I found that gcc-built clang 3.9 does not have the ABI issue anymore, so I'm just using that now.
Dmitry Babokin <[email protected]> schrieb am So., 4. Sep. 2016 um 19:53 Uhr: > Not sure if this is gcc 5 ABI Tag issue, but it's definetly some kind of > ABI incompatibility between gcc and clang. So if you see this issue - > either use clang self build or use gcc for ispc build. > > SKX patch is not need for 3.9, it was back ported from 3.9. > > On Sun, Sep 4, 2016 at 1:39 PM, Aristid Breitkreuz <[email protected]> > wrote: > >> Do I understand it correctly that it's the GCC 5 ABI Tag issue? >> >> If so, using Clang 3.9 should also work, right? >> >> Also, is the SKX patch for LLVM 3.8 still applicable to 3.9? >> >> >> Cheers, >> >> Aristid >> >> >> On Saturday, July 9, 2016 at 3:47:23 PM UTC+2, Dmitry Babokin wrote: >>> >>> I've reproduced the problem. >>> >>> It comes from some incompatibility of clang (which is used by default to >>> build ispc) and gcc (which is used to build the clang itself). To fix the >>> problem you have to either build clang with clang (basically add >>> --selfbuild to alloy.py command line) or build ispc with gcc (i.e. run >>> "make gcc", not just "make"). This is something new, it used to work well. >>> >>> >>> >>> On Sat, Jul 9, 2016 at 12:06 AM, steve heistand <[email protected]> >>> wrote: >>> >>>> fyi with the release of 1.9, I grabbed the 1.9.1 tree (using git >>>> https://github.com/ispc/ispc.git returns 1.9.2dev btw) >>>> >>>> still had the same problem, as did 1.9.2dev. When I picked llvm version >>>> 3.8. >>>> >>>> if I use alloy.py --version=trunk both 1.9 versions compile fine. >>>> >>>> 1.8.X fails on trying to build llvm with the trunk option. >>>> >>>> so Im good at this point. you might want to update what make in the >>>> ispc directory returns when >>>> >>>> llvm_home is defined not appropriately. >>>> >>>> >>>> thanks >>>> >>>> >>>> s >>>> >>>> >>>> <https://github.com/ispc/ispc.git> >>>> >>>> On 07/08/2016 10:38 AM, Dmitry Babokin wrote: >>>> > Everything looks perfectly correct. And your ispc sources are based >>>> on fresh https://github.com/ispc/ispc.git, right? >>>> > >>>> > On Fri, Jul 8, 2016 at 8:16 PM, Steve Heistand <[email protected] >>>> <mailto:[email protected]>> wrote: >>>> > >>>> > cd /usr/local/src/ispc >>>> > export ISPC_HOME=`pwd` >>>> > export LLVM_HOME=`pwd` >>>> > ./alloy.py -b --version=3.8 >>>> > export PATH=$LLVM_HOME/bin-3.8/bin:$PATH >>>> > which clang >>>> > /usr/local/src/ispc/bin-3.8/bin/clang >>>> > ./check_env.py >>>> > === in PATH: === >>>> > Tools: >>>> > m4 (GNU M4) 1.4.17 >>>> > bison (GNU Bison) 3.0.4 >>>> > flex 2.6.0 >>>> > Tools' versions are ok >>>> > >>>> > SDE: >>>> > Warning: you don't have sde >>>> > >>>> > ISPC: >>>> > Warning: you don't have ispc >>>> > >>>> > C/C++ compilers: >>>> > clang version 3.8.0 (tags/RELEASE_380/final 274880) >>>> > gcc (Ubuntu 5.3.1-14ubuntu2.1) 5.3.1 20160413 >>>> > Warning: you don't have icc >>>> > >>>> > === in ISPC specific environment variables: === >>>> > Your LLVM_HOME:/usr/local/src/ispc >>>> > Your ISPC_HOME:/usr/local/src/ispc >>>> > Warning: you don't have ISPC in your ISPC_HOME >>>> > Warning: You have no SDE_HOME >>>> > >>>> > >>>> > On 07/08/2016 10:00 AM, Dmitry Babokin wrote: >>>> > > That's weird. Can you run check_env.py and copy its output here? >>>> > > >>>> > > I was going to build a docker image with ispc build environment >>>> (including clang), this may probably help debugging your issue. I think >>>> I'll >>>> > > build it over the weekend. >>>> > > >>>> > > On Fri, Jul 8, 2016 at 7:52 PM, Steve Heistand < >>>> [email protected] <mailto:[email protected]> <mailto: >>>> [email protected] <mailto:[email protected]>>> wrote: >>>> > > >>>> > > I do get the clang (and llvm) that is built with the >>>> alloy.py script. >>>> > > even trying older versions of llvm/clang/ispc result in the >>>> end with the same error. >>>> > > At the moment I using a build script that has worked for >>>> other, its using a really >>>> > > old 3.6.2 version of llvm and the latest ispc. will see how >>>> that goes. >>>> > > >>>> > > s >>>> > > >>>> > > >>>> > > On 07/08/2016 09:49 AM, Dmitry Babokin wrote: >>>> > > > Hi, >>>> > > > >>>> > > > Alloy.py build should work, but not sure about other >>>> options. Just to make sure - have you checked that you correctly added >>>> clang built by >>>> > > > alloy.py to your path? I.e, "which clang" points to newly >>>> built clang. >>>> > > > >>>> > > > Dmitry. >>>> > > > >>>> > > > On Fri, Jul 8, 2016 at 7:29 AM, Steve Heistand < >>>> [email protected] <mailto:[email protected]> <mailto: >>>> [email protected] <mailto:[email protected]>> <mailto: >>>> [email protected] <mailto:[email protected]> <mailto: >>>> [email protected] <mailto:[email protected]>>>> wrote: >>>> > > > >>>> > > > Hi folks, >>>> > > > Im running into some issues trying to compile ispc on >>>> a linux machine. >>>> > > > the ispc build process goes nicely up until: >>>> > > > ... >>>> > > > Compiling objs/lex.cpp >>>> > > > Creating ispc executable >>>> > > > objs/module.o: In function >>>> `Module::execPreprocessor(char const*, llvm::raw_string_ostream*) const': >>>> > > > module.cpp:(.text+0xc36): undefined reference to >>>> `llvm::sys::getDefaultTargetTriple()' >>>> > > > objs/ctx.o: In function >>>> `FunctionEmitContext::BroadcastValue(llvm::Value*, llvm::Type*, char >>>> const*)': >>>> > > > ctx.cpp:(.text+0x71b6): undefined reference to >>>> `llvm::Twine::str() const' >>>> > > > objs/ispc.o: In function `Target::GetTripleString() >>>> const': >>>> > > > ispc.cpp:(.text+0x2211): undefined reference to >>>> `llvm::sys::getDefaultTargetTriple()' >>>> > > > objs/opt.o: In function >>>> `IntrinsicsOpt::runOnBasicBlock(llvm::BasicBlock&)': >>>> > > > opt.cpp:(.text+0x1a6a): undefined reference to >>>> `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> llvm::ArrayRef<llvm::Type*>)' >>>> > > > opt.cpp:(.text+0x1af6): undefined reference to >>>> `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> llvm::ArrayRef<llvm::Type*>)' >>>> > > > opt.cpp:(.text+0x1bd2): undefined reference to >>>> `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> llvm::ArrayRef<llvm::Type*>)' >>>> > > > opt.cpp:(.text+0x1c5e): undefined reference to >>>> `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> llvm::ArrayRef<llvm::Type*>)' >>>> > > > opt.cpp:(.text+0x1d22): undefined reference to >>>> `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> llvm::ArrayRef<llvm::Type*>)' >>>> > > > objs/opt.o:opt.cpp:(.text+0x1dd7): more undefined >>>> references to `llvm::Intrinsic::getName(llvm::Intrinsic::ID, >>>> > > > llvm::ArrayRef<llvm::Type*>)' follow >>>> > > > clang-3.8: error: linker command failed with exit >>>> code 1 (use -v to see invocation) >>>> > > > Makefile:264: recipe for target 'ispc' failed >>>> > > > >>>> > > > Ive tried this using the llvm/clang build with the >>>> alloy.py script, Ive tried with the binary versions apt-get installs. Ive >>>> build >>>> > > > llvm/clang from the tarballs >>>> > > > all give the same result. All are 3.8 versions if it >>>> matters. and version 5 of gcc which might matter more then anything. I >>>> have Intel >>>> > > > compilers on the machine >>>> > > > but they fail in the build process way earlier. >>>> > > > >>>> > > > thoughts/suggestions? >>>> > > > >>>> > > > thanks >>>> > > > >>>> > > > steve >>>> > > > >>>> > > > >>>> > > > -- >>>> > > > You received this message because you are subscribed >>>> to the Google Groups "Intel SPMD Program Compiler Users" group. >>>> > > > To unsubscribe from this group and stop receiving >>>> emails from it, send an email to [email protected] >>>> <mailto:ispc-users%[email protected]> <mailto: >>>> ispc-users%[email protected] <mailto: >>>> ispc-users%[email protected]>> >>>> > > > <mailto:[email protected] >>>> <mailto:ispc-users%[email protected]> <mailto: >>>> ispc-users%[email protected] <mailto: >>>> ispc-users%[email protected]>>>. >>>> > > > For more options, visit >>>> https://groups.google.com/d/optout. >>>> > > > >>>> > > > >>>> > > > -- >>>> > > > You received this message because you are subscribed to >>>> the Google Groups "Intel SPMD Program Compiler Users" group. >>>> > > > To unsubscribe from this group and stop receiving emails >>>> from it, send an email to [email protected] <mailto: >>>> ispc-users%[email protected]> <mailto: >>>> ispc-users%[email protected] <mailto: >>>> ispc-users%[email protected]>> >>>> > > > <mailto:[email protected] <mailto: >>>> ispc-users%[email protected]> <mailto: >>>> ispc-users%[email protected] <mailto: >>>> ispc-users%[email protected]>>>. >>>> > > > For more options, visit >>>> https://groups.google.com/d/optout. >>>> > > >>>> > > -- >>>> > > "Why is it so hot inside this handbasket?" >>>> > > -- >>>> > > Steve Heistand [email protected] <mailto: >>>> [email protected]> <mailto:[email protected] <mailto: >>>> [email protected]>> >>>> > > -- >>>> > > >>>> > > -- >>>> > > You received this message because you are subscribed to the >>>> Google Groups "Intel SPMD Program Compiler Users" group. >>>> > > To unsubscribe from this group and stop receiving emails >>>> from it, send an email to [email protected] <mailto: >>>> ispc-users%[email protected]> >>>> > > <mailto:ispc-users%[email protected] <mailto: >>>> ispc-users%[email protected]>>. >>>> > > For more options, visit https://groups.google.com/d/optout. >>>> > > >>>> > > >>>> > > -- >>>> > > You received this message because you are subscribed to the >>>> Google Groups "Intel SPMD Program Compiler Users" group. >>>> > > To unsubscribe from this group and stop receiving emails from >>>> it, send an email to [email protected] <mailto: >>>> ispc-users%[email protected]> >>>> > > <mailto:[email protected] <mailto: >>>> ispc-users%[email protected]>>. >>>> > > For more options, visit https://groups.google.com/d/optout. >>>> > >>>> > -- >>>> > "Why is it so hot inside this handbasket?" >>>> > -- >>>> > Steve Heistand [email protected] <mailto: >>>> [email protected]> >>>> > -- >>>> > >>>> > -- >>>> > You received this message because you are subscribed to the >>>> Google Groups "Intel SPMD Program Compiler Users" group. >>>> > To unsubscribe from this group and stop receiving emails from it, >>>> send an email to [email protected] <mailto: >>>> ispc-users%[email protected]>. >>>> > For more options, visit https://groups.google.com/d/optout. >>>> > >>>> > >>>> > -- >>>> > You received this message because you are subscribed to the Google >>>> Groups "Intel SPMD Program Compiler Users" group. >>>> > To unsubscribe from this group and stop receiving emails from it, >>>> send an email to [email protected] <mailto: >>>> [email protected]>. >>>> > For more options, visit https://groups.google.com/d/optout. >>>> >>>> -- >>>> -- >>>> "Why is it so hot inside this handbasket?" >>>> -- >>>> Steve Heistand [email protected] >>>> >>>> >>>> -- >>>> You received this message because you are subscribed to the Google >>>> Groups "Intel SPMD Program Compiler Users" group. >>>> To unsubscribe from this group and stop receiving emails from it, send >>>> an email to [email protected]. >>>> For more options, visit https://groups.google.com/d/optout. >>>> >>> >>> -- >> You received this message because you are subscribed to the Google Groups >> "Intel SPMD Program Compiler Users" group. >> > To unsubscribe from this group and stop receiving emails from it, send an >> email to [email protected]. >> > >> For more options, visit https://groups.google.com/d/optout. >> > -- > You received this message because you are subscribed to a topic in the > Google Groups "Intel SPMD Program Compiler Users" group. > To unsubscribe from this topic, visit > https://groups.google.com/d/topic/ispc-users/luWwKgGE1M0/unsubscribe. > To unsubscribe from this group and all its topics, send an email to > [email protected]. > For more options, visit https://groups.google.com/d/optout. > -- You received this message because you are subscribed to the Google Groups "Intel SPMD Program Compiler Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
