[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #27 from tejohnson at gcc dot gnu.org --- Author: tejohnson Date: Tue Nov 26 03:48:06 2013 New Revision: 205378 URL: http://gcc.gnu.org/viewcvs?rev=205378root=gccview=rev Log: Backport a variety of profile maintanance fixes from trunk to google/4_8: r203041 r203126 r203152 r203395 r203823 r203830 r204178 r204704 r204756 r197595 198416 202128 r203041 | law | 2013-09-30 13:09:25 -0700 (Mon, 30 Sep 2013) | 7 lines 2013-09-30 Teresa Johnson tejohn...@google.com * tree-ssa-threadupdate.c (ssa_fix_duplicate_block_edges): Update redirected out edge count in joiner case. (ssa_redirect_edges): Common the joiner and non-joiner cases so that joiner case gets profile updates. r203126 | tejohnson | 2013-10-02 12:18:17 -0700 (Wed, 02 Oct 2013) | 18 lines 2013-10-02 Teresa Johnson tejohn...@google.com * dojump.c (do_jump_1): Divide probability between both conditions of a TRUTH_ORIF_EXPR. r203152 | tejohnson | 2013-10-02 22:06:05 -0700 (Wed, 02 Oct 2013) | 8 lines 2013-10-02 Teresa Johnson tejohn...@google.com * predict.c (probably_never_executed): New function. (probably_never_executed_bb_p): Invoke probably_never_executed. (probably_never_executed_edge_p): Ditto. * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges): Treat profile insanities conservatively. r203395 | tejohnson | 2013-10-10 13:30:08 -0700 (Thu, 10 Oct 2013) | 9 lines 2013-10-10 Teresa Johnson tejohn...@google.com * predict.c (tree_estimate_probability): Add new parameter for estimate_bb_frequencies. (estimate_bb_frequencies): Add new parameter to force estimation. (rebuild_frequencies): When max frequency in function is small, recompute counts from frequencies. * predict.h (estimate_bb_frequencies): New parameter. r203823 | tejohnson | 2013-10-18 07:36:53 -0700 (Fri, 18 Oct 2013) | 5 lines 2013-10-18 Teresa Johnson tejohn...@google.com * tree-ssa-tail-merge.c (replace_block_by): Update edge weights during merging. r203830 | tejohnson | 2013-10-18 10:25:44 -0700 (Fri, 18 Oct 2013) | 6 lines 2013-10-18 Teresa Johnson tejohn...@google.com * predict.c (probably_never_executed): Compare frequency-based count to number of training runs. * params.def (UNLIKELY_BB_COUNT_FRACTION): New parameter. r204178 | tejohnson | 2013-10-29 13:39:49 -0700 (Tue, 29 Oct 2013) | 6 lines 2013-10-29 Teresa Johnson tejohn...@google.com PR ipa/58862 * tree-ssa-tail-merge.c (replace_block_by): Tolerate profile insanities when updating probabilities. r204704 | tejohnson | 2013-11-12 07:03:28 -0800 (Tue, 12 Nov 2013) | 12 lines 2013-11-12 Teresa Johnson tejohn...@google.com Jan Hubicka j...@suse.cz * predict.c (drop_profile): New function. (handle_missing_profiles): Ditto. (counts_to_freqs): Don't overwrite estimated frequencies when function has no profile counts. * predict.h (handle_missing_profiles): Declare. * tree-inline.c (freqs_to_counts): New function. (copy_cfg_body): Invoke freqs_to_counts as needed. * tree-profile.c (tree_profiling): Invoke handle_missing_profiles. r204756 | tejohnson | 2013-11-13 13:51:44 -0800 (Wed, 13 Nov 2013) | 6 lines 2013-11-13 Teresa Johnson tejohn...@google.com PR ipa/58862 * predict.c (drop_profile): Error is currently too strict. (handle_missing_profiles): Pass call_count to drop_profile. r197595 | tejohnson | 2013-04-08 10:39:10 -0700 (Mon, 08 Apr 2013) | 60 lines 2013-04-08 Teresa Johnson
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #24 from tejohnson at gcc dot gnu.org --- Author: tejohnson Date: Wed Nov 13 21:48:59 2013 New Revision: 204755 URL: http://gcc.gnu.org/viewcvs?rev=204755root=gccview=rev Log: 2013-11-13 Teresa Johnson tejohn...@google.com PR ipa/58862 * ipa-inline.c (edge_badness): Fix overflow. Modified: trunk/gcc/ChangeLog trunk/gcc/ipa-inline.c
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #25 from tejohnson at gcc dot gnu.org --- Author: tejohnson Date: Wed Nov 13 21:51:44 2013 New Revision: 204756 URL: http://gcc.gnu.org/viewcvs?rev=204756root=gccview=rev Log: 2013-11-13 Teresa Johnson tejohn...@google.com PR ipa/58862 * predict.c (drop_profile): Error is currently too strict. (handle_missing_profiles): Pass call_count to drop_profile. Modified: trunk/gcc/ChangeLog trunk/gcc/predict.c
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #26 from Uroš Bizjak ubizjak at gmail dot com --- Fixed.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #19 from Teresa Johnson tejohnson at google dot com --- On Mon, Nov 4, 2013 at 11:11 AM, tejohnson at google dot com gcc-bugzi...@gcc.gnu.org wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #18 from Teresa Johnson tejohnson at google dot com --- Just hit this same error with cpu2006 bzip2. The .i and .gcda are attached. Reproduce with: gcc -c -fprofile-use -O2 blocksort.i blocksort.c:1136:1: internal compiler error: in edge_badness, at ipa-inline.c:1011 } ^ 0xe13caa edge_badness /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1011 0xe16935 update_edge_key /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1024 0xe16935 inline_small_functions /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1610 0xe16935 ipa_inline /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:2015 0xe16935 execute /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:2385 Please submit a full bug report, Teresa Since this was blocking my SPEC testing I took a look. The issue is that the new edge_count variable is declared as an int and is overflowing. It should be gcov_type. Here is the patch that fixed the bzip2 failure. I will run regression and lto bootstrap testing on it: 2013-11-12 Teresa Johnson tejohn...@google.com PR ipa/58862 * ipa-inline.c (edge_badness): Fix overflow. Index: ipa-inline.c === --- ipa-inline.c(revision 204703) +++ ipa-inline.c(working copy) @@ -909,7 +909,7 @@ edge_badness (struct cgraph_edge *edge, bool dump) /* Capping edge-count to max_count. edge-count can be larger than max_count if an inline adds new edges which increase max_count after max_count is computed. */ - int edge_count = edge-count max_count ? max_count : edge-count; + gcov_type edge_count = edge-count max_count ? max_count : edge-count; sreal_init (relbenefit_real, relbenefit, 0); sreal_init (growth_real, growth, 0); Teresa -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #20 from Paulo J. Matos pa...@matos-sorge.com --- Thanks for fixing this.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #21 from Uroš Bizjak ubizjak at gmail dot com --- Created attachment 31200 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=31200action=edit libiberty patch to push lto-profiledbootstrap further Additional patch to libiberty to fix some c++ casting warnings/errors.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #22 from Uroš Bizjak ubizjak at gmail dot com --- (In reply to Teresa Johnson from comment #19) Since this was blocking my SPEC testing I took a look. The issue is that the new edge_count variable is declared as an int and is overflowing. It should be gcov_type. Here is the patch that fixed the bzip2 failure. I will run regression and lto bootstrap testing on it: You will need additional patch to libiberty to push the lto-profiledbootstrap further. Unfortunately, the lto profiledbootstrap will later die in stagefeedback with: /ssd/uros/gcc-build-profiled/./prev-gcc/xgcc -B/ssd/uros/gcc-build-profiled/./prev-gcc/ -B/usr/local/x86_64-unknown-linux-gnu/bin/ -B/usr/local/x86_64-unknown-linux-gnu/bin/ -B/usr/local/x86_64-unknown-linux-gnu/lib/ -isystem /usr/local/x86_64-unknown-linux-gnu/include -isystem /usr/local/x86_64-unknown-linux-gnu/sys-include-c -DHAVE_CONFIG_H -g -O2 -flto=jobserver -frandom-seed=1 -fprofile-use -I. -I/home/uros/gcc-svn/trunk/libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fpic /home/uros/gcc-svn/trunk/libiberty/pex-unix.c -o pic/pex-unix.o; \ else true; fi if [ x-fpic != x ]; then \ /home/uros/gcc-svn/trunk/libiberty/pex-unix.c:790:1: error: Missing counts for called function pex_child_error.isra.1/75 } ^ gmake[3]: *** [pex-unix.o] Error 1 gmake[3]: *** Waiting for unfinished jobs gmake[2]: *** [all-stagefeedback-libiberty] Error 2 gmake[2]: *** Waiting for unfinished jobs gmake[2]: Leaving directory `/ssd/uros/gcc-build-profiled' gmake[1]: *** [stagefeedback-bubble] Error 2 gmake[1]: Leaving directory `/ssd/uros/gcc-build-profiled' gmake: *** [profiledbootstrap] Error 2
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #23 from Teresa Johnson tejohnson at google dot com --- On Tue, Nov 12, 2013 at 11:06 AM, ubizjak at gmail dot com gcc-bugzi...@gcc.gnu.org wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #22 from Uroš Bizjak ubizjak at gmail dot com --- (In reply to Teresa Johnson from comment #19) Since this was blocking my SPEC testing I took a look. The issue is that the new edge_count variable is declared as an int and is overflowing. It should be gcov_type. Here is the patch that fixed the bzip2 failure. I will run regression and lto bootstrap testing on it: You will need additional patch to libiberty to push the lto-profiledbootstrap further. Unfortunately, the lto profiledbootstrap will later die in stagefeedback with: /ssd/uros/gcc-build-profiled/./prev-gcc/xgcc -B/ssd/uros/gcc-build-profiled/./prev-gcc/ -B/usr/local/x86_64-unknown-linux-gnu/bin/ -B/usr/local/x86_64-unknown-linux-gnu/bin/ -B/usr/local/x86_64-unknown-linux-gnu/lib/ -isystem /usr/local/x86_64-unknown-linux-gnu/include -isystem /usr/local/x86_64-unknown-linux-gnu/sys-include-c -DHAVE_CONFIG_H -g -O2 -flto=jobserver -frandom-seed=1 -fprofile-use -I. -I/home/uros/gcc-svn/trunk/libiberty/../include -W -Wall -Wwrite-strings -Wc++-compat -Wstrict-prototypes -pedantic -fpic /home/uros/gcc-svn/trunk/libiberty/pex-unix.c -o pic/pex-unix.o; \ else true; fi if [ x-fpic != x ]; then \ /home/uros/gcc-svn/trunk/libiberty/pex-unix.c:790:1: error: Missing counts for called function pex_child_error.isra.1/75 } ^ gmake[3]: *** [pex-unix.o] Error 1 gmake[3]: *** Waiting for unfinished jobs gmake[2]: *** [all-stagefeedback-libiberty] Error 2 gmake[2]: *** Waiting for unfinished jobs gmake[2]: Leaving directory `/ssd/uros/gcc-build-profiled' gmake[1]: *** [stagefeedback-bubble] Error 2 gmake[1]: Leaving directory `/ssd/uros/gcc-build-profiled' gmake: *** [profiledbootstrap] Error 2 Yep, I just hit this myself while testing the badness overflow fix. It is due to my own patch and I just sent a fix to gcc-patches for it - this should be made into a warning for now. Teresa -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #16 from Teresa Johnson tejohnson at google dot com --- Created attachment 31154 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=31154action=edit blocksort.i
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #17 from Teresa Johnson tejohnson at google dot com --- Created attachment 31155 -- http://gcc.gnu.org/bugzilla/attachment.cgi?id=31155action=edit blocksort.gcda
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #18 from Teresa Johnson tejohnson at google dot com --- Just hit this same error with cpu2006 bzip2. The .i and .gcda are attached. Reproduce with: gcc -c -fprofile-use -O2 blocksort.i blocksort.c:1136:1: internal compiler error: in edge_badness, at ipa-inline.c:1011 } ^ 0xe13caa edge_badness /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1011 0xe16935 update_edge_key /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1024 0xe16935 inline_small_functions /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:1610 0xe16935 ipa_inline /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:2015 0xe16935 execute /usr/local/google/home/tejohnson/gcc_trunk_1/gcc/ipa-inline.c:2385 Please submit a full bug report, Teresa
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #15 from Uroš Bizjak ubizjak at gmail dot com --- On Sun, Nov 3, 2013 at 6:54 AM, pa...@matos-sorge.com gcc-bugzi...@gcc.gnu.org wrote: --- Comment #14 from Paulo J. Matos pa...@matos-sorge.com --- (In reply to Uroš Bizjak from comment #13) (In reply to Paulo J. Matos from comment #9) Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating. Have you been able to trigger the ICE following the instructions in Comment 12? No, not yet. Is that through a profiledbootstrap on HEAD? Yes, but I was able to trigger the ICE using a couple of days old normal lto1 on the object files.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #13 from Uroš Bizjak ubizjak at gmail dot com --- (In reply to Paulo J. Matos from comment #9) Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating. Have you been able to trigger the ICE following the instructions in Comment 12?
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #14 from Paulo J. Matos pa...@matos-sorge.com --- (In reply to Uroš Bizjak from comment #13) (In reply to Paulo J. Matos from comment #9) Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating. Have you been able to trigger the ICE following the instructions in Comment 12? No, not yet. Is that through a profiledbootstrap on HEAD?
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #9 from Paulo J. Matos pa...@matos-sorge.com --- I didn't manage to reproduce the bug yet. With the git sha before my commit 4bc0f16, I get the following on a profiledbootstrap on x64: insn-opinit.c: In function 'void init_all_optabs(target_optabs*)': insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1437-2606 66380 init_all_optabs (struct target_optabs *optabs) ^ insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1427-2598 66380 insn-opinit.c:1234:1: internal compiler error: verify_flow_info failed Please submit a full bug report, with preprocessed source if appropriate. See http://gcc.gnu.org/bugs.html for instructions. Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #10 from Teresa Johnson tejohnson at google dot com --- On Fri, Nov 1, 2013 at 10:01 AM, pa...@matos-sorge.com gcc-bugzi...@gcc.gnu.org wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #9 from Paulo J. Matos pa...@matos-sorge.com --- I didn't manage to reproduce the bug yet. With the git sha before my commit 4bc0f16, I get the following on a profiledbootstrap on x64: insn-opinit.c: In function 'void init_all_optabs(target_optabs*)': insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1437-2606 66380 init_all_optabs (struct target_optabs *optabs) ^ insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1427-2598 66380 insn-opinit.c:1234:1: internal compiler error: verify_flow_info failed Please submit a full bug report, with preprocessed source if appropriate. See http://gcc.gnu.org/bugs.html for instructions. The above was introduced by r203823, and fixed by r204178. Doesn't show up with a normal profiledbootstrap, but apparently does with an lto-profiledbootstrap (going by the above report, haven't confirmed this myself). Thanks, Teresa Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating. -- You are receiving this mail because: You are on the CC list for the bug.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #11 from Teresa Johnson tejohnson at google dot com --- On Fri, Nov 1, 2013 at 10:14 AM, Teresa Johnson tejohn...@google.com wrote: On Fri, Nov 1, 2013 at 10:01 AM, pa...@matos-sorge.com gcc-bugzi...@gcc.gnu.org wrote: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #9 from Paulo J. Matos pa...@matos-sorge.com --- I didn't manage to reproduce the bug yet. With the git sha before my commit 4bc0f16, I get the following on a profiledbootstrap on x64: insn-opinit.c: In function 'void init_all_optabs(target_optabs*)': insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1437-2606 66380 init_all_optabs (struct target_optabs *optabs) ^ insn-opinit.c:1234:1: error: verify_flow_info: Wrong probability of edge 1427-2598 66380 insn-opinit.c:1234:1: internal compiler error: verify_flow_info failed Please submit a full bug report, with preprocessed source if appropriate. See http://gcc.gnu.org/bugs.html for instructions. The above was introduced by r203823, and fixed by r204178. Doesn't show up with a normal profiledbootstrap, but apparently does with an lto-profiledbootstrap (going by the above report, haven't confirmed this myself). On re-reading this I wasn't very clear. I mean that the current failure on edge badness doesn't show up with a normal profiledbootstrap, but apparently does with an lto-profiledbootstrap (going by the above report, haven't confirmed this myself). The probability issue occurred even with a normal profiledbootstrap, but is fixed now. Teresa Thanks, Teresa Unfortunately running one of these takes a long time so it's a slow process to check it out since as far as I am aware it's not possible to use a parallel build. Do let me know if there's a fast way to build it. I will keep investigating. -- You are receiving this mail because: You are on the CC list for the bug. -- Teresa Johnson | Software Engineer | tejohn...@google.com | 408-460-2413
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #12 from Uroš Bizjak ubizjak at gmail dot com --- Untar the file from the link [1] to a temp dir, go to the gcc directory in the temp dir and type: path-to-gcc-build-dir/gcc/lto1 @ccBysEru.args Reading object files: build/genmddeps.o build/read-md.o build/errors.o .././libiberty/libiberty.a@0x183c82 .././libiberty/libiberty.a@0x1f8142 .././libiberty/libiberty.a@0x207f4e .././libiberty/libiberty.a@0x22e3be .././libiberty/libiberty.a@0x286fce .././libiberty/libiberty.a@0x345dbe .././libiberty/libiberty.a@0x34e386 .././libiberty/libiberty.a@0x3538de .././libiberty/libiberty.a@0x34394a Reading the callgraph Merging declarations Reading summaries Materializing decls: whole-program profile_estimate cp cdtor inlinelto1: internal compiler error: in edge_badness, at ipa-inline.c:1011 0xd598ea edge_badness /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1011 0xd5c575 update_edge_key /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1024 0xd5c575 inline_small_functions /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1610 0xd5c575 ipa_inline /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:2015 0xd5c575 execute /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:2385 Please submit a full bug report, Using gdb: #2 0x00d598eb in edge_badness (edge=0x71a45e38, dump=optimized out) at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1011 1011 gcc_assert (badness = INT_MIN); (gdb) bt #0 internal_error (gmsgid=gmsgid@entry=0x114f198 in %s, at %s:%d) at /home/uros/gcc-svn/trunk/gcc/diagnostic.c:1123 #1 0x00dc7514 in fancy_abort (file=file@entry=0x110da18 /home/uros/gcc-svn/trunk/gcc/ipa-inline.c, line=line@entry=1011, function=function@entry=0x110e2ca edge_badness(cgraph_edge*, bool)::__FUNCTION__ edge_badness) at /home/uros/gcc-svn/trunk/gcc/diagnostic.c:1183 #2 0x00d598eb in edge_badness (edge=0x71a45e38, dump=optimized out) at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1011 #3 0x00d5c576 in update_edge_key (edge=optimized out, heap=optimized out) at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1024 #4 inline_small_functions () at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:1610 #5 ipa_inline () at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:2015 #6 (anonymous namespace)::pass_ipa_inline::execute (this=0x114f198) at /home/uros/gcc-svn/trunk/gcc/ipa-inline.c:2385 #7 0x007fd2ea in execute_one_pass (pass=pass@entry=0x15cdc00) at /home/uros/gcc-svn/trunk/gcc/passes.c:2215 #8 0x007fdb7b in execute_ipa_pass_list (pass=0x15cdc00) at /home/uros/gcc-svn/trunk/gcc/passes.c:2579 (gdb) p badness $1 = -31503367557611520 [1] https://drive.google.com/file/d/0BzMiXQxzb9IONVd2UWcxUDBrblE/edit?usp=sharing
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2013-10-30 Ever confirmed|0 |1 --- Comment #6 from Uroš Bizjak ubizjak at gmail dot com --- There is still lto-profiledbootstrap failure reported at r204180+ [1], after the patch was committed to SVN. [1] http://gcc.gnu.org/ml/gcc-regression/2013-10/msg00317.html
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Teresa Johnson tejohnson at google dot com changed: What|Removed |Added CC||pmatos at gcc dot gnu.org --- Comment #7 from Teresa Johnson tejohnson at google dot com --- This looks like a separate issue from the edge probability issue that I fixed. The edge probability issue was introduced earlier. Almost certainly due to the following change, given that you noticed it at r203899 and it is an error about edge badness: r203897 | pmatos | 2013-10-21 08:41:46 -0700 (Mon, 21 Oct 2013) | 4 lines Changed paths: M /trunk/gcc/ChangeLog M /trunk/gcc/ipa-inline.c * ipa-inline.c (edge_badness): Cap edge-count at max_count for badness calculations. Adding author to cc. Teresa
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Paulo J. Matos pa...@matos-sorge.com changed: What|Removed |Added CC||pa...@matos-sorge.com --- Comment #8 from Paulo J. Matos pa...@matos-sorge.com --- (In reply to Teresa Johnson from comment #7) This looks like a separate issue from the edge probability issue that I fixed. The edge probability issue was introduced earlier. Almost certainly due to the following change, given that you noticed it at r203899 and it is an error about edge badness: r203897 | pmatos | 2013-10-21 08:41:46 -0700 (Mon, 21 Oct 2013) | 4 lines Changed paths: M /trunk/gcc/ChangeLog M /trunk/gcc/ipa-inline.c * ipa-inline.c (edge_badness): Cap edge-count at max_count for badness calculations. Adding author to cc. Teresa Thanks for adding me to CC. I will try to confirm if it was my patch.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Teresa Johnson tejohnson at google dot com changed: What|Removed |Added CC||tejohnson at gcc dot gnu.org, ||tejohnson at google dot com --- Comment #3 from Teresa Johnson tejohnson at google dot com --- I hit the verify_flow_info: Wrong probability of edge error in a profiledbootstrap. I triaged this down to the following commit, made by me: r203823 | tejohnson | 2013-10-18 07:36:53 -0700 (Fri, 18 Oct 2013) | 5 lines Changed paths: M /trunk/gcc/ChangeLog M /trunk/gcc/tree-ssa-tail-merge.c 2013-10-18 Teresa Johnson tejohn...@google.com * tree-ssa-tail-merge.c (replace_block_by): Update edge weights during merging. Investigating right now. Looks like there were profile weight insanities going into this code, that were magnified by the fixes being performed by my change. To avoid the error, I need to adjust the change I made to ensure that the profile insanities don't get propagated into the edge probabilities. In the case I looked at, the edge weight insanities were originally introduced during jump threading. I had previously made a fix to the profile updates being done by that optimization (r203041, committed by law), but there have been a number of changes to that code since then so this needs to be revisited. I will get a fix for the failure though hopefully today. Thanks, Teresa
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862
--- Comment #4 from Teresa Johnson tejohnson at google dot com ---
On Tue, Oct 29, 2013 at 8:05 AM, tejohnson at google dot com
gcc-bugzi...@gcc.gnu.org wrote:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862
Teresa Johnson tejohnson at google dot com changed:
What|Removed |Added
CC||tejohnson at gcc dot gnu.org,
||tejohnson at google dot com
--- Comment #3 from Teresa Johnson tejohnson at google dot com ---
I hit the verify_flow_info: Wrong probability of edge error in a
profiledbootstrap. I triaged this down to the following commit, made by me:
r203823 | tejohnson | 2013-10-18 07:36:53 -0700 (Fri, 18 Oct 2013) | 5 lines
Changed paths:
M /trunk/gcc/ChangeLog
M /trunk/gcc/tree-ssa-tail-merge.c
2013-10-18 Teresa Johnson tejohn...@google.com
* tree-ssa-tail-merge.c (replace_block_by): Update edge
weights during merging.
Investigating right now. Looks like there were profile weight insanities going
into this code, that were magnified by the fixes being performed by my change.
To avoid the error, I need to adjust the change I made to ensure that the
profile insanities don't get propagated into the edge probabilities.
In the case I looked at, the edge weight insanities were originally introduced
during jump threading. I had previously made a fix to the profile updates
being
done by that optimization (r203041, committed by law), but there have been a
number of changes to that code since then so this needs to be revisited.
I will get a fix for the failure though hopefully today.
This patch fixes the profiledbootstrap errors I was looking at.
Testing will a full profiledbootstrap and a regular
bootstrap/regression test run.
2013-10-29 Teresa Johnson tejohn...@google.com
* tree-ssa-tail-merge.c (replace_block_by): Tolerate profile
insanities when updating probabilities.
Index: tree-ssa-tail-merge.c
===
--- tree-ssa-tail-merge.c (revision 204166)
+++ tree-ssa-tail-merge.c (working copy)
@@ -1467,7 +1467,7 @@ static void
replace_block_by (basic_block bb1, basic_block bb2)
{
edge pred_edge;
- edge e1;
+ edge e1, e2;
edge_iterator ei;
unsigned int i;
gimple bb2_phi;
@@ -1502,16 +1502,22 @@ replace_block_by (basic_block bb1, basic_block bb2
bb2-count += bb1-count;
/* Merge the outgoing edge counts from bb1 onto bb2. */
+ gcov_type out_sum = 0;
FOR_EACH_EDGE (e1, ei, bb1-succs)
{
- edge e2;
e2 = find_edge (bb2, e1-dest);
gcc_assert (e2);
e2-count += e1-count;
- /* Recompute the probability from the new merged edge count (bb2-count
- was updated above). */
- e2-probability = GCOV_COMPUTE_SCALE (e2-count, bb2-count);
+ out_sum += e2-count;
}
+ /* Recompute the edge probabilities from the new merged edge count.
+ Use the sum of the new merged edge counts computed above instead
+ of bb2's merged count, in case there are profile count insanities
+ making the bb count inconsistent with the edge weights. */
+ FOR_EACH_EDGE (e2, ei, bb2-succs)
+{
+ e2-probability = GCOV_COMPUTE_SCALE (e2-count, out_sum);
+}
/* Do updates that use bb1, before deleting bb1. */
release_last_vdef (bb1);
Thanks,
Teresa
--
You are receiving this mail because:
You are on the CC list for the bug.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 --- Comment #5 from tejohnson at gcc dot gnu.org --- Author: tejohnson Date: Tue Oct 29 20:39:49 2013 New Revision: 204178 URL: http://gcc.gnu.org/viewcvs?rev=204178root=gccview=rev Log: 2013-10-29 Teresa Johnson tejohn...@google.com PR ipa/58862 * tree-ssa-tail-merge.c (replace_block_by): Tolerate profile insanities when updating probabilities. Modified: trunk/gcc/ChangeLog trunk/gcc/tree-ssa-tail-merge.c
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Uroš Bizjak ubizjak at gmail dot com changed: What|Removed |Added CC||hubicka at gcc dot gnu.org Summary|LTO profiledbootstrap |[4.9 Regression] LTO |failure: lto1: ICE in |profiledbootstrap failure: |edge_badness, at|lto1: ICE in edge_badness, |ipa-inline.c:1008 |at ipa-inline.c:1008 --- Comment #2 from Uroš Bizjak ubizjak at gmail dot com --- The failure perist as of revision 204007. Adding CC.
[Bug ipa/58862] [4.9 Regression] LTO profiledbootstrap failure: lto1: ICE in edge_badness, at ipa-inline.c:1008
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58862 Richard Biener rguenth at gcc dot gnu.org changed: What|Removed |Added Target Milestone|--- |4.9.0
