diff --git a/src/ctrl/ctrl_core.c b/src/ctrl/ctrl_core.c
index 03c60e83..ccb3076d 100644
--- a/src/ctrl/ctrl_core.c
+++ b/src/ctrl/ctrl_core.c
@@ -4147,6 +4147,63 @@ ctrl_thread__next_dmn_event(Arena *arena, DMN_CtrlCtx *ctrl_ctx, CTRL_Msg *msg,
   }
   ctrl_c2u_push_events(&evts);
   
+  //- rjf: when a new process is loaded, for the first module, pre-emptively try
+  // to open all adjacent debug infos. with debug events, we learn about loaded
+  // modules serially, and we need to completely load debug info before continuing.
+  // for massive projects, this is a problem, because completely loading debug info
+  // isn't a trivial cost, and there are often 100s of DLLs.
+  //
+  // an imperfect but usually reasonable heuristic is to look at adjacent debug info
+  // files, in the same directory as the initially loaded, and pre-emptively convert
+  // all of them (which for us is the heaviest part of debug info loading, if native
+  // RDI is not used).
+  if(event->kind == DMN_EventKind_LoadModule)
+  {
+    CTRL_Handle process_handle = ctrl_handle_make(CTRL_MachineID_Local, event->process);
+    CTRL_Handle loaded_module_handle = ctrl_handle_make(CTRL_MachineID_Local, event->module);
+    CTRL_Entity *process = ctrl_entity_from_handle(ctrl_state->ctrl_thread_entity_store, process_handle);
+    CTRL_Entity *loaded_module = ctrl_entity_from_handle(ctrl_state->ctrl_thread_entity_store, loaded_module_handle);
+    B32 is_first = 1;
+    for(CTRL_Entity *child = process->first; child != &ctrl_entity_nil; child = child->next)
+    {
+      if(child->kind == CTRL_EntityKind_Module && child != loaded_module)
+      {
+        is_first = 0;
+        break;
+      }
+    }
+    if(is_first)
+    {
+      DI_Key loaded_di_key = ctrl_dbgi_key_from_module(loaded_module);
+      String8 loaded_di_name = str8_skip_last_slash(loaded_di_key.path);
+      String8 containing_folder_path = str8_chop_last_slash(loaded_di_key.path);
+      if(containing_folder_path.size < loaded_di_key.path.size)
+      {
+        String8 debug_info_ext = str8_skip_last_dot(loaded_di_key.path);
+        OS_FileIter *it = os_file_iter_begin(scratch.arena, containing_folder_path, OS_FileIterFlag_SkipFolders);
+        DI_KeyList preemptively_loaded_keys = {0};
+        for(OS_FileInfo info = {0}; os_file_iter_next(scratch.arena, it, &info);)
+        {
+          if(str8_match(str8_skip_last_dot(info.name), debug_info_ext, StringMatchFlag_CaseInsensitive) &&
+             !str8_match(loaded_di_name, info.name, StringMatchFlag_CaseInsensitive))
+          {
+            DI_Key key = {push_str8f(scratch.arena, "%S/%S", containing_folder_path, info.name), info.props.modified};
+            di_open(&key);
+            di_key_list_push(scratch.arena, &preemptively_loaded_keys, &key);
+          }
+        }
+        for(DI_KeyNode *n = preemptively_loaded_keys.first; n != 0; n = n->next)
+        {
+          DI_Scope *di_scope = di_scope_open();
+          RDI_Parsed *rdi = di_rdi_from_key(di_scope, &n->v, max_U64);
+          di_scope_close(di_scope);
+          di_close(&n->v);
+        }
+        os_file_iter_end(it);
+      }
+    }
+  }
+  
   //- rjf: clear process memory cache, if we've just started a lone process
   if(event->kind == DMN_EventKind_CreateProcess && ctrl_state->process_counter == 1)
   {
@@ -5797,6 +5854,7 @@ ctrl_u2ms_dequeue_req(CTRL_Handle *out_process, Rng1U64 *out_vaddr_range, B32 *o
 internal void
 ctrl_mem_stream_thread__entry_point(void *p)
 {
+  ThreadNameF("[ctrl] mem stream thread #%I64u", (U64)p);
   CTRL_ProcessMemoryCache *cache = &ctrl_state->process_memory_cache;
   for(;;)
   {
@@ -5806,6 +5864,7 @@ ctrl_mem_stream_thread__entry_point(void *p)
     B32 zero_terminated = 0;
     ctrl_u2ms_dequeue_req(&process, &vaddr_range, &zero_terminated);
     U128 key = ctrl_calc_hash_store_key_from_process_vaddr_range(process, vaddr_range, zero_terminated);
+    ProfBegin("memory stream request");
     
     //- rjf: unpack process memory cache key
     U64 process_hash = ctrl_hash_from_string(str8_struct(&process));
@@ -5956,5 +6015,6 @@ ctrl_mem_stream_thread__entry_point(void *p)
     
     //- rjf: broadcast changes
     os_condition_variable_broadcast(process_stripe->cv);
+    ProfEnd();
   }
 }
diff --git a/src/font_cache/font_cache.c b/src/font_cache/font_cache.c
index ae23c3ec..14390a25 100644
--- a/src/font_cache/font_cache.c
+++ b/src/font_cache/font_cache.c
@@ -575,46 +575,48 @@ fnt_push_run_from_string(Arena *arena, FNT_Tag tag, F32 size, F32 base_align_px,
   FNT_Hash2StyleRasterCacheNode *hash2style_node = fnt_hash2style_from_tag_size_flags(tag, size, flags);
   
   //- rjf: decode string & produce run pieces
-  B32 first = 1;
   FNT_PieceChunkList piece_chunks = {0};
   Vec2F32 dim = {0};
   B32 font_handle_mapped_on_miss = 0;
   FP_Handle font_handle = {0};
   U64 piece_substring_start_idx = 0;
-  for(U64 idx = 0; idx < string.size; first = 0)
+  U64 piece_substring_end_idx = 0;
+  for(U64 idx = 0; idx <= string.size;)
   {
     //- rjf: decode next codepoint & get piece substring, or continuation rule
-    String8 piece_substring;
+    U8 byte = (idx < string.size ? string.str[idx] : 0);
     B32 need_another_codepoint = 0;
-    switch(utf8_class[string.str[idx]>>3])
+    if(byte == 0)
+    {
+      idx += 1;
+    }
+    else switch(utf8_class[byte>>3])
     {
       case 1:
       {
-        piece_substring.str = &string.str[idx];
-        piece_substring.size = 1;
         idx += 1;
+        piece_substring_end_idx += 1;
+        need_another_codepoint = 0;
       }break;
       default:
       {
         UnicodeDecode decode = utf8_decode(string.str+idx, string.size-idx);
         idx += decode.inc;
-        if(decode.inc == 0) { break; }
-        piece_substring.str = string.str + piece_substring_start_idx;
-        piece_substring.size = decode.inc;
-        // NOTE(rjf): assuming 1 codepoint per piece for now.
+        piece_substring_end_idx += decode.inc;
+        need_another_codepoint = 0;
       }break;
     }
     
-    //- rjf: need another codepoint? -> continue
-    if(need_another_codepoint)
+    //- rjf: need another codepoint, or have no substring? -> continue
+    if(need_another_codepoint || piece_substring_end_idx == piece_substring_start_idx)
     {
       continue;
     }
     
-    //- rjf: do not need another codepoint? -> bump piece start idx
-    {
-      piece_substring_start_idx = idx;
-    }
+    //- rjf: do not need another codepoint? -> grab substring, bump piece start idx
+    String8 piece_substring = str8_substr(string, r1u64(piece_substring_start_idx, piece_substring_end_idx));
+    piece_substring_start_idx = idx;
+    piece_substring_end_idx = idx;
     
     //- rjf: determine if this piece is a tab - if so, use space info to draw
     B32 is_tab = (piece_substring.size == 1 && piece_substring.str[0] == '\t');
diff --git a/src/raddbg/raddbg_widgets.c b/src/raddbg/raddbg_widgets.c
index 145e13af..1bda7c4a 100644
--- a/src/raddbg/raddbg_widgets.c
+++ b/src/raddbg/raddbg_widgets.c
@@ -2697,6 +2697,7 @@ rd_help_label(String8 string)
 internal DR_FancyStringList
 rd_fancy_string_list_from_code_string(Arena *arena, F32 alpha, B32 indirection_size_change, Vec4F32 base_color, String8 string)
 {
+  ProfBeginFunction();
   Temp scratch = scratch_begin(&arena, 1);
   DR_FancyStringList fancy_strings = {0};
   TXT_TokenArray tokens = txt_token_array_from_string__c_cpp(scratch.arena, 0, string);
@@ -2857,6 +2858,7 @@ rd_fancy_string_list_from_code_string(Arena *arena, F32 alpha, B32 indirection_s
     indirection_counter = ClampBot(0, indirection_counter);
   }
   scratch_end(scratch);
+  ProfEnd();
   return fancy_strings;
 }
 
@@ -2877,6 +2879,8 @@ rd_code_label(F32 alpha, B32 indirection_size_change, Vec4F32 base_color, String
 internal UI_Signal
 rd_line_edit(RD_LineEditFlags flags, S32 depth, FuzzyMatchRangeList *matches, TxtPt *cursor, TxtPt *mark, U8 *edit_buffer, U64 edit_buffer_size, U64 *edit_string_size_out, B32 *expanded_out, String8 pre_edit_value, String8 string)
 {
+  ProfBeginFunction();
+  
   //- rjf: unpack visual metrics
   F32 expander_size_px = ui_top_font_size()*2.f;
   
@@ -3282,6 +3286,7 @@ rd_line_edit(RD_LineEditFlags flags, S32 depth, FuzzyMatchRangeList *matches, Tx
   if(is_auto_focus_hot) { ui_pop_focus_hot(); }
   if(is_auto_focus_active) { ui_pop_focus_active(); }
   
+  ProfEnd();
   return sig;
 }
 
diff --git a/src/text_cache/text_cache.c b/src/text_cache/text_cache.c
index 08e6af17..3830b205 100644
--- a/src/text_cache/text_cache.c
+++ b/src/text_cache/text_cache.c
@@ -160,7 +160,9 @@ txt_token_array_from_list(Arena *arena, TXT_TokenList *list)
 internal TXT_TokenArray
 txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, String8 string)
 {
+  ProfBeginFunction();
   Temp scratch = scratch_begin(&arena, 1);
+  U64 chunk_size = Clamp(8, string.size/8, 4096);
   
   //- rjf: generate token list
   TXT_TokenChunkList tokens = {0};
@@ -234,7 +236,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
         else
         {
           TXT_Token token = {TXT_TokenKind_Error, r1u64(idx, idx+1)};
-          txt_token_chunk_list_push(scratch.arena, &tokens, 4096, &token);
+          txt_token_chunk_list_push(scratch.arena, &tokens, chunk_size, &token);
         }
       }
       
@@ -420,7 +422,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
               break;
             }
           }
-          txt_token_chunk_list_push(scratch.arena, &tokens, 4096, &token);
+          txt_token_chunk_list_push(scratch.arena, &tokens, chunk_size, &token);
         }
         
         // rjf: split symbols by maximum-munch-rule
@@ -462,7 +464,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
                 found = 1;
                 next_off = off + c_cpp_multichar_symbol_strings[idx].size;
                 TXT_Token token = {TXT_TokenKind_Symbol, r1u64(active_token_start_idx+off, active_token_start_idx+next_off)};
-                txt_token_chunk_list_push(scratch.arena, &tokens, 4096, &token);
+                txt_token_chunk_list_push(scratch.arena, &tokens, chunk_size, &token);
                 break;
               }
             }
@@ -470,7 +472,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
             {
               next_off = off+1;
               TXT_Token token = {TXT_TokenKind_Symbol, r1u64(active_token_start_idx+off, active_token_start_idx+next_off)};
-              txt_token_chunk_list_push(scratch.arena, &tokens, 4096, &token);
+              txt_token_chunk_list_push(scratch.arena, &tokens, chunk_size, &token);
             }
           }
         }
@@ -478,7 +480,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
         // rjf: all other tokens
         else
         {
-          txt_token_chunk_list_push(scratch.arena, &tokens, 4096, &token);
+          txt_token_chunk_list_push(scratch.arena, &tokens, chunk_size, &token);
         }
         
         // rjf: increment by ender padding
@@ -497,6 +499,7 @@ txt_token_array_from_string__c_cpp(Arena *arena, U64 *bytes_processed_counter, S
   //- rjf: token list -> token array
   TXT_TokenArray result = txt_token_array_from_chunk_list(arena, &tokens);
   scratch_end(scratch);
+  ProfEnd();
   return result;
 }