SWDEV-548892 - Always declare used ocml and ockl device libs functions (#2230)

Ignore __CLANG_HIP_RUNTIME_WRAPPER_INCLUDED__. This should not be relying
on declarations from the clang builtin headers. There is no issue declaring
the same intrinsics multiple times. This will enable removal of declarations
from the clang builtin headers.

projects/clr/hipamd/include/hip/amd_detail/hip_fp16_math_fwd.h

@@ -28,7 +28,6 @@ THE SOFTWARE.
 #if !defined(__HIPCC_RTC__)
 #include "host_defines.h"
 #endif
-#ifndef __CLANG_HIP_RUNTIME_WRAPPER_INCLUDED__
 extern "C" {
 __device__ __attribute__((const)) _Float16 __ocml_ceil_f16(_Float16);
 __device__ _Float16 __ocml_cos_f16(_Float16);
@@ -79,7 +78,6 @@ __device__ __attribute__((const)) _Float16 __ocml_cvtrtn_f16_f32(float);
 __device__ __attribute__((const)) _Float16 __ocml_cvtrtp_f16_f32(float);
 __device__ __attribute__((const)) _Float16 __ocml_cvtrtz_f16_f32(float);
 }
-#endif  // !__CLANG_HIP_RUNTIME_WRAPPER_INCLUDED__
 // TODO: remove these after they get into clang header __clang_hip_libdevice_declares.h'
 extern "C" {
 __device__ __attribute__((const)) _Float16 __ocml_fmax_f16(_Float16, _Float16);

projects/clr/hipamd/include/hip/amd_detail/math_fwd.h

@@ -55,7 +55,6 @@ __device__ __attribute__((const)) unsigned int __ockl_udot8(unsigned int, unsign
                                                             unsigned int, bool);
 #endif
 
-#if !__CLANG_HIP_RUNTIME_WRAPPER_INCLUDED__
 // BEGIN FLOAT
 __device__ __attribute__((const)) float __ocml_acos_f32(float);
 __device__ __attribute__((pure)) float __ocml_acosh_f32(float);
@@ -282,8 +281,6 @@ __device__ __attribute__((const)) double __ocml_fma_rtz_f64(double, double, doub
 // END INTRINSICS
 // END DOUBLE
 
-#endif  // !__CLANG_HIP_RUNTIME_WRAPPER_INCLUDED__
-
 #if defined(__cplusplus)
 }  // extern "C"
 #endif