similar to: Dovecot LTO build failure

(repost the reply using my personal account -- previous reply to the list got hold up) On Thu, Dec 25, 2014 at 11:55 PM, Adve, Vikram Sadanand <vadve at illinois.edu> wrote: > Diego, Teresa, David, > > Sorry for my delayed reply; I left for vacation right after sending my message about this. > > Diego, it wasn't explicit from your message whether LLVM LTO can handle

On Fri, Dec 12, 2014 at 1:59 PM, Diego Novillo <dnovillo at google.com> wrote: > On 12/12/14 15:56, Adve, Vikram Sadanand wrote: >> >> I've been asked how LTO in LLVM compares to equivalent capabilities >> in GCC. How do the two compare in terms of scalability? And >> robustness for large applications? > > > Neither GCC nor LLVM can handle our

Distro: OpenSuSE Tumbleweed for x86_64 Failing version: dovecot23-2.3.11.3-1.1.x86_64 Install Date: 2020-08-18 Reverting to previous version works: dovecot23-2.3.10.1-2.3.x86_64 (Packages downgraded coordinately: dovecot23 dovecot23-backend-sqlite) How to make it fail: As the user, execute doveadm expunge mailbox Spam37 savedbefore 3day #User's actual cmd doveadm who #The simplest

The Apple ld (shipped with Xcode, open-source) saves the 3 files if -save-temps is applied: - xxx.lto.bc : the merged bit-code before IPA - xxx.lto.opt.bc: the optimized bit-code of the *.lto.bc - xxx.lto.o : the object file of xxx.lto.bc I think it is better way to go, often time we need IR both after and after the transformation. On 6/4/13 3:37 AM, Jonas Wagner wrote: > >

Hi, I have a question regarding the behavior of weak symbol resolution in LTO: Suppose there are weak definitions in both the source code and some native lib. In non-LTO path, we will use the version from source code. In LTO path, LLVM may discard the definition as it has "linkeonce" linkage type. So the native version will be selected by linker. Now, non-LTO and LTO build may have

On Wed, Jul 01, 2020 at 06:07:25PM +0100, Dave P Martin wrote: > On Tue, Jun 30, 2020 at 06:37:34PM +0100, Will Deacon wrote: > > When building with LTO, there is an increased risk of the compiler > > converting an address dependency headed by a READ_ONCE() invocation > > into a control dependency and consequently allowing for harmful > > reordering by the CPU. > >

I'm not completely sure what you are asking. Are you looking for performance benchmarks to use for LTO and ThinLTO testing? Or are you asking how to build/run with LTO and ThinLTO? Are you asking how to run LLVM's performance test-suite with LTO and ThinLTO? Teresa On Fri, May 11, 2018 at 5:21 AM Siddharth Shankar Swain < h2015096 at pilani.bits-pilani.ac.in> wrote: > Hi, >

What is going on here? I'm trying to commit a change, but it (repeatedly) fails, so apparently there is something wrong: > Commit failed (details follow): > Base checksum mismatch on '/branches/Testing/packaging/opensuse/nut.spec.in': > expected: c7ab1364e6e2f7263633c3fc4847619c > actual: 208ae1d7c1fe7075b00cbd7b94731815 I have no idea what is wrong and therefor, no idea

Due to the LTO modifications to the build system it is not possible to do old non LTO builds that involve some old PIC behaviour, this means Cygwin and possible other platforms old behaviours is broken, and they are not LTO'ed yet either. Would it be possible to have both the old PIC behaviour and the new LTO based behavious coincide via a flags like --enable-lto=(yes,no) and for older PIC

On Tue, Jun 30, 2020 at 03:57:54PM -0700, Sami Tolvanen wrote: > On Tue, Jun 30, 2020 at 12:47 PM Marco Elver <elver at google.com> wrote: > > > > On Tue, 30 Jun 2020 at 19:39, Will Deacon <will at kernel.org> wrote: > > > > > > When building with LTO, there is an increased risk of the compiler > > > converting an address dependency headed by a

> > > I often use the gold plugin's also-emit-llvm option to verify the final > LLVM bitcode of a program. However, I realized that the gold plugin generates >> the >> bitcode file before running LTO optimizations, such that the changes >> performed >> by LTO are not reflected in the bitcode file. >> >> What is the reason for this behavior?

It looks the problem lies in how your compiler generates debug info. LLVM doesn't expect DIDerivedType scope to be an instance of DICompileUnit. Here is a quick fix: DIE *DwarfUnit::getOrCreateContextDIE(const DIScope *Context) { - if (!Context || isa<DIFile>(Context)) + if (!Context || isa<DIFile>(Context) || isa<DICompileUnit>(Context)) However, I suggest talking to

LTOCodeGenerator::addModule (which is wrapped by lto_codegen_add_module) takes an LTOModule as an argument and links the latter's llvm::Module into its own. It does not change the latter's llvm::Module. The lto_module_dispose API call destroys an LTOModule. This frees the LTOModule's llvm::Module, but also invalidates strings returned by, e.g., lto_module_get_symbol_name. I propose

> The Apple ld (shipped with Xcode, open-source) saves the 3 files if > > -save-temps is applied: > > - xxx.lto.bc : the merged bit-code before IPA > > - xxx.lto.opt.bc: the optimized bit-code of the *.lto.bc > > - xxx.lto.o : the object file of xxx.lto.bc > > > > I think it is better way to go, often time we need IR both after and > after >

On 4 June 2013 13:36, Shuxin Yang <shuxin.llvm at gmail.com> wrote: > The Apple ld (shipped with Xcode, open-source) saves the 3 files if > -save-temps is applied: > - xxx.lto.bc : the merged bit-code before IPA > - xxx.lto.opt.bc: the optimized bit-code of the *.lto.bc > - xxx.lto.o : the object file of xxx.lto.bc > > I think it is better way to go, often time we

yeah so when LTO runs, in the linker, it uses the target to produce object code. In the sense of code generation for target arch. Thanks, Siddharth On Mon, Jan 29, 2018 at 11:03 PM, Teresa Johnson <tejohnson at google.com> wrote: > Can you be more specific? LTO/ThinLTO are target independent. You just > need to be using a linker that supports *LTO (gold, lld, ld64, eg). > Teresa

I've only measured performance on x86. There are some old results for SPEC cpu2006 in the blog post here: http://blog.llvm.org/2016/06/thinlto-scalable-and-incremental-lto.html The benefit is benchmark dependent, e.g. small benchmarks often don't improve much as they don't require whole program optimizations. We've improved the ThinLTO optimizations since then, but I don't

I run the tests again. This time with only the more interesting candidates and with the constructor alias optimization enabled. The gcc used is the one with fedora 13 (4.4.4) The size of the final clang binary in (KiB) when compiled on linux x86-64 was gcc -Os 23073 clang -Os 25251 clang -Os lto 18295 Time to "link" the final clang binary with lto when the gold plugin was

Hi everyone, We are implementing a profiling instrumentation tool as a part of clang's LTO. The tool inserts code at compile time that makes calls to a runtime library (written as a C file) during program execution. The runtime library can collect different kinds of profile information depending on user input. Previously, the user had to manually link the correct profiling library to the

On Mon, Jul 06, 2020 at 05:00:23PM +0100, Dave Martin wrote: > On Thu, Jul 02, 2020 at 08:23:02AM +0100, Will Deacon wrote: > > On Wed, Jul 01, 2020 at 06:07:25PM +0100, Dave P Martin wrote: > > > Also, can you illustrate code that can only be unsafe with Clang LTO? > > > > I don't have a concrete example, but it's an ongoing concern over on the LTO > >