[SYCL][LIBCLC][NATIVECPU] Fix ldexp and popcount on Native CPU (#15687)

`ldexp` in `generic` (for `half`) and `popcount` in `native_cpu` are
implemented using clang builtins. This PR makes it so we call the
correct "overload" of the builtin:
* `__builtin_ldexpf16` for `half`
* `__builtin_popcountg` for unsigned integer types
* for signed integer types the implicit cast to `int` performed by
calling `__builtin_popcount` doesn't work due to sign extension, so we
explicitly cast to `unsigned char` or `unsigned short`, and use
`__builtin_popcountg`.

Author: PietroGhg

libclc/generic/include/math/clc_ldexp.h

@@ -7,5 +7,5 @@ _CLC_DEF _CLC_OVERLOAD double __clc_ldexp(double, int);
 
 #ifdef cl_khr_fp16
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
-_CLC_DEF _CLC_OVERLOAD float __clc_ldexp(half, int);
+_CLC_DEF _CLC_OVERLOAD half __clc_ldexp(half, int);
 #endif

libclc/generic/libspirv/math/clc_ldexp.cl

@@ -135,6 +135,6 @@ _CLC_DEF _CLC_OVERLOAD double __clc_ldexp(double x, int n) {
 
 #pragma OPENCL EXTENSION cl_khr_fp16 : enable
 
-_CLC_DEFINE_BINARY_BUILTIN(half, __clc_ldexp, __builtin_ldexp, half, int)
+_CLC_DEFINE_BINARY_BUILTIN(half, __clc_ldexp, __builtin_ldexpf16, half, int)
 
 #endif

libclc/native_cpu-unknown-linux/libspirv/integer/popcount.cl

@@ -2,12 +2,26 @@
 #include <clcmacro.h>
 #include <spirv/spirv.h>
 
+// We can't use __builtin_popcountg because it supports only unsigned
+// types, and we can't use __builtin_popcount because the implicit cast
+// to int doesn't work due to sign extension, so we use type punning to
+// preserve the bit pattern and avoid sign extension.
+
+#define DEF_POPCOUNT_HELPER(TYPE, UTYPE) \
+_CLC_OVERLOAD TYPE __popcount_helper(TYPE c) { \
+  return __builtin_popcountg(*(UTYPE*)&c); \
+}
+
+DEF_POPCOUNT_HELPER(char, unsigned char)
+DEF_POPCOUNT_HELPER(schar, unsigned char)
+DEF_POPCOUNT_HELPER(short, unsigned short)
+
 _CLC_DEFINE_UNARY_BUILTIN(int, __spirv_ocl_popcount, __builtin_popcount, int)
 _CLC_DEFINE_UNARY_BUILTIN(uint, __spirv_ocl_popcount, __builtin_popcount, uint)
-_CLC_DEFINE_UNARY_BUILTIN(short, __spirv_ocl_popcount, __builtin_popcount, short)
-_CLC_DEFINE_UNARY_BUILTIN(ushort, __spirv_ocl_popcount, __builtin_popcount, ushort)
-_CLC_DEFINE_UNARY_BUILTIN(long, __spirv_ocl_popcount, __builtin_popcount, long)
-_CLC_DEFINE_UNARY_BUILTIN(ulong, __spirv_ocl_popcount, __builtin_popcount, ulong)
-_CLC_DEFINE_UNARY_BUILTIN(char, __spirv_ocl_popcount, __builtin_popcount, char)
-_CLC_DEFINE_UNARY_BUILTIN(uchar, __spirv_ocl_popcount, __builtin_popcount, uchar)
-_CLC_DEFINE_UNARY_BUILTIN(schar, __spirv_ocl_popcount, __builtin_popcount, schar)
+_CLC_DEFINE_UNARY_BUILTIN(short, __spirv_ocl_popcount, __popcount_helper, short)
+_CLC_DEFINE_UNARY_BUILTIN(ushort, __spirv_ocl_popcount, __builtin_popcountg, ushort)
+_CLC_DEFINE_UNARY_BUILTIN(long, __spirv_ocl_popcount, __builtin_popcountl, long)
+_CLC_DEFINE_UNARY_BUILTIN(ulong, __spirv_ocl_popcount, __builtin_popcountl, ulong)
+_CLC_DEFINE_UNARY_BUILTIN(char, __spirv_ocl_popcount, __popcount_helper, char)
+_CLC_DEFINE_UNARY_BUILTIN(uchar, __spirv_ocl_popcount, __builtin_popcountg, uchar)
+_CLC_DEFINE_UNARY_BUILTIN(schar, __spirv_ocl_popcount, __popcount_helper, schar)