https://bugs.winehq.org/show_bug.cgi?id=55190
Bug ID: 55190 Summary: Conspiracy's Clean Slate 64K demo crashes with 'Assertion failed: iff->condition.node->data_type->dimx == 1' Product: vkd3d Version: 1.8 Hardware: x86-64 URL: https://web.archive.org/web/20210807072016/https://con spiracy.hu/files/cns%21cns-AMDfix.zip OS: Linux Status: NEW Keywords: download Severity: normal Priority: P2 Component: hlsl Assignee: wine-bugs@winehq.org Reporter: gijsvrm@gmail.com Distribution: ---
Created attachment 74788 --> https://bugs.winehq.org/attachment.cgi?id=74788 output
Continuation of bug 54926 & bug 54484.
https://bugs.winehq.org/show_bug.cgi?id=55190
Fabian Maurer dark.shadow4@web.de changed:
What |Removed |Added ---------------------------------------------------------------------------- CC| |dark.shadow4@web.de
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #1 from Nikolay Sivov bunglehead@gmail.com --- Created attachment 74790 --> https://bugs.winehq.org/attachment.cgi?id=74790 shader
Attaching affected shader. Profile is ps_5_0, entry point "p".
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #2 from Nikolay Sivov bunglehead@gmail.com --- Hard to tell from the log, but the issue now is that fmod() does not support vector arguments properly, producing conditional variables with dimx > 1.
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #3 from Nikolay Sivov bunglehead@gmail.com --- This works for me https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/268.
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #4 from Gijs Vermeulen gijsvrm@gmail.com --- (In reply to Nikolay Sivov from comment #3)
This works for me https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/268.
I just tried this one and it seems I'm still hitting this assertion. I made sure to rerun 'autogen.sh' and used 'make dlls/ntdll/clean' to make sure that the new vkd3d build was picked up. Any ideas what I could be doing wrong?
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #5 from Nikolay Sivov bunglehead@gmail.com --- Not sure. I usually test with vkd3d-compiler directly, using standalone shaders. Using correct profile is also important.
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #6 from Nikolay Sivov bunglehead@gmail.com --- This is now comitted, https://source.winehq.org/git/vkd3d.git/commit/fc2aaee224806a97cea4e4d229fe6.... We should retest with the next import.
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #7 from Nikolay Sivov bunglehead@gmail.com --- (In reply to Gijs Vermeulen from comment #4)
(In reply to Nikolay Sivov from comment #3)
This works for me https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/268.
I just tried this one and it seems I'm still hitting this assertion. I made sure to rerun 'autogen.sh' and used 'make dlls/ntdll/clean' to make sure that the new vkd3d build was picked up. Any ideas what I could be doing wrong?
I see what you mean now. Same assertion happens for float modulus, because we generate a conditional there for vectors, and it should be using per-component move instead.
Example:
--- float4 f, m;
float4 main() : sv_target { return f % m; } ---
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #8 from Nikolay Sivov bunglehead@gmail.com --- Here is another one https://gitlab.winehq.org/wine/vkd3d/-/merge_requests/336, now attached shader compiles for me.
https://bugs.winehq.org/show_bug.cgi?id=55190
Nikolay Sivov bunglehead@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |RESOLVED Fixed by SHA1| |6d1ba83856abceafc9facf3c3d1 | |b3d790a93157a Resolution|--- |FIXED
--- Comment #9 from Nikolay Sivov bunglehead@gmail.com --- This should be fully fixed now, https://source.winehq.org/git/vkd3d.git/commit/6d1ba83856abceafc9facf3c3d1b3....
https://bugs.winehq.org/show_bug.cgi?id=55190
--- Comment #10 from Gijs Vermeulen gijsvrm@gmail.com --- Fixed indeed, thanks!
It now crashes on:
0024:err:d3dcompiler:D3DCompile2 <anonymous>:4:485: E5017: Aborting due to not yet implemented feature: Dereference with non-constant offset of type HLSL_IR_EXPR.
https://bugs.winehq.org/show_bug.cgi?id=55190
Nikolay Sivov bunglehead@gmail.com changed:
What |Removed |Added ---------------------------------------------------------------------------- Status|RESOLVED |CLOSED
--- Comment #11 from Nikolay Sivov bunglehead@gmail.com --- Closing bugs fixed in 1.10.