[all-commits] [llvm/llvm-project] 61aeba: [flang][FIRToMemRef] Fix lowering of complex array...

[Susan Tan (ス-ザン タン) via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 61aebacee7487ab168d2cf6757697ae178762dd8
      https://github.com/llvm/llvm-project/commit/61aebacee7487ab168d2cf6757697ae178762dd8
  Author: Susan Tan (ス-ザン　タン) <zujunt at nvidia.com>
  Date:   2026-04-14 (Tue, 14 Apr 2026)

  Changed paths:
    M flang/lib/Optimizer/Transforms/FIRToMemRef.cpp
    A flang/test/Transforms/FIRToMemRef/slice-projected.mlir

  Log Message:
  -----------
  [flang][FIRToMemRef] Fix lowering of complex array component slices (z%re, z%im) (#191846)

fir.slice with a path component (z%re, z%im) was silently dropped by
FIRToMemRef. Since memref.reinterpret_cast cannot change element type,
layout must come from the projected box descriptor via
fir.box_dims/fir.box_elesize rather than the triplets. Only
complex-array projections are handled here —
sizeof(complex<T>)/sizeof(T) = 2 is always exact for divsi. Derived-type
component projections bail out to downstream FIR-to-LLVM lowering where
strides can be non-integer.



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