Reduce perf impact of redundant ittnotify calls

[tlwilmar]

Summary

Improved performance of ittnotify calls by request from ittnotify
owner: calls to __itt_string_handle_create made unique (it was called multiple times).

Patch by Andrey Churbanov

[AndreyChurbanov]

LGTM

runtime/src/kmp_itt.h

const int KMP_MAX_FRAME_DOMAINS = 512; // Maximum number of frame domains to use (maps to
                                            // different OpenMP regions in the user source code).
     extern kmp_int32 __kmp_barrier_domain_count;
     extern kmp_int32 __kmp_region_domain_count;
     extern __itt_domain* __kmp_itt_barrier_domains[KMP_MAX_FRAME_DOMAINS];
     extern __itt_domain* __kmp_itt_region_domains[KMP_MAX_FRAME_DOMAINS];
     extern __itt_domain* __kmp_itt_imbalance_domains[KMP_MAX_FRAME_DOMAINS];
     extern kmp_int32 __kmp_itt_region_team_size[KMP_MAX_FRAME_DOMAINS];
     extern __itt_domain * metadata_domain;
+    extern __itt_string_handle * string_handle_imbl;
+    extern __itt_string_handle * string_handle_loop;
+    extern __itt_string_handle * string_handle_sngl;
 
 #else
 
 // Null definitions of the synchronization tracing functions.
 # define KMP_FSYNC_PREPARE(   obj )        ((void)0)
 # define KMP_FSYNC_CANCEL(    obj )        ((void)0)
 # define KMP_FSYNC_ACQUIRED(  obj )        ((void)0)
 # define KMP_FSYNC_RELEASING( obj )        ((void)0)

runtime/src/kmp_itt.c

#if USE_ITT_NOTIFY
 
     kmp_int32 __kmp_barrier_domain_count;
     kmp_int32 __kmp_region_domain_count;
     __itt_domain* __kmp_itt_barrier_domains[KMP_MAX_FRAME_DOMAINS];
     __itt_domain* __kmp_itt_region_domains[KMP_MAX_FRAME_DOMAINS];
     __itt_domain* __kmp_itt_imbalance_domains[KMP_MAX_FRAME_DOMAINS];
     kmp_int32 __kmp_itt_region_team_size[KMP_MAX_FRAME_DOMAINS];
     __itt_domain * metadata_domain = NULL;
+    __itt_string_handle * string_handle_imbl = NULL;
+    __itt_string_handle * string_handle_loop = NULL;
+    __itt_string_handle * string_handle_sngl = NULL;
 
     #include "kmp_version.h"
     #include "kmp_i18n.h"
     #include "kmp_str.h"
 
     KMP_BUILD_ASSERT( sizeof( kmp_itt_mark_t ) == sizeof( __itt_mark_type ) );
 
     /*

runtime/src/kmp_itt.inl

 LINKAGE void
 __kmp_itt_metadata_imbalance( int gtid, kmp_uint64 begin, kmp_uint64 end, kmp_uint64 imbalance, kmp_uint64 reduction ) {
 #if USE_ITT_NOTIFY
     if( metadata_domain == NULL) {
         __kmp_acquire_bootstrap_lock( & metadata_lock );
         if( metadata_domain == NULL) {
             __itt_suppress_push(__itt_suppress_memory_errors);
             metadata_domain = __itt_domain_create( "OMP Metadata" );
+            string_handle_imbl = __itt_string_handle_create( "omp_metadata_imbalance");
+            string_handle_loop = __itt_string_handle_create( "omp_metadata_loop");
+            string_handle_sngl = __itt_string_handle_create( "omp_metadata_single");
             __itt_suppress_pop();
         }
         __kmp_release_bootstrap_lock( & metadata_lock );
     }
 
-    __itt_string_handle * string_handle = __itt_string_handle_create( "omp_metadata_imbalance");
-
     kmp_uint64 imbalance_data[ 4 ];
     imbalance_data[ 0 ] = begin;
     imbalance_data[ 1 ] = end;
     imbalance_data[ 2 ] = imbalance;
     imbalance_data[ 3 ] = reduction;
 
-    __itt_metadata_add(metadata_domain, __itt_null, string_handle, __itt_metadata_u64, 4, imbalance_data);
+    __itt_metadata_add(metadata_domain, __itt_null, string_handle_imbl, __itt_metadata_u64, 4, imbalance_data);
 #endif
 } // __kmp_itt_metadata_imbalance
 
 // -------------------------------------------------------------------------------------------------
 
 LINKAGE void
 __kmp_itt_metadata_loop( ident_t * loc, kmp_uint64 sched_type, kmp_uint64 iterations, kmp_uint64 chunk ) {
 #if USE_ITT_NOTIFY
     if( metadata_domain == NULL) {
         __kmp_acquire_bootstrap_lock( & metadata_lock );
         if( metadata_domain == NULL) {
             __itt_suppress_push(__itt_suppress_memory_errors);
             metadata_domain = __itt_domain_create( "OMP Metadata" );
+            string_handle_imbl = __itt_string_handle_create( "omp_metadata_imbalance");
+            string_handle_loop = __itt_string_handle_create( "omp_metadata_loop");
+            string_handle_sngl = __itt_string_handle_create( "omp_metadata_single");
             __itt_suppress_pop();
         }
         __kmp_release_bootstrap_lock( & metadata_lock );
     }
 
-    __itt_string_handle * string_handle = __itt_string_handle_create( "omp_metadata_loop");
-
     // Parse line and column from psource string: ";file;func;line;col;;"
     char * s_line;
     char * s_col;
     KMP_DEBUG_ASSERT(loc->psource);
 #ifdef __cplusplus
     s_line = strchr((char*)loc->psource, ';');
 #else
     s_line = strchr(loc->psource, ';');
 #endif
     KMP_DEBUG_ASSERT(s_line);
     s_line = strchr(s_line + 1, ';');   // 2-nd semicolon
     KMP_DEBUG_ASSERT(s_line);
     s_line = strchr(s_line + 1, ';');   // 3-rd semicolon
     KMP_DEBUG_ASSERT(s_line);
     s_col = strchr(s_line + 1, ';');    // 4-th semicolon
     KMP_DEBUG_ASSERT(s_col);
 
     kmp_uint64 loop_data[ 5 ];
     loop_data[ 0 ] = atoi(s_line + 1); // read line
     loop_data[ 1 ] = atoi(s_col + 1);  // read column
     loop_data[ 2 ] = sched_type;
     loop_data[ 3 ] = iterations;
     loop_data[ 4 ] = chunk;
 
-    __itt_metadata_add(metadata_domain, __itt_null, string_handle, __itt_metadata_u64, 5, loop_data);
+    __itt_metadata_add(metadata_domain, __itt_null, string_handle_loop, __itt_metadata_u64, 5, loop_data);
 #endif
 } // __kmp_itt_metadata_loop
 
 // -------------------------------------------------------------------------------------------------
 
 LINKAGE void
 __kmp_itt_metadata_single( ident_t * loc ) {
 #if USE_ITT_NOTIFY
     if( metadata_domain == NULL) {
         __kmp_acquire_bootstrap_lock( & metadata_lock );
         if( metadata_domain == NULL) {
             __itt_suppress_push(__itt_suppress_memory_errors);
             metadata_domain = __itt_domain_create( "OMP Metadata" );
+            string_handle_imbl = __itt_string_handle_create( "omp_metadata_imbalance");
+            string_handle_loop = __itt_string_handle_create( "omp_metadata_loop");
+            string_handle_sngl = __itt_string_handle_create( "omp_metadata_single");
             __itt_suppress_pop();
         }
         __kmp_release_bootstrap_lock( & metadata_lock );
     }
 
-    __itt_string_handle * string_handle = __itt_string_handle_create( "omp_metadata_single");
     kmp_str_loc_t str_loc = __kmp_str_loc_init( loc->psource, 1 );
     kmp_uint64 single_data[ 2 ];
     single_data[ 0 ] = str_loc.line;
     single_data[ 1 ] = str_loc.col;
 
     __kmp_str_loc_free( &str_loc );
 
-    __itt_metadata_add(metadata_domain, __itt_null, string_handle, __itt_metadata_u64, 2, single_data);
+    __itt_metadata_add(metadata_domain, __itt_null, string_handle_sngl, __itt_metadata_u64, 2, single_data);
 #endif
 } // __kmp_itt_metadata_single
 
 // -------------------------------------------------------------------------------------------------
 
 LINKAGE void
 __kmp_itt_region_starting( int gtid ) {
 #if USE_ITT_NOTIFY