[PATCH] Add support for armv6k subtarget

Vinicius Tinti viniciustinti at gmail.com
Sat Mar 7 19:57:37 PST 2015


In http://reviews.llvm.org/D8127#135999, @rengolin wrote:

> You don't seem to be dealing with v6Z or v6KZ at all, but LLVM does and your tests do, too. Shouldn't you make sure the arch/cpu relationships maintained in LLVM to be the same here?


I am not dealing with v6Z and v6KZ indeed. But there are several tests that are marked with FIXME instead of RUN.

Before this patches the backend was only dealing with v6, v6m and v6t2. There was references for v6Z and v6KZ but I don't think they were being used.
Now it deals with v6k plus all the previous it was dealing before.

That is why I am caring more about v6k.

Besides I took care to keep the same processor subtarget mapping on LLVM and Clang as we can check in LLVM:lib/Support/Triple.cpp and CLANG:lib/Basic/Targets.cpp  and CLANG:lib/Driver/ToolChains.cpp

The only thing I did not like was the netbsd triple changing from armv6 to armv6k but this is expected since arm1176jzf-s moved from v6 to v6k and is the default netbsd subtarget for EABIHF as Renato and I discussed.


================
Comment at: lib/Driver/ToolChains.cpp:113
@@ -112,2 +112,3 @@
   return llvm::StringSwitch<const char*>(Value)
-    .Case("armv6k", "armv6")
+    .Case("armv6", "armv6")
+    .Case("armv6k", "armv6k")
----------------
rengolin wrote:
> Isn't there a default to keep the same name?
I will test.

But since there was an armv6m -> armv6m right below I believed it was needed.

http://reviews.llvm.org/D8127

EMAIL PREFERENCES
  http://reviews.llvm.org/settings/panel/emailpreferences/






More information about the cfe-commits mailing list