Optimizations, have return ask if it is a heap value, move lexer up

.clocignore

@@ -0,0 +1,2 @@
+.ccls-cache/
+build/

.gitignore

@@ -108,3 +108,5 @@ build/
 .gdb_history
 .vscode
 .clangd
+.directory
+perf.*

Makefile

@@ -44,8 +44,8 @@ ifeq ($(BUILD_MODE), release)
     LDFLAGS         += -s
     TARGET_SUFFIX   := -release
 else
-    CORE_CFLAGS     += -g -DDEBUG
-    PLATFORM_CFLAGS += -g -DDEBUG
+    CORE_CFLAGS     += -O0 -ggdb3 -fno-omit-frame-pointer -DDEBUG
+    PLATFORM_CFLAGS += -O0 -ggdb3 -fno-omit-frame-pointer -DDEBUG
     TARGET_SUFFIX   := -debug
 endif
 
@@ -91,14 +91,14 @@ ifeq ($(BUILD_MODE), release)
 	PLATFORM_SOURCE := $(ARCH_DIR)/main.c \
 		$(ARCH_DIR)/devices.c\
 		$(SRC_DIR)/tools/lexer.c \
-		$(SRC_DIR)/tools/assembler/assembler.c  \
-		$(SRC_DIR)/tools/compiler/compiler.c
+		$(SRC_DIR)/tools/assembler/assembler.c  
+		# $(SRC_DIR)/tools/compiler/compiler.c
 else
 	PLATFORM_SOURCE := $(ARCH_DIR)/main.c \
 		$(ARCH_DIR)/devices.c \
 		$(SRC_DIR)/tools/lexer.c \
-		$(SRC_DIR)/tools/assembler/assembler.c \
-		$(SRC_DIR)/tools/compiler/compiler.c 
+		$(SRC_DIR)/tools/assembler/assembler.c 
+		# $(SRC_DIR)/tools/compiler/compiler.c 
 endif
 
 # --- OBJECT FILES ---

src/arch/linux/main.c

@@ -1,4 +1,4 @@
-#include "../../tools/compiler/compiler.h"
+//#include "../../tools/compiler/compiler.h"
 #include "../../tools/assembler/assembler.h"
 #include "../../vm/vm.h"
 #include "devices.h"
@@ -225,8 +225,8 @@ bool compileAndSave(const char *source_file, const char *output_file, VM *vm) {
   fclose(f);
 
   ScopeTable table = {0};
-  symbol_table_init(&table);
-  compile(vm, &table, source);
+  /* symbol_table_init(&table); */
+  /* compile(vm, &table, source); */
 #ifndef STATIC
   free(table.scopes);
 #endif
@@ -340,7 +340,9 @@ i32 main(i32 argc, char *argv[]) {
                      &console_device_ops, 4);
 
   if (terminal_only_mode) {
-    while (step_vm(&vm));
+    while (step_vm(&vm)) {
+      //printf("code[%d] = %s\n", vm.pc, opcode_to_string(vm.code[vm.pc]));
+    }    
     return 0;
   }
 

src/tools/assembler/assembler.c

@@ -163,7 +163,7 @@ void asm_emit_u32(VM *vm, u32 value) {
 }
 
 Symbol *asm_symbol_table_lookup(ScopeTable *table, const char *name, u32 length,
-                            i32 scope_ref) {
+                                i32 scope_ref) {
   SymbolTable st = table->scopes[scope_ref];
   for (u32 i = 0; i < st.count; i++) {
     if (st.symbols[i].name_length == length) {
@@ -214,7 +214,8 @@ u8 asm_symbol_table_add(ScopeTable *table, Symbol s) {
     printf("code[%d] = %s\n", s.ref, s.name);
   }
 #endif
-  table->scopes[table->scope_ref].symbols[table->scopes[table->scope_ref].count] = s;
+  table->scopes[table->scope_ref]
+      .symbols[table->scopes[table->scope_ref].count] = s;
   u8 index = table->scopes[table->scope_ref].count;
   table->scopes[table->scope_ref].count++;
   return index;
@@ -602,11 +603,8 @@ void asm_define_branch(VM *vm, ScopeTable *st) {
 
 int asm_get_instruction_byte_size(const char *opname) {
 
-  if (strcmp(opname, "return") == 0) {
-    return 2;
-  }
-
-  if (strcmp(opname, "neg_int") == 0 || strcmp(opname, "abs_int") == 0 ||
+  if (strcmp(opname, "return") == 0 ||
+      strcmp(opname, "neg_int") == 0 || strcmp(opname, "abs_int") == 0 ||
       strcmp(opname, "neg_nat") == 0 || strcmp(opname, "abs_nat") == 0 ||
       strcmp(opname, "neg_real") == 0 || strcmp(opname, "abs_real") == 0 ||
       strcmp(opname, "int_to_string") == 0 ||
@@ -699,9 +697,9 @@ int asm_get_instruction_byte_size(const char *opname) {
         token = next_token();                                                  \
       }                                                                        \
       /*printf("code[%d]=%s\n %d + %d = %d\n", vm->cp, op,                     \
-       * asm_get_instruction_byte_size(op), vm->cp, vm->cp +                       \
-       * asm_get_instruction_byte_size(op)); */                                    \
-      vm->cp += asm_get_instruction_byte_size(op);                                 \
+       * asm_get_instruction_byte_size(op), vm->cp, vm->cp +                   \
+       * asm_get_instruction_byte_size(op)); */                                \
+      vm->cp += asm_get_instruction_byte_size(op);                             \
     } while (0);
 
 /**
@@ -729,7 +727,8 @@ void asm_build_symbol_table(VM *vm, char *source, ScopeTable *st) {
       if (token.type == TOKEN_RBRACE) {
         i32 current_scope = st->scope_ref;
         i32 parent = st->scopes[current_scope].parent;
-        if (parent < 0) parent = 0;
+        if (parent < 0)
+          parent = 0;
         st->scope_ref = parent;
         continue;
       }
@@ -773,6 +772,15 @@ void asm_build_symbol_table(VM *vm, char *source, ScopeTable *st) {
 
         asm_get_reg(next, st);
         vm->cp++;
+
+        next = next_token();
+        if (next.type == TOKEN_SEMICOLON) {
+          /* assume it is not a heap value */
+          vm->cp++;
+          continue;
+        }
+
+	      vm->cp++;
         asm_next_token_is(TOKEN_SEMICOLON);
         continue;
       }
@@ -959,7 +967,8 @@ void asm_emit_bytecode(VM *vm, char *source, ScopeTable *st) {
       if (token.type == TOKEN_RBRACE) {
         i32 current_scope = st->scope_ref;
         i32 parent = st->scopes[current_scope].parent;
-        if (parent < 0) parent = 0;
+        if (parent < 0)
+          parent = 0;
         st->scope_ref = parent;
         continue;
       }
@@ -1012,12 +1021,40 @@ void asm_emit_bytecode(VM *vm, char *source, ScopeTable *st) {
           /* put 0xFF as return register */
           asm_emit_byte(vm, 0xFF);
           vm->cp++;
+
+          /* if no return then also no heap */
+          asm_emit_byte(vm, false);
+          vm->cp++;
           continue;
         }
 
         u32 reg = asm_get_reg(next, st);
         asm_emit_byte(vm, reg);
         vm->cp++;
+
+        next = next_token();
+        if (next.type == TOKEN_SEMICOLON) {
+          /* assume it is not a heap value */
+          asm_emit_byte(vm, false);
+          vm->cp++;
+          continue;
+        }
+
+        switch (next.type) {
+        case TOKEN_KEYWORD_TRUE: {
+          asm_emit_byte(vm, true);
+          break;
+        }
+        case TOKEN_KEYWORD_FALSE: {
+          asm_emit_byte(vm, false);
+          break;
+        }
+        default:
+          printf("Error: must be true or false for if it is a heap value");
+          exit(1);
+        }
+	      vm->cp++;
+
         asm_next_token_is(TOKEN_SEMICOLON);
         continue;
       }

src/tools/codegen.h

@@ -2,6 +2,7 @@
 #define UNDAR_CODEGEN_H
 
 #include "../vm/common.h"
+#include "lexer.h"
 
 typedef enum { GLOBAL, LOCAL, VAR } ScopeType;
 typedef enum {
@@ -66,6 +67,7 @@ struct symbol_s {
   char name[MAX_SYMBOL_NAME_LENGTH];
   u8 name_length;
   SymbolType type;
+  TokenType return_type;
   ScopeType scope;
   u32 ref;  // vm->mp if global, vm->pc local, register if var 
   u32 size; // size of symbol

src/tools/lexer.c

@@ -30,7 +30,7 @@ static char advance() {
   return lexer.current[-1];
 }
 
-static char peek() { return *lexer.current; }
+char peek() { return *lexer.current; }
 
 static char peek_next() {
   if (is_at_end())

src/tools/lexer.h

@@ -86,5 +86,6 @@ typedef struct {
 void init_lexer(const char *source);
 Token next_token();
 const char* token_type_to_string(TokenType type);
+char peek();
 
 #endif

src/vm/opcodes.h

@@ -99,7 +99,6 @@ typedef enum {
 
 #define MAX_LOCALS 256
 typedef struct frame_s {
-  u32 heap_mask[8];
   u32 locals[MAX_LOCALS]; /* $0-$255 */
   u32 start;              /* start of memory block */
   u32 end;                /* end of memory block */

src/vm/vm.c

@@ -16,7 +16,7 @@ const char radix_set[11] = "0123456789";
     u8 src1, src2;                                                             \
     type value;                                                                \
     type value2;                                                               \
-    target = read_u32(vm, code, vm->pc);                                       \
+    target = *code_u32;                                                         \
     vm->pc += 4;                                                               \
     src1 = read_u8(vm, code, vm->pc);                                          \
     vm->pc++;                                                                  \
@@ -58,23 +58,6 @@ const char radix_set[11] = "0123456789";
     return true;                                                               \
   } while (0)
 
-/* Set heap status for a register in current frame */
-void set_heap_status(VM *vm, u8 reg, bool is_heap) {
-  u32 index = reg / 32;
-  u32 mask = 1 << (reg % 32);
-
-  if (is_heap) {
-    vm->frames[vm->fp].heap_mask[index] |= mask;
-  } else {
-    vm->frames[vm->fp].heap_mask[index] &= ~mask;
-  }
-}
-
-/* Check if register contains heap pointer */
-bool is_heap_value(VM *vm, u8 reg) {
-  u32 index = reg / 32;
-  return (vm->frames[vm->fp].heap_mask[index] >> (reg % 32)) & 1;
-}
 
 u32 str_alloc(VM *vm, Frame *frame, const char *str, u32 length) {
   u32 str_addr = vm->mp;
@@ -95,20 +78,20 @@ u32 str_alloc(VM *vm, Frame *frame, const char *str, u32 length) {
  */
 bool step_vm(VM *vm) {
   /* Get current instruction & Advance to next instruction */
-  u8 opcode = vm->code[vm->pc++];
   Frame *frame = &vm->frames[vm->fp];
+  u8 opcode = vm->code[vm->pc++];
+  u32 *code_u32 = (u32*)(&vm->code[vm->pc]);
 
   switch (opcode) {
   case OP_EXIT: {
-    vm->flag = read_u32(vm, code, vm->pc);
+    vm->flag = *code_u32;
     return false;
   }
   case OP_CALL: {
     u8 N, return_reg, src_reg, args[MAX_LOCALS];
     Frame *child;
-    u32 jmp, mask, i;
-
-    jmp = read_u32(vm, code, vm->pc);
+    u32 jmp, i;
+    jmp = *code_u32;
     vm->pc += 4;
     N = vm->code[vm->pc++];
 
@@ -135,16 +118,13 @@ bool step_vm(VM *vm) {
     for (i = 0; i < N; i++) {
       src_reg = args[i];
       child->locals[i] = frame->locals[src_reg];
-      mask = 1 << (src_reg % 32);
-      if (frame->heap_mask[src_reg / 32] & mask) {
-        child->heap_mask[i / 32] |= 1 << (i % 32);
-      }
     }
 
     vm->pc = jmp;
     return true;
   }
   case OP_RETURN: {
+    bool is_heap_value;
     u8 child_return_reg;
     u32 value;
     u32 ptr;
@@ -154,13 +134,14 @@ bool step_vm(VM *vm) {
     Frame *parent;
 
     child_return_reg = vm->code[vm->pc++];
+    is_heap_value = vm->code[vm->pc++];
     child = frame;
     parent = &vm->frames[vm->fp - 1];
 
-    if (child_return_reg != 0xFF && parent->return_reg != 0xFF) {
+    if (child_return_reg != 0xFF) {
       value = child->locals[child_return_reg];
 
-      if (is_heap_value(vm, child_return_reg)) {
+      if (is_heap_value) {
         ptr = value;
         size = *(u32 *)(vm->memory + ptr - 4);
         new_ptr = parent->end;
@@ -171,11 +152,8 @@ bool step_vm(VM *vm) {
         memcopy(vm->memory + new_ptr + 4, vm->memory + ptr + 4, size);
         parent->end += size + 4;
         parent->locals[parent->return_reg] = new_ptr;
-        parent->heap_mask[parent->return_reg / 32] |= (1 << parent->return_reg);
       } else {
         parent->locals[parent->return_reg] = value;
-        parent->heap_mask[parent->return_reg / 32] &=
-            ~(1 << parent->return_reg);
       }
     }
 
@@ -187,7 +165,7 @@ bool step_vm(VM *vm) {
   case OP_LOAD_IMM: {
     u32 v;
     u8 dest;
-    v = read_u32(vm, code, vm->pc);
+    v = *code_u32;
     vm->pc += 4;
     dest = read_u8(vm, code, vm->pc);
     vm->pc++;
@@ -205,13 +183,12 @@ bool step_vm(VM *vm) {
     size = frame->locals[src1];
     write_u32(vm, memory, vm->mp, size);
     vm->mp += (size + 4);
-    set_heap_status(vm, dest, true); /* Mark as heap pointer */
     return true;
   }
   case OP_LOAD_ABS_32: {
     u32 v, ptr;
     u8 dest;
-    ptr = read_u32(vm, code, vm->pc);
+    ptr = *code_u32;
     vm->pc += 4;
     v = read_u32(vm, memory, ptr);
     dest = read_u8(vm, code, vm->pc);
@@ -222,7 +199,7 @@ bool step_vm(VM *vm) {
   case OP_LOAD_ABS_16: {
     u32 v, ptr;
     u8 dest;
-    ptr = read_u32(vm, code, vm->pc);
+    ptr = *code_u32;
     vm->pc += 4;
     v = read_u16(vm, memory, ptr);
     dest = read_u8(vm, code, vm->pc);
@@ -233,7 +210,7 @@ bool step_vm(VM *vm) {
   case OP_LOAD_ABS_8: {
     u32 v, ptr;
     u8 dest;
-    ptr = read_u32(vm, code, vm->pc);
+    ptr = *code_u32;
     vm->pc += 4;
     v = read_u8(vm, memory, ptr);
     dest = read_u8(vm, code, vm->pc);
@@ -544,23 +521,16 @@ bool step_vm(VM *vm) {
     dest = read_u8(vm, code, vm->pc);
     vm->pc++;
     frame->locals[dest] = frame->locals[src1];
-
-    if (is_heap_value(vm, src1)) {
-      set_heap_status(vm, dest, true);
-    } else {
-      set_heap_status(vm, dest, false);
-    }
-
     return true;
   }
   case OP_JMP: {
-    u32 jmp = read_u32(vm, code, vm->pc);
+    u32 jmp = *code_u32;
     vm->pc = jmp; /* Jump to address */
     return true;
   }
   case OP_JMPF: { /* error handling for syscall, jump if flag == 0 */
     u32 mask;
-    u32 jmp = read_u32(vm, code, vm->pc);
+    u32 jmp = *code_u32;
     mask = -(u32)(vm->flag == 0);
     vm->pc = (jmp & mask) | (vm->pc & ~mask);
     return true;
@@ -568,7 +538,7 @@ bool step_vm(VM *vm) {
   case OP_SYSCALL: {
     u32 syscall_id;
 
-    syscall_id = read_u32(vm, code, vm->pc);
+    syscall_id = *code_u32;
     vm->pc += 4;
 
     switch (syscall_id) {
@@ -966,7 +936,6 @@ bool step_vm(VM *vm) {
       buffer[--i] = '0';
 
     frame->locals[dest] = str_alloc(vm, frame, buffer + i, MAX_LEN_INT32 - i);
-    set_heap_status(vm, dest, true); /* Mark as heap pointer */
     return true;
   }
   case OP_NAT_TO_STRING: {
@@ -991,7 +960,6 @@ bool step_vm(VM *vm) {
       buffer[--i] = '0';
     /* Copy from buffer[i] to buffer + MAX_LEN_INT32 */
     frame->locals[dest] = str_alloc(vm, frame, buffer + i, MAX_LEN_INT32 - i);
-    set_heap_status(vm, dest, true); /* Mark as heap pointer */
     return true;
   }
   case OP_REAL_TO_STRING: {
@@ -1036,7 +1004,6 @@ bool step_vm(VM *vm) {
       buffer[--i] = '-';
 
     frame->locals[dest] = str_alloc(vm, frame, buffer + i, MAX_LEN_REAL32 - i);
-    set_heap_status(vm, dest, true); /* Mark as heap pointer */
     return true;
   }
   case OP_STRLEN: {