reimplement HIP_INIT as hip_impl::hip_init(), add hip_init() to some of the inlined API (#966)

* reimplement HIP_INIT as a function, expose it as hip_impl::hip_init()
so that it could be called from hipLaunchKernelGGL and other inlined
HIP functions

* Don't call hip_init from ihipPreLaunchKernel

include/hip/hcc_detail/functional_grid_launch.hpp

@@ -189,6 +189,7 @@ inline
 void hipLaunchKernelGGL(F kernel, const dim3& numBlocks, const dim3& dimBlocks,
                         std::uint32_t sharedMemBytes, hipStream_t stream,
                         Args... args) {
+    hip_impl::hip_init();
     auto kernarg = hip_impl::make_kernarg(
         kernel, std::tuple<Args...>{std::move(args)...});
     std::size_t kernarg_size = kernarg.size();
@@ -212,4 +213,4 @@ inline void hipLaunchKernel(F kernel, const dim3& numBlocks, const dim3& dimBloc
                             std::uint32_t groupMemBytes, hipStream_t stream, Args... args) {
     hipLaunchKernelGGL(kernel, numBlocks, dimBlocks, groupMemBytes, stream, hipLaunchParm{},
                        std::move(args)...);
-}
+}

include/hip/hcc_detail/hip_runtime_api.h

@@ -78,6 +78,10 @@ THE SOFTWARE.
   #define __dparm(x)
 #endif
 
+namespace hip_impl {
+hipError_t hip_init();
+}  // namespace hip_impl
+
 // Structure definitions:
 #ifdef __cplusplus
 extern "C" {
@@ -1396,7 +1400,7 @@ inline
 __attribute__((visibility("hidden")))
 hipError_t hipGetSymbolAddress(void** devPtr, const void* symbolName) {
     //HIP_INIT_API(hipGetSymbolAddress, devPtr, symbolName);
-
+    hip_impl::hip_init();
     size_t size = 0;
     return hipModuleGetGlobal(devPtr, &size, 0, (const char*)symbolName);
 }
@@ -1416,7 +1420,7 @@ inline
 __attribute__((visibility("hidden")))
 hipError_t hipGetSymbolSize(size_t* size, const void* symbolName) {
     // HIP_INIT_API(hipGetSymbolSize, size, symbolName);
-
+    hip_impl::hip_init();
     void* devPtr = nullptr;
     return hipModuleGetGlobal(&devPtr, size, 0, (const char*)symbolName);
 }