[Bug c++/84197] Segmentation fault when setting -g
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84197 Richard Biener changed: What|Removed |Added Keywords||ice-on-valid-code Status|UNCONFIRMED |NEW Last reconfirmed||2018-02-05 Ever confirmed|0 |1 Known to fail||7.3.1, 8.0 --- Comment #1 from Richard Biener --- Confirmed.
[Bug c++/84196] [6/7/8 Regression] invalid call to a function template with a vector argument silently accepted
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84196 Richard Biener changed: What|Removed |Added Target Milestone|--- |6.5
[Bug c++/70401] [c++1z on mingw]compile variadic template failed
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70401 Paolo Carlini changed: What|Removed |Added CC||jyong at gcc dot gnu.org --- Comment #2 from Paolo Carlini --- Let's add Jonathan in CC.
[Bug c++/84192] [7/8 Regression] ICE with statement expression
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84192 Richard Biener changed: What|Removed |Added Version|unknown |7.3.1 Target Milestone|--- |7.4
[Bug c/84190] [7/8 Regression] double arithmetic on x86 no longer rounds to nearest
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84190 Richard Biener changed: What|Removed |Added Keywords||wrong-code Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2018-02-05 Target Milestone|--- |7.4 Summary|[7 Regression] double |[7/8 Regression] double |arithmetic on x86 no longer |arithmetic on x86 no longer |rounds to nearest |rounds to nearest Ever confirmed|0 |1 --- Comment #6 from Richard Biener --- Confirmed. We somehow lose the volatile qualifiers. I'll have a look.
[Bug c++/82782] ICE: nested template alias and specialized template with auto template parameter
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82782 Paolo Carlini changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED Target Milestone|--- |8.0 --- Comment #3 from Paolo Carlini --- Fixed for 8.1.0.
[Bug c++/82782] ICE: nested template alias and specialized template with auto template parameter
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82782 --- Comment #2 from paolo at gcc dot gnu.org --- Author: paolo Date: Mon Feb 5 11:15:55 2018 New Revision: 257388 URL: https://gcc.gnu.org/viewcvs?rev=257388=gcc=rev Log: 2018-02-05 Paolo CarliniPR c++/82782 * g++.dg/cpp1z/inline-var4.C: New. Added: trunk/gcc/testsuite/g++.dg/cpp1z/inline-var4.C Modified: trunk/gcc/testsuite/ChangeLog
[Bug tree-optimization/84188] assume non-null malloc pointers are distinct
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84188 Richard Biener changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |DUPLICATE --- Comment #2 from Richard Biener --- Dup. *** This bug has been marked as a duplicate of bug 13962 ***
[Bug tree-optimization/13962] [tree-ssa] make "fold" use alias information to optimize pointer comparisons
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=13962 --- Comment #14 from Richard Biener --- *** Bug 84188 has been marked as a duplicate of this bug. ***
[Bug c/84179] -save-temps breaks implicit-fallthrough comment heuristic
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84179 Richard Biener changed: What|Removed |Added Keywords||diagnostic Version|unknown |7.2.0 --- Comment #1 from Richard Biener --- I think we have a dup for this.
[Bug target/27855] [6/7/8 regression] reassociation causes the RA to be confused
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 Aldy Hernandez changed: What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED --- Comment #58 from Aldy Hernandez --- (In reply to rguent...@suse.de from comment #57) > > IMO, "manually optimized" doesn't seem to merit spending more time on this. > > As > > you mention, not real world enough :). May I suggest WONTFIX, but willing > > to > > reopen if someone cares enough to un-obfuscate the code? > > I'd say FIXED - that sounds nicer ;) My pleasure!
[Bug target/27827] [4.0 Regression] gcc 4 produces worse x87 code on all platforms than gcc 3
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27827 Bug 27827 depends on bug 27855, which changed state. Bug 27855 Summary: [6/7/8 regression] reassociation causes the RA to be confused https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 What|Removed |Added Status|NEW |RESOLVED Resolution|--- |FIXED
[Bug target/84172] option "-O3" create slower code
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84172 Richard Biener changed: What|Removed |Added Status|UNCONFIRMED |RESOLVED Resolution|--- |FIXED --- Comment #10 from Richard Biener --- Indeed fixed in GCC 7.
[Bug c++/84171] [8 Regression] ICE with -Wsign-compare
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84171 Richard Biener changed: What|Removed |Added Priority|P3 |P4
[Bug target/27855] [6/7/8 regression] reassociation causes the RA to be confused
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 --- Comment #57 from rguenther at suse dot de --- On Mon, 5 Feb 2018, aldyh at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 > > --- Comment #56 from Aldy Hernandez --- > (In reply to Richard Biener from comment #55) > > > Note the original report used -O and Aldhy used -O2 but we are talking > > about a benchmark and when you use -ffast-math you also use -O3. > > Thanks, I will keep a note of this for next time. FWIW, I used the same flags > as the last time this was reconfirmed, which was comment #52. I also saw a > bunch of runs at -O2 by Uros, which likely influenced me. > > > The benchmark is somewhat badly written (manually "optimized") so our > > vectorization attempts fail. > > > > Overall conclusion is I'm unsure if it's worth pursuing this bug further? > > There is a register pressure issue left but the testcase maybe not > > real-world enough? That is, I would usually recommend to first un-obfuscate > > the manually optimized code. > > IMO, "manually optimized" doesn't seem to merit spending more time on this. > As > you mention, not real world enough :). May I suggest WONTFIX, but willing to > reopen if someone cares enough to un-obfuscate the code? I'd say FIXED - that sounds nicer ;)
[Bug ipa/57218] [6/7/8 Regression] Excessive inlining even at -Os
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57218 --- Comment #13 from Aldy Hernandez --- (In reply to Jan Hubicka from comment #12) > Set component as IPA so it stays at my radar. It is probably too late for > GCC 8 but I will try to finally take a look next stage1. Inlining those > constructions is often beneficial because things subsequently simplifies. > We would probably need bit more context dependent metrics than just single > call window we do now. This however is hard to do becuase that makes number > of inlining decisions to explode. Thanks Jan. What's the protocol here, do we move the target milestone to GCC 9, or do we leave it as is? Thanks.
[Bug lto/81004] [7/8 Regression] linking failed with -flto and static libboost_program_options
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81004 --- Comment #20 from Martin Liška --- It's really fixed on trunk since r257023.
[Bug target/27855] [6/7/8 regression] reassociation causes the RA to be confused
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 --- Comment #56 from Aldy Hernandez --- (In reply to Richard Biener from comment #55) > Note the original report used -O and Aldhy used -O2 but we are talking > about a benchmark and when you use -ffast-math you also use -O3. Thanks, I will keep a note of this for next time. FWIW, I used the same flags as the last time this was reconfirmed, which was comment #52. I also saw a bunch of runs at -O2 by Uros, which likely influenced me. > The benchmark is somewhat badly written (manually "optimized") so our > vectorization attempts fail. > > Overall conclusion is I'm unsure if it's worth pursuing this bug further? > There is a register pressure issue left but the testcase maybe not > real-world enough? That is, I would usually recommend to first un-obfuscate > the manually optimized code. IMO, "manually optimized" doesn't seem to merit spending more time on this. As you mention, not real world enough :). May I suggest WONTFIX, but willing to reopen if someone cares enough to un-obfuscate the code?
[Bug sanitizer/84210] New: __ubsan_handle_builtin_unreachable shoun't be const
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84210 Bug ID: 84210 Summary: __ubsan_handle_builtin_unreachable shoun't be const Product: gcc Version: 8.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: sanitizer Assignee: unassigned at gcc dot gnu.org Reporter: ryabinin.a.a at gmail dot com CC: dodji at gcc dot gnu.org, dvyukov at gcc dot gnu.org, jakub at gcc dot gnu.org, kcc at gcc dot gnu.org, marxin at gcc dot gnu.org Target Milestone: --- Compiling the linux kernel with gcc-8 gives this warning: lib/ubsan.c:432:1: error: ignoring attribute 'noreturn' in declaration of a built-in function '__ubsan_handle_builtin_unreachable' because it conflicts with attribute 'const' [-Werror=attributes] It seems that GCC defines __ubsan_handle_builtin_unreachable with noreturn and const attributes: DEF_SANITIZER_BUILTIN(BUILT_IN_UBSAN_HANDLE_BUILTIN_UNREACHABLE, "__ubsan_handle_builtin_unreachable", BT_FN_VOID_PTR, ATTR_COLD_CONST_NORETURN_NOTHROW_LEAF_LIST) And const doesn't make any sense for function that returns void or doesn't return at all. Shouldn't it be just ATTR_COLD_NORETURN_NOTHROW_LEAF_LIST ?
[Bug target/84209] [avr] Don't split SP in split2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84209 Georg-Johann Lay changed: What|Removed |Added Keywords||wrong-code Target||avr Priority|P3 |P4 Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2018-02-05 Assignee|unassigned at gcc dot gnu.org |gjl at gcc dot gnu.org Ever confirmed|0 |1
[Bug target/84209] New: [avr] Don't split SP in split2
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84209 Bug ID: 84209 Summary: [avr] Don't split SP in split2 Product: gcc Version: 7.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: target Assignee: unassigned at gcc dot gnu.org Reporter: gjl at gcc dot gnu.org Target Milestone: --- Since r242907 there is a split for HI moved that might also split SP. This might lead to wrong code. Using -fdisable-rtl-split2 might do as a work-around.
[Bug lto/70929] [6/7/8 regression] Cross-module inlining for functions having argument passed by reference is no longer working.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70929 --- Comment #8 from rguenther at suse dot de --- On Mon, 5 Feb 2018, hubicka at gcc dot gnu.org wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70929 > > --- Comment #7 from Jan Hubicka --- > Hmm, this actually looks reasonable for me too. Shall I give it a try and > figure out what incompatibilities we get here? That would be nice. Maybe throw it at firefox and dump somewhere when the two variants decide differently? > What about K prototypes? No idea what types_compatible_p does with them...
[Bug tree-optimization/39612] [6/7/8 Regression] LIM inserts loads from uninitialized local memory
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=39612 Richard Biener changed: What|Removed |Added Last reconfirmed|2009-10-05 20:09:44 |2018-2-5 --- Comment #30 from Richard Biener --- Yes, the warning is gone because we no longer warn in the late VRP pass... re-confirmed for the "missed optimization".
[Bug tree-optimization/38785] [6/7/8 Regression] huge performance regression on EEMBC bitmnp01
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=38785 Richard Biener changed: What|Removed |Added Status|REOPENED|NEW Last reconfirmed|2010-02-19 12:17:54 |2018-2-5 --- Comment #49 from Richard Biener --- The testcase from comment#3 still reproduces the issue for me: [local count: 955630223]: # cstore_55 = PHI <2147483647(4), 0(5)> # prephitmp_87 = PHI <3221225470(4), 1073741823(5)> # prephitmp_88 = PHI <3937053352(4), 1789569705(5)> ... many more... # prephitmp_116 = PHI <2934894317(4), 787410670(5)> # prephitmp_117 = PHI <2505397588(4), 357913941(5)> MEM[(void *)_3] = cstore_55; _7 = *_4; ... There's the old issue of PRE lacking a cost model (and undoing PRE being impossible for these cases).
[Bug gcov-profile/84137] Typo in gcov online documentation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84137 Martin Liška changed: What|Removed |Added Known to work||8.0 Known to fail||6.4.0, 7.3.0 --- Comment #3 from Martin Liška --- Fixed on trunk, queued for backports.
[Bug gcov-profile/84137] Typo in gcov online documentation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84137 --- Comment #2 from Martin Liška --- Author: marxin Date: Mon Feb 5 09:59:16 2018 New Revision: 257384 URL: https://gcc.gnu.org/viewcvs?rev=257384=gcc=rev Log: Fix GCOV documentation (PR gcov-profile/84137). 2018-02-05 Martin LiskaPR gcov-profile/84137 * doc/gcov.texi: Fix typo in documentation. Modified: trunk/gcc/ChangeLog trunk/gcc/doc/gcov.texi
[Bug target/27855] [6/7/8 regression] reassociation causes the RA to be confused
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=27855 --- Comment #55 from Richard Biener --- I think the original report is about x87 math vs. SSE math. It's a bit hard to benchmark this through the releases given changes in tuning and vector feature sets (-march=native is out of the question). So I use -O3 -ffast-math -DREPS=10 -m32 as base and see ISA4.3.6 4.6.4 4.8.5 7.2 -mno-sse 1855 6930 4618 5623 -msse2 -mfpmath=sse 1967 6945 4744 6472 -m64 2977 6917 4935 6205 note I edited the benchmark and put noinline,noclone attributes on the gemm_atlas function. I benchmarked on a broadwell system with minimal CPU frequency boosting but still varying REPS varies the reported MFLOPS _a lot_ (but individual runs are somewhat stable, for the last reported number 6205 I also can get 6331 or 6186). I used the attached benchmark, the cited URL doesn't work anymore. So there's still an appearant regression, the trunk numbers aren't very different from the 7.2 results, the 4.6.4 variant is still fastest and we recovered to current levels with 4.9.4 already (just checked -m64 across all releases). With -march=native I get to new heights obviously because we use things like FMA, AVX, etc. if I add just -mavx to 4.6.4 it's not faster than without but 7.2 improves to 6628 for example (4.6.4 doesn't know AVX2 and -mfma results in bogus assembler being generated...). If I look at the generated code for -m64 (with just SSE) we no longer spill a lot in the inner loop (only once) and we don't vectorize. 4.6.4 manages to avoid any spilling in the computation (even in the outer loop). So the original analysis (RA sucks) still holds. Note the original report used -O and Aldhy used -O2 but we are talking about a benchmark and when you use -ffast-math you also use -O3. Note the biggest regression we still see is with x87 math - I think we can reasonably disregard that now. The benchmark is somewhat badly written (manually "optimized") so our vectorization attempts fail. Overall conclusion is I'm unsure if it's worth pursuing this bug further? There is a register pressure issue left but the testcase maybe not real-world enough? That is, I would usually recommend to first un-obfuscate the manually optimized code.
[Bug lto/81004] [7/8 Regression] linking failed with -flto and static libboost_program_options
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81004 --- Comment #19 from Jan Hubicka --- Weird, for me I see the problem reproducing with $ ~/8-install/bin/g++ --version g++ (GCC) 8.0.1 20180124 (experimental) Copyright (C) 2018 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. but not with $ ~/8-install-slow/bin/g++ --version g++ (GCC) 8.0.1 20180205 (experimental) Copyright (C) 2018 Free Software Foundation, Inc. This is free software; see the source for copying conditions. There is NO warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. so it seems the problem went away in those few days but I do not recall anything related being commit. Martin, would it be possible to bisect this? (hope it is not --enable-checking=release used to build first compiler)
[Bug c++/81917] [6/7/8 Regression] internal compiler error: in finish_member_declaration, at cp/semantics.c:3004
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81917 --- Comment #5 from Paolo Carlini --- This variant causes the same ICE and it's a tad cleaner: template using a = void; templatestruct b { typedef int c; }; template class b ; template ::c> class f; template class g { }; template class h { class i; typedef g j; class i { j k; }; }; typename h ::
[Bug lto/81004] [7/8 Regression] linking failed with -flto and static libboost_program_options
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81004 Jan Hubicka changed: What|Removed |Added CC||hubicka at gcc dot gnu.org --- Comment #18 from Jan Hubicka --- Let me try to understand the problem. In .res file we have: 328 e6522ac3d089e792 PREVAILING_DEF _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED1Ev 333 e6522ac3d089e792 PREVAILING_DEF_IRONLY _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED2Ev We end up privatizing _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED2Ev and keep _ZNSt7__cxx1112basic_stringIcSt11char_traitsIcESaIcEED1Ev external. This seems correct but for some reason we do not emit .globl at the time we produce the alias. This is because TREE_PUBLIC of alias->decl is not set. Wonder why that happens. do_assemble_alias contains: /* Make name accessible from other files, if appropriate. */ if (TREE_PUBLIC (decl) || TREE_PUBLIC (orig_decl)) { globalize_decl (decl); maybe_assemble_visibility (decl); } which looks like it is impossible to produce non-public alias of public symbol that looks quite wrong (we do make use of local aliases to public symbols and we do not want those exported). Perhaps we never realy set TREE_PUBLIC of those aliases as expected?
[Bug gcov-profile/83879] __gcov_dump doesn't work with dlopen-ed libraries
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83879 Martin Liška changed: What|Removed |Added Status|NEW |ASSIGNED Known to work||8.0 Known to fail||6.4.0, 7.3.0 --- Comment #7 from Martin Liška --- Fixed on trunk, queued for backports.
[Bug gcov-profile/83879] __gcov_dump doesn't work with dlopen-ed libraries
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83879 --- Comment #6 from Martin Liška --- Author: marxin Date: Mon Feb 5 09:19:18 2018 New Revision: 257383 URL: https://gcc.gnu.org/viewcvs?rev=257383=gcc=rev Log: Document --dynamic-list-data option for --coverage usage. 2018-02-05 Martin LiskaPR gcov-profile/83879 * doc/gcov.texi: Document necessity of --dynamic-list-data when using dlopen functionality. Modified: trunk/gcc/ChangeLog trunk/gcc/doc/gcov.texi
[Bug target/68028] [6/7/8 regression] Compilation error "lto1: error: target attribute or pragma changes single precision floating point" with LTO on PowerPC
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68028 --- Comment #10 from Richard Biener --- (In reply to Nick Clifton from comment #9) > Created attachment 43318 [details] > Proposed patch > > Hi Guys, > > Richard was right - there is code that sets extra flags based upon the > setting of -mcpu. In fact it is just before the code he mentioned: > > /* For the E500 family of cores, reset the single/double FP flags to let us > check that they remain constant across attributes or pragmas. */ > > switch (rs6000_cpu) > { > case PROCESSOR_PPC8540: > case PROCESSOR_PPC8548: > case PROCESSOR_PPCE500MC: > case PROCESSOR_PPCE500MC64: > case PROCESSOR_PPCE5500: > case PROCESSOR_PPCE6500: > rs6000_single_float = 0; > rs6000_double_float = 0; > break; > > default: > break; > } > > This has lead me to propose the attached patch. Basically what it does is to > tell the rs6000 backend that when it is operating in LTO mode, it should > trust > the function attributes and not the command line. > > My assumption here is that if there are any discrepancies between real > function attributes (ie not ones generated for LTO) and the command line, > then these will have been detected and reported during the original > compilation. > > What do people think ? Is the patch OK ? If the backend doesn't support mixing of -msingle-float/-mno-single-float within a compilation unit then this will only work if the user didn't mix TUs with conflicting setting at link-time. So the following will not be diagnosed but will instead have conflicting IL accepted silently with your patch (with unknown effect): > gcc -c t1.c -flto -msingle-float > gcc -c t2.c -flto -mdouble-float > gcc t1.o t2.o -flto so I think it's better to have rs6000_single_float/rs6000_double_float initialized to for example -1 and allow changing that a _single_ time, verifying all following uses will match. But I don't see why mixing shouldn't be possible? Maybe the target wants to instead limit inlining of functions with differing settings instead? If the caller/callee uses FP? (see i386.c:ix86_can_inline_p use of cgraph_node::get (callee))->fp_expressions to guard similar things).
[Bug tree-optimization/82518] [8 regression] gfortran.fortran-torture/execute/in-pack.f90 fails on armeb since r252917
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82518 --- Comment #11 from Christophe Lyon --- My setup uses armeb-none-linux-gnueabihf (as opposed to armeb-eabi as you report). I have never tried armeb-eabi. I am also using qemu as simulator (in user mode, not in system mode). The failure in initialise_monitor_handles indicates a problem in the startup code, while initializing the semi-hosting interface. Can you try qemu?
[Bug gcov-profile/84137] Typo in gcov online documentation
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84137 Martin Liška changed: What|Removed |Added Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2018-02-05 Assignee|unassigned at gcc dot gnu.org |marxin at gcc dot gnu.org Ever confirmed|0 |1 --- Comment #1 from Martin Liška --- Thanks for the report, I'll fix it soon.
[Bug gcov-profile/84107] indirect call profiling broken with multiple DSOs
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84107 Martin Liška changed: What|Removed |Added Status|UNCONFIRMED |ASSIGNED Last reconfirmed||2018-02-05 Assignee|unassigned at gcc dot gnu.org |marxin at gcc dot gnu.org Ever confirmed|0 |1 --- Comment #1 from Martin Liška --- Thanks Alexander for filling that.
[Bug tree-optimization/84106] loop distribution cost-model needs work
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84106 Richard Biener changed: What|Removed |Added Status|UNCONFIRMED |NEW Last reconfirmed||2018-02-05 CC||amker at gcc dot gnu.org Summary|gcc is not able to |loop distribution |vectorize code for 1D |cost-model needs work |array, but does so for 2D | |array of the same size | Ever confirmed|0 |1 --- Comment #5 from Richard Biener --- Confirmed.
[Bug ipa/65797] [6/7/8 regression] IPA ICF causes function to be emitted with no debug line info
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65797 Jan Hubicka changed: What|Removed |Added Assignee|hubicka at ucw dot cz |unassigned at gcc dot gnu.org --- Comment #16 from Jan Hubicka --- We need someone who understand debug info to decide what to do. So I am unasigning myself :)
[Bug ipa/80726] [7/8 Regression] Destructor not inlined anymore (regression)
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80726 Jan Hubicka changed: What|Removed |Added Status|NEW |ASSIGNED CC||marxin at gcc dot gnu.org Component|tree-optimization |ipa Assignee|unassigned at gcc dot gnu.org |hubicka at gcc dot gnu.org --- Comment #5 from Jan Hubicka --- This really looks like a conflict with function splitting and comdats. We obviously can't introduce call to comdat local function but we indeed can bring function out of the comdat group. For that we, of course, would need to check how big that function is (account cross-module unsharing) into cost of the inline decision which will need bit of work. ipa-sra is currently bringing functions out of comdats "on wild" (i.e. without bounds on how big the function can be). I will play with this. Probably bit too involved for stage4 though.
[Bug lto/70929] [6/7/8 regression] Cross-module inlining for functions having argument passed by reference is no longer working.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=70929 --- Comment #7 from Jan Hubicka --- Hmm, this actually looks reasonable for me too. Shall I give it a try and figure out what incompatibilities we get here? What about K prototypes?
[Bug middle-end/84200] r256888 causes 30% performance regression of 519.lbm_r at -Ofast generic tuning on Zen
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84200 Sebastian Peryt changed: What|Removed |Added CC||sebastian.peryt at intel dot com --- Comment #1 from Sebastian Peryt --- I'm not sure if that can be treated as duplicate but that performance degradation looks like is related to PR84149.
[Bug ipa/57218] [6/7/8 Regression] Excessive inlining even at -Os
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=57218 Jan Hubicka changed: What|Removed |Added CC||marxin at gcc dot gnu.org Component|tree-optimization |ipa --- Comment #12 from Jan Hubicka --- Set component as IPA so it stays at my radar. It is probably too late for GCC 8 but I will try to finally take a look next stage1. Inlining those constructions is often beneficial because things subsequently simplifies. We would probably need bit more context dependent metrics than just single call window we do now. This however is hard to do becuase that makes number of inlining decisions to explode.
[Bug sanitizer/84208] fsanitize-address-use-after-scope Not working for ARM
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84208 --- Comment #2 from Akhilesh Kumar --- > Does it work on non-changed gcc 7.2 on arm? Not yet verified because unable to cross compile gcc 7.2. > And with arm do mean arm-linux-gnueabi as the target or aarch64-linux-gnu? I am using arm-*-gnueabi target