updated odin's str8_fmt_kt1l

C/watl.v0.llvm.lottes_hybrid.c

@@ -401,7 +401,7 @@ void farena_allocator_proc(AllocatorProc_In in, AllocatorProc_Out_R out);
 #define farena_push_mem(arena, amount, ...) farena__push(arena, amount, 1, opt_args(Opts_farena, lit(stringify(B1)), __VA_ARGS__))
 
 #define farena_push(arena, type, ...) \
-cast(type*R_, farena__push(arena, size_of(type), 1, opt_args(Opts_farena, lit(stringify(type)), __VA_ARGS__))).ptr
+cast(type*, farena__push(arena, size_of(type), 1, opt_args(Opts_farena, lit(stringify(type)), __VA_ARGS__))).ptr
 
 #define farena_push_array(arena, type, amount, ...) \
 (Slice ## type){ farena__push(arena, size_of(type), amount, opt_args(Opts_farena, lit(stringify(type)), __VA_ARGS__)).ptr, amount }
@@ -504,7 +504,7 @@ void arena_allocator_proc(AllocatorProc_In in, AllocatorProc_Out_R out);
 #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*R_, arena__push(arena, 1, size_of(type), opt_args(Opts_arena, lit(stringify(type)), __VA_ARGS__) ).ptr)
+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 }
@@ -1677,38 +1677,38 @@ Str8 str8__fmt_kt1l(AllocatorInfo ainfo, Slice_Mem*R_ _buffer, KT1L_Str8 table,
 	UTF8_R cursor_buffer    = cast(UTF8_R, buffer.ptr);
 	U8     buffer_remaining = buffer.len;
 
-	UTF8   curr_code  = fmt_template.ptr[0];
 	UTF8_R cursor_fmt = fmt_template.ptr;
 	U8     left_fmt   = fmt_template.len;
 	while (left_fmt && buffer_remaining)
 	{
 		// Forward until we hit the delimiter '<' or the template's contents are exhausted.
-		while (curr_code && curr_code != '<' && cursor_fmt != slice_end(fmt_template)) {
+		U8 copy_offset = 0;
-			cursor_buffer[0] = cursor_fmt[0];
+		while (cursor_fmt[copy_offset] != '<' && (cursor_fmt + copy_offset) < slice_end(fmt_template)) {
-			++ cursor_buffer;
+			++ copy_offset;
-			++ cursor_fmt;
-			-- buffer_remaining;
-			-- left_fmt;
-			curr_code = cursor_fmt[0];
 		}
-		if (curr_code == '<')
+		memory_copy(u8_(cursor_buffer), u8_(cursor_fmt), copy_offset);
+		buffer_remaining -= copy_offset;
+		left_fmt         -= copy_offset;
+		cursor_buffer    += copy_offset;
+		cursor_fmt       += copy_offset;
+
+		if (cursor_fmt[0] == '<')
 		{
-			UTF8_R cursor_potential_token = cursor_fmt + 1;
+			UTF8_R potential_token_cursor = cursor_fmt + 1;
-			U8     potential_token_length = 0;
+			U8     potential_token_len    = 0;
 			B4     fmt_overflow           = false;
 			for (;;) {
-				UTF8_R cursor       = cursor_potential_token + potential_token_length;
+				UTF8_R cursor       = potential_token_cursor + potential_token_len;
 				fmt_overflow        = cursor >= slice_end(fmt_template);
-				B4 found_terminator = (cursor_potential_token + potential_token_length)[0] == '>';
+				B4 found_terminator = potential_token_cursor[potential_token_len] == '>';
 				if (fmt_overflow || found_terminator) { break; }
-				++ potential_token_length;
+				++ potential_token_len;
 			}
 			if (fmt_overflow) continue;
 			// Hashing the potential token and cross checking it with our token table
-			U8     key   = 0; hash64_djb8(& key, slice_mem(u8_(cursor_potential_token), potential_token_length));
+			U8     key   = 0; hash64_djb8(& key, slice_mem(u8_(potential_token_cursor), potential_token_len));
 			Str8_R value = nullptr;
-			for (slice_iter(table, token))
+			for (slice_iter(table, token)) {
-			{
 				// We do a linear iteration instead of a hash table lookup because the user should be never substiuting with more than 100 unqiue tokens..
 				if (token->key == key) {
 					value = & token->value;
@@ -1718,32 +1718,26 @@ Str8 str8__fmt_kt1l(AllocatorInfo ainfo, Slice_Mem*R_ _buffer, KT1L_Str8 table,
 			if (value)
 			{
 				// We're going to appending the string, make sure we have enough space in our buffer.
-				if (ainfo.proc != nullptr && (buffer_remaining - potential_token_length) <= 0) {
+				if (ainfo.proc != nullptr && (buffer_remaining - potential_token_len) <= 0) {
-					buffer            = mem_grow(ainfo, buffer, buffer.len + potential_token_length);
+					buffer            = mem_grow(ainfo, buffer, buffer.len + potential_token_len);
-					buffer_remaining += potential_token_length;
+					buffer_remaining += potential_token_len;
-				}
-				U8   left         = value->len;
-				U1_R cursor_value = value->ptr;
-
-				while (left && buffer_remaining) {
-					cursor_buffer[0] = cursor_value[0];
-					++ cursor_buffer;
-					++ cursor_value;
-					-- buffer_remaining;
-					-- left;
 				}
+				assert((buffer_remaining - potential_token_len) > 0);
+				memory_copy(u8_(cursor_buffer), u8_(value->ptr), value->len);
 				// Sync cursor format to after the processed token
-				cursor_fmt = cursor_potential_token + potential_token_length + 1;
+				cursor_buffer    += value->len;
-				curr_code  = cursor_fmt[0];
+				buffer_remaining -= value->len;
-				left_fmt  -= potential_token_length + 2; // The 2 here are the '<' & '>' delimiters being omitted.
+				cursor_fmt        = potential_token_cursor + potential_token_len + 1;
+				left_fmt         -= potential_token_len + 2; // The 2 here are the '<' & '>' delimiters being omitted.
 				continue;
 			}
+			// If not a subsitution, we do a single copy for the '<' and continue.
 			cursor_buffer[0] = cursor_fmt[0];
 			++ cursor_buffer;
 			++ cursor_fmt;
 			-- buffer_remaining;
 			-- left_fmt;
-			curr_code = cursor_fmt[0];
+			continue;
 		}
 	}
 	_buffer[0] = buffer;
@@ -2277,7 +2271,7 @@ int main(void)
 	os_init();
 
 	VArena_R vm_file = varena_make(.reserve_size = giga(4), .flags = VArenaFlag_NoLargePages);
-	FileOpInfo file  = file_read_contents(lit("watl.v0.msvc.c"), .backing = ainfo_varena(vm_file));
+	FileOpInfo file  = file_read_contents(lit("watl.v0.llvm.lottes_hybrid.c"), .backing = ainfo_varena(vm_file));
 	slice_assert(file.content);
 
 	Arena_R a_msgs = arena_make();

C/watl.v0.msvc.c

@@ -1585,7 +1585,7 @@ Str8 str8__fmt_kt1l(AllocatorInfo ainfo, Slice_Byte* _buffer, KT1L_Str8 table, S
 	{
 		SSIZE copy_offset = 0;
 		// Forward until we hit the delimiter '<' or the template's contents are exhausted.
-		while (cursor_fmt[copy_offset] != cast(UTF8, '<') && (cursor_fmt + copy_offset) != slice_end(fmt_template)) {
+		while (cursor_fmt[copy_offset] != cast(UTF8, '<') && (cursor_fmt + copy_offset) < slice_end(fmt_template)) {
 			++ copy_offset;
 		}
 		memory_copy(cursor_buffer, cursor_fmt, copy_offset);
@@ -1602,7 +1602,7 @@ Str8 str8__fmt_kt1l(AllocatorInfo ainfo, Slice_Byte* _buffer, KT1L_Str8 table, S
 			for (;;) {
 				UTF8* cursor         = potential_token_cursor + potential_token_len;
 				fmt_overflow         = cursor >= slice_end(fmt_template);
-				B32 found_terminator = * (potential_token_cursor + potential_token_len) == '>';
+				B32 found_terminator = potential_token_cursor[potential_token_len] == '>';
 				if (fmt_overflow || found_terminator) { break; }
 				++ potential_token_len;
 			}
@@ -1633,7 +1633,7 @@ Str8 str8__fmt_kt1l(AllocatorInfo ainfo, Slice_Byte* _buffer, KT1L_Str8 table, S
 				left_fmt         -= potential_token_len + 2; // The 2 here are the '<' & '>' delimiters being omitted.
 				continue;
 			}
-			// If not a value, we do a single copy for the '<' and continue.
+			// If not a subsitution, we do a single copy for the '<' and continue.
 			*  cursor_buffer = * cursor_fmt;
 			++ cursor_buffer;
 			++ cursor_fmt;

Odin/watl.v0.win32.odin

@@ -2,7 +2,7 @@
 WATL Exercise
 Version:   0 (From Scratch, 1-Stage Compilation, WinAPI Only, Win CRT Multi-threaded Static Linkage)
 Host:      Windows 11 (x86-64)
-Toolchain: odin-lang/Odin dev-2025-07
+Toolchain: odin-lang/Odin dev-2025-09
 */
 package odin
 
@@ -31,9 +31,9 @@ copy :: proc {
 	slice_copy,
 	string_copy,
 }
-copy_non_overlapping :: proc {
+copy_overlapping :: proc {
-	memory_copy_non_overlapping,
+	memory_copy_overlapping,
-	slice_copy_non_overlapping,
+	slice_copy_overlapping,
 }
 cursor :: proc {
 	ptr_cursor,
@@ -109,15 +109,33 @@ memory_zero_explicit :: proc "contextless" (data: rawptr, len: int) -> rawptr {
 	intrinsics.atomic_thread_fence(.Seq_Cst) // Prevent reordering
 	return data
 }
-memory_copy :: proc "contextless" (dst, src: rawptr, len: int) -> rawptr {
+memory_copy_overlapping :: proc "contextless" (dst, src: rawptr, len: int) -> rawptr {
 	intrinsics.mem_copy(dst, src, len)
 	return dst
 }
-memory_copy_non_overlapping :: proc "contextless" (dst, src: rawptr, len: int) -> rawptr {
+memory_copy :: proc "contextless" (dst, src: rawptr, len: int) -> rawptr {
 	intrinsics.mem_copy_non_overlapping(dst, src, len)
 	return dst
 }
 
+sll_stack_push_n :: proc "contextless" (curr, n, n_link: ^^$Type) {
+    (n_link ^) = (curr ^)
+    (curr   ^) = (n    ^)
+}
+sll_queue_push_nz :: proc "contextless" (first: ^$ParentType, last, n: ^^$Type, nil_val: ^Type) {
+	if (first ^) == nil_val {
+		(first ^) = n^
+		(last  ^) = n^
+		n^.next = nil_val
+	}
+	else {
+		(last ^).next = n^
+		(last ^)      = n^
+		n^.next        = nil_val
+	}
+}
+sll_queue_push_n :: #force_inline proc "contextless" (first: $ParentType, last, n: ^^$Type) { sll_queue_push_nz(first, last, n, nil) }
+
 SliceByte :: struct {
 	data: [^]byte,
 	len: int
@@ -139,37 +157,19 @@ slice_end :: #force_inline proc "contextless" (s : $SliceType / []$Type) -> ^Typ
 
 slice_zero :: proc "contextless" (data: $SliceType / []$Type) { memory_zero(raw_data(data), size_of(Type) * len(data)) }
 slice_copy :: proc "contextless" (dst, src: $SliceType / []$Type) -> int {
-	n := max(0, min(len(dst), len(src)))
-	if n > 0 {
-		intrinsics.mem_copy(raw_data(dst), raw_data(src), n * size_of(Type))
-	}
-	return n
-}
-slice_copy_non_overlapping :: proc "contextless" (dst, src: $SliceType / []$Type) -> int {
 	n := max(0, min(len(dst), len(src)))
 	if n > 0 {
 		intrinsics.mem_copy_non_overlapping(raw_data(dst), raw_data(src), n * size_of(Type))
 	}
 	return n
 }
-
+slice_copy_overlapping :: proc "contextless" (dst, src: $SliceType / []$Type) -> int {
-sll_stack_push_n :: proc "contextless" (curr, n, n_link: ^^$Type) {
+	n := max(0, min(len(dst), len(src)))
-    (n_link ^) = (curr ^)
+	if n > 0 {
-    (curr   ^) = (n    ^)
+		intrinsics.mem_copy(raw_data(dst), raw_data(src), n * size_of(Type))
-}
-sll_queue_push_nz :: proc "contextless" (first: ^$ParentType, last, n: ^^$Type, nil_val: ^Type) {
-	if (first ^) == nil_val {
-		(first ^) = n^
-		(last  ^) = n^
-		n^.next = nil_val
-	}
-	else {
-		(last ^).next = n^
-		(last ^)      = n^
-		n^.next        = nil_val
 	}
+	return n
 }
-sll_queue_push_n :: #force_inline proc "contextless" (first: $ParentType, last, n: ^^$Type) { sll_queue_push_nz(first, last, n, nil) }
 //endregion Memory
 
 //region Allocator Interface
@@ -1220,35 +1220,36 @@ str8_fmt_kt1l :: proc(ainfo: AllocatorInfo, _buffer: ^[]byte, table: []KT1L_Slot
 	cursor_buffer    := cursor(buffer)
 	buffer_remaining := len(buffer)
 
-	curr_code  := fmt_template[0]
 	cursor_fmt := cursor(transmute([]u8) fmt_template)
 	left_fmt   := len(fmt_template)
 	for ; left_fmt > 0 && buffer_remaining > 0;
 	{
 		// Forward until we hit the delimiter '<' or the template's contents are exhausted.
-		for ; curr_code != '<' && cursor_fmt != end(fmt_template); {
+		copy_offset : int = 0
-			cursor_buffer[0]  = cursor_fmt   [0]
+		for ; cursor_fmt[copy_offset] != '<' && cursor_fmt[copy_offset:] != end(fmt_template); {
-			cursor_buffer     = cursor_buffer[1:]
+			copy_offset += 1
-			cursor_fmt        = cursor_fmt   [1:]
-			curr_code         = cursor_fmt   [0]
-			buffer_remaining -= 1
-			left_fmt         -= 1
 		}
-		if curr_code == '<'
+		copy(cursor_buffer, cursor_fmt, copy_offset)
+		buffer_remaining -= copy_offset
+		left_fmt         -= copy_offset
+		cursor_buffer     = cursor_buffer[copy_offset:]
+		cursor_fmt        = cursor_fmt   [copy_offset:]
+
+		if cursor_fmt[0] == '<'
 		{
-			cursor_potential_token := cursor_fmt[1:]
+			potential_token_cursor := cursor_fmt[1:]
-			potential_token_length := 0
+			potential_token_len    := 0
 			fmt_overflow           := b32(false)
 			for ;; {
-				cursor      := cursor_potential_token[potential_token_length:]
+				cursor      := potential_token_cursor[potential_token_len:]
 				fmt_overflow = cursor >= end(fmt_template)
-				found_terminator := cast(b32) (cursor_potential_token[potential_token_length] == '>')
+				found_terminator := cast(b32) (potential_token_cursor[potential_token_len] == '>')
 				if fmt_overflow || found_terminator do break
-				potential_token_length += 1
+				potential_token_len += 1
 			}
 			if fmt_overflow do continue
 			// Hashing the potential token and cross checking it with our token table
-			key   : u64     = 0; hash64_djb8(& key, slice(cursor_potential_token, potential_token_length))
+			key   : u64     = 0; hash64_djb8(& key, slice(potential_token_cursor, potential_token_len))
 			value : ^string = nil
 			for & token in table 
 			{
@@ -1261,32 +1262,26 @@ str8_fmt_kt1l :: proc(ainfo: AllocatorInfo, _buffer: ^[]byte, table: []KT1L_Slot
 			if value != nil 
 			{
 				// We're going to appending the string, make sure we have enough space in our buffer.
-				if ainfo.procedure != nil && (buffer_remaining - potential_token_length) <= 0 {
+				if ainfo.procedure != nil && (buffer_remaining - potential_token_len) <= 0 {
-					buffer            = mem_grow(ainfo, buffer, len(buffer) + potential_token_length)
+					buffer            = mem_grow(ainfo, buffer, len(buffer) + potential_token_len)
-					buffer_remaining += potential_token_length
+					buffer_remaining += potential_token_len
-				}
-				left         := len(value)
-				cursor_value := cursor(transmute([]u8) value^)
-				for ; left > 0 && buffer_remaining > 0; {
-					cursor_buffer[0]  = cursor_value [0]
-					cursor_buffer     = cursor_buffer[1:]
-					cursor_value      = cursor_value [1:]
-					cursor_fmt        = cursor_fmt   [1:]
-					buffer_remaining -= 1
-					left             -= 1
 				}
+				assert((buffer_remaining - potential_token_len) > 0)
+				copy(cursor_buffer, cursor(value ^), len(value))
 				// Sync cursor format to after the processed token
-				cursor_fmt = cursor_potential_token[potential_token_length + 1:]
+				cursor_buffer     = cursor_buffer[len(value):]
-				curr_code  = cursor_fmt[0]
+				buffer_remaining -= len(value)
-				left_fmt  -= potential_token_length + 2 // The 2 here are the '<' & '>' delimiters being omitted.
+				cursor_fmt        = potential_token_cursor[potential_token_len + 1:]
+				left_fmt         -= potential_token_len + 2 // The 2 here are the '<' & '>' delimiters being omitted.
 				continue
 			}
+			// If not a subsitution, we do a single copy for the '<' and continue.
 			cursor_buffer[0]  = cursor_fmt   [0]
 			cursor_buffer     = cursor_buffer[1:]
 			cursor_fmt        = cursor_fmt   [1:]
-			curr_code         = cursor_fmt   [0]
 			buffer_remaining -= 1
 			left_fmt         -= 1
+			continue
 		}
 	}
 	_buffer ^ = buffer

scripts/build.c.ps1

@@ -42,8 +42,8 @@ $flag_link_win_machine_64          = '/MACHINE:X64'
 $flag_link_win_path_output         = '/OUT:'
 $flag_link_win_rt_dll              = '/MD'
 $flag_link_win_rt_dll_debug        = '/MDd'
-$flag_link_win_rt_static           = '/MT:STATIC'
+$flag_link_win_rt_static           = '/MT'
-$flag_link_win_rt_static_debug     = '/MTd:STATIC'
+$flag_link_win_rt_static_debug     = '/MTd'
 $flag_link_win_subsystem_console   = '/SUBSYSTEM:CONSOLE'
 $flag_link_win_subsystem_windows   = '/SUBSYSTEM:WINDOWS'
 $flag_no_optimization              = '/Od'

scripts/build.c_lottes_hybrid.ps1

@@ -126,7 +126,7 @@ $compiler_args += ($flag_define + 'BUILD_DEBUG')
 $compiler_args += $flag_debug
 $compiler_args += ( $flag_path_debug + $path_build + '\' )
 # Use the static, multithreaded, debug runtime library
-$compiler_args += $flag_link_win_rt_static_debug
+# $compiler_args += $flag_link_win_rt_static_debug
 
 # Include setup
 $compiler_args += ($flag_include + $path_root)

scripts/build.odin.odin

@@ -85,7 +85,6 @@ msvc_link_default_base_address :: 0x180000000
 //endregion Script Grime
 
 build :: proc(working_dir : string, args : []string) -> (stdout : string, stderr : string) { 
-	fmt.println("Building:", args)
 	res, errs : []byte; _, res, errs, _ = os.process_exec({ working_dir = working_dir, command = args}, context.allocator)
 	return transmute(string)res, transmute(string)errs;
 }
@@ -106,8 +105,8 @@ main :: proc() {
 		file_source,
 		flag_file,
 		join_str(flag_output_path, file_exe),
-		// flag_optimize_none,
+		flag_optimize_none,
-		falg_optimize_aggressive,
+		// falg_optimize_aggressive,
 		flag_default_allocator_nil,
 		flag_debug,
 		flag_microarch_zen5,
@@ -117,6 +116,4 @@ main :: proc() {
 		flag_use_lld,
 		join_str(flag_subsystem, "console"),
 	})
-	fmt.println(res)
-	fmt.println(errs)
 }

scripts/build.odin.ps1

@@ -44,5 +44,5 @@ $build_args += $flag_dyn_map_calls
 $build_args += $flag_default_allocator_nil
 $build_args += $flag_output_path + $exe
 if ($need_rebuild) { & $odin $build_args }
-pop-location
 & $exe
+pop-location