[ARM] Fix arm_neon.h with -flax-vector-conversions=none, part 3

[efriedma]

Summary

It's completely impossible to check that I've actually found all the issues, due to the use of macros in arm_neon.h, but hopefully this time it'll take more than a few hours for someone to find another issue.

I have no idea why, but apparently there's a rule that some, but not all, builtins which should take an fp16 vector actually take an int8 vector as an argument. Fix this, and add test coverage.

[rsmith]

Thanks again! In my testing, this is enough to get the clang testsuite to pass with the default changed to -flax-vector-conversions=integer.

test/CodeGen/aarch64-v8.2a-neon-intrinsics.c

149–170 ↗

(On Diff #224489)

I already committed these three fixes in r374457.

[efriedma]

Ping

[SjoerdMeijer]

Yep, thanks again!

clang/test/CodeGen/aarch64-v8.2a-neon-intrinsics.c

 // RUN: %clang_cc1 -triple arm64-none-linux-gnu -target-feature +neon -target-feature +fullfp16 -target-feature +v8.2a\
-// RUN: -fallow-half-arguments-and-returns -S -disable-O0-optnone -emit-llvm -o - %s \
+// RUN: -fallow-half-arguments-and-returns -flax-vector-conversions=none -S -disable-O0-optnone -emit-llvm -o - %s \
 // RUN: | opt -S -mem2reg \
 // RUN: | FileCheck %s
 
 // REQUIRES: aarch64-registered-target
 
 #include <arm_neon.h>
 
 // CHECK-LABEL: test_vabs_f16

clang/utils/TableGen/NeonEmitter.cpp

if (hasSplat() && I == getSplatIdx()) {
       Arg += "}";
 
       CastToType = BaseType;
     } else {
       Arg = V.getName();
     }
 
     // Check if an explicit cast is needed.
-    if (CastToType.isVector() && LocalCK == ClassB) {
+    if (CastToType.isVector() &&
+        (LocalCK == ClassB || (T.isHalf() && !T.isScalarForMangling()))) {
       CastToType.makeInteger(8, true);
       Arg = "(" + CastToType.str() + ")" + Arg;
     } else if (CastToType.isVector() && LocalCK == ClassI) {
       CastToType.makeSigned();
       Arg = "(" + CastToType.str() + ")" + Arg;
     }
 
     S += Arg + ", ";