[clang] [Clang][AMDGPU] Use unsigned data type for `__builtin_amdgcn_raw_buffer_store_*` (PR #99546)

[Shilei Tian via cfe-commits]

================
@@ -3,12 +3,12 @@
 
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
 
-typedef char i8;
-typedef short i16;
-typedef int i32;
-typedef int i64 __attribute__((ext_vector_type(2)));
-typedef int i96 __attribute__((ext_vector_type(3)));
-typedef int i128 __attribute__((ext_vector_type(4)));
+typedef unsigned char i8;
+typedef unsigned short i16;
+typedef unsigned int i32;
+typedef unsigned int i64 __attribute__((ext_vector_type(2)));
+typedef unsigned int i96 __attribute__((ext_vector_type(3)));
+typedef unsigned int i128 __attribute__((ext_vector_type(4)));
----------------
shiltian wrote:

That is because we said to use `<N x i32>` for consistency when the builtins were introduced.

https://github.com/llvm/llvm-project/pull/99546