[Bug]: Combining --no-inline and CHPL_LOCALE_MODEL=gpu results in a internal error

[jabraham17]

Attempting to compile any Chapel program with --no-inline and CHPL_LOCALE_MODEL=gpu results in an internal error. This occurs in any Chapel program.

;

Compiling as CHPL_LOCALE_MODEL=gpu CHPL_GPU=cpu chpl bug.chpl --no-inline hits the error. This is due to an incorrect assumption in extractIndicesAndLowerBounds in gpuTransforms. This also gets hit with real gpus (CHPL_GPU=amd for example).

This is not a particularly high priority bug as I think its uncommon for --no-inline to be thrown by users, but I find it useful when debugging (which is how I hit this problem)

Chapel env

CHPL_TARGET_PLATFORM: darwin
CHPL_TARGET_COMPILER: llvm
CHPL_TARGET_ARCH: arm64
CHPL_TARGET_CPU: native *
CHPL_LOCALE_MODEL: gpu *
  CHPL_GPU: cpu *
CHPL_COMM: none
CHPL_TASKS: qthreads
CHPL_LAUNCHER: none
CHPL_TIMERS: generic
CHPL_UNWIND: none
CHPL_MEM: jemalloc
CHPL_ATOMICS: cstdlib
CHPL_GMP: system *
CHPL_HWLOC: system *
CHPL_RE2: bundled *
CHPL_LLVM: system
CHPL_AUX_FILESYS: none

[e-kayrakli]

Nice reproducer!