[clang] [Clang] Honor -flax-vector-conversions=none on some tests (PR #153433)

[Mikołaj Piróg via cfe-commits]

================
@@ -1,16 +1,16 @@
 // REQUIRES: x86-registered-target
-// RUN: %clang_cc1 %s -triple=i686-apple-darwin -target-feature +ssse3 -O1 -S -o - | FileCheck %s
+// RUN: %clang_cc1 %s -triple=i686-apple-darwin -target-feature +ssse3 -O1 -S -flax-vector-conversions=none -o - | FileCheck %s
 
 #define _mm_alignr_epi8(a, b, n) (__builtin_ia32_palignr128((a), (b), (n)))
-typedef __attribute__((vector_size(16))) int int4;
+typedef char __v16qi __attribute__((__vector_size__(16)));
 
 // CHECK: palignr $15, %xmm1, %xmm0
-int4 align1(int4 a, int4 b) { return _mm_alignr_epi8(a, b, 15); }
+__v16qi align1(__v16qi a, __v16qi b) { return _mm_alignr_epi8(a, b, 15); }
----------------
mikolaj-pirog wrote:

It wouldn't compile, since the arguments won't match the type. I think it's clearer to just change the type to what the macro expects 

https://github.com/llvm/llvm-project/pull/153433