I've had a closer look at this, and ICC definitely supports the gas format assembly and the constraints and modifiers that gas does.
But taking a closer look, you say that you are experiencing this with the Intel Compiler for Win32 on an x64 machine. The compiler and assembler need to be matching, so if you only have ICC for Win32 then you have to compile with a 32 bit build project in MSVC. This is exactly what the q modifier is for. It tells the assembler to use 64 bits. But as you say, your assembler is a 32 bit assembler, so it cannot do this. If there are further issues, you could see if Brian can help you with this, now that we know what the problem is. Bill. On 11 June 2015 at 00:15, Bill Hart <goodwillh...@googlemail.com> wrote: > We aren't having any issues with MSYS2 or MSVC, so this appears to be an > ICC only problem. > > Since I don't have a copy of ICC, I would advise taking the issue up with > Intel. > > Their compiler defines __GNUC__ as it is basically supposed to do what GCC > will do. Who knows, they might consider it to be a bug. > > Bill. > > On 9 June 2015 at 17:05, Bill Hart <goodwillh...@googlemail.com> wrote: > >> Actually, I just realised that you are doing the build under MSVC and on >> Windows. So running configure obviously isn't possible. >> >> MSVC is probably using specially prepared versions of longlong.h for MSVC >> on Windows 32. >> >> I'm guessing MSVC just ignores code for GCC. But ICC pretends to be GCC. >> >> Brian is probably right. The code is probably invalid for Windows 32. We >> should be pulling in the longlong.h used by MinGW say, on 32 bit Windows, >> not whatever is currently being pulled in. >> >> This probably means MPIR has also broken on MinGW32 since MPIR-2.6.0. >> >> I will try to build under MinGW32 at some point and see if I can find the >> problem. >> >> In the mean time, if you could send me a copy of longlong.h from both >> MPIR-2.6.0 and MPIR-2.7.0-alpha as it is constituted on this machine by the >> MSVC build process, that would be helpful. >> >> Bill. >> >> On 9 June 2015 at 16:51, Bill Hart <goodwillh...@googlemail.com> wrote: >> >>> longlong.h hasn't changed between 2.6.0 and master. >>> >>> The file longlong.h is also constructed by configure on the machine in >>> question. So if there are x64 constraints in longlong.h, it is because it >>> wasn't configured for the correct machine. >>> >>> If you could compile MPIR from the current master branch (or the >>> MPIR-2.7.0-alpha will do) on the machine in question, from a clean clone of >>> the repository or tarball and send me the following, that would be very >>> helpful: >>> >>> * The exact error messages output by ICC >>> * A copy of config.log >>> * A copy of your longlong.h *after* configure is run on this machine >>> >>> It would also help very much if you could send me a copy of config.log >>> and longlong.h from MPIR-2.6.0 after running configure on the same machine. >>> >>> I don't promise to fix it, but I'll have a go. Note I do not have access >>> to ICC, so will need some assistance with testing possible fixes, etc. >>> >>> Thanks, >>> >>> Bill. >>> >>> On 1 March 2015 at 15:06, Pavel Holoborodko <pa...@holoborodko.com> >>> wrote: >>> >>>> Yes, same ICC builds 2.6.0 in win32 configuration without any issues. >>>> >>>> Actually 2.7.0 also compiles fine in x64 by the same compiler. >>>> Issue appears only in win32 mode. >>>> >>>> As I see from the Brian scripts, x64 uses different longlong.h (?) >>>> >>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "mpir-devel" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to mpir-devel+unsubscr...@googlegroups.com. >>>> To post to this group, send email to mpir-devel@googlegroups.com. >>>> Visit this group at http://groups.google.com/group/mpir-devel. >>>> For more options, visit https://groups.google.com/d/optout. >>>> >>> >>> >> > -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to mpir-devel+unsubscr...@googlegroups.com. To post to this group, send email to mpir-devel@googlegroups.com. Visit this group at http://groups.google.com/group/mpir-devel. For more options, visit https://groups.google.com/d/optout.
