progress on lottes.c

C/watl.v0.llvm.lottes.c

@@ -4,13 +4,14 @@ Version:   0 (From Scratch, 1-Stage Compilation, LLVM & WinAPI Only, Win CRT Mul
 Host:      Windows 11 (x86-64)
 Toolchain: LLVM (2025-08-30), C-Stanard: 11
 
-Following strictly (mostly): Neokineogfx - Fixing C
+Following strictly: Neokineogfx - Fixing C
 https://youtu.be/RrL7121MOeA
-
-Unlike lottes_hybrid this file will be entirely untyped for any pointer addressing.
-Win CRT imports will also be typeless signatures.
 */
 
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wpre-c11-compat"
+// #pragma clang diagnostic ignored "-Wc++-keyword"
+#pragma clang diagnostic ignored "-Wcast-qual"
 #pragma clang diagnostic ignored "-Wunused-const-variable"
 #pragma clang diagnostic ignored "-Wunused-but-set-variable"
 #pragma clang diagnostic ignored "-Wswitch"
@@ -24,12 +25,10 @@ Win CRT imports will also be typeless signatures.
 #pragma clang diagnostic ignored "-Wkeyword-macro"
 #pragma clang diagnostic ignored "-Wc23-compat"
 #pragma clang diagnostic ignored "-Wreserved-identifier"
-#pragma clang diagnostic ignored "-Wpre-c11-compat"
 #pragma clang diagnostic ignored "-Wc23-extensions"
 #pragma clang diagnostic ignored "-Wunused-macros"
 #pragma clang diagnostic ignored "-Wdeclaration-after-statement"
 #pragma clang diagnostic ignored "-Wunsafe-buffer-usage"
-#pragma clang diagnostic ignored "-Wc++-keyword"
 #pragma clang diagnostic ignored "-Wimplicit-function-declaration"
 #pragma clang diagnostic ignored "-Wcast-align"
 #pragma clang diagnostic ignored "-Wunused-parameter"
@@ -54,12 +53,12 @@ Win CRT imports will also be typeless signatures.
 #define V_      volatile
 #define W_      __attribute((__stdcall__)) __attribute__((__force_align_arg_pointer__))
 
-#define reg     register
+// #define reg     register
 
 #define glue_impl(A, B)    A ## B
 #define glue(A, B)         glue_impl(A, B)
 #define stringify_impl(S)  #S
-#define stringify(S)       stringify_impl(S)
+#define stringify(S)       cast(UTF8*, stringify_impl(S))
 #define tmpl(prefix, type) prefix ## _ ## type
 
 #define static_assert      _Static_assert
@@ -131,7 +130,7 @@ enum { false = 0, true  = 1, true_overflow, };
 #define pcast(type, data)                   cast(type*, & (data))[0]
 #define nullptr                             cast(void*, 0)
 #define null                                cast(U8,    0)
-#define offset_of(type, member)             cast(U8, & (((type*) 0)->member))
+#define soff(type, member)                  cast(U8, & (((type*) 0)->member)) // offset_of
 #define size_of(data)                       cast(U8, sizeof(data))
 
 #define r_(ptr)                             cast(typeof_ptr(ptr)*R_, ptr)
@@ -198,8 +197,8 @@ typedef def_struct(Slice_Str8) { U8 ptr; U8 len; };
 	{                        \
 		assert_handler(        \
 			stringify(cond),     \
-			__FILE__,            \
-			__func__,            \
+			(UTF8*)__FILE__,     \
+			(UTF8*)__func__,     \
 			cast(S4, __LINE__),  \
 			msg,                 \
 			## __VA_ARGS__);     \
@@ -230,6 +229,8 @@ I_ B4 mem_zero            (U8 dest,           U8 len) { if (dest == 0) return fa
 #define struct_copy(type, dest, src) mem_copy(dest, src, sizeof(type))
 #define struct_zero(type, dest)      mem_zero(dest,      sizeof(type))
 
+#define struct_assign(type, dest, src) cast(type*R_, dest)[0] = cast(type*R_, src)[0]
+
 I_ U8 align_pow2(U8 x, U8 b) {
     assert(b != 0);
     assert((b & (b - 1)) == 0);  // Check power of 2
@@ -276,7 +277,7 @@ typedef def_Slice(B1);
 #define slice_to_bytes(slice) ((Slice_B1){cast(B1*, (slice).ptr), (slice).len * size_of_slice_type(slice)})
 #define slice_fmem(mem)       slice_mem(u8_(mem), size_of(mem))
 
-I_ void slice__zero(Slice_B1 mem, U8 typewidth) { slice_assert(mem); memory_zero(u8_(mem.ptr), mem.len); }
+I_ void slice__zero(Slice_B1 mem, U8 typewidth) { slice_assert(mem); mem_zero(u8_(mem.ptr), mem.len); }
 #define slice_zero(slice) slice__zero(slice_mem_s(slice), size_of_slice_type(slice))
 
 I_ void slice__copy(Slice_B1 dest, U8 dest_typewidth, Slice_B1 src, U8 src_typewidth) {
@@ -293,6 +294,19 @@ I_ void slice__copy(Slice_B1 dest, U8 dest_typewidth, Slice_B1 src, U8 src_typew
 #define slice_iter(container, iter)     (typeof((container).ptr) iter = (container).ptr; iter != slice_end(container); ++ iter)
 #define slice_arg_from_array(type, ...) & (tmpl(Slice,type)) { .ptr = farray_init(type, __VA_ARGS__), .len = farray_len( farray_init(type, __VA_ARGS__)) }
 
+I_ void slice_assign(U8 dest, U8 src) {
+	u8_r(dest + soff(Slice_Mem, ptr))[0] = u8_r(src + soff(Slice_Mem, ptr))[0];
+	u8_r(dest + soff(Slice_Mem, len))[0] = u8_r(src + soff(Slice_Mem, len))[0];
+}
+I_ void slice_assign_comp(U8 dest, U8 ptr, U8 len) {
+	u8_r(dest + soff(Slice_Mem, ptr))[0] = ptr;
+	u8_r(dest + soff(Slice_Mem, len))[0] = len;
+}
+I_ void slice_clear(U8 base) {
+	u8_r(base + soff(Slice_Mem, ptr))[0] = 0;
+	u8_r(base + soff(Slice_Mem, len))[0] = 0;
+}
+
 #define span_iter(type, iter, m_begin, op, m_end) \
 (                                   \
 	tmpl(Iter_Span,type) iter = {     \
@@ -442,7 +456,7 @@ typedef def_struct(FArena) {
 I_ void farena_init__u  (U8 arena, U8 mem_ptr, U8 mem_len);
    void farena__push__u (U8 arena, U8 amount, U8 type_width, U8 alignment, U8 slice_addr);
 I_ void farena_reset__u (U8 arena);
-I_ void farena_rewind__u(U8 arena, U8 sp_type_sig, U8 sp_slot);
+I_ void farena_rewind__u(U8 arena, U8 sp_slot);
 I_ void farena_save__u  (U8 arena, U8 sp);
 
 I_ FArena      farena_make  (Slice_Mem mem);
@@ -472,17 +486,18 @@ typedef def_struct(OS_Windows_State) { OS_SystemInfo system_info; };
 global OS_Windows_State os__windows_info;
 
 I_ OS_SystemInfo* os_system_info(void);
-I_ void           os_init       (void);
+   void           os_init       (void);
 
-I_ U8   os_vmem_reserve__u(       U8 size, U4 no_large_pages, U8 base_addr);
-I_ B4   os_vmem_commit__u (U8 vm, U8 size);
+I_ U8   os_vmem_reserve__u(       U8 size, B4 no_large_pages, U8 base_addr);
+I_ B4   os_vmem_commit__u (U8 vm, U8 size, B4 no_large_pages);
 I_ void os_vmem_release__u(U8 vm, U8 size);
 
 I_ U8   os__vmem_reserve(       U8 size, Opts_vmem_R opts);
-I_ B4   os_vmem_commit  (U8 vm, U8 size);
+I_ B4   os__vmem_commit (U8 vm, U8 size, Opts_vmem_R opts);
 I_ void os_vmem_release (U8 vm, U8 size);
 
-#define os_vmem_reserve(size, ...)    os__vmem_reserve(size, opt_args(Opts_vmem, __VA_ARGS__))
+#define os_vmem_commit(vm, size, ...) os__vmem_commit (vm, size, opt_args(Opts_vmem, __VA_ARGS__))
+#define os_vmem_reserve(size, ...)    os__vmem_reserve(    size, opt_args(Opts_vmem, __VA_ARGS__))
 #pragma endregion OS
 
 #pragma region VArena (Virtual Address Space Arena)
@@ -505,10 +520,10 @@ typedef def_struct(Opts_varena_make) {
 	VArenaFlags flags;
 };
 
-I_ U8   varena__make__u  (U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr);
+   U8   varena__make__u  (U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr);
 I_ void varena_release__u(U8 arena);
 I_ void varena_reset__u  (U8 arena);
-I_ void varena_rewind__u (U8 arena,  U8 sp_type_sig, U8 sp_slot);
+I_ void varena_rewind__u (U8 arena,  U8 sp_slot);
 I_ void varena_save__u   (U8 arena,  U8 sp_addr);
    void varena__push__u  (U8 arena,  U8 amount, U8 type_width, U8 alignment, U8 slice_addr);
    void varena__grow__u  (U8 result, U8 arena,  U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment, B4 should_zero);
@@ -522,6 +537,8 @@ I_ void         varena_rewind (VArena_R arena, AllocatorSP save_point);
 I_ Slice_Mem    varena__shrink(VArena_R arena, Slice_Mem old_allocation, U8 requested_size, Opts_varena*R_ opts);
 I_ AllocatorSP  varena_save   (VArena_R arena);
 
+#define varena_make(...) varena__make(opt_args(Opts_varena_make, __VA_ARGS__))
+
 void varena_allocator_proc(U8 data, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out);
 
 #define ainfo_varena(arena) (AllocatorInfo){ .proc = varena_allocator_proc, .data = u8_(arena) }
@@ -535,6 +552,111 @@ cast(type*, varena__push(arena, size_of(type), 1, opt_args(Opts_varena, __VA_ARG
 (tmpl(Slice,type)){ varena__push(arena, size_of(type), amount, opt_args(Opts_varena, __VA_ARGS__)).ptr, amount }
 #pragma endregion VArena
 
+#pragma region Arena
+typedef Opts_varena Opts_arena;
+typedef def_enum(U4, ArenaFlags) {
+	ArenaFlag_NoLargePages = (1 << 0),
+	ArenaFlag_NoChain      = (1 << 1),
+};
+typedef def_struct(Arena) {
+	VArena*    backing;
+	Arena*     prev;
+	Arena*     current;
+	U8         base_pos;
+	U8         pos;
+	ArenaFlags flags;
+	A4_B1 _PAD_;
+};
+
+U8   arena_make__u   (U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr);
+void arena__push__u  (U8 arena, U8 amount, U8 type_width, U8 alignemnt, U8 out_mem);
+void arena_release__u(U8 arena);
+void arena_reset__u  (U8 arena);
+void arena_rewind__u (U8 arena, U8 slot);
+void arena_save__u   (U8 arena, U8 out_sp);
+
+typedef Opts_varena_make Opts_arena_make;
+   Arena*      arena__make  (Opts_arena_make*R_ opts);
+   Slice_Mem   arena__push  (Arena_R arena, U8 amount, U8 type_width, Opts_arena*R_ opts);
+I_ void        arena_release(Arena_R arena);
+I_ void        arena_reset  (Arena_R arena);
+   void        arena_rewind (Arena_R arena, AllocatorSP save_point);
+I_ AllocatorSP arena_save   (Arena_R arena);
+
+void arena_allocator_proc(U8 data, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out);
+#define ainfo_arena(arena) (AllocatorInfo){ .proc = & arena_allocator_proc, .data = u8_(arena) }
+
+#define arena_make(...) arena__make(opt_args(Opts_arena_make, __VA_ARGS__))
+
+#define arena_push_mem(arena, amount, ...) arena__push(arena, amount, 1, opt_args(Opts_arena, lit(stringify(B1)), __VA_ARGS__))
+
+#define arena_push(arena, type, ...) \
+cast(type*, arena__push(arena, 1, size_of(type), opt_args(Opts_arena, lit(stringify(type)), __VA_ARGS__) ).ptr)
+
+#define arena_push_array(arena, type, amount, ...) \
+(tmpl(Slice,type)){ arena__push(arena, size_of(type), amount, opt_args(Opts_arena, lit(stringify(type)), __VA_ARGS__)).ptr, amount }
+#pragma endregion Arena
+
+#pragma region Hashing
+I_ void hash64_fnv1a__u(U8 hash, U8 data_ptr, U8 data_len, U8 seed) {
+	local_persist U8 const default_seed = 0xcbf29ce484222325; 
+	if (seed != 0) { u8_r(hash)[0] = seed; }
+	else           { u8_r(hash)[0] = default_seed; }
+	U8 elem = data_ptr;
+loop:
+	if (elem == data_ptr + data_len) goto end;
+	u8_r(hash)[0] ^= u1_r(elem)[0];
+	u8_r(hash)[0] *= 0x100000001b3;
+	elem += 1;
+	goto loop;
+end: 
+	return;
+}
+typedef def_struct(Opts_hash64_fnv1a) { U8 seed; };
+I_ void hash64__fnv1a(U8_R hash, Slice_Mem data, Opts_hash64_fnv1a*R_ opts) {
+	assert(opts != nullptr);
+	hash64_fnv1a__u(u8_(hash), data.ptr, data.len, opts->seed);
+}
+#define hash64_fnv1a(hash, data, ...) hash64__fnv1a(hash, data, opt_args(Opts_hash64_fnv1a, __VA_ARGS__))
+#pragma endregion Hashing
+
+#pragma region Key Table Linear (KTL)
+#define def_KTL_Slot(type)        \
+def_struct(tmpl(KTL_Slot,type)) { \
+	U8   key;   \
+	type value; \
+}
+#define def_KTL(type)             \
+	def_Slice(tmpl(KTL_Slot,type)); \
+	typedef tmpl(Slice_KTL_Slot,type) tmpl(KTL,type)
+
+typedef Slice_Mem  KTL_Byte;
+typedef def_struct(KTL_Meta) {
+	U8   slot_size;
+	U8   kt_value_offset;
+	U8   type_width;
+	Str8 type_name;
+};
+
+typedef def_farray(Str8, 2);
+typedef def_Slice(A2_Str8);
+typedef def_KTL_Slot(Str8);
+typedef def_KTL(Str8);
+void ktl_populate_slice_a2_str8(U8 kt, U8 backing_proc, U8 backing_data, U8 values);
+#pragma endregion KTL
+
+#pragma region Key Table 1-Layer Chained-Chunked-Cells (KT1CX)
+#pragma endregion KT1CX
+
+#pragma region String Operations
+#pragma endregion String Operations
+
+#pragma region FIle System
+#pragma endregion FIle System
+
+#pragma region WATL
+#pragma endregion WATL
+
 #pragma endregion Header
 
 #pragma region Implementation
@@ -560,29 +682,29 @@ I_ void mem_save_point__u(U8 proc, U8 data, U8 sp) {
 	assert(proc != null);
 	uvar(AllocatorProc_Out, out) = {0}; 
 	cast(AllocatorProc*, proc)(data, 0, 0, 0, 0, AllocatorOp_SavePoint, u8_(out));
-	struct_copy(AllocatorSP, sp, (U8) out + offset_of(AllocatorProc_Out, save_point));
+	struct_assign(AllocatorSP, sp, (U8) out + soff(AllocatorProc_Out, save_point));
 }
 I_ void mem__alloc__u(U8 out_mem, U8 proc, U8 data, U8 size, U8 alignment, B4 no_zero) {
 	assert(proc != null);
 	uvar(AllocatorProc_Out, out) = {0};
 	cast(AllocatorProc*, proc)(data, size, alignment, 0, 0, no_zero ? AllocatorOp_Alloc_NoZero : AllocatorOp_Alloc, u8_(out));
-	struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
+	slice_assign(out_mem, (U8) out + soff(AllocatorProc_Out, allocation));
 }
 I_ void mem__grow__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_actual) {
 	assert(proc != null);
 	uvar(AllocatorProc_Out, out) = {0};
 	cast(AllocatorProc*, proc)(data, size, alignment, old_ptr, old_len, no_zero ? AllocatorOp_Grow_NoZero : AllocatorOp_Grow, u8_(out));
-	if (give_actual == false) { u8_r(out + offset_of(AllocatorProc_Out, allocation) + offset_of(Slice_Mem, len))[0] = size; }
-	struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
+	if (give_actual == false) { u8_r(out + soff(AllocatorProc_Out, allocation) + soff(Slice_Mem, len))[0] = size; }
+	slice_assign(out_mem, (U8) out + soff(AllocatorProc_Out, allocation));
 }
 I_ void mem__shrink__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment) {
 	assert(proc != null);
 	uvar(AllocatorProc_Out, out) = {0};
 	cast(AllocatorProc*, proc)(data, size, alignment, old_ptr, old_len, AllocatorOp_Shrink, u8_(out));
-	struct_copy(Slice_Mem, out_mem, (U8) out + offset_of(AllocatorProc_Out, allocation));
+	slice_assign(out_mem, (U8) out + soff(AllocatorProc_Out, allocation));
 }
 I_ void mem__resize__u(U8 out_mem, U8 proc, U8 data, U8 old_ptr, U8 old_len, U8 size, U8 alignment, B4 no_zero, B4 give_acutal) { 
-	if (old_len == size) { struct_copy(Slice_Mem, out_mem, (U8)& slice_mem(old_ptr, old_len)); }
+	if (old_len == size) { slice_assign_comp(out_mem, old_ptr, old_len); }
 	if (old_len <  size) { mem__grow__u  (out_mem, proc, data, old_ptr, old_len, size, alignment, no_zero, give_acutal); }
 	else                 { mem__shrink__u(out_mem, proc, data, old_ptr, old_len, size, alignment); }
 }
@@ -621,87 +743,86 @@ I_ Slice_Mem mem__shrink(AllocatorInfo ainfo, Slice_Mem mem, U8 size, Opts_mem_s
 #pragma region FArena (Fixed-Sized Arena)
 I_ void farena_init__u(U8 arena, U8 mem_ptr, U8 mem_len) {
 	assert(arena != null);
-	u8_r(arena + offset_of(FArena, start)   )[0] = mem_ptr;
-	u8_r(arena + offset_of(FArena, capacity))[0] = mem_len;
-	u8_r(arena + offset_of(FArena, used)    )[0] = 0;
+	u8_r(arena + soff(FArena, start)   )[0] = mem_ptr;
+	u8_r(arena + soff(FArena, capacity))[0] = mem_len;
+	u8_r(arena + soff(FArena, used)    )[0] = 0;
 }
 inline void farena__push__u(U8 arena, U8 amount, U8 type_width, U8 alignment, U8 result) {
 	if (amount == 0) { struct_zero(Slice_Mem, result); }
-	U8   reg desired   = type_width * amount;
-	U8   reg to_commit = align_pow2(desired, alignment ?  alignment : MEMORY_ALIGNMENT_DEFAULT);
-	U8_R reg used      = u8_r(arena + offset_of(FArena, used));
-	U8   reg unused    = u8_r(arena + offset_of(FArena, capacity))[0] - used[0]; assert(to_commit <= unused);
-	U8   reg ptr       = u8_r(arena + offset_of(FArena, start)   )[0] + used[0];
+	U8   desired   = type_width * amount;
+	U8   to_commit = align_pow2(desired, alignment ?  alignment : MEMORY_ALIGNMENT_DEFAULT);
+	U8_R used      = u8_r(arena + soff(FArena, used));
+	U8   unused    = u8_r(arena + soff(FArena, capacity))[0] - used[0]; assert(to_commit <= unused);
+	U8   ptr       = u8_r(arena + soff(FArena, start)   )[0] + used[0];
 	used[0] += to_commit;
-	struct_copy(Slice_Mem, result, (U8)& slice_mem(ptr, desired));
+	slice_assign_comp(result, ptr, desired);
 }
 inline void farena__grow__u(U8 result, U8 arena, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment, B4 should_zero) {
 	assert(result != null);
 	assert(arena  != null);
-	U8_R reg used = u8_r(arena + offset_of(FArena, used));
+	U8_R used = u8_r(arena + soff(FArena, used));
 	/*Check if the allocation is at the end of the arena*/{
-		U8 reg alloc_end = old_ptr + old_len;
-		U8 reg arena_end = u8_r(arena + offset_of(FArena, start))[0] + used[0];
+		U8 alloc_end = old_ptr + old_len;
+		U8 arena_end = u8_r(arena + soff(FArena, start))[0] + used[0];
 		if (alloc_end != arena_end) {
 			// Not at the end, can't grow in place
-			struct_zero(Slice_Mem, result);
+			slice_clear(result);
 			return;
 		}
 	}
 	// Calculate growth
-	U8 reg grow_amount  = requested_size - old_len;
-	U8 reg aligned_grow = align_pow2(grow_amount, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
-	U8 reg unused       = u8_r(arena + offset_of(FArena, capacity))[0] - used[0];
+	U8 grow_amount  = requested_size - old_len;
+	U8 aligned_grow = align_pow2(grow_amount, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
+	U8 unused       = u8_r(arena + soff(FArena, capacity))[0] - used[0];
 	if (aligned_grow > unused) {
 		// Not enough space
-		struct_zero(Slice_Mem, result);
+		slice_clear(result);
 		return;
 	}
 	used[0] += aligned_grow;
-	struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, aligned_grow + requested_size));
-	memory_zero(old_ptr + old_len, grow_amount * cast(U8, should_zero));
+	slice_assign_comp(result, old_ptr, aligned_grow + requested_size);
+	mem_zero(old_ptr + old_len, grow_amount * cast(U8, should_zero));
 }
 inline void farena__shrink__u(U8 result, U8 arena, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment) {
 	assert(result != null);
 	assert(arena  != null);
-	U8_R reg used = u8_r(arena + offset_of(FArena, used));
+	U8_R used = u8_r(arena + soff(FArena, used));
 	/*Check if the allocation is at the end of the arena*/ {
-		U8 reg alloc_end = old_ptr + old_len;
-		U8 reg arena_end = u8_r(arena + offset_of(FArena, start))[0] + used[0];
+		U8 alloc_end = old_ptr + old_len;
+		U8 arena_end = u8_r(arena + soff(FArena, start))[0] + used[0];
 		if (alloc_end != arena_end) {
 			// Not at the end, can't shrink but return adjusted size
-			struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, requested_size));
+			slice_assign_comp(result, old_ptr, requested_size);
 			return;
 		}
 	}
-	U8 reg aligned_original = align_pow2(old_len, MEMORY_ALIGNMENT_DEFAULT);
-	U8 reg aligned_new      = align_pow2(requested_size, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
+	U8 aligned_original = align_pow2(old_len, MEMORY_ALIGNMENT_DEFAULT);
+	U8 aligned_new      = align_pow2(requested_size, alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT);
 	used[0] -= (aligned_original - aligned_new);
-	struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, requested_size));
+	slice_assign_comp(result, old_ptr, requested_size);
 }
-I_ void farena_reset__u(U8 arena) { u8_r(arena + offset_of(FArena, used))[0] = 0; }
-I_ void farena_rewind__u(U8 arena, U8 sp_type_sig, U8 sp_slot) {
-	assert(sp_type_sig == (U8)& farena_allocator_proc);
-	U8   reg start = u8_r(arena + offset_of(FArena, start))[0];
-	U8_R reg used  = u8_r(arena + offset_of(FArena, used));
-	U8   reg end   = start + used[0]; assert_bounds(sp_slot, start, end);
+I_ void farena_reset__u(U8 arena) { u8_r(arena + soff(FArena, used))[0] = 0; }
+I_ void farena_rewind__u(U8 arena, U8 sp_slot) {
+	U8   start = u8_r(arena + soff(FArena, start))[0];
+	U8_R used  = u8_r(arena + soff(FArena, used));
+	U8   end   = start + used[0]; assert_bounds(sp_slot, start, end);
 	used[0] -= sp_slot - start;
 }
 I_ void farena_save__u(U8 arena, U8 sp) {
-	u8_r(sp + offset_of(AllocatorSP, type_sig))[0] = (U8)& farena_allocator_proc;
-	u8_r(sp + offset_of(AllocatorSP, slot    ))[0] = u8_r(arena + offset_of(FArena, used))[0];
+	u8_r(sp + soff(AllocatorSP, type_sig))[0] = (U8)& farena_allocator_proc;
+	u8_r(sp + soff(AllocatorSP, slot    ))[0] = u8_r(arena + soff(FArena, used))[0];
 }
 void farena_allocator_proc(U8 arena, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, /*AllocatorProc_Out*/U8 out)
 {
 	assert(out   != null);
 	assert(arena != null);
-	U8 reg allocation = arena + offset_of(AllocatorProc_Out, allocation);
+	U8 allocation = arena + soff(AllocatorProc_Out, allocation);
 	switch (op)
 	{
 	case AllocatorOp_Alloc:
 	case AllocatorOp_Alloc_NoZero:
 		farena__push__u(arena, requested_size, 1, alignment, allocation);
-			memory_zero(u8_r(allocation + offset_of(Slice_Mem, ptr))[0], u8_r(allocation + offset_of(Slice_Mem, len))[0] * op);
+		mem_zero(u8_r(allocation + soff(Slice_Mem, ptr))[0], u8_r(allocation + soff(Slice_Mem, len))[0] * op);
 	break;
 	case AllocatorOp_Free:                       break;
 	case AllocatorOp_Reset: farena_reset__u(arena); break;
@@ -714,21 +835,21 @@ void farena_allocator_proc(U8 arena, U8 requested_size, U8 alignment, U8 old_ptr
 		farena__shrink__u(allocation, arena, old_ptr, old_len, requested_size, alignment);
 	break;
 
-		case AllocatorOp_Rewind:    farena_rewind__u(arena, old_ptr, old_len); break;
+	case AllocatorOp_Rewind:    farena_rewind__u(arena, old_len); break;
 	case AllocatorOp_SavePoint: farena_save__u(arena, allocation);         break;
 
 	case AllocatorOp_Query:
-			u4_r(out + offset_of(AllocatorQueryInfo, features))[0] =
+		u4_r(out + soff(AllocatorQueryInfo, features))[0] =
 			AllocatorQuery_Alloc
 		| AllocatorQuery_Reset
 		| AllocatorQuery_Resize
 		| AllocatorQuery_Rewind
 		;
-			U8 reg max_alloc = u8_r(arena + offset_of(FArena, capacity))[0] - u8_r(arena + offset_of(FArena, used))[0];
-			u8_r(out + offset_of(AllocatorQueryInfo, max_alloc))[0] = max_alloc;
-			u8_r(out + offset_of(AllocatorQueryInfo, min_alloc))[0] = 0;
-			u8_r(out + offset_of(AllocatorQueryInfo, left     ))[0] = max_alloc;
-			farena_save__u(arena, out + offset_of(AllocatorQueryInfo, save_point));
+		U8 max_alloc = u8_r(arena + soff(FArena, capacity))[0] - u8_r(arena + soff(FArena, used))[0];
+		u8_r(out + soff(AllocatorQueryInfo, max_alloc))[0] = max_alloc;
+		u8_r(out + soff(AllocatorQueryInfo, min_alloc))[0] = 0;
+		u8_r(out + soff(AllocatorQueryInfo, left     ))[0] = max_alloc;
+		farena_save__u(arena, out + soff(AllocatorQueryInfo, save_point));
 	break;
 	}
 	return;
@@ -782,7 +903,7 @@ typedef def_struct(MS_TOKEN_PRIVILEGES)    { MS_DWORD PrivilegeCount; MS_LUID_AN
 W_ MS_BOOL   ms_close_handle(MS_HANDLE hObject) __asm__("CloseHandle");
 W_ MS_BOOL   ms_adjust_token_privleges(MS_HANDLE TokenHandle, MS_BOOL DisableAllPrivileges, MS_PTOKEN_PRIVILEGES NewState, MS_DWORD BufferLength, MS_PTOKEN_PRIVILEGES PreviousState, MS_PDWORD ReturnLength) __asm__("AdjustTokenPrivileges");
 W_ MS_HANDLE ms_get_current_process(void) __asm__("GetCurrentProcess");
-W_ U8        ms_get_larg_page_minimum(void) __asm__("GetCurrentProcess");
+W_ U8        ms_get_larg_page_minimum(void) __asm__("GetLargePageMinimum");
 W_ MS_BOOL   ms_lookup_priviledge_value_w(MS_LPWSTR lpSystemName, MS_LPWSTR lpName, MS_PLUID lpLuid) __asm__("LookupPrivilegeValueW");
 W_ MS_BOOL   ms_open_process_token(MS_HANDLE ProcessHandle, MS_DWORD DesiredAccess, MS_PHANDLE TokenHandle) __asm__("OpenProcessToken");
 W_ MS_LPVOID ms_virtual_alloc(MS_LPVOID lpAddress, U8 dwSize, MS_DWORD flAllocationType, MS_DWORD flProtect) __asm__("VirtualAlloc");
@@ -806,117 +927,133 @@ I_ void os__enable_large_pages(void) {
 		ms_close_handle(token);
 	}
 }
-I_ void os_init(void) {
-	os__enable_large_pages();
+inline 
+void os_init(void) {
+	// os__enable_large_pages();
 	os_system_info()->target_page_size = ms_get_larg_page_minimum();
 }
 I_ U8 os_vmem_reserve__u(U8 size, B4 no_large_pages, U8 base_addr) {
-	return cast(U8, VirtualAlloc(cast(MS_LPVOID, base_addr), size, MS_MEM_RESERVE,
+	return cast(U8, ms_virtual_alloc(cast(MS_LPVOID, base_addr), size, MS_MEM_RESERVE,
 		MS_PAGE_READWRITE /* | (opts->no_large_pages ? 0 : MS_MEM_LARGE_PAGES) */)
 	);
 }
-I_ B4   os_vmem_commit__u (U8 vm, U8 size) { return ms_virtual_alloc(cast(MS_LPVOID, vm), size, MS_MEM_COMMIT, MS_PAGE_READWRITE) != null; }
+I_ B4   os_vmem_commit__u (U8 vm, U8 size, B4 no_large_pages) { 
+	// if (no_large_pages == false ) { return 1; }
+	return ms_virtual_alloc(cast(MS_LPVOID, vm), size, MS_MEM_COMMIT, MS_PAGE_READWRITE) != null; 
+}
 I_ void os_vmem_release__u(U8 vm, U8 size) { ms_virtual_free(cast(MS_LPVOID, vm), 0, MS_MEM_RESERVE); }
 
 I_ U8   os__vmem_reserve( U8 size, Opts_vmem_R opts) { 
 	assert(opts != nullptr);
 	return os_vmem_reserve__u(size, opts->no_large_pages, opts->base_addr); 
 }
-I_ B4   os_vmem_commit (U8 vm, U8 size) { return os_vmem_commit__u(vm, size); }
+I_ B4 os__vmem_commit (U8 vm, U8 size, Opts_vmem_R opts) { 
+	assert(opts != nullptr);
+	return os_vmem_commit__u(vm, size, opts->no_large_pages); 
+}
 I_ void os_vmem_release(U8 vm, U8 size) { os_vmem_release__u(vm, size); }
 #pragma endregion OS
 
 #pragma region VArena (Virtual Address Space Arena)
 I_ U8 varena_header_size(void) { return align_pow2(size_of(VArena), MEMORY_ALIGNMENT_DEFAULT); }
 
-I_ U8 varena__make__u(U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr) {
+inline U8 varena__make__u(U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr) {
 	if (reserve_size == 0) { reserve_size = mega(64); }
 	if (commit_size  == 0) { commit_size  = mega(64); }
-	U8 reg page       = os_system_info()->target_page_size;
-	U8 reg reserve_sz = align_pow2(reserve_size, page);
-	U8 reg commit_sz  = align_pow2(commit_size,  page);
-	B4 reg no_large   = (flags & VArenaFlag_NoLargePages) != 0;
+	U8 page       = os_system_info()->target_page_size;
+	U8 reserve_sz = align_pow2(reserve_size, page);
+	U8 commit_sz  = align_pow2(commit_size,  page);
+	B4 no_large   = (flags & VArenaFlag_NoLargePages) != 0;
 	U8 base       = os_vmem_reserve__u(reserve_sz, no_large, base_addr); assert(base != 0);
-	B4 ok         = os_vmem_commit__u(base, commit_sz);	                 assert(ok   != 0);
+	B4 ok         = os_vmem_commit__u(base, commit_sz, no_large);	       assert(ok   != 0);
 	U8 header     = varena_header_size();
 	U8 data_start = base + header;
-	u8_r(base + offset_of(VArena, reserve_start))[0] = data_start;
-	u8_r(base + offset_of(VArena, reserve      ))[0] = reserve_sz;
-	u8_r(base + offset_of(VArena, commit_size  ))[0] = commit_sz;
-	u8_r(base + offset_of(VArena, committed    ))[0] = commit_sz;
-	u8_r(base + offset_of(VArena, commit_used  ))[0] = header;
-	u4_r(base + offset_of(VArena, flags        ))[0] = flags;
+	u8_r(base + soff(VArena, reserve_start))[0] = data_start;
+	u8_r(base + soff(VArena, reserve      ))[0] = reserve_sz;
+	u8_r(base + soff(VArena, commit_size  ))[0] = commit_sz;
+	u8_r(base + soff(VArena, committed    ))[0] = commit_sz;
+	u8_r(base + soff(VArena, commit_used  ))[0] = header;
+	u4_r(base + soff(VArena, flags        ))[0] = flags;
 	return base;
 }
 inline
 void varena__push__u(U8 vm, U8 amount, U8 type_width, U8 alignment, U8 result) {
 	assert(result != null);
 	assert(vm  != null);
-	if (amount == 0) { struct_zero(Slice_Mem, result); return; }
-	U8       align          = alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT;
+	if (amount == 0) { slice_clear(result); return; }
+	alignment = alignment == 0 ? alignment : MEMORY_ALIGNMENT_DEFAULT;
 	U8   requested_size = amount * type_width;
-	U8   reg aligned_size   = align_pow2(requested_size, align);
-	U8_R reg commit_used    = u8_r(vm + offset_of(VArena, commit_used  ));
-	U8   reg reserve_left   = u8_r(vm + offset_of(VArena, reserve      ))[0] - commit_used[0];
-	if (aligned_size > reserve_left) { struct_zero(Slice_Mem, result); return; }
-	U8 reg committed   = u8_r(vm + offset_of(VArena, committed  ))[0];
+	U8   aligned_size   = align_pow2(requested_size, alignment);
+	U8_R commit_used    = u8_r(vm + soff(VArena, commit_used  ));
+	U8   to_be_used     = commit_used[0] + aligned_size;
+	U8   reserve_left   = u8_r(vm + soff(VArena, reserve      ))[0] - commit_used[0];
+	U8   committed      = u8_r(vm + soff(VArena, committed  ))[0];
 	U8   commit_left    = committed - commit_used[0];
-	if (commit_left < aligned_size) {
-		U8 reg commit_size = u8_r(vm + offset_of(VArena, commit_size))[0];
-		U8 reg next_commit = reserve_left > aligned_size ? max(commit_size, aligned_size) : reserve_left;
-		if (next_commit != 0) {
-			B4 no_large = (u4_r(vm + offset_of(VArena, flags))[0] & VArenaFlag_NoLargePages) != 0;
-			U8 reg next_commit_start = vm + committed;
-			if (os_vmem_commit__u(next_commit_start, next_commit) == false) {
+	assert(to_be_used< reserve_left);
+	if (/*exhausted?*/commit_left < aligned_size) {
+		U8 commit_size      = u8_r(vm + soff(VArena, commit_size))[0];
+		U8 next_commit_size = reserve_left > aligned_size ? max(commit_size, aligned_size) : reserve_left;
+		if (next_commit_size != 0) {
+			B4 no_large_pages = (u4_r(vm + soff(VArena, flags))[0] & VArenaFlag_NoLargePages) != 0;
+			U8 next_commit_start = vm + committed;
+			if (os_vmem_commit__u(next_commit_start, next_commit_size, no_large_pages) == false) {
 				struct_zero(Slice_Mem, result);
 				return;
 			}
-			committed += next_commit;
-			u8_r(vm + offset_of(VArena, committed))[0] = committed;
+			committed += next_commit_size;
+			u8_r(vm + soff(VArena, committed))[0] = committed;
 		}
 	}
-	commit_used[0] += aligned_size; {
-		U8 reg current_offset = u8_r(vm + offset_of(VArena, reserve_start))[0] + commit_used[0];
-		struct_copy(Slice_Mem, result, (U8)& slice_mem(current_offset, requested_size));
-	}
+	commit_used[0] += aligned_size;
+	U8 current_offset = u8_r(vm + soff(VArena, reserve_start))[0] + commit_used[0];
+	slice_assign(result, (U8)& slice_mem(current_offset, requested_size));
 }
 inline
 void varena__grow__u(U8 result, U8 vm, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment, B4 should_zero) {
 	assert(vm     != null);
 	assert(result != null);
-	U8 reg grow_amount = requested_size - old_len;
-	if (grow_amount == 0) { struct_copy(Slice_Mem, result, (U8)& slice_mem(old_ptr, old_len)); return; }
-	U8 reg current_offset = u8_r(vm + offset_of(VArena, reserve_start))[0] + u8_r(vm + offset_of(VArena, commit_used))[0];
+	U8 grow_amount = requested_size - old_len;
+	if (grow_amount == 0) { slice_assign(result, (U8)& slice_mem(old_ptr, old_len)); return; }
+	U8 current_offset = u8_r(vm + soff(VArena, reserve_start))[0] + u8_r(vm + soff(VArena, commit_used))[0];
 	// Growing when not the last allocation not allowed
 	assert(old_ptr == current_offset); 
 	uvar(Slice_Mem, allocation); varena__push__u(vm, grow_amount, 1, alignment, u8_(allocation));
-
+	U8 a_ptr = u8_r(allocation + soff(Slice_Mem, ptr))[0];
+	U8 a_len = u8_r(allocation + soff(Slice_Mem, len))[0];
+	assert(a_ptr != 0);
+	mem_zero(a_ptr, a_len * should_zero);
+	slice_assign(result, (U8)& slice_mem(old_ptr, old_len + a_len));
 }
+inline
 void varena__shrink__u(U8 result, U8 vm, U8 old_ptr, U8 old_len, U8 requested_size, U8 alignment) {
 	assert(vm     != null);
 	assert(result != null);
-
+	U8 shrink_amount = old_len - requested_size;
+	if (lt_s(shrink_amount, 0)) { slice_assign(result, (U8)& slice_mem(old_ptr, old_len)); return; }
+	U8_R commit_used    = u8_r(vm + soff(VArena, commit_used));
+	U8   current_offset = u8_r(vm + soff(VArena, reserve_start))[0] + commit_used[0]; assert(old_ptr == current_offset);
+	commit_used[0]   -= shrink_amount;
+	slice_assign(result, (U8)& slice_mem(old_ptr, requested_size));
 }
 I_ void varena_release__u(U8 vm) {
 	assert(vm != null);
-	os_vmem_release__u(vm, u8_r(vm + offset_of(VArena, reserve))[0]);
+	os_vmem_release__u(vm, u8_r(vm + soff(VArena, reserve))[0]);
 }
 I_ void varena_reset__u(U8 vm) {
 	assert(vm != null);
-	u8_r(vm + offset_of(VArena, commit_used))[0] = 0;
+	u8_r(vm + soff(VArena, commit_used))[0] = 0;
 }
-I_ void varena_rewind__u(U8 vm, U8 sp_type_sig, U8 sp_slot) {
+I_ void varena_rewind__u(U8 vm, U8 sp_slot) {
 	assert(vm       != null);
-	assert(sp_type_sig == (U8) varena_allocator_proc);
-	U8 reg header = varena_header_size();
+	U8 header = varena_header_size();
 	if (sp_slot < header) { sp_slot = header; }
-	u8_r(vm + offset_of(VArena, commit_used))[0] = sp_slot;
+	u8_r(vm + soff(VArena, commit_used))[0] = sp_slot;
 }
 I_ void varena_save__u(U8 vm, U8 sp_addr) {
 	assert(vm   != null);
 	assert(sp_addr != null);
-	u8_r(sp_addr + offset_of(AllocatorSP, type_sig))[0] = (U8) varena_allocator_proc;
-	u8_r(sp_addr + offset_of(AllocatorSP, slot    ))[0] = u8_r(vm + offset_of(VArena, commit_used))[0];
+	u8_r(sp_addr + soff(AllocatorSP, type_sig))[0] = (U8) varena_allocator_proc;
+	u8_r(sp_addr + soff(AllocatorSP, slot    ))[0] = u8_r(vm + soff(VArena, commit_used))[0];
 }
 
 I_ VArena* varena__make(Opts_varena_make*R_ opts) {
@@ -938,7 +1075,8 @@ I_ void varena_release(VArena_R vm) { varena_release__u(u8_(vm)); }
 I_ void varena_reset  (VArena_R vm) { varena_reset__u  (u8_(vm)); }
 
 I_ void varena_rewind (VArena_R vm, AllocatorSP save_point) { 
-	varena_rewind__u(u8_(vm), u8_(save_point.type_sig), save_point.slot);
+	assert(save_point.type_sig == varena_allocator_proc);
+	varena_rewind__u(u8_(vm), save_point.slot);
 }
 I_ AllocatorSP varena_save(VArena_R vm) { AllocatorSP sp; varena_save__u(u8_(vm), u8_(& sp)); return sp; }
 
@@ -946,16 +1084,16 @@ void varena_allocator_proc(U8 vm, U8 requested_size, U8 alignment, U8 old_ptr, U
 {
 	assert(vm       != null);
 	assert(out_addr != null);
-	U8 out_allocation = out_addr ? out_addr + offset_of(AllocatorProc_Out, allocation) : 0;
+	U8 out_allocation = out_addr ? out_addr + soff(AllocatorProc_Out, allocation) : 0;
 	switch (op)
 	{
 	case AllocatorOp_Alloc:
 	case AllocatorOp_Alloc_NoZero:
 		varena__push__u(vm, requested_size, 1, alignment, out_allocation);
 		if (op == AllocatorOp_Alloc) {
-			U8 ptr = u8_r(out_allocation + offset_of(Slice_Mem, ptr))[0];
-			U8 len = u8_r(out_allocation + offset_of(Slice_Mem, len))[0];
-			if (ptr && len) { memory_zero(ptr, len); }
+			U8 ptr = u8_r(out_allocation + soff(Slice_Mem, ptr))[0];
+			U8 len = u8_r(out_allocation + soff(Slice_Mem, len))[0];
+			if (ptr && len) { mem_zero(ptr, len); }
 		}
 	break;
 
@@ -970,33 +1108,254 @@ void varena_allocator_proc(U8 vm, U8 requested_size, U8 alignment, U8 old_ptr, U
 		varena__shrink__u(out_allocation, vm, old_ptr, old_len, requested_size, alignment);
 	break;
 
-	case AllocatorOp_Rewind:    varena_rewind__u(vm, old_ptr, old_len);                                    break;
-	case AllocatorOp_SavePoint: varena_save__u  (vm, out_addr + offset_of(AllocatorProc_Out, save_point)); break;
+	case AllocatorOp_Rewind:    varena_rewind__u(vm, old_len);                               break;
+	case AllocatorOp_SavePoint: varena_save__u  (vm, out_addr + soff(AllocatorProc_Out, save_point)); break;
 
 	case AllocatorOp_Query:
-			u4_r(out_addr + offset_of(AllocatorQueryInfo, features))[0] =
+			u4_r(out_addr + soff(AllocatorQueryInfo, features))[0] =
 			  AllocatorQuery_Alloc
 			| AllocatorQuery_Reset
 			| AllocatorQuery_Resize
 			| AllocatorQuery_Rewind;
-			U8 reserve   = u8_r(vm + offset_of(VArena, reserve  ))[0];
-			U8 committed = u8_r(vm + offset_of(VArena, committed))[0];
+			U8 reserve   = u8_r(vm + soff(VArena, reserve  ))[0];
+			U8 committed = u8_r(vm + soff(VArena, committed))[0];
 			U8 max_alloc = (reserve > committed) ? (reserve - committed) : 0;
-			u8_r(out_addr + offset_of(AllocatorQueryInfo, max_alloc))[0] = max_alloc;
-			u8_r(out_addr + offset_of(AllocatorQueryInfo, min_alloc))[0] = kilo(4);
-			u8_r(out_addr + offset_of(AllocatorQueryInfo, left     ))[0] = max_alloc;
-			AllocatorSP sp = { .type_sig = varena_allocator_proc, .slot = u8_r(vm + offset_of(VArena, commit_used))[0] };
-			struct_copy(AllocatorSP, out_addr + offset_of(AllocatorQueryInfo, save_point), (U8)& sp);
+			u8_r(out_addr + soff(AllocatorQueryInfo, max_alloc))[0] = max_alloc;
+			u8_r(out_addr + soff(AllocatorQueryInfo, min_alloc))[0] = kilo(4);
+			u8_r(out_addr + soff(AllocatorQueryInfo, left     ))[0] = max_alloc;
+			AllocatorSP sp = { .type_sig = varena_allocator_proc, .slot = u8_r(vm + soff(VArena, commit_used))[0] };
+			struct_assign(AllocatorSP, out_addr + soff(AllocatorQueryInfo, save_point), (U8)& sp);
 	break;
 	}
 }
 #pragma endregion VArena
 
+#pragma region Arena
+I_ U8 arena_header_size(void) { return align_pow2(size_of(Arena), MEMORY_ALIGNMENT_DEFAULT); }
+
+U8 arena_make__u(U8 reserve_size, U8 commit_size, U4 flags, U8 base_addr) {
+	U8 header_size = arena_header_size();
+	U8 current     = varena__make__u(reserve_size, commit_size, flags, base_addr); assert(current != null);
+	U8 arena; varena__push__u(current, header_size, 1, MEMORY_ALIGNMENT_DEFAULT, (U8)& arena);
+	u8_r(arena + soff(Arena, backing ))[0] = current;
+	u8_r(arena + soff(Arena, prev    ))[0] = null;
+	u8_r(arena + soff(Arena, current ))[0] = arena;
+	u8_r(arena + soff(Arena, base_pos))[0] = 0;
+	u8_r(arena + soff(Arena, pos     ))[0] = header_size;
+	u8_r(arena + soff(Arena, flags   ))[0] = flags;
+	return arena;
+}
+void arena__push__u(U8 arena, U8 amount, U8 type_width, U8 alignment, U8 out_mem) {
+	assert(arena != null);
+	U8 active          = u8_r(arena + soff(Arena, current ))[0];
+	U8 size_requested  = amount * type_width;
+	   alignment       = alignment ? alignment : MEMORY_ALIGNMENT_DEFAULT;
+	U8 size_aligned    = align_pow2(size_requested, alignment);
+	U8 pos_pre         = u8_r(active + soff(Arena, pos))[0];
+	U8 pos_pst         = pos_pre + size_aligned;
+	U8 backing         = active + soff(Arena, backing);
+	U8 reserve         = u8_r(backing + soff(VArena, reserve))[0];
+	B4 should_chain    = 
+     ((u8_r(arena + soff(Arena, flags))[0] & ArenaFlag_NoChain) == 0)
+	&& reserve < pos_pst;
+	if (should_chain)
+	{
+		U8 current   = arena + soff(Arena, current);
+		U8 new_arena = arena_make__u(
+			reserve, 
+			u8_r(backing + soff(VArena, commit_size))[0], 
+			u4_r(backing + soff(VArena, flags      ))[0], 
+			0
+		);
+		u8_r(new_arena + soff(Arena, base_pos))[0] = u8_r(active + soff(Arena, base_pos))[0] + reserve;
+		u8_r(new_arena + soff(Arena, prev    ))[0] = u8_r(current)[0];
+		u8_r(current)[0]                           = new_arena;
+		active                                     = u8_r(current)[0];
+	}
+	U8 result = active + pos_pre;
+	varena__push__u(u8_r(backing)[0], size_aligned, 1, alignment, out_mem);
+	assert(u8_r(out_mem + soff(Slice_Mem, ptr))[0] == result);
+	assert(u8_r(out_mem + soff(Slice_Mem, len))[0] >  0);
+	u8_r(active + soff(Arena, pos))[0] = pos_pst;
+}
+I_ void arena_release__u(U8 arena) {
+	assert(arena != null);
+	U8 curr = arena + soff(Arena, current); 
+	U8 prev = null;
+	for (; u8_r(curr)[0] != null; curr = prev) {
+		u8_r(prev)[0] = u8_r(curr + soff(Arena, prev))[0];
+		varena_release__u(u8_r(curr)[0]);
+	}
+}
+I_ void arena_reset__u(U8 arena) { arena_rewind__u(arena, 0); }
+void arena_rewind__u(U8 arena, U8 slot) {
+	assert(arena != null);
+	U8 header_size = arena_header_size();
+	U8 curr        = arena + soff(Arena, current);
+	U8 big_pos     = clamp_bot(header_size, slot);
+	for (U8 prev = null; u8_r(curr, soff(Arena, base_pos))[0] >= big_pos; u8_r(curr) = prev) {
+		prev = u8_r(curr + soff(Arena, prev))[0];
+		varena_release__u(u8_r(curr + soff(Arena, backing))[0]);
+	}
+	u8_r(arena + soff(Arena, current))[0] = u8_r(curr)[0];
+	U8 new_pos = big_pos - u8_r(curr + soff(Arena, base_pos))[0]; assert(new_pos <= u8_r(curr + soff(Arena, pos))[0]);
+	u8_r(curr + soff(Arena, pos))[0] = new_pos;
+}
+I_ void arena_save__u(U8 arena, U8 out_sp) {
+	u8_r(out_sp + soff(AllocatorSP, type_sig))[0] = arena_allocator_proc;
+	u8_r(out_sp + soff(AllocatorSP, slot    ))[0] = 
+	    u8_r(arena + soff(Arena, base_pos)                  )[0] 
+		+ u8_r(arena + soff(Arena, current) + soff(Arena, pos))[0];
+}
+void arena_allocator_proc(U8 arena, U8 requested_size, U8 alignment, U8 old_ptr, U8 old_len, U4 op, U8 out_addr)
+{
+	assert(out_addr != null);
+	assert(arena    != null);
+	switch (op)
+	{
+	case AllocatorOp_Alloc:
+	case AllocatorOp_Alloc_NoZero:
+	break;
+
+	case AllocatorOp_Free: break;
+	case AllocatorOp_Reset: break;
+
+	case AllocatorOp_Grow:
+	case AllocatorOp_Grow_NoZero: break;
+	case AllocatorOp_Shrink: break;
+
+	case AllocatorOp_Rewind: break;
+	case AllocatorOp_SavePoint: break;
+
+	case AllocatorOp_Query:
+	break;
+	}
+}
+#pragma endregion Arena
+
+#pragma region Key Table Linear (KTL)
+I_ void ktl_populate_slice_a2_str8(U8 kt, U8 backing_ptr, U8 backing_len, U8 values) {
+	assert(kt != null);
+	U8 values_len = u8_r(values + soff(Slice_A2_Str8, len))[0];
+	if (values_len == 0) return;
+	mem__alloc__u(kt, backing_ptr, backing_len, size_of(KTL_Slot_Str8) * values_len, 0, false);
+	for (U8 id = 0; id < values_len; ++id) { 
+		U8 kt_slot = kt + soff(KTL_Str8, ptr) * id; 
+		U8 value   = u8_r(values + soff(Slice_A2_Str8, ptr))[0] + size_of(A2_Str8) * id;
+		mem_copy        (kt_slot + soff(KTL_Slot_Str8, value), value + size_of(Str8) * 1, size_of(Str8));
+		hash64__fnv1a__u(kt_slot + soff(KTL_Slot_Str8, key),   value);
+	}
+}
+#pragma endregion KTL
+
+#pragma region Key Table 1-Layer Chained-Chunked_Cells (KT1CX)
+#pragma endregion Key Table
+
+#pragma region String Operations
+#pragma endregion String Operations
+
+#pragma region File System
+#pragma endregion File System
+
+#pragma region Debug
+#if defined(BUILD_DEBUG)
+// #include <stdio.h>
+#define MS_CRT_INTERNAL_LOCAL_PRINTF_OPTIONS (*__local_stdio_printf_options())
+#define MS_stderr                          (__acrt_iob_func(2))
+#define MS__crt_va_start_a(ap, x)          ((void)(__va_start(&ap, x)))
+#define MS__crt_va_arg(ap, t)                                          \
+	((sizeof(t) > sizeof(__int64) || (sizeof(t) & (sizeof(t) - 1)) != 0) \
+		? **(t**)((ap += sizeof(__int64)) - sizeof(__int64))               \
+		:  *(t* )((ap += sizeof(__int64)) - sizeof(__int64)))
+#define MS__crt_va_end(ap)           ((void)(ap = (va_list)0))
+#define va_start(ap, x)              MS__crt_va_start_a(ap, x)
+#define va_arg                       MS__crt_va_arg
+#define va_end                       MS__crt_va_end
+#define va_copy(destination, source) ((destination) = (source))
+typedef def_struct(__crt_locale_pointers) { struct __crt_locale_data* locinfo; struct __crt_multibyte_data* mbcinfo; };
+typedef __crt_locale_pointers* _locale_t;
+typedef char*                  va_list;
+MS_FILE* __cdecl __acrt_iob_func(unsigned _Ix);
+N_
+U8* __cdecl __local_stdio_printf_options(void) {
+	// NOTE(CRT): This function must not be inlined into callers to avoid ODR violations.  The
+	// static local variable has different names in C and in C++ translation units.
+	local_persist U8 _OptionsStorage; return &_OptionsStorage;
+}
+int __cdecl __stdio_common_vfprintf_s(
+	U8               _Options,
+	MS_FILE*         _Stream,
+	char const*      _Format,
+	_locale_t        _Locale,
+	va_list          _ArgList
+);
+void __cdecl __va_start(va_list* , ...);
+I_ int printf_err(char const* fmt, ...) {
+	int result;
+	va_list args;
+	va_start(args, fmt);
+	result = __stdio_common_vfprintf_s(MS_CRT_INTERNAL_LOCAL_PRINTF_OPTIONS, MS_stderr, fmt, nullptr, args);
+	va_end(args);
+	return result;
+}
+void assert_handler( UTF8*R_ condition, UTF8*R_ file, UTF8*R_ function, S4 line, UTF8*R_ msg, ... ) {
+	printf_err( "%s - %s:(%d): Assert Failure: ", file, function, line );
+	if ( condition )
+		printf_err( "`%s` \n", condition );
+	if ( msg ) {
+		va_list va = {0};
+		va_start( va, msg );
+		__stdio_common_vfprintf_s(MS_CRT_INTERNAL_LOCAL_PRINTF_OPTIONS, MS_stderr, msg, nullptr, va);
+		va_end( va );
+	}
+	printf_err( "%s", "\n" );
+}
+#endif
+#pragma endregion Debug
+
+#pragma region WATL
+#pragma endregion WATL
+
 #pragma endregion Implementation
 
 int main(void)
 {
-	return 0;
+	os_init();
+	VArena_R vm_file = varena_make(.reserve_size = giga(4), .flags = VArenaFlag_NoLargePages);
+	// FileOpInfo file  = file_read_contents(lit("watl.v0.llvm.lottes.c"), .backing = ainfo_varena(vm_file));
+	// slice_assert(file.content);
+
+	Arena_R a_msgs = arena_make();
+	Arena_R a_toks = arena_make();
+	// WATL_LexInfo lex_res = watl_lex(pcast(Str8, file.content),
+	// 	.ainfo_msgs = ainfo_arena(a_msgs),
+	// 	.ainfo_toks = ainfo_arena(a_toks),
+	// );
+	// assert((lex_res.signal & WATL_LexStatus_MemFail_SliceConstraintFail) == 0);
+
+	// Arena_R str_cache_kt1_ainfo = arena_make();
+	// Str8Cache str_cache = str8cache_make(
+	// 	.str_reserve    = ainfo_arena(arena_make(.reserve_size = mega(256))),
+	// 	.cell_reserve   = ainfo_arena(str_cache_kt1_ainfo),
+	// 	.tbl_backing    = ainfo_arena(str_cache_kt1_ainfo),
+	// 	.cell_pool_size = kilo(8),
+	// 	.table_size     = kilo(64),
+	// );
+
+	// Arena_R a_lines = arena_make();
+	// WATL_ParseInfo parse_res = watl_parse(lex_res.toks,
+	// 	.ainfo_msgs  = ainfo_arena(a_msgs),
+	// 	.ainfo_nodes = ainfo_arena(a_toks),
+	// 	.ainfo_lines = ainfo_arena(a_lines),
+	// 	.str_cache   = & str_cache
+	// );
+	// assert((parse_res.signal & WATL_ParseStatus_MemFail_SliceConstraintFail) == 0);
+
+	// arena_reset(a_msgs);
+	// arena_reset(a_toks);
+	// Str8 listing = watl_dump_listing(ainfo_arena(a_msgs), parse_res.lines);
+	// file_write_str8(lit("watl.v0.lottes.c.listing.txt"), listing);
+	// return 0;
 }
 
 #pragma clang diagnostic pop

C/watl.v0.llvm.lottes_hybrid.c

@@ -9,6 +9,9 @@ https://youtu.be/RrL7121MOeA
 */
 
 #pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wpre-c11-compat"
+// #pragma clang diagnostic ignored "-Wc++-keyword"
+#pragma clang diagnostic ignored "-Wcast-qual"
 #pragma clang diagnostic ignored "-Wunused-const-variable"
 #pragma clang diagnostic ignored "-Wunused-but-set-variable"
 #pragma clang diagnostic ignored "-Wswitch"
@@ -20,14 +23,12 @@ https://youtu.be/RrL7121MOeA
 #pragma clang diagnostic ignored "-W#pragma-messages"
 #pragma clang diagnostic ignored "-Wstatic-in-inline"
 #pragma clang diagnostic ignored "-Wkeyword-macro"
-#pragma clang diagnostic ignored "-Wc23-compat"
 #pragma clang diagnostic ignored "-Wreserved-identifier"
-#pragma clang diagnostic ignored "-Wpre-c11-compat"
+#pragma clang diagnostic ignored "-Wc23-compat"
 #pragma clang diagnostic ignored "-Wc23-extensions"
 #pragma clang diagnostic ignored "-Wunused-macros"
 #pragma clang diagnostic ignored "-Wdeclaration-after-statement"
 #pragma clang diagnostic ignored "-Wunsafe-buffer-usage"
-#pragma clang diagnostic ignored "-Wc++-keyword"
 #pragma clang diagnostic ignored "-Wimplicit-function-declaration"
 #pragma clang diagnostic ignored "-Wcast-align"
 #pragma clang diagnostic ignored "-Wunused-parameter"
@@ -39,10 +40,14 @@ https://youtu.be/RrL7121MOeA
 #pragma region Header
 
 #pragma region DSL
+#define local_persist      static
+#define global             static
+#define internal           static
+
 #define align_(value)     __attribute__((aligned (value)))               // for easy alignment
 #define expect_(x, y)     __builtin_expect(x, y)                         // so compiler knows the common path
-#define finline           static inline __attribute__((always_inline)) // force inline
-#define no_inline         static        __attribute__((noinline))      // force no inline [used in thread api]
+#define finline           internal inline __attribute__((always_inline)) // force inline
+#define no_inline         internal        __attribute__((noinline))      // force no inline [used in thread api]
 #define R_                __restrict                                     // pointers are either restricted or volatile and nothing else 
 #define V_                volatile                                       // pointers are either restricted or volatile and nothing else
 #define W_                __attribute((__stdcall__)) __attribute__((__force_align_arg_pointer__))
@@ -50,13 +55,9 @@ https://youtu.be/RrL7121MOeA
 #define glue_impl(A, B)    A ## B
 #define glue(A, B)         glue_impl(A, B)
 #define stringify_impl(S)  #S
-#define stringify(S)       stringify_impl(S)
+#define stringify(S)       cast(UTF8*, stringify_impl(S))
 #define tmpl(prefix, type) prefix ## _ ## type
 
-#define local_persist      static
-#define global             static
-#define internal           static
-
 #define static_assert      _Static_assert
 #define typeof             __typeof__
 #define typeof_ptr(ptr)    typeof(ptr[0])
@@ -176,8 +177,8 @@ typedef def_struct(Slice_Str8) { Str8* ptr; U8 len; };
 	{                        \
 		assert_handler(        \
 			stringify(cond),     \
-			__FILE__,            \
-			__func__,            \
+			(UTF8*)__FILE__,     \
+			(UTF8*)__func__,     \
 			cast(S4, __LINE__),  \
 			msg,                 \
 			## __VA_ARGS__);     \
@@ -1100,7 +1101,6 @@ inline void  os_vmem_release(U8 vm, U8 size) { VirtualFree(cast(MS_LPVOID, vm),
 
 #pragma region VArena (Virutal Address Space Arena)
 finline U8 varena_header_size(void) { return align_pow2(size_of(VArena), MEMORY_ALIGNMENT_DEFAULT); }
-
 inline
 VArena* varena__make(Opts_varena_make*R_ opts) {
 	assert(opts != nullptr);
@@ -1130,13 +1130,11 @@ Slice_Mem varena__push(VArena_R vm, U8 amount, U8 type_width, Opts_varena*R_ opt
 	U8 alignment      = opts->alignment ? opts->alignment : MEMORY_ALIGNMENT_DEFAULT;
 	U8 requested_size = amount * type_width;
 	U8 aligned_size   = align_pow2(requested_size, alignment);
-	U8 current_offset = vm->reserve_start + vm->commit_used;
 	U8 to_be_used     = vm->commit_used   + aligned_size;
 	U8 reserve_left   = vm->reserve       - vm->commit_used;
 	U8 commit_left    = vm->committed     - vm->commit_used;
-	B4 exhausted      = commit_left < to_be_used; assert(to_be_used < reserve_left);
-	if (exhausted)
-	{
+	assert(to_be_used < reserve_left);
+	if (/*exhausted?*/commit_left < to_be_used) {
 		U8 next_commit_size = reserve_left > 0 ?
 			max(vm->commit_size, to_be_used)
 		:	align_pow2( reserve_left, os_system_info()->target_page_size);
@@ -1149,17 +1147,17 @@ Slice_Mem varena__push(VArena_R vm, U8 amount, U8 type_width, Opts_varena*R_ opt
 		}
 	}
 	vm->commit_used = to_be_used;
+	U8 current_offset = vm->reserve_start + vm->commit_used;
 	return (Slice_Mem){.ptr = current_offset, .len = requested_size};
 }
 inline
-Slice_Mem varena__grow(VArena_R vm, Slice_Mem old_allocation, U8 requested_size, U8 alignment, B4 no_zero) {
+Slice_Mem varena__grow(VArena_R vm, Slice_Mem old_allocation, U8 requested_size, U8 alignment, B4 should_zero) {
 	U8 grow_amount = requested_size - old_allocation.len;
 	if (grow_amount == 0) { return old_allocation; }                    // Growing when not the last allocation not allowed
 	U8 current_offset = vm->reserve_start + vm->commit_used;            assert(old_allocation.ptr == current_offset); 
 	Slice_Mem allocation = varena_push_mem(vm, grow_amount, alignment); assert(allocation.ptr != 0);
-	Slice_Mem result     = (Slice_Mem){ old_allocation.ptr, requested_size + allocation.len }; 
-	mem_zero(result.ptr, result.len * no_zero);
-	return result;
+	mem_zero(allocation.ptr, allocation.len * should_zero);
+	return (Slice_Mem){ old_allocation.ptr, old_allocation.len + allocation.len }; 
 }
 finline void varena_release(VArena_R arena) { os_vmem_release(u8_(arena), arena->reserve); }
 inline
@@ -1219,12 +1217,12 @@ void varena_allocator_proc(AllocatorProc_In in, AllocatorProc_Out* out)
 #pragma endregion VArena
 
 #pragma region Arena (Chained Arena)
+finline U8 arena_header_size(void) { return align_pow2(size_of(Arena), MEMORY_ALIGNMENT_DEFAULT); }
 inline
 Arena* arena__make(Opts_arena_make*R_ opts) {
 	assert(opts != nullptr);
-	U8 header_size  = align_pow2(size_of(Arena), MEMORY_ALIGNMENT_DEFAULT);
-	VArena_R current = varena__make(opts);
-	assert(current != nullptr);
+	U8 header_size = arena_header_size();
+	VArena_R current = varena__make(opts); assert(current != nullptr);
 	Arena*   arena   = varena_push(current, Arena); r_(arena)[0] = (Arena){
 		.backing  = current,
 		.prev     = nullptr,
@@ -1288,8 +1286,7 @@ void arena_rewind(Arena_R arena, AllocatorSP save_point) {
 		varena_release(curr->backing);
 	}
 	arena->current = curr;
-	U8 new_pos = big_pos - curr->base_pos;
-	assert(new_pos <= curr->pos);
+	U8 new_pos     = big_pos - curr->base_pos; assert(new_pos <= curr->pos);
 	curr->pos      = new_pos;
 	varena_rewind(curr->backing, (AllocatorSP){varena_allocator_proc, curr->pos + size_of(VArena)});
 }

C/watl.v0.msvc.c

@@ -334,9 +334,9 @@ typedef def_struct(Opts_vmem) {
 void os_init(void);
 OS_SystemInfo* os_system_info(void);
 
-inline B32   os__vmem_commit(void* vm, SSIZE size, Opts_vmem* opts);
-inline Byte* os__vmem_reserve(SSIZE size, Opts_vmem* opts);
-inline void  os_vmem_release(void* vm, SSIZE size);
+inline B32   os__vmem_commit (void* vm, SSIZE size, Opts_vmem* opts);
+inline Byte* os__vmem_reserve(          SSIZE size, Opts_vmem* opts);
+inline void  os_vmem_release (void* vm, SSIZE size);
 
 #define os_vmem_reserve(size, ...)    os__vmem_reserve(   size, opt_args(Opts_vmem, __VA_ARGS__))
 #define os_vmem_commit(vm, size, ...) os__vmem_commit(vm, size, opt_args(Opts_vmem, __VA_ARGS__))
@@ -844,7 +844,7 @@ Slice_Byte farena__push(FArena* arena, SSIZE amount, SSIZE type_width, Opts_fare
 	return (Slice_Byte){ptr, desired};
 }
 inline
-Slice_Byte farena__grow(FArena* arena, SSIZE requested_size, Slice_Byte old_allocation, SSIZE alignment, B32 no_zero) {
+Slice_Byte farena__grow(FArena* arena, SSIZE requested_size, Slice_Byte old_allocation, SSIZE alignment, B32 should_zero) {
 	// Check if the allocation is at the end of the arena
 	Byte* alloc_end = old_allocation.ptr + old_allocation.len;
 	Byte* arena_end = cast(Byte*, cast(SSIZE, arena->start) + arena->used);
@@ -861,9 +861,8 @@ Slice_Byte farena__grow(FArena* arena, SSIZE requested_size, Slice_Byte old_allo
 		return (Slice_Byte){0};
 	}
 	arena->used += aligned_grow;
-	Slice_Byte result = (Slice_Byte){old_allocation.ptr, requested_size};
-	memory_zero(old_allocation.ptr + old_allocation.len, grow_amount * cast(SSIZE, no_zero));
-	return result;
+	memory_zero(old_allocation.ptr + old_allocation.len, grow_amount * cast(SSIZE, should_zero));
+	return (Slice_Byte){old_allocation.ptr, requested_size};
 }
 inline
 Slice_Byte farena__shrink(FArena* arena, Slice_Byte old_allocation, SSIZE requested_size, SSIZE alignment) {
@@ -890,8 +889,7 @@ void farena_rewind(FArena* arena, AllocatorSP save_point) {
 }
 inline
 AllocatorSP farena_save (FArena  arena) {
-	AllocatorSP sp = { .type_sig = & farena_allocator_proc, .slot = cast(SSIZE, arena.used) };
-	return sp;
+	return (AllocatorSP){ .type_sig = & farena_allocator_proc, .slot = cast(SSIZE, arena.used) };
 }
 void farena_allocator_proc(AllocatorProc_In in, AllocatorProc_Out* out)
 {
@@ -1071,7 +1069,6 @@ Slice_Byte varena__push(VArena* vm, SSIZE amount, SSIZE type_width, Opts_varena*
 	SSIZE alignment      = opts->alignment ? opts->alignment : MEMORY_ALIGNMENT_DEFAULT;
 	SSIZE requested_size = amount * type_width;
 	SSIZE aligned_size   = align_pow2(requested_size, alignment);
-	SSIZE current_offset = vm->reserve_start + vm->commit_used;
 	SSIZE to_be_used     = vm->commit_used   + aligned_size;
 	SSIZE reserve_left   = vm->reserve       - vm->commit_used;
 	SSIZE commit_left    = vm->committed     - vm->commit_used;
@@ -1089,18 +1086,18 @@ Slice_Byte varena__push(VArena* vm, SSIZE amount, SSIZE type_width, Opts_varena*
 		}
 	}
 	vm->commit_used = to_be_used;
+	SSIZE current_offset = vm->reserve_start + vm->commit_used;
 	return (Slice_Byte){.ptr = cast(Byte*, current_offset), .len = requested_size};
 }
 inline
-Slice_Byte varena__grow(VArena* vm, SSIZE requested_size, Slice_Byte old_allocation, SSIZE alignment, B32 no_zero) {
+Slice_Byte varena__grow(VArena* vm, SSIZE requested_size, Slice_Byte old_allocation, SSIZE alignment, B32 should_zero) {
 	assert(vm != nullptr);
 	SSIZE grow_amount = requested_size - old_allocation.len;
 	if (grow_amount == 0) { return old_allocation; }                             // Growing when not the last allocation not allowed
 	SSIZE current_offset = vm->reserve_start + vm->commit_used;                  assert(old_allocation.ptr == cast(Byte*, current_offset));
 	Slice_Byte allocation = varena_push_array(vm, Byte, grow_amount, alignment); assert(allocation.ptr != nullptr);
-	Slice_Byte result     = (Slice_Byte){ old_allocation.ptr, requested_size };
-	memory_zero(allocation.ptr, allocation.len * no_zero);
-	return result;
+	memory_zero(allocation.ptr, allocation.len * should_zero);
+	return (Slice_Byte){ old_allocation.ptr, old_allocation.len + allocation.len };
 }
 inline Slice_Byte varena__shrink(VArena* vm, Slice_Byte old_allocation, SSIZE requested_size) {
 	SSIZE current_offset = vm->reserve_start + vm->commit_used;