similar to: profiling JIT compiled code with perf

Have you seen https://reviews.llvm.org/D44892? we are using it in Julia to use perf on jitted code. -Valentin On Wed, 18 Jul 2018 at 17:13 Frank Tetzel via llvm-dev < llvm-dev at lists.llvm.org> wrote: > > Hello, > > > > is there any support in LLVM for the jitdump format [1] of perf? > > > > It enables perf report to also "zoom in" and annotate the

Hi, On 2018-07-19 14:57:42 +0200, Frank Tetzel via llvm-dev wrote: > > Have you seen https://reviews.llvm.org/D44892? we are using it in > > Julia to use perf on jitted code. > > No, I did not see this patch before. Thanks a lot for the pointer. > > It seems to be doing exactly what I want. Let's see if I get it working. FWIW, I just merged this. Did you have any

Hi, On 2020-03-09 21:20:44 +0100, Frank Tetzel via llvm-dev wrote: > I think, debugging and profiling support is very important for a JIT > engine. I could never get it to work with older LLVM versions. Is there > example code somewhere available? Since I added the perf listener I probably can help you with that. What exactly was the problem you were hitting? I don't have isolated

Having something like this available in tree would definitely be useful. For simplicity, why don't we start with support for the second style? This is the long term useful one and would be a good starting point for getting the code in tree. Can you give a pointer to the patch so that I can assess the rough complexity? If it's simple enough, I'd be happy to help get it reviewed

Hi, On 2016-12-29 13:17:50 -0800, Philip Reames wrote: > Having something like this available in tree would definitely be > useful. Cool. > For simplicity, why don't we start with support for the second style? This > is the long term useful one and would be a good starting point for getting > the code in tree. Works for me. > Can you give a pointer to the patch so that

I'm trying to use PerfJITEventListener with llvm::orc::LLJITBuilder: 1. perf record -o /tmp/perf.data -- <my_binary_with_event_listener> 2. perf inject -j -v -i /tmp/perf.data -o /tmp/perf.data.jit *jit marker found: ~.debug/jit/llvm-IR-jit-20200417-3c2242/jit-149849.dump* *injecting: ~/.debug/jit/llvm-IR-jit-20200417-3c2242/jit-149849.dump* *write ELF image

Hi all, I open sourced a small project of mine recently. It is an EDSL for C++ which makes just-in-time compilation much easier to write and maintain. One of its backends is LLVM's OrcJIT (LLVM 7, still). https://github.com/tetzank/coat Here is a blog post of mine introducing the concept: https://tetzank.github.io/posts/coat-edsl-for-codegen/ You can see it as an abstraction layer on top

Hi, On 2017-02-01 23:20:40 -0800, Andres Freund wrote: > > Can you give a pointer to the patch so that I can assess the rough > > complexity? If it's simple enough, I'd be happy to help get it > > reviewed and in. If it's more complicated, I probably won't have the > > time to assist. > > Patch (and a prerequisite) attached. Took me a while to get

Hi All, Just a couple of updates this week: First: there is a preview patch up at https://reviews.llvm.org/D75838 to enable use of JITEventListeners in RTDyldObjectLinkingLayer. I've only done very preliminary testing on it, but was able to debug simple JIT'd programs on Linux using the GDB registration listener. If you've been wanting to move to ORC but held up by lack of debugger /

I've been trying to get oprofile results for jitted code without success. I built an 3.5.0 llvm with oprofile enabled, and tested it with lli on a small test case. I built the latest oprofile from the git repository. Debugging I can see that lli is registering the listener and making the oprofile calls to the libopagent api to specify the names and address ranges of jit'd routines, and

Mi Maurice, A follow up to Andy's comments: MCJIT can find line numbers for ELF files, at least in limited circumstances. I know because I broke the regression test for it while doing some cleanup recently. ;) Looking at lib/ExecutionEngine/OProfileJIT/OProfileJITEventListener.cpp, I see the line: // TODO: support line number info (similar to IntelJITEventListener.cpp) >From a glance at

Hi, I am quite new to R and I have two questions regarding ROCR. 1. I have tried to understand how to extract area-under-curve value by looking at the ROCR document and googling. Still I am not sure if I am doing the right thing. Here is my code, is "auc1" the auc value? " pred1 <- prediction(resp1,label1) perf1 <- performance(pred1,"tpr","fpr") plot(

Hi, Under linux a large portion of the profiling these days happens with perf, but there's no support for it from LLVM's JITs. For a while perf could associate address+size to symbols by writing a /tmp/perf-$pid.map file: http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/tools/perf/Documentation/jit-interface.txt A year or so perf also gained the ability to actually

I believe the perf tool cannot profile/analyze the JITed code in MCJIT model. Can you please confirm this ? I was working on a patch to fix this. Another question, is the there any support to map the llvm IR with x86 generated assembly ? so its easier to analyze the code generator. -------------- next part -------------- An HTML attachment was scrubbed... URL:

hi Ingo, please consider pulling thanks, jirka The following changes since commit 3617660e4e1618a888a2e3a4067224534302cb33: Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf into perf/core (2014-05-01 08:24:59 +0200) are available in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git

hi Ingo, please consider pulling thanks, jirka The following changes since commit 3617660e4e1618a888a2e3a4067224534302cb33: Merge tag 'perf-core-for-mingo' of git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf into perf/core (2014-05-01 08:24:59 +0200) are available in the git repository at: git://git.kernel.org/pub/scm/linux/kernel/git/jolsa/perf.git

> Last time I tried it was some months ago with LLVM 9. I'll try it > shortly with LLVM trunk/master and report back. Finally, I found the time to try it out again. And it works fine in LLVM master. I haven't tried any other LLVM version. Profiling with perf on assembly level works fine. And so does gdb. One can set pending breakpoints on the to be generated function and execution

Hi, I get an internal server error when accessing llvm.org/perf. This error causes buildbot failures as they can not properly report their performance findings. Chris, could you have a look? Thanks, Tobias

The LNT Perf reporting website is down for a few days now, should I disable my perf bot to avoid noise while that gets fixed? cheers, --renato

Hi Chris, llvm.org/perf gives again error 500s. I remember you had a look the last time. Would you mind looking again? Also, is there something which we could do to make LNT more robust. Do you have an idea what the last issue was about? Thanks, Tobias -------- Original Message -------- Subject: buildbot failure in LLVM on polly-perf-O3-polly-scev-codegen-isl Date: Fri, 21 Feb 2014 17:08:44