[PATCH] D71843: [ARM][MVE] MVE-I should not be disabled by -mfpu=none

[Simon Tatham via Phabricator via llvm-commits]

simon_tatham added inline comments.


================
Comment at: clang/test/Driver/arm-mfpu.c:410
+// CHECK-MVEFP-FPUNONE-DAG: "-target-feature" "-mve.fp"
+// CHECK-MVEFP-FPUNONE-DAG: "-target-feature" "-fpregs"
+
----------------
I'm confused by this.

A set of target-feature options that turn off `fpregs`, when parsed by LLVM which knows the target-feature dependencies, will surely also disable even //integer// MVE, because the FP regs are also the MVE vector regs.

But if I start with full MVE (both integer and float), and I say `-mfpu=none` to disable //float// MVE, then surely I should expect //integer// MVE to still be enabled, and therefore `fpregs` should still be there to support it?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D71843/new/

https://reviews.llvm.org/D71843