Speed up SSA generation and clang compilation

examples/game.odin

@@ -6,9 +6,7 @@ TWO_HEARTS :: #rune "💕"
 
 win32_perf_count_freq := GetQueryPerformanceFrequency()
 time_now :: proc() -> f64 {
-	if win32_perf_count_freq == 0 {
-		debug_trap()
-	}
+	assert(win32_perf_count_freq != 0)
 
 	counter: i64
 	_ = QueryPerformanceCounter(^counter)
@@ -25,10 +23,10 @@ win32_print_last_error :: proc() {
 }
 
 // Yuk!
-to_c_string :: proc(s: string) -> ^u8 {
-	c_str: ^u8 = alloc(len(s)+1)
-	memory_copy(c_str, ^s[0], len(s))
-	ptr_offset(c_str, len(s))^ = 0
+to_c_string :: proc(s: string) -> []u8 {
+	c_str := new_slice(u8, len(s)+1)
+	_ = copy(c_str, s as []byte)
+	c_str[len(s)] = 0
 	return c_str
 }
 
@@ -39,7 +37,7 @@ Window :: type struct {
 	dc:                 HDC
 	hwnd:               HWND
 	opengl_context, rc: HGLRC
-	c_title:            ^u8
+	c_title:            []u8
 }
 
 make_window :: proc(title: string, msg, height: int, window_proc: WNDPROC) -> (Window, bool) {
@@ -65,7 +63,7 @@ make_window :: proc(title: string, msg, height: int, window_proc: WNDPROC) -> (W
 	}
 
 	w.hwnd = CreateWindowExA(0,
-	                         c_class_name, w.c_title,
+	                         c_class_name, ^w.c_title[0],
 	                         WS_VISIBLE | WS_OVERLAPPED | WS_CAPTION | WS_SYSMENU | WS_MINIMIZEBOX,
 	                         CW_USEDEFAULT, CW_USEDEFAULT,
 	                         w.width as i32, w.height as i32,
@@ -114,7 +112,7 @@ make_window :: proc(title: string, msg, height: int, window_proc: WNDPROC) -> (W
 }
 
 destroy_window :: proc(w: ^Window) {
-	heap_free(w.c_title)
+	delete(w.c_title)
 }
 
 display_window :: proc(w: ^Window) {

examples/runtime.odin

@@ -1,15 +1,25 @@
 #load "win32.odin"
 
-debug_trap :: proc() #foreign "llvm.debugtrap"
+assume :: proc(cond: bool) #foreign "llvm.assume"
+
+__debug_trap           :: proc()           #foreign "llvm.debugtrap"
+__trap                 :: proc()           #foreign "llvm.trap"
+read_cycle_counter     :: proc() -> u64    #foreign "llvm.readcyclecounter"
+
+bit_reverse16 :: proc(b: u16) -> u16 #foreign "llvm.bitreverse.i16"
+bit_reverse32 :: proc(b: u32) -> u32 #foreign "llvm.bitreverse.i32"
+bit_reverse64 :: proc(b: u64) -> u64 #foreign "llvm.bitreverse.i64"
+
+byte_swap16 :: proc(b: u16) -> u16 #foreign "llvm.bswap.i16"
+byte_swap32 :: proc(b: u32) -> u32 #foreign "llvm.bswap.i32"
+byte_swap64 :: proc(b: u64) -> u64 #foreign "llvm.bswap.i64"
+
+fmuladd_f32 :: proc(a, b, c: f32) -> f32 #foreign "llvm.fmuladd.f32"
+fmuladd_f64 :: proc(a, b, c: f64) -> f64 #foreign "llvm.fmuladd.f64"
 
 // TODO(bill): make custom heap procedures
-heap_alloc :: proc(len: int) -> rawptr {
-	return HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, len)
-}
-heap_free :: proc(ptr: rawptr) {
-	_ = HeapFree(GetProcessHeap(), 0, ptr)
-}
-
+heap_alloc   :: proc(len: int)   -> rawptr #foreign "malloc"
+heap_dealloc :: proc(ptr: rawptr)          #foreign "free"
 
 memory_zero :: proc(data: rawptr, len: int) {
 	d := slice_ptr(data as ^byte, len)
@@ -288,10 +298,10 @@ Allocator :: type struct {
 
 
 Context :: type struct {
-	thread_id: int
+	thread_ptr: rawptr
 
-	user_index: int
 	user_data:  rawptr
+	user_index: int
 
 	allocator: Allocator
 }
@@ -305,6 +315,10 @@ __check_context :: proc() {
 	if context.allocator.procedure == null {
 		context.allocator = __default_allocator()
 	}
+	if context.thread_ptr == null {
+		// TODO(bill):
+		// context.thread_ptr = current_thread_pointer()
+	}
 }
 
 
@@ -368,13 +382,11 @@ __default_allocator_proc :: proc(allocator_data: rawptr, mode: Allocation_Mode,
 	using Allocation_Mode
 	match mode {
 	case ALLOC:
-		data := heap_alloc(size)
-		memory_zero(data, size)
-		return data
+		return heap_alloc(size)
 	case RESIZE:
 		return default_resize_align(old_memory, old_size, size, alignment)
 	case DEALLOC:
-		heap_free(old_memory)
+		heap_dealloc(old_memory)
 	case DEALLOC_ALL:
 		// NOTE(bill): Does nothing
 	}
@@ -394,5 +406,7 @@ __default_allocator :: proc() -> Allocator {
 
 __assert :: proc(msg: string) {
 	file_write(file_get_standard(File_Standard.ERROR), msg as []byte)
-	debug_trap()
+	// TODO(bill): Which is better?
+	// __trap()
+	__debug_trap()
 }