-
Notifications
You must be signed in to change notification settings - Fork 99
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Bugreport for complex codebase (probably some math problem) #1613
Comments
i have tested both release version and from main branch |
Can you post what the error message is, assuming that you mean there is a
compile time error? If a runtime error can you describe what the intended
result should be?
…On Wed, Jan 17, 2024, 8:49 AM Vladimir ***@***.***> wrote:
i have tested both release version and from main branch
—
Reply to this email directly, view it on GitHub
<#1613 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAJTUXABUT3I6LEIYHGIWITYO7JIHAVCNFSM6AAAAABB6SMAAGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQOJVHA2TEMJRG4>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
|
[ 8%] Built target miniz
#0 0x00007f5bb53ae6f6 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) build-llvm/tools/clang/stage2-bins/llvm/lib/Support/Unix/Signals.inc:567:13 PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT: make[2]: *** [CMakeFiles/hydra.dir/build.make:90: CMakeFiles/hydra.dir/diff_render/integrator_dr.cpp.o] Error 1
|
The txt file with message is also in the archive on google disk https://drive.google.com/file/d/1NbskYqtnwYr1u4n_sppsGNw-Q8YVkrhy/view?usp=sharing Its definitely compile time |
@FROL256 I tried setting up your repo to debug, but it looks like it wasn't committed to the repo?
|
srry, forgot about that: git submodule init |
@FROL256 so I retried your codebase on the lastest main and it appears to compile successfully. I don't have the file you need but otherwise it seems fine?
|
Yep, this is because i have disabled all materials except Lambert in main branch. Please consider those two branches:
You can clone target branch with sub-modules in this way: |
So I am on the Can you rerun and confirm what error message you see? Specifically I was on
|
Ok, so i have made some updates. It seems enzyme_101 AD behaves much better. Here is the full way to reproduce the bug from scratch:
Illegal orIn: Integer right: Float@float PointerIntSame=0
PLEASE ATTACH THE FOLLOWING FILES TO THE BUG REPORT: make[2]: *** [CMakeFiles/hydra.dir/build.make:90: CMakeFiles/hydra.dir/diff_render/integrator_dr.cpp.o] Error 1 |
If there is something in my code which is unsupported and cause 'Illegal orIn: Integer right: Float@float PointerIntSame=0', could you please give me and advice what i have to seek for. I'll try to remove it. Thanks! |
@wsmoses , do you have any ideas about 'Illegal orIn: Integer right: Float@float PointerIntSame=0' ? ) thx ) |
I have finally restructured my code by making less nesting and less in-out parameters, and on v103 it works. |
Greetings!
I have a project on differential rendering (path tracer) which uses Enzyme.
While working with it I encounter some Enzyme bugs and i'm not sure how to report them correctly.
The problem is that i can't reproduce bug on simple code base, and simplifying the complex code base didn't helped me too -- it arises in different places related to math functions.
I have put files from clang bug report to this archive
https://drive.google.com/file/d/1NbskYqtnwYr1u4n_sppsGNw-Q8YVkrhy/view?usp=sharing
The buggy branch of my project is here
https://github.com/Ray-Tracing-Systems/HydraCore3/tree/ad_bug
It has a build instruction, but probably you will need to change some path related to clang and enzyme there ...
The problem arises on call gltfEval->FrDielectricPBRT
Inside FrDielectricPBRT different instructions are lead to bug, however the same function in separate project works
https://github.com/FROL256/enzyme_autodiff_examples
Suppose there is some problems with std::sqrt call because code complies if i remove them.
Please let me know if i could do something to make bug-report useful if it is not in that state currently)
Tnanks!
The text was updated successfully, but these errors were encountered: