replace extra gen parameters in artifact creation hook; replace with explicit cancellation signal, passed by caller

src/artifact_cache/artifact_cache.c

@@ -161,8 +161,8 @@ ac_artifact_from_key_(Access *access, String8 key, AC_ArtifactParams *params, U6
           }
           n->v.key = str8_copy(req_batch->arena, key);
           n->v.gen = params->gen;
-          n->v.last_requested_gen = &node->last_requested_gen;
           n->v.create = params->create;
+          n->v.cancel_signal = params->cancel_signal;
         }
         cond_var_broadcast(async_tick_start_cond_var);
         ins_atomic_u32_eval_assign(&async_loop_again, 1);
@@ -356,7 +356,7 @@ ac_async_tick(void)
         
         // rjf: compute val
         B32 retry = 0;
-        AC_Artifact val = r->create(r->key, r->gen, r->last_requested_gen, &retry);
+        AC_Artifact val = r->create(r->key, r->cancel_signal, &retry);
         
         // rjf: retry? -> resubmit request
         if(retry && lane_idx() == 0)
@@ -458,7 +458,7 @@ ac_async_tick(void)
         
         // rjf: compute val
         B32 retry = 0;
-        AC_Artifact val = r->create(r->key, r->gen, r->last_requested_gen, &retry);
+        AC_Artifact val = r->create(r->key, r->cancel_signal, &retry);
         
         // rjf: retry? -> resubmit request
         if(retry)

src/artifact_cache/artifact_cache.h

@@ -16,7 +16,7 @@ struct AC_Artifact
 ////////////////////////////////
 //~ rjf: Artifact Computation Function Types
 
-typedef AC_Artifact AC_CreateFunctionType(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+typedef AC_Artifact AC_CreateFunctionType(String8 key, B32 *cancel_signal, B32 *retry_out);
 typedef void AC_DestroyFunctionType(AC_Artifact artifact);
 
 typedef U32 AC_Flags;
@@ -37,6 +37,7 @@ struct AC_ArtifactParams
   U64 gen;
   U64 evict_threshold_us;
   B32 *stale_out;
+  B32 *cancel_signal;
   AC_Flags flags;
 };
 
@@ -48,7 +49,7 @@ struct AC_Request
 {
   String8 key;
   U64 gen;
-  U64 *last_requested_gen;
+  B32 *cancel_signal;
   AC_CreateFunctionType *create;
 };
 
@@ -138,6 +139,11 @@ global AC_Shared *ac_shared = 0;
 
 internal void ac_init(void);
 
+////////////////////////////////
+//~ rjf: Helpers
+
+internal B32 ac_cancelled(void);
+
 ////////////////////////////////
 //~ rjf: Cache Lookups
 

src/ctrl/ctrl_core.c

@@ -6134,7 +6134,7 @@ ctrl_thread__single_step(DMN_CtrlCtx *ctrl_ctx, CTRL_Msg *msg)
 //~ rjf: Process Memory Artifact Cache Hooks / Lookups
 
 internal AC_Artifact
-ctrl_memory_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+ctrl_memory_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   AC_Artifact artifact = {0};
   {
@@ -6503,7 +6503,7 @@ ctrl_process_write(CTRL_Handle process, Rng1U64 range, void *src)
 //~ rjf: Call Stack Artifact Cache Hooks / Lookups
 
 internal AC_Artifact
-ctrl_call_stack_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+ctrl_call_stack_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   AC_Artifact artifact = {0};
   {
@@ -6699,7 +6699,7 @@ ctrl_call_stack_from_thread(Access *access, CTRL_Handle thread_handle, B32 high_
 //~ rjf: Call Stack Tree Artifact Cache Hooks / Lookups
 
 internal AC_Artifact
-ctrl_call_stack_tree_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+ctrl_call_stack_tree_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   Temp scratch = scratch_begin(0, 0);
   Access *access = access_open();

src/ctrl/ctrl_core.h

@@ -1002,7 +1002,7 @@ internal void ctrl_thread__single_step(DMN_CtrlCtx *ctrl_ctx, CTRL_Msg *msg);
 ////////////////////////////////
 //~ rjf: Process Memory Artifact Cache Hooks / Lookups
 
-internal AC_Artifact ctrl_memory_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact ctrl_memory_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void ctrl_memory_artifact_destroy(AC_Artifact artifact);
 internal C_Key ctrl_key_from_process_vaddr_range(CTRL_Handle process, Rng1U64 vaddr_range, B32 zero_terminated, U64 endt_us, B32 *out_is_stale);
 
@@ -1017,14 +1017,14 @@ internal B32 ctrl_process_write(CTRL_Handle process, Rng1U64 range, void *src);
 ////////////////////////////////
 //~ rjf: Call Stack Artifact Cache Hooks / Lookups
 
-internal AC_Artifact ctrl_call_stack_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact ctrl_call_stack_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void ctrl_call_stack_artifact_destroy(AC_Artifact artifact);
 internal CTRL_CallStack ctrl_call_stack_from_thread(Access *access, CTRL_Handle thread_handle, B32 high_priority, U64 endt_us);
 
 ////////////////////////////////
 //~ rjf: Call Stack Tree Artifact Cache Hooks / Lookups
 
-internal AC_Artifact ctrl_call_stack_tree_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact ctrl_call_stack_tree_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void ctrl_call_stack_tree_artifact_destroy(AC_Artifact artifact);
 internal CTRL_CallStackTree ctrl_call_stack_tree(Access *access, U64 endt_us);
 

src/dbg_info/dbg_info2.c

@@ -861,7 +861,7 @@ di2_conversion_completion_signal_receiver_thread_entry_point(void *p)
 //~ rjf: Search Artifact Cache Hooks / Lookups
 
 internal AC_Artifact
-di2_search_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+di2_search_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   ProfBeginFunction();
   Access *access = access_open();
@@ -1065,9 +1065,144 @@ di2_search_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_
     }
     lane_sync();
     
-    //- rjf: flatten into array
+    //- rjf: produce sort records
+    typedef struct SortRecord SortRecord;
+    struct SortRecord
+    {
+      U64 key;
+      DI2_SearchItem *item;
+    };
+    U64 sort_records_count = all_items->total_count;
+    SortRecord *sort_records = 0;
+    SortRecord *sort_records__swap = 0;
+    ProfScope("produce sort records")
+    {
+      if(lane_idx() == 0)
+      {
+        sort_records = push_array(scratch.arena, SortRecord, sort_records_count);
+      }
+      if(lane_idx() == lane_from_task_idx(1))
+      {
+        sort_records__swap = push_array(scratch.arena, SortRecord, sort_records_count);
+      }
+      lane_sync_u64(&sort_records, 0);
+      lane_sync_u64(&sort_records__swap, lane_from_task_idx(1));
+      for EachNode(n, DI2_SearchItemChunk, all_items->first)
+      {
+        Rng1U64 range = lane_range(n->count);
+        U64 dst_idx = n->base_idx + range.min;
+        for EachInRange(n_idx, range)
+        {
+          DI2_SearchItem *item = &n->v[n_idx];
+          sort_records[dst_idx].item = item;
+          sort_records[dst_idx].key = (((item->missed_size & 0xffffffffull) << 32) | (u64_hash_from_seed_str8(item->idx, str8_struct(&key)) & 0xffffffffull));
+          dst_idx += 1;
+        }
+      }
+    }
+    lane_sync();
+    
+    //- rjf: sort records
+    ProfScope("sort records")
+    {
+      //- rjf: set up common data
+      U64 bits_per_digit = 8;
+      U64 digits_count = 64 / bits_per_digit;
+      U64 num_possible_values_per_digit = 1 << bits_per_digit;
+      U32 **lanes_digit_counts = 0;
+      U32 **lanes_digit_offsets = 0;
+      if(lane_idx() == 0)
+      {
+        lanes_digit_counts = push_array(scratch.arena, U32 *, lane_count());
+        lanes_digit_offsets = push_array(scratch.arena, U32 *, lane_count());
+      }
+      lane_sync_u64(&lanes_digit_counts, 0);
+      lane_sync_u64(&lanes_digit_offsets, 0);
+      
+      //- rjf: set up this lane
+      lanes_digit_counts[lane_idx()] = push_array(scratch.arena, U32, num_possible_values_per_digit);
+      lanes_digit_offsets[lane_idx()] = push_array(scratch.arena, U32, num_possible_values_per_digit);
+      SortRecord *src = sort_records;
+      SortRecord *dst = sort_records__swap;
+      U64 count = sort_records_count;
+      
+      //- rjf: do all per-digit sorts
+      for EachIndex(digit_idx, digits_count)
+      {
+        // rjf: count digit value occurrences per-lane
+        {
+          U32 *digit_counts = lanes_digit_counts[lane_idx()];
+          MemoryZero(digit_counts, sizeof(digit_counts[0])*num_possible_values_per_digit);
+          Rng1U64 range = lane_range(count);
+          for EachInRange(idx, range)
+          {
+            SortRecord *r = &src[idx];
+            U16 digit_value = (U16)(U8)(r->key >> (digit_idx*bits_per_digit));
+            digit_counts[digit_value] += 1;
+          }
+        }
+        lane_sync();
+        
+        // rjf: compute thread * digit value *relative* offset table
+        {
+          Rng1U64 range = lane_range(num_possible_values_per_digit);
+          for EachInRange(value_idx, range)
+          {
+            U64 layout_off = 0;
+            for EachIndex(lane_idx, lane_count())
+            {
+              lanes_digit_offsets[lane_idx][value_idx] = layout_off;
+              layout_off += lanes_digit_counts[lane_idx][value_idx];
+            }
+          }
+        }
+        lane_sync();
+        
+        // rjf: convert relative offsets -> absolute offsets
+        if(lane_idx() == 0)
+        {
+          U64 last_off = 0;
+          U64 num_of_nonzero_digit = 0;
+          for EachIndex(value_idx, num_possible_values_per_digit)
+          {
+            for EachIndex(lane_idx, lane_count())
+            {
+              lanes_digit_offsets[lane_idx][value_idx] += last_off;
+            }
+            last_off = lanes_digit_offsets[lane_count()-1][value_idx] + lanes_digit_counts[lane_count()-1][value_idx];
+          }
+          // NOTE(rjf): required that: (last_off == element_count)
+        }
+        lane_sync();
+        
+        // rjf: move
+        {
+          U32 *lane_digit_offsets = lanes_digit_offsets[lane_idx()];
+          Rng1U64 range = lane_range(count);
+          for EachInRange(idx, range)
+          {
+            SortRecord *src_r = &src[idx];
+            U16 digit_value = (U16)(U8)(src_r->key >> (digit_idx*bits_per_digit));
+            U64 dst_off = lane_digit_offsets[digit_value];
+            lane_digit_offsets[digit_value] += 1;
+            MemoryCopyStruct(&dst[dst_off], src_r);
+          }
+        }
+        lane_sync();
+        
+        // rjf: swap
+        {
+          SortRecord *swap = src;
+          src = dst;
+          dst = swap;
+        }
+      }
+    }
+    lane_sync();
+    
+    //- rjf: produce final array
     DI2_SearchItemArray items = {0};
-    ProfScope("flatten into array")
+    ProfScope("produce final array")
     {
       if(lane_idx() == 0)
       {
@@ -1076,19 +1211,15 @@ di2_search_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_
       }
       lane_sync_u64(&items.count, 0);
       lane_sync_u64(&items.v, 0);
-      for EachNode(n, DI2_SearchItemChunk, all_items->first)
+      Rng1U64 range = lane_range(sort_records_count);
+      for EachInRange(idx, range)
       {
-        Rng1U64 range = lane_range(n->count);
-        U64 dst_idx = n->base_idx + range.min;
-        MemoryCopy(&items.v[dst_idx], n->v, sizeof(n->v[0]) * dim_1u64(range));
+        SortRecord *record = &sort_records[idx];
+        DI2_SearchItem *dst_item = &items.v[idx];
+        MemoryCopyStruct(dst_item, record->item);
       }
     }
-    
-    //- rjf: sort items
-    ProfScope("sort items")
-    {
-      
-    }
+    lane_sync();
     
     //- rjf: bundle as artifact
     artifact.u64[0] = (U64)arena;

src/dbg_info/dbg_info2.h

@@ -274,7 +274,7 @@ internal void di2_conversion_completion_signal_receiver_thread_entry_point(void
 ////////////////////////////////
 //~ rjf: Search Artifact Cache Hooks / Lookups
 
-internal AC_Artifact di2_search_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact di2_search_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void di2_search_artifact_destroy(AC_Artifact artifact);
 internal DI2_SearchItemArray di2_search_item_array_from_target_query(Access *access, RDI_SectionKind target, String8 query, U64 endt_us);
 

src/disasm/disasm.c

@@ -266,7 +266,7 @@ struct DASM_Artifact
 };
 
 internal AC_Artifact
-dasm_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+dasm_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   DASM_Artifact *artifact = 0;
   if(lane_idx() == 0)

src/disasm/disasm.h

@@ -197,7 +197,7 @@ internal U64 dasm_line_array_code_off_from_idx(DASM_LineArray *array, U64 idx);
 ////////////////////////////////
 //~ rjf: Artifact Cache Hooks / Lookups
 
-internal AC_Artifact dasm_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact dasm_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void dasm_artifact_destroy(AC_Artifact artifact);
 internal DASM_Info dasm_info_from_hash_params(Access *access, U128 hash, DASM_Params *params);
 internal DASM_Info dasm_info_from_key_params(Access *access, C_Key key, DASM_Params *params, U128 *hash_out);

src/file_stream/file_stream.c

@@ -32,7 +32,7 @@ fs_change_gen(void)
 //~ rjf: Cache Interaction
 
 internal AC_Artifact
-fs_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+fs_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   ProfBeginFunction();
   Temp scratch = scratch_begin(0, 0);

src/file_stream/file_stream.h

@@ -55,7 +55,7 @@ internal U64 fs_change_gen(void);
 ////////////////////////////////
 //~ rjf: Artifact Cache Hooks / Accessing API
 
-internal AC_Artifact fs_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact fs_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void fs_artifact_destroy(AC_Artifact artifact);
 
 internal C_Key fs_key_from_path_range(String8 path, Rng1U64 range, U64 endt_us);

src/raddbg/raddbg_views.c

@@ -3749,7 +3749,7 @@ struct RD_BitmapCanvasBoxDrawData
 };
 
 internal AC_Artifact
-rd_bitmap_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+rd_bitmap_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   Access *access = access_open();
   
@@ -4363,7 +4363,7 @@ struct RD_Geo3DBoxDrawData
 };
 
 internal AC_Artifact
-rd_geo3d_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+rd_geo3d_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   Access *access = access_open();
   U128 hash = {0};

src/text/text.c

@@ -1970,7 +1970,7 @@ struct TXT_ArtifactCreateShared
 };
 
 internal AC_Artifact
-txt_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out)
+txt_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out)
 {
   ProfBeginFunction();
   Temp scratch = scratch_begin(0, 0);

src/text/text.h

@@ -203,7 +203,7 @@ internal TXT_ScopeNode *txt_scope_node_from_info_pt(TXT_TextInfo *info, TxtPt pt
 ////////////////////////////////
 //~ rjf: Artifact Cache Hooks / Lookups
 
-internal AC_Artifact txt_artifact_create(String8 key, U64 gen, U64 *requested_gen, B32 *retry_out);
+internal AC_Artifact txt_artifact_create(String8 key, B32 *cancel_signal, B32 *retry_out);
 internal void txt_artifact_destroy(AC_Artifact artifact);
 internal TXT_TextInfo txt_text_info_from_hash_lang(Access *access, U128 hash, TXT_LangKind lang);
 internal TXT_TextInfo txt_text_info_from_key_lang(Access *access, C_Key key, TXT_LangKind lang, U128 *hash_out);