Optimize calling UR functions (#20776)

_checkUrResult_ function was made as short as possible by moving away
all error handling to a separate non-inlined function and keeping common
success path tiny and additionally optimized by compiler
__builtin_expect. Also a few more simple inlines were added.

Author: lslusarczyk

source/loader/ur_lib.cpp

@@ -28,15 +28,9 @@
 #include <stdlib.h>
 
 namespace ur_lib {
-///////////////////////////////////////////////////////////////////////////////
-context_t *getContext() { return context_t::get_direct(); }
-
 ///////////////////////////////////////////////////////////////////////////////
 context_t::context_t() { parseEnvEnabledLayers(); }
 
-///////////////////////////////////////////////////////////////////////////////
-context_t::~context_t() {}
-
 void context_t::parseEnvEnabledLayers() {
   auto maybeEnableEnvVarVec = getenv_to_vec("UR_ENABLE_LAYERS");
   if (!maybeEnableEnvVarVec.has_value()) {

source/loader/ur_lib.hpp

@@ -61,7 +61,6 @@ class __urdlllocal context_t : public AtomicSingleton<context_t> {
 #endif
 
   context_t();
-  ~context_t();
 
   std::once_flag initOnce;
 
@@ -117,7 +116,7 @@ class __urdlllocal context_t : public AtomicSingleton<context_t> {
   void tearDownLayers() const;
 };
 
-context_t *getContext();
+inline context_t *getContext() { return context_t::get_direct(); }
 
 ur_result_t urLoaderConfigCreate(ur_loader_config_handle_t *phLoaderConfig);
 ur_result_t urLoaderConfigRetain(ur_loader_config_handle_t hLoaderConfig);