similar to: Avoiding function addresses in llvm_prf_data when value profiling is disabled

LLD revision is r331862. To add, I had initially tried it on r328903, which also reproduced the issue. On Wed, May 9, 2018 at 9:26 AM Pirama Arumuga Nainar <pirama at google.com> wrote: > Hi Teresa, > > Thanks for looking into this. I hadn't initially tried ToT, but it > reproduces in ToT as well when I tried. > > $ ./clang --version > > clang version 7.0.0

Adding Peter to comment on the linker resolution issue. >From adding save-temps, it looks like lld and gold are giving different resolutions to the symbols, which is presumably creating this issue: (first file is with lld and second is with gold) $ diff a.out.resolution.txt gold/ 4c4 < -r=a.o,__llvm_profile_raw_version,plx --- > -r=a.o,__llvm_profile_raw_version,l 8,9c8,9 <

Sorry, operator error. I can reproduce now. Interestingly, this does not reproduce using gold, and they utilize the same underlying LTO API. Let me dig a little using save-temps and see where they diverge. Teresa On Wed, May 9, 2018 at 9:28 AM Pirama Arumuga Nainar <pirama at google.com> wrote: > LLD revision is r331862. To add, I had initially tried it on r328903, > which also

On Wed, May 9, 2018 at 6:43 AM Teresa Johnson <tejohnson at google.com> wrote: > Hi Pirama, > > I can't reproduce with either lld or gold, using a compiler built from > head. What version is your clang? > (and your lld) > > Thanks, > Teresa > > On Tue, May 8, 2018 at 7:50 PM Pirama Arumuga Nainar via llvm-dev < > llvm-dev at lists.llvm.org>

Thanks Peter, your patch does fix the reproducer. I filed https://bugs.llvm.org/show_bug.cgi?id=37422 to track this bug. I have no clue on how to resolve the tests - whether further cleanup is required in the code or in the tests. But if Teresa or you cannot get to it, I can, with some help, take a crack at fixing the tests. On Wed, May 9, 2018 at 11:26 AM Peter Collingbourne <peter at

The problem is that ThinLTO is not dropping the non-prevailing definitions, and they end up being emitted into the object file for b.o. $ ../ra/bin/llvm-dis -o - b.o0.0.preopt.bc | grep __llvm_prof $__llvm_profile_raw_version = comdat any $__llvm_profile_filename = comdat any @__llvm_profile_raw_version = constant i64 72057594037927940, comdat @__llvm_profile_filename = constant [19 x i8]

Hi Teresa, Thanks for looking into this. I hadn't initially tried ToT, but it reproduces in ToT as well when I tried. $ ./clang --version clang version 7.0.0 (trunk 331879) (llvm/trunk 331888) Target: x86_64-unknown-linux-gnu Thread model: posix InstalledDir: /ssd2/pirama/llvm-upstream/llvm-release/bin/. $ ./ld.lld --version LLD 7.0.0 (https://git.llvm.org/git/lld.git

Can you upload a reproducer? You can create it using the /linkrepro flag. Peter On Thu, Jun 7, 2018 at 2:50 PM, Reid Kleckner via llvm-dev < llvm-dev at lists.llvm.org> wrote: > GCC does comdats completely differently from the spec. Since you contacted > me about this off of the mailing list, I started investigating what they > do, and it is completely different. It's

The duplicate symbol errors are for __llvm_profile_filename and __llvm_profile_raw_version. IIUC, these are supposed to be weak symbols but Thin LTO seems to break this in some way. This does't happen with gold or no LTO or full LTO. $ cat > a.c extern int foo(); int main() { return foo(); } $ cat > b.c int foo() { return 0; } $ clang a.c -fprofile-generate -flto=thin -c $ clang

The gold plugin supports many of the backend options. But Clang doesn't pass all of them to the plugin as it does to the backend. For example, to support -emulated-tls with LTO, Clang needs to pass -emulated-tls to the LTO backend. Shall I just change the driver to add -plugin-opt=-emulated-tls or is the best practice to embed this in the IR? -------------- next part -------------- An HTML

I encountered a segfault when using lld to cross-compile for Windows (+MinGW) from Linux. The problem happens with objects built by gcc. The problem is that ObjFile::CreateRegular considers a PendingComdat to be valid (and later causes an illegal pointer dereference). The following patch fixes the crash: diff --git a/COFF/InputFiles.cpp b/COFF/InputFiles.cpp index 9e2345b0a..f47d612df 100644

Hi Pirama, I can't reproduce with either lld or gold, using a compiler built from head. What version is your clang? Thanks, Teresa On Tue, May 8, 2018 at 7:50 PM Pirama Arumuga Nainar via llvm-dev < llvm-dev at lists.llvm.org> wrote: > The duplicate symbol errors are for __llvm_profile_filename and __llvm_profile_raw_version. > IIUC, these are supposed to be weak symbols but

I'm an idiot and sent to llvm-commits instead of llvm-dev. Fixing. On 10/17/17, 5:09 PM, "llvm-commits on behalf of Shoaib Meenai via llvm-commits" <llvm-commits-bounces at lists.llvm.org on behalf of llvm-commits at lists.llvm.org> wrote: Hi all (CC beanz for cmake advice), I'm running into a cmake problem when I try to cross-compile a

On Fri, Jan 18, 2019 at 3:56 PM Manman Ren <manman.ren at gmail.com> wrote: > Some background information first, then a quick summary of what we have > discussed so far! > > Background: Facebook app is one of the biggest iOS apps. Because of this, > we want the instrumentation to be as lightweight as possible in terms of > binary size, profile data size, and runtime

On Fri, Jan 18, 2019 at 9:10 PM Manman Ren <manman.ren at gmail.com> wrote: > > > On Fri, Jan 18, 2019 at 4:11 PM Xinliang David Li <davidxl at google.com> > wrote: > >> >> >> On Fri, Jan 18, 2019 at 3:56 PM Manman Ren <manman.ren at gmail.com> wrote: >> >>> Some background information first, then a quick summary of what we have

I would love to see this kind of order profiling support. Using dtrace to generate function orders is actually really problematic because dtrace made tradeoffs in implementation allowing it to ignore probe execution if the performance impact is too great on the system. This can result in dtrace being non-deterministic which is not ideal for generating optimization data. Additionally if order

Hi- In August we made an announcement of "FC: A new fortran front end" [1]. At that time to get an end-to-end solution, we made FC to emit LLVM IR directly. At present, we have upgraded FC to emit MLIR. Currently the language supported is close to Fortran-95. Apart from 400+ unit test cases, out framework passes two SPEC-2017 benchmarks successfully. Currently we are cleaning up the

Agreed! Is the code already available? What are your plans for it, and are you interested in collaboration with the rest of the LLVM community? -Chris > On Jan 11, 2020, at 11:58 AM, Finkel, Hal J. via llvm-dev <llvm-dev at lists.llvm.org> wrote: > > Hi, Prashanth, > > That's great news! It sounds like you've made a lot of progress, and I certainly hope that you

Hi Gang Server, two interfaces, routing to two different networks. Two transports defined, each bound to the corresponding ip assigned to the interface. But still, especially when an 183 message is sent, the Contact header does contain the wrong IP Address. Is this a known issue 13.18.3? Or is there a way to make absolutely sure the IP addresses within the Contact header is corresponding to

Hello all, There is one enhancement request open for dwarfv5, http://dwarfstd.org/ShowIssue.php?issue=191107.1 The request is for addition of dwarf expression operator to swap the top of the dwarf stack, the response seems positive but it may take some time till v6. I like to add that operator as vendor extension but I am not sure how to go about it for llvm/lldb. Currently I am using it as