The reproducer: $ cat foo.c void foo () { #pragma omp target ; } $ cat bar.c void bar () { #pragma omp target ; } $ cat main.c extern void foo (); extern void bar (); int main () { foo (); bar (); return 0; } $ gcc -c -fopenmp -flto foo.c $ gcc -c -fopenmp bar.c main.c $ gcc -fopenmp foo.o bar.o main.o main.o: In function `main': main.c:(.text+0x14): undefined reference to `bar' collect2: error: ld returned 1 exit status This happens because the linker plugin in claim_file_handler claims bar.o, and linker just drops it, because linker considers bar.o as LTO object. Without offload it claims only LTO objects, but now it claims objects with any IR. (Yes, offloading misuses lto-plugin and lto-wrapper a bit.) And even worse, it fails with -foffload=disable, because we decided to stream- out offload IR unconditionally: https://gcc.gnu.org/ml/gcc-patches/2014-10/msg00628.html So -foffload=disable only disables compilation of target images in lto-wrapper, but objects are handled by linker plugin before that. The first solution that comes to mind - do not claim objects, which contain offload IR without LTO IR. But this will cause run-time error: "libgomp: Cannot map target functions or variables (expected 1, have 2)", because lto-wrapper will surround only *.ltrans.o (derived from foo.o) with crtoffload{begin,end}.o; and bar.o will be added at the end of the list of objects from lto-wrapper. But we need this order to get correct host table: "crtoffloadbegin.o, *.ltrans.o, bar.o, crtoffloadend.o". Here is a bit more about tables: https://gcc.gnu.org/wiki/Offloading#Address_mapping_tables Or maybe we should implement new linker offload-plugin with its offload-wrapper, but I don't know how difficult it would be to support 2 plugins in the linkers, and it really doesn't solve the issue with crtoffload{begin,end}.o placement. Or maybe just print an error during linking that offloading doesn't support mixing LTO and non-LTO objects (even if some of them doesn't have offload)?
> Or maybe just print an error during linking that offloading doesn't support > mixing LTO and non-LTO objects (even if some of them doesn't have offload)? That's the worst solution - having non-LTO objects is the whole point of linker-plugin support. I presume the same issue exists for GCC 5.
> I presume the same issue exists for GCC 5. Yes. It seems that we can fix this issue by passing a new option to lto-wrapper, which will contain a list of object files with offload (or a filename with the list). It also will allow to remove some hacky code from lto-wrapper, like this comparison: if (strncmp (argv[i], "-fresolution=", sizeof ("-fresolution=") ... E.g., if there are 4 objects: * obj1.o - non-LTO, offload; * obj2.o - LTO, non-offload; * obj3.o - non-LTO, non-offload; * obj4.o - LTO, offload; then linker plugin will claim only obj2.o and obj4.o, as it was intended. So it will call lto-wrapper by passing obj2.o and obj4.o as argv. But additionally linker plugin will pass something like: -foffload_objects="obj1.o,obj4.o". lto-wrapper will perform LTO on objects from argv as usually, and additionally compile target images using offload IR from obj1.o and obj4.o. The tables also should match, because host table will consist of: pieces from all LTO objects with offload + pieces from non-LTO objects with offload. Just need to reorder offload_objects correspondingly before passing them to the targer compiler (obj4.o,obj1.o). However in this case both obj1.o and obj4.o cannot be surrounded by crtoffload{begin,end}.o, because lto-wrapper cannot place crtoffload* before or after obj1.o, because it is unclaimed. But I guess this can be fixed by something like linker script, which will place sections from crtoffload* at the begin/end of the final joint section.
On Mon, 30 Nov 2015, iverbin at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68463 > > iverbin at gcc dot gnu.org changed: > > What |Removed |Added > ---------------------------------------------------------------------------- > Status|UNCONFIRMED |NEW > Last reconfirmed| |2015-11-30 > CC| |bernds at gcc dot gnu.org, > | |hubicka at gcc dot gnu.org, > | |jakub at gcc dot gnu.org, > | |rguenth at gcc dot gnu.org > Ever confirmed|0 |1 > > --- Comment #2 from iverbin at gcc dot gnu.org --- > > I presume the same issue exists for GCC 5. > Yes. > > It seems that we can fix this issue by passing a new option to lto-wrapper, > which will contain a list of object files with offload (or a filename with the > list). It also will allow to remove some hacky code from lto-wrapper, like > this > comparison: if (strncmp (argv[i], "-fresolution=", sizeof ("-fresolution=") ... > > E.g., if there are 4 objects: > * obj1.o - non-LTO, offload; > * obj2.o - LTO, non-offload; > * obj3.o - non-LTO, non-offload; > * obj4.o - LTO, offload; > > then linker plugin will claim only obj2.o and obj4.o, as it was intended. So > it > will call lto-wrapper by passing obj2.o and obj4.o as argv. But additionally > linker plugin will pass something like: -foffload_objects="obj1.o,obj4.o". > lto-wrapper will perform LTO on objects from argv as usually, and additionally > compile target images using offload IR from obj1.o and obj4.o. > The tables also should match, because host table will consist of: pieces from > all LTO objects with offload + pieces from non-LTO objects with offload. Just > need to reorder offload_objects correspondingly before passing them to the > targer compiler (obj4.o,obj1.o). > However in this case both obj1.o and obj4.o cannot be surrounded by > crtoffload{begin,end}.o, because lto-wrapper cannot place crtoffload* before or > after obj1.o, because it is unclaimed. But I guess this can be fixed by > something like linker script, which will place sections from crtoffload* at the > begin/end of the final joint section. Sounds like a sensible plan to me. Not sure if I like -foffload-objects=a,b,c,d I'd have used a temporary list file and -foffload-objects=/tmp/ccxxxha to avoid quoting issues. Richard.
Author: iverbin Date: Thu Feb 25 12:23:52 2016 New Revision: 233712 URL: https://gcc.gnu.org/viewcvs?rev=233712&root=gcc&view=rev Log: gcc/ PR driver/68463 * config/gnu-user.h (CRTOFFLOADBEGIN): Define. Add crtoffloadbegin.o if offloading is enabled and -fopenacc or -fopenmp is specified. (CRTOFFLOADEND): Likewise. (GNU_USER_TARGET_STARTFILE_SPEC): Add CRTOFFLOADBEGIN. (GNU_USER_TARGET_ENDFILE_SPEC): Add CRTOFFLOADEND. * lto-wrapper.c (offloadbegin, offloadend): Remove static vars. (offload_objects_file_name): New static var. (tool_cleanup): Remove offload_objects_file_name file. (find_offloadbeginend): Replace with ... (find_crtoffloadtable): ... this. (run_gcc): Remove offload_argc and offload_argv. Get offload_objects_file_name from -foffload-objects=... option. Read names of object files with offload from this file, pass them to compile_images_for_offload_targets. Don't call find_offloadbeginend and don't pass offloadbegin and offloadend to the linker. Don't pass offload non-LTO files to the linker, because now they're not claimed. libgcc/ PR driver/68463 * Makefile.in (crtoffloadtable$(objext)): New rule. * configure.ac (extra_parts): Add crtoffloadtable$(objext) if enable_offload_targets is not empty. * configure: Regenerate. * offloadstuff.c: Move __OFFLOAD_TABLE__ from crtoffloadend to crtoffloadtable. libgomp/ PR driver/68463 * testsuite/libgomp.oacc-c-c++-common/parallel-dims-2.c: Remove. lto-plugin/ PR driver/68463 * lto-plugin.c (struct plugin_offload_file): New. (offload_files): Change type. (offload_files_last, offload_files_last_obj): New. (offload_files_last_lto): New. (free_2): Adjust accordingly. (all_symbols_read_handler): Don't add offload files to lto_arg_ptr. Don't call free_1 for offload_files. Write names of object files with offloading to the temporary file. Add new option to lto_arg_ptr. (claim_file_handler): Don't claim file if it contains offload sections without LTO sections. If it contains offload sections, add to the list. Removed: trunk/libgomp/testsuite/libgomp.oacc-c-c++-common/parallel-dims-2.c Modified: trunk/gcc/ChangeLog trunk/gcc/config/gnu-user.h trunk/gcc/lto-wrapper.c trunk/libgcc/ChangeLog trunk/libgcc/Makefile.in trunk/libgcc/configure trunk/libgcc/configure.ac trunk/libgcc/offloadstuff.c trunk/libgomp/ChangeLog trunk/lto-plugin/ChangeLog trunk/lto-plugin/lto-plugin.c
Fixed in GCC 6.
Per <http://news.gmane.org/find-root.php?message_id=%3C573200B5.7050302%40codesourcery.com%3E>: PowerPC GNU/Linux target fixed on trunk in r236098; still needs to be fixed on gcc-6-branch.
Author: jnorris Date: Mon May 16 13:47:47 2016 New Revision: 236287 URL: https://gcc.gnu.org/viewcvs?rev=236287&root=gcc&view=rev Log: Backport from mainline r236098. 2016-05-10 James Norris <jnorris@codesourcery.com> PR driver/68463 * config/rs6000/sysv4.h (CRTOFFLOADBEGIN): Define. Add crtoffloadbegin.o if offloading is enabled and -fopenacc or -fopenmp is specified. (CRTOFFLOADEND): Likewise. (STARTFILE_LINUX_SPEC): Add CRTOFFLOADBEGIN. (ENDFILE_LINUX_SPEC): Add CRTOFFLOADEND. Modified: branches/gcc-6-branch/gcc/ChangeLog branches/gcc-6-branch/gcc/config/rs6000/sysv4.h
Author: jnorris Date: Mon Jun 13 13:17:22 2016 New Revision: 237379 URL: https://gcc.gnu.org/viewcvs?rev=237379&root=gcc&view=rev Log: Backport from mainline r236098. 2016-05-10 James Norris <jnorris@codesourcery.com> PR driver/68463 * config/rs6000/sysv4.h (CRTOFFLOADBEGIN): Define. Add crtoffloadbegin.o if offloading is enabled and -fopenacc or -fopenmp is specified. (CRTOFFLOADEND): Likewise. (STARTFILE_LINUX_SPEC): Add CRTOFFLOADBEGIN. (ENDFILE_LINUX_SPEC): Add CRTOFFLOADEND. Modified: branches/gomp-4_0-branch/gcc/ChangeLog.gomp branches/gomp-4_0-branch/gcc/config/rs6000/sysv4.h
Complete backporting.