[all-commits] [llvm/llvm-project] 30434b: [Reland][llubi] Add support for load/store/lifetim...

[Yingwei Zheng via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 30434bab1a85451b0a85fab94a835a7a40818ffa
      https://github.com/llvm/llvm-project/commit/30434bab1a85451b0a85fab94a835a7a40818ffa
  Author: Yingwei Zheng <dtcxzyw2333 at gmail.com>
  Date:   2026-03-08 (Sun, 08 Mar 2026)

  Changed paths:
    M llvm/docs/CommandGuide/llubi.rst
    A llvm/test/tools/llubi/loadstore_be.ll
    A llvm/test/tools/llubi/loadstore_le.ll
    A llvm/test/tools/llubi/loadstore_misaligned.ll
    A llvm/test/tools/llubi/loadstore_null.ll
    A llvm/test/tools/llubi/loadstore_oob1.ll
    A llvm/test/tools/llubi/loadstore_overaligned.ll
    A llvm/test/tools/llubi/loadstore_poison.ll
    A llvm/test/tools/llubi/loadstore_uaf.ll
    A llvm/test/tools/llubi/store_dead.ll
    M llvm/tools/llubi/lib/Context.cpp
    M llvm/tools/llubi/lib/Context.h
    M llvm/tools/llubi/lib/Interpreter.cpp
    M llvm/tools/llubi/lib/Value.h
    M llvm/tools/llubi/llubi.cpp

  Log Message:
  -----------
  [Reland][llubi] Add support for load/store/lifetime markers (#185196)

This patch relands https://github.com/llvm/llvm-project/pull/182532. The
original version causes test failures related undef bits since it
incorrectly assumes `std::uniform_int_distribution` yields the same
results across different stdlib vendors. This patch simply uses low bits
to avoid the issue. I am not sure whether it still generates uniformly
distributed random numbers. But abseil also uses this trick:
https://github.com/abseil/abseil-cpp/blob/e72b94a2f257ba069ec0b99e557e9f1f6b9c1a3e/absl/random/uniform_int_distribution.h#L203-L206

I have confirmed all tests passed with libstdc++ and libc++.

Original PR description:
Bytes are adjusted to respect the incoming byte type proposed in
https://discourse.llvm.org/t/rfc-add-a-new-byte-type-to-llvm-ir/89522.

Note that the current implementation of constant folding doesn't handle
bitcasts with weird types like `<8 x i3> to <3 x i8>`:

https://github.com/llvm/llvm-project/blob/49de34459c99b33c27aae4596ed5e67ce3d89eae/llvm/lib/Analysis/ConstantFolding.cpp#L223-L232
So I take the result from Alive2 as reference:
https://alive2.llvm.org/ce/z/ZGA_xP



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications