XChy
XChy
Same with 0.10.3 still
Thanks. That's OK. But to note, this patch just makes it compilable, not free of bugs. Such integration with latest LLVM is untested, so we may need to fix it...
I may not clarify it well, what I really want is to control the "strength" of forward/backward slicing. Take a llvm-ir as example (because I work only on llvm-ir, instead...
> Any instruction can be a slicing criterion: https://github.com/mchalupa/dg/blob/master/doc/llvm-slicer.md#slicing-criteria-the--sc-option It seems to require emitting debug instructions, but now I only have IR. I prefer to taking the name of Value/Instruction...
Thanks for your explanation and assistance. That's really helpful to me, your guide on where to hack it is highly appreciated!
I would like to add the support for the flags in SCCP.
> Since I see vim up there: https://savannah.gnu.org/git/?group=emacs Also favor Neovim here: https://github.com/neovim/neovim :)
@leewei05 , it seems that @Abhinkop is not working on this, and feel free to fix it. Please take a look at [guide](https://llvm.org/docs/InstCombineContributorGuide.html) before submitting your patch.
@leewei05 I'm sorry for my fault in this old issue. You're right, `(sext(a) & sext(c1)) == c2` doesn't match the motivating example. A correct and more general pattern should be...
Formal alive2 proof: https://alive2.llvm.org/ce/z/iQUQfE