[PATCH] D113026: [ARM] reject -mtp=cp15 if target subarch does not support it

[Ard Biesheuvel via Phabricator via cfe-commits]

ardb marked 2 inline comments as done.
ardb added inline comments.


================
Comment at: clang/lib/Driver/ToolChains/Arch/ARM.cpp:160-161
+    if (ThreadPointer == ReadTPMode::Cp15 &&
+        getARMSubArchVersionNumber(Triple) < 7 &&
+        llvm::ARM::parseArch(Triple.getArchName()) != llvm::ARM::ArchKind::ARMV6T2) {
+          D.Diag(diag::err_target_unsupported_tp_hard) << Triple.getArchName();
----------------
nickdesaulniers wrote:
> The last two checks seems redundant (`getARMSubArchVersionNumber(Triple) < 7 && llvm::ARM::parseArch(Triple.getArchName()) != llvm::ARM::ArchKind::ARMV6T2`).  If the `ArchKind` was `ARMV6T2` then `getARMSubArchVersionNumber` should return `6`.
> 
> Or do we want to support ARMv6T2, in which case the patch description should be edited:
> 
> > ensure that -mtp=cp15 is rejected for ARMv6T2 or earlier.
> 
> s/ARMv6T2 or earlier/ARMv6 (except ARMv6T2) or earlier/
> The last two checks seems redundant (`getARMSubArchVersionNumber(Triple) < 7 && llvm::ARM::parseArch(Triple.getArchName()) != llvm::ARM::ArchKind::ARMV6T2`).  If the `ArchKind` was `ARMV6T2` then `getARMSubArchVersionNumber` should return `6`.
> 
> Or do we want to support ARMv6T2, in which case the patch description should be edited:
> 
> > ensure that -mtp=cp15 is rejected for ARMv6T2 or earlier.
> 
> s/ARMv6T2 or earlier/ARMv6 (except ARMv6T2) or earlier/

Yeah, the latter. ARMv6T2 is the first core that implements Thumb2, 




CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D113026/new/

https://reviews.llvm.org/D113026