[all-commits] [llvm/llvm-project] ca0a0b: [flang] optionally lower scalar and explicit shape...

[jeanPerier via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: ca0a0bf9638c9705862ef2684aba192376c7d7cb
      https://github.com/llvm/llvm-project/commit/ca0a0bf9638c9705862ef2684aba192376c7d7cb
  Author: Jean Perier <jperier at nvidia.com>
  Date:   2022-10-20 (Thu, 20 Oct 2022)

  Changed paths:
    M flang/include/flang/Lower/SymbolMap.h
    M flang/include/flang/Optimizer/Builder/FIRBuilder.h
    M flang/lib/Lower/ConvertVariable.cpp
    M flang/lib/Lower/SymbolMap.cpp
    M flang/lib/Optimizer/Builder/FIRBuilder.cpp
    A flang/test/Lower/HLFIR/convert-variable.f90
    M flang/test/Lower/HLFIR/expr-addr.f90

  Log Message:
  -----------
  [flang] optionally lower scalar and explicit shape with fir.declare

Lower scalar and explicit shape arrays to fir.declare under the -hlfir option.
Update the SymMap so that it can hold fir::FortranVariableInterface.

The plan is to go towards a SymMap that only contains fir::FortranVariableInterface
once current expression lowering can be replaced. This should make the SymMap lighter
than it is today (SymBox/ExtendedValue are above 256 bytes).

Assumed shape, allocatable and pointer are left TODOs for now. Anything with a
specification expression that is not a constant expression will only be able to
be lowered when the HLFIR expression lowering skeleton is added.

Differential Revision: https://reviews.llvm.org/D136252