32-gcc -std=gnu99 -shared >>>>> .libs\\libmpir.la.lnkscript -O2 -m64 -Wl,--export-all-symbols >>>>> -Wl,--output-def -Wl,.libs/libmpir-3.dll.def >>>>> -o .libs/libmpir-7.dll -Wl,--enable-auto-image-base -Xlinker >>>>> --out-implib >>>>> -Xlinker .libs/libmpir.dll.a >>>>> Creating library file: .libs/libmpir.dll.a >>>>> *mpn\\.libs\\addmul_2.o:addmul_2.as:(.text+0x0): multiple definition of >>>>> `__gmpn_addmul_2'* >>>>> *mpn\\.libs\\redc_2.o:redc_2.c:(.text+0x0): first defined here* >>>>> collect2: ld returned 1 exit status >>>>> >>>>> *** >>>>> Interestingly, 2.5.0 compiles fine under the same settings. >>>>> >>>>> I would appreciate any help in this regard. >>>>> >>>>> Pavel Holoborodko. >>>>> >>>>> -- >>>>> Multiprecision Computing Toolbox for MATLABhttp://www.advanpix.com/ >>>> >>>> -- >>>> 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. >>>> >>> >>> -- >>> 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. >> >> -- >> 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. >> > > -- > 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. > -- 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.