[llvm] [TLI] Use AArch64 vector calling convention for ArmPL routines (PR #135790)

[Paul Walker via llvm-commits]

================
@@ -1337,6 +1337,8 @@ static const VecDesc VecFuncs_ArmPL[] = {
 #define TLI_DEFINE_ARMPL_VECFUNCS
 #define TLI_DEFINE_VECFUNC(SCAL, VEC, VF, MASK, VABI_PREFIX)                   \
   {SCAL, VEC, VF, MASK, VABI_PREFIX},
+#define TLI_DEFINE_VECFUNC_CC(SCAL, VEC, VF, MASK, VABI_PREFIX, CONV)          \
+  {SCAL, VEC, VF, MASK, VABI_PREFIX, CONV},
----------------
paulwalker-arm wrote:

Is a new macro required? It looks like TLI_DEFINE_VECFUNC is already library specific with some having a mask parameter and some not so I suspect you can just add the calling convention to the variant used by the one library you care about.

Whilst this kind of breaks the intent of this style of include, that boat looks to have sailed with the aforementioned mask change.

https://github.com/llvm/llvm-project/pull/135790