Merge branch 'master' into sys-linux-additions

16 files changed, 615 insertions, 190 deletions

diff --git a/src/build_settings.cpp b/src/build_settings.cpp
index ffb276d1e..94807a852 100644
--- a/src/build_settings.cpp
+++ b/src/build_settings.cpp
@@ -82,6 +82,23 @@ gb_global String target_arch_names[TargetArch_COUNT] = {
 	str_lit("wasm64p32"),
 };
 
+gb_global String target_microarch_list[TargetArch_COUNT] = {
+	// TargetArch_Invalid,
+	str_lit("Invalid!"),
+	// TargetArch_amd64,
+	str_lit("alderlake,amdfam10,athlon-fx,athlon64,athlon64-sse3,atom_sse4_2,atom_sse4_2_movbe,barcelona,bdver1,bdver2,bdver3,bdver4,broadwell,btver1,btver2,cannonlake,cascadelake,cooperlake,core-avx-i,core-avx2,core2,core_2_duo_sse4_1,core_2_duo_ssse3,core_2nd_gen_avx,core_3rd_gen_avx,core_4th_gen_avx,core_4th_gen_avx_tsx,core_5th_gen_avx,core_5th_gen_avx_tsx,core_aes_pclmulqdq,core_i7_sse4_2,corei7,corei7-avx,generic,goldmont,goldmont-plus,goldmont_plus,grandridge,graniterapids,graniterapids-d,graniterapids_d,haswell,icelake-client,icelake-server,icelake_client,icelake_server,ivybridge,k8,k8-sse3,knl,knm,meteorlake,mic_avx512,native,nehalem,nocona,opteron,opteron-sse3,penryn,raptorlake,rocketlake,sandybridge,sapphirerapids,sierraforest,silvermont,skx,skylake,skylake-avx512,skylake_avx512,slm,tigerlake,tremont,westmere,x86-64,x86-64-v2,x86-64-v3,x86-64-v4,znver1,znver2,znver3,znver4"),
+	// TargetArch_i386,
+	str_lit("athlon,athlon-4,athlon-mp,athlon-tbird,athlon-xp,atom,bonnell,c3,c3-2,generic,geode,i386,i486,i586,i686,k6,k6-2,k6-3,lakemont,native,pentium,pentium-m,pentium-mmx,pentium2,pentium3,pentium3m,pentium4,pentium4m,pentium_4,pentium_4_sse3,pentium_ii,pentium_iii,pentium_iii_no_xmm_regs,pentium_m,pentium_mmx,pentium_pro,pentiumpro,prescott,winchip-c6,winchip2,yonah"),
+	// TargetArch_arm32,
+	str_lit("arm1020e,arm1020t,arm1022e,arm10e,arm10tdmi,arm1136j-s,arm1136jf-s,arm1156t2-s,arm1156t2f-s,arm1176jz-s,arm1176jzf-s,arm710t,arm720t,arm7tdmi,arm7tdmi-s,arm8,arm810,arm9,arm920,arm920t,arm922t,arm926ej-s,arm940t,arm946e-s,arm966e-s,arm968e-s,arm9e,arm9tdmi,cortex-a12,cortex-a15,cortex-a17,cortex-a32,cortex-a35,cortex-a5,cortex-a53,cortex-a55,cortex-a57,cortex-a7,cortex-a710,cortex-a72,cortex-a73,cortex-a75,cortex-a76,cortex-a76ae,cortex-a77,cortex-a78,cortex-a78c,cortex-a8,cortex-a9,cortex-m0,cortex-m0plus,cortex-m1,cortex-m23,cortex-m3,cortex-m33,cortex-m35p,cortex-m4,cortex-m55,cortex-m7,cortex-m85,cortex-r4,cortex-r4f,cortex-r5,cortex-r52,cortex-r7,cortex-r8,cortex-x1,cortex-x1c,cyclone,ep9312,exynos-m3,exynos-m4,exynos-m5,generic,iwmmxt,krait,kryo,mpcore,mpcorenovfp,native,neoverse-n1,neoverse-n2,neoverse-v1,sc000,sc300,strongarm,strongarm110,strongarm1100,strongarm1110,swift,xscale"),
+	// TargetArch_arm64,
+	str_lit("a64fx,ampere1,ampere1a,apple-a10,apple-a11,apple-a12,apple-a13,apple-a14,apple-a15,apple-a16,apple-a7,apple-a8,apple-a9,apple-latest,apple-m1,apple-m2,apple-s4,apple-s5,carmel,cortex-a34,cortex-a35,cortex-a510,cortex-a53,cortex-a55,cortex-a57,cortex-a65,cortex-a65ae,cortex-a710,cortex-a715,cortex-a72,cortex-a73,cortex-a75,cortex-a76,cortex-a76ae,cortex-a77,cortex-a78,cortex-a78c,cortex-r82,cortex-x1,cortex-x1c,cortex-x2,cortex-x3,cyclone,exynos-m3,exynos-m4,exynos-m5,falkor,generic,kryo,native,neoverse-512tvb,neoverse-e1,neoverse-n1,neoverse-n2,neoverse-v1,neoverse-v2,saphira,thunderx,thunderx2t99,thunderx3t110,thunderxt81,thunderxt83,thunderxt88,tsv110"),
+	// TargetArch_wasm32,
+	str_lit("generic"),
+	// TargetArch_wasm64p32,
+	str_lit("generic"),
+};
+
 gb_global String target_endian_names[TargetEndian_COUNT] = {
 	str_lit("little"),
 	str_lit("big"),
@@ -109,8 +126,6 @@ gb_global TargetEndianKind target_endians[TargetArch_COUNT] = {
 
 gb_global String const ODIN_VERSION = str_lit(ODIN_VERSION_RAW);
 
-
-
 struct TargetMetrics {
 	TargetOsKind   os;
 	TargetArchKind arch;
@@ -623,7 +638,6 @@ gb_internal TargetArchKind get_target_arch_from_string(String str) {
 	return TargetArch_Invalid;
 }
 
-
 gb_internal bool is_excluded_target_filename(String name) {
 	String original_name = name;
 	name = remove_extension_from_path(name);
diff --git a/src/check_builtin.cpp b/src/check_builtin.cpp
index 77ba6b435..c0061a397 100644
--- a/src/check_builtin.cpp
+++ b/src/check_builtin.cpp
@@ -2088,6 +2088,8 @@ gb_internal bool check_builtin_procedure(CheckerContext *c, Operand *operand, As
 		t = default_type(t);
 
 		add_type_info_type(c, t);
+		GB_ASSERT(t_type_info_ptr != nullptr);
+		add_type_info_type(c, t_type_info_ptr);
 
 		if (is_operand_value(o) && is_type_typeid(t)) {
 			add_package_dependency(c, "runtime", "__type_info_of");
@@ -5117,6 +5119,202 @@ gb_internal bool check_builtin_procedure(CheckerContext *c, Operand *operand, As
 		}
 		break;
 
+	case BuiltinProc_type_union_tag_type:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+			
+			operand->mode = Addressing_Type;
+			operand->type = union_tag_type(u);
+		}
+		break;
+
+	case BuiltinProc_type_union_tag_offset:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+			
+			// NOTE(jakubtomsu): forces calculation of variant_block_size
+			type_size_of(u);
+			i64 tag_offset = u->Union.variant_block_size;
+			GB_ASSERT(tag_offset > 0);
+			
+			operand->mode = Addressing_Constant;
+			operand->type = t_untyped_integer;
+			operand->value = exact_value_i64(tag_offset);
+		}
+		break;
+
+	case BuiltinProc_type_union_base_tag_value:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+			
+			operand->mode = Addressing_Constant;
+			operand->type = t_untyped_integer;
+			operand->value = exact_value_i64(u->Union.kind == UnionType_no_nil ? 0 : 1);
+		} break;
+
+	case BuiltinProc_type_union_variant_count:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+			
+			operand->mode = Addressing_Constant;
+			operand->type = t_untyped_integer;
+			operand->value = exact_value_i64(u->Union.variants.count);
+		} break;
+
+	case BuiltinProc_type_variant_type_of:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+			Operand x = {};
+			check_expr_or_type(c, &x, ce->args[1]);
+			if (!is_type_integer(x.type) || x.mode != Addressing_Constant) {
+				error(call, "Expected a constant integer for '%.*s", LIT(builtin_name));
+				operand->mode = Addressing_Type;
+				operand->type = t_invalid;
+				return false;
+			}
+			
+			i64 index = big_int_to_i64(&x.value.value_integer);
+			if (index < 0 || index >= u->Union.variants.count) {
+				error(call, "Variant tag out of bounds index for '%.*s", LIT(builtin_name));
+				operand->mode = Addressing_Type;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			operand->mode = Addressing_Type;
+			operand->type = u->Union.variants[index];
+		}
+		break;
+	
+	case BuiltinProc_type_variant_index_of:
+		{
+			if (operand->mode != Addressing_Type) {
+				error(operand->expr, "Expected a type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *u = operand->type;
+
+			if (!is_type_union(u)) {
+				error(operand->expr, "Expected a union type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+
+			Type *v = check_type(c, ce->args[1]);
+			u = base_type(u);
+			GB_ASSERT(u->kind == Type_Union);
+
+			i64 index = -1;			
+			for_array(i, u->Union.variants) {
+				Type *vt = u->Union.variants[i];
+				if (union_variant_index_types_equal(v, vt)) {
+					index = i64(i);
+					break;
+				}
+			}
+			
+			if (index < 0) {
+				error(operand->expr, "Expected a variant type for '%.*s'", LIT(builtin_name));
+				operand->mode = Addressing_Invalid;
+				operand->type = t_invalid;
+				return false;
+			}
+			
+			operand->mode = Addressing_Constant;
+			operand->type = t_untyped_integer;
+			operand->value = exact_value_i64(index);
+		}
+		break;
+
 	case BuiltinProc_type_struct_field_count:
 		operand->value = exact_value_i64(0);
 		if (operand->mode != Addressing_Type) {
diff --git a/src/check_expr.cpp b/src/check_expr.cpp
index 2ee761b9c..71accfb81 100644
--- a/src/check_expr.cpp
+++ b/src/check_expr.cpp
@@ -2339,7 +2339,7 @@ gb_internal void check_unary_expr(CheckerContext *c, Operand *o, Token op, Ast *
 			ast_node(ue, UnaryExpr, node);
 			if (ast_node_expect(ue->expr, Ast_IndexExpr)) {
 				ast_node(ie, IndexExpr, ue->expr);
-				Type *soa_type = type_of_expr(ie->expr);
+				Type *soa_type = type_deref(type_of_expr(ie->expr));
 				GB_ASSERT(is_type_soa_struct(soa_type));
 				o->type = alloc_type_soa_pointer(soa_type);
 			} else {
@@ -2768,6 +2768,11 @@ gb_internal void check_shift(CheckerContext *c, Operand *x, Operand *y, Ast *nod
 					gb_string_free(to_type);
 					x->mode = Addressing_Invalid;
 				}
+			} else if (!is_type_integer(x->type)) {
+				gbString x_str = expr_to_string(x->expr);
+				error(node, "Non-integer shifted operand '%s' is not allowed", x_str);
+				gb_string_free(x_str);
+				x->mode = Addressing_Invalid;
 			}
 			// x->value = x_val;
 			return;
@@ -3558,6 +3563,30 @@ gb_internal void check_binary_expr(CheckerContext *c, Operand *x, Ast *node, Typ
 		return;
 	}
 
+	switch (op.kind) {
+	case Token_Quo:
+	case Token_Mod:
+	case Token_ModMod:
+	case Token_QuoEq:
+	case Token_ModEq:
+	case Token_ModModEq:
+		if (is_type_integer(y->type) && !is_type_untyped(y->type) &&
+		    is_type_float(x->type)   &&  is_type_untyped(x->type)) {
+		    	char const *suggestion = "\tSuggestion: Try explicitly casting the constant value for clarity";
+
+		    	gbString t = type_to_string(y->type);
+			if (x->value.kind != ExactValue_Invalid) {
+				gbString s = exact_value_to_string(x->value);
+				warning(node, "Dividing an untyped float '%s' by '%s' will perform integer division\n%s", s, t, suggestion);
+				gb_string_free(s);
+			} else {
+				warning(node, "Dividing an untyped float by '%s' will perform integer division\n%s", t, suggestion);
+			}
+			gb_string_free(t);
+		}
+		break;
+	}
+
 	convert_to_typed(c, x, y->type);
 	if (x->mode == Addressing_Invalid) {
 		return;
@@ -5577,9 +5606,6 @@ gb_internal CallArgumentError check_call_arguments_internal(CheckerContext *c, A
 	for (isize i = 0; i < pt->param_count; i++) {
 		if (!visited[i]) {
 			Entity *e = pt->params->Tuple.variables[i];
-			if (is_blank_ident(e->token)) {
-				continue;
-			}
 			if (e->kind == Entity_Variable) {
 				if (e->Variable.param_value.kind != ParameterValue_Invalid) {
 					ordered_operands[i].mode = Addressing_Value;
@@ -5623,6 +5649,14 @@ gb_internal CallArgumentError check_call_arguments_internal(CheckerContext *c, A
 			} else {
 				if (show_error) {
 					check_assignment(c, o, param_type, str_lit("procedure argument"));
+
+					Type *src = base_type(o->type);
+					Type *dst = base_type(param_type);
+					if (is_type_slice(src) && are_types_identical(src->Slice.elem, dst)) {
+						gbString a = expr_to_string(o->expr);
+						error_line("\tSuggestion: Did you mean to pass the slice into the variadic parameter with ..%s?\n\n", a);
+						gb_string_free(a);
+					}
 				}
 				err = CallArgumentError_WrongTypes;
 			}
@@ -7391,7 +7425,7 @@ gb_internal bool check_set_index_data(Operand *o, Type *t, bool indirection, i64
 				*max_count = t->Struct.soa_count;
 			}
 			o->type = t->Struct.soa_elem;
-			if (o->mode == Addressing_SoaVariable || o->mode == Addressing_Variable) {
+			if (o->mode == Addressing_SoaVariable || o->mode == Addressing_Variable || indirection) {
 				o->mode = Addressing_SoaVariable;
 			} else {
 				o->mode = Addressing_Value;
diff --git a/src/checker.cpp b/src/checker.cpp
index 29f22bd9c..60b2bbb90 100644
--- a/src/checker.cpp
+++ b/src/checker.cpp
@@ -2211,9 +2211,14 @@ gb_internal void add_min_dep_type_info(Checker *c, Type *t) {
 				Entity *e = entry.value;
 				switch (bt->Struct.soa_kind) {
 				case StructSoa_Dynamic:
+					add_min_dep_type_info(c, t_type_info_ptr); // append_soa
+
 					add_min_dep_type_info(c, t_allocator);
 					/*fallthrough*/
 				case StructSoa_Slice:
+					add_min_dep_type_info(c, t_int);
+					add_min_dep_type_info(c, t_uint);
+					/*fallthrough*/
 				case StructSoa_Fixed:
 					add_min_dep_type_info(c, alloc_type_pointer(e->type));
 					break;
@@ -4733,7 +4738,7 @@ gb_internal void check_add_foreign_import_decl(CheckerContext *ctx, Ast *decl) {
 	}
 
 	if (has_asm_extension(fullpath)) {
-		if (build_context.metrics.arch != TargetArch_amd64) {
+		if (build_context.metrics.arch != TargetArch_amd64 && build_context.metrics.os != TargetOs_darwin) {
 			error(decl, "Assembly files are not yet supported on this platform: %.*s_%.*s",
 			      LIT(target_os_names[build_context.metrics.os]), LIT(target_arch_names[build_context.metrics.arch]));
 		}
@@ -6091,9 +6096,6 @@ gb_internal void check_parsed_files(Checker *c) {
 	TIME_SECTION("calculate global init order");
 	calculate_global_init_order(c);
 
-	TIME_SECTION("check test procedures");
-	check_test_procedures(c);
-
 	TIME_SECTION("add type info for type definitions");
 	add_type_info_for_type_definitions(c);
 	check_merge_queues_into_arrays(c);
@@ -6104,6 +6106,11 @@ gb_internal void check_parsed_files(Checker *c) {
 	TIME_SECTION("generate minimum dependency set");
 	generate_minimum_dependency_set(c, c->info.entry_point);
 
+	// NOTE(laytan): has to be ran after generate_minimum_dependency_set,
+	// because that collects the test procedures.
+	TIME_SECTION("check test procedures");
+	check_test_procedures(c);
+
 	TIME_SECTION("check bodies have all been checked");
 	check_unchecked_bodies(c);
 
diff --git a/src/checker_builtin_procs.hpp b/src/checker_builtin_procs.hpp
index c89ab2429..3bab16293 100644
--- a/src/checker_builtin_procs.hpp
+++ b/src/checker_builtin_procs.hpp
@@ -260,6 +260,12 @@ BuiltinProc__type_simple_boolean_end,
 	BuiltinProc_type_is_specialization_of,
 
 	BuiltinProc_type_is_variant_of,
+	BuiltinProc_type_union_tag_type,
+	BuiltinProc_type_union_tag_offset,
+	BuiltinProc_type_union_base_tag_value,
+	BuiltinProc_type_union_variant_count,
+	BuiltinProc_type_variant_type_of,
+	BuiltinProc_type_variant_index_of,
 
 	BuiltinProc_type_struct_field_count,
 
@@ -557,7 +563,13 @@ gb_global BuiltinProc builtin_procs[BuiltinProc_COUNT] = {
 
 	{STR_LIT("type_is_specialization_of"), 2, false, Expr_Expr, BuiltinProcPkg_intrinsics},
 
-	{STR_LIT("type_is_variant_of"), 2, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_is_variant_of"),          2, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_union_tag_type"),         1, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_union_tag_offset"),       1, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_union_base_tag_value"),   1, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_union_variant_count"),    1, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_variant_type_of"),        2, false, Expr_Expr, BuiltinProcPkg_intrinsics},
+	{STR_LIT("type_variant_index_of"),       2, false, Expr_Expr, BuiltinProcPkg_intrinsics},
 
 	{STR_LIT("type_struct_field_count"),   1, false, Expr_Expr, BuiltinProcPkg_intrinsics},
 
diff --git a/src/linker.cpp b/src/linker.cpp
index eb3687ae2..c3ede0f55 100644
--- a/src/linker.cpp
+++ b/src/linker.cpp
@@ -337,20 +337,34 @@ gb_internal i32 linker_stage(LinkerData *gen) {
 							obj_format = str_lit("elf32");
 						}
 #endif // GB_ARCH_*_BIT
-						// Note(bumbread): I'm assuming nasm is installed on the host machine.
-						// Shipping binaries on unix-likes gets into the weird territorry of
-						// "which version of glibc" is it linked with.
-						result = system_exec_command_line_app("nasm",
-							"nasm \"%.*s\" "
-							"-f \"%.*s\" "
-							"-o \"%.*s\" "
-							"%.*s "
-							"",
-							LIT(asm_file),
-							LIT(obj_format),
-							LIT(obj_file),
-							LIT(build_context.extra_assembler_flags)
-						);						
+
+						if (is_osx) {
+							// `as` comes with MacOS.
+							result = system_exec_command_line_app("as",
+								"as \"%.*s\" "
+								"-o \"%.*s\" "
+								"%.*s "
+								"",
+								LIT(asm_file),
+								LIT(obj_file),
+								LIT(build_context.extra_assembler_flags)
+							);
+						} else {
+							// Note(bumbread): I'm assuming nasm is installed on the host machine.
+							// Shipping binaries on unix-likes gets into the weird territorry of
+							// "which version of glibc" is it linked with.
+							result = system_exec_command_line_app("nasm",
+								"nasm \"%.*s\" "
+								"-f \"%.*s\" "
+								"-o \"%.*s\" "
+								"%.*s "
+								"",
+								LIT(asm_file),
+								LIT(obj_format),
+								LIT(obj_file),
+								LIT(build_context.extra_assembler_flags)
+							);						
+						}
 						array_add(&gen->output_object_paths, obj_file);
 					} else {
 						if (string_set_update(&libs, lib)) {
diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index 513b40a4d..f61c297bd 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -21,6 +21,25 @@
 #include "llvm_backend_stmt.cpp"
 #include "llvm_backend_proc.cpp"
 
+String get_default_microarchitecture() {
+	String default_march = str_lit("generic");
+	if (build_context.metrics.arch == TargetArch_amd64) {
+		// NOTE(bill): x86-64-v2 is more than enough for everyone
+		//
+		// x86-64: CMOV, CMPXCHG8B, FPU, FXSR, MMX, FXSR, SCE, SSE, SSE2
+		// x86-64-v2: (close to Nehalem) CMPXCHG16B, LAHF-SAHF, POPCNT, SSE3, SSE4.1, SSE4.2, SSSE3
+		// x86-64-v3: (close to Haswell) AVX, AVX2, BMI1, BMI2, F16C, FMA, LZCNT, MOVBE, XSAVE
+		// x86-64-v4: AVX512F, AVX512BW, AVX512CD, AVX512DQ, AVX512VL
+		if (ODIN_LLVM_MINIMUM_VERSION_12) {
+			if (build_context.metrics.os == TargetOs_freestanding) {
+				default_march = str_lit("x86-64");
+			} else {
+				default_march = str_lit("x86-64-v2");
+			}
+		}
+	}
+	return default_march;
+}
 
 gb_internal void lb_add_foreign_library_path(lbModule *m, Entity *e) {
 	if (e == nullptr) {
@@ -1827,7 +1846,7 @@ cgscc(
 		function-attrs,
 		function(
 			require<should-not-run-function-passes>
-		),
+		)
 	)
 ),
 deadargelim,
@@ -2490,34 +2509,27 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 		code_mode = LLVMCodeModelKernel;
 	}
 
-	char const *host_cpu_name = LLVMGetHostCPUName();
-	char const *llvm_cpu = "generic";
+	String      host_cpu_name = copy_string(permanent_allocator(), make_string_c(LLVMGetHostCPUName()));
+	String      llvm_cpu      = get_default_microarchitecture();
 	char const *llvm_features = "";
 	if (build_context.microarch.len != 0) {
 		if (build_context.microarch == "native") {
 			llvm_cpu = host_cpu_name;
 		} else {
-			llvm_cpu = alloc_cstring(permanent_allocator(), build_context.microarch);
+			llvm_cpu = copy_string(permanent_allocator(), build_context.microarch);
 		}
-		if (gb_strcmp(llvm_cpu, host_cpu_name) == 0) {
+		if (llvm_cpu == host_cpu_name) {
 			llvm_features = LLVMGetHostCPUFeatures();
 		}
-	} else if (build_context.metrics.arch == TargetArch_amd64) {
-		// NOTE(bill): x86-64-v2 is more than enough for everyone
-		//
-		// x86-64: CMOV, CMPXCHG8B, FPU, FXSR, MMX, FXSR, SCE, SSE, SSE2
-		// x86-64-v2: (close to Nehalem) CMPXCHG16B, LAHF-SAHF, POPCNT, SSE3, SSE4.1, SSE4.2, SSSE3
-		// x86-64-v3: (close to Haswell) AVX, AVX2, BMI1, BMI2, F16C, FMA, LZCNT, MOVBE, XSAVE
-		// x86-64-v4: AVX512F, AVX512BW, AVX512CD, AVX512DQ, AVX512VL
-		if (ODIN_LLVM_MINIMUM_VERSION_12) {
-			if (build_context.metrics.os == TargetOs_freestanding) {
-				llvm_cpu = "x86-64";
-			} else {
-				llvm_cpu = "x86-64-v2";
-			}
-		}
 	}
 
+	// NOTE(Jeroen): Uncomment to get the list of supported microarchitectures.
+	/*
+	if (build_context.microarch == "?") {
+		string_set_add(&build_context.target_features_set, str_lit("+cpuhelp"));
+	}
+	*/
+
 	if (build_context.target_features_set.entries.count != 0) {
 		llvm_features = target_features_set_to_cstring(permanent_allocator(), false);
 	}
@@ -2566,7 +2578,7 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 
 	for (auto const &entry : gen->modules) {
 		LLVMTargetMachineRef target_machine = LLVMCreateTargetMachine(
-			target, target_triple, llvm_cpu,
+			target, target_triple, (const char *)llvm_cpu.text,
 			llvm_features,
 			code_gen_level,
 			reloc_mode,
diff --git a/src/llvm_backend.hpp b/src/llvm_backend.hpp
index 5894dd38a..4e193bcea 100644
--- a/src/llvm_backend.hpp
+++ b/src/llvm_backend.hpp
@@ -339,6 +339,8 @@ struct lbProcedure {
 	bool             in_multi_assignment;
 	Array<LLVMValueRef> raw_input_parameters;
 
+	LLVMValueRef temp_callee_return_struct_memory;
+
 	Ast *curr_stmt;
 
 	Array<Scope *>       scope_stack;
@@ -550,6 +552,7 @@ gb_internal LLVMValueRef lb_call_intrinsic(lbProcedure *p, const char *name, LLV
 gb_internal void lb_mem_copy_overlapping(lbProcedure *p, lbValue dst, lbValue src, lbValue len, bool is_volatile=false);
 gb_internal void lb_mem_copy_non_overlapping(lbProcedure *p, lbValue dst, lbValue src, lbValue len, bool is_volatile=false);
 gb_internal LLVMValueRef lb_mem_zero_ptr_internal(lbProcedure *p, LLVMValueRef ptr, LLVMValueRef len, unsigned alignment, bool is_volatile);
+gb_internal LLVMValueRef lb_mem_zero_ptr_internal(lbProcedure *p, LLVMValueRef ptr, usize len, unsigned alignment, bool is_volatile);
 
 gb_internal gb_inline i64 lb_max_zero_init_size(void) {
 	return cast(i64)(4*build_context.int_size);
diff --git a/src/llvm_backend_expr.cpp b/src/llvm_backend_expr.cpp
index bde975d96..d1176f896 100644
--- a/src/llvm_backend_expr.cpp
+++ b/src/llvm_backend_expr.cpp
@@ -684,7 +684,8 @@ gb_internal lbValue lb_emit_matrix_flatten(lbProcedure *p, lbValue m, Type *type
 	Type *mt = base_type(m.type);
 	GB_ASSERT(mt->kind == Type_Matrix);
 
-	if (lb_is_matrix_simdable(mt)) {
+	// TODO(bill): Determine why this fails on Windows sometimes
+	if (false && lb_is_matrix_simdable(mt)) {
 		LLVMValueRef vector = lb_matrix_to_trimmed_vector(p, m);
 		return lb_matrix_cast_vector_to_type(p, vector, type);
 	}
@@ -693,13 +694,29 @@ gb_internal lbValue lb_emit_matrix_flatten(lbProcedure *p, lbValue m, Type *type
 
 	i64 row_count = mt->Matrix.row_count;
 	i64 column_count = mt->Matrix.column_count;
+	TEMPORARY_ALLOCATOR_GUARD();
+
+	auto srcs = array_make<lbValue>(temporary_allocator(), 0, row_count*column_count);
+	auto dsts = array_make<lbValue>(temporary_allocator(), 0, row_count*column_count);
+
 	for (i64 j = 0; j < column_count; j++) {
 		for (i64 i = 0; i < row_count; i++) {
 			lbValue src = lb_emit_matrix_ev(p, m, i, j);
+			array_add(&srcs, src);
+		}
+	}
+
+	for (i64 j = 0; j < column_count; j++) {
+		for (i64 i = 0; i < row_count; i++) {
 			lbValue dst = lb_emit_array_epi(p, res.addr, i + j*row_count);
-			lb_emit_store(p, dst, src);
+			array_add(&dsts, dst);
 		}
 	}
+
+	GB_ASSERT(srcs.count == dsts.count);
+	for_array(i, srcs) {
+		lb_emit_store(p, dsts[i], srcs[i]);
+	}
 	return lb_addr_load(p, res);
 }
 
@@ -1069,6 +1086,7 @@ gb_internal lbValue lb_emit_arith(lbProcedure *p, TokenKind op, lbValue lhs, lbV
 			args[1] = rhs;
 
 			switch (type_size_of(ft)) {
+			case 2: return lb_emit_runtime_call(p, "quo_complex32", args);
 			case 4: return lb_emit_runtime_call(p, "quo_complex64", args);
 			case 8: return lb_emit_runtime_call(p, "quo_complex128", args);
 			default: GB_PANIC("Unknown float type"); break;
@@ -1145,6 +1163,7 @@ gb_internal lbValue lb_emit_arith(lbProcedure *p, TokenKind op, lbValue lhs, lbV
 			args[1] = rhs;
 
 			switch (8*type_size_of(ft)) {
+			case 16: return lb_emit_runtime_call(p, "mul_quaternion64", args);
 			case 32: return lb_emit_runtime_call(p, "mul_quaternion128", args);
 			case 64: return lb_emit_runtime_call(p, "mul_quaternion256", args);
 			default: GB_PANIC("Unknown float type"); break;
@@ -1157,6 +1176,7 @@ gb_internal lbValue lb_emit_arith(lbProcedure *p, TokenKind op, lbValue lhs, lbV
 			args[1] = rhs;
 
 			switch (8*type_size_of(ft)) {
+			case 16: return lb_emit_runtime_call(p, "quo_quaternion64", args);
 			case 32: return lb_emit_runtime_call(p, "quo_quaternion128", args);
 			case 64: return lb_emit_runtime_call(p, "quo_quaternion256", args);
 			default: GB_PANIC("Unknown float type"); break;
@@ -2919,6 +2939,12 @@ gb_internal lbValue lb_build_unary_and(lbProcedure *p, Ast *expr) {
 	} else if (is_type_soa_pointer(tv.type)) {
 		ast_node(ie, IndexExpr, ue_expr);
 		lbValue addr = lb_build_addr_ptr(p, ie->expr);
+
+		if (is_type_pointer(type_deref(addr.type))) {
+			addr = lb_emit_load(p, addr);
+		}
+		GB_ASSERT(is_type_pointer(addr.type));
+
 		lbValue index = lb_build_expr(p, ie->index);
 
 		if (!build_context.no_bounds_check) {
diff --git a/src/llvm_backend_general.cpp b/src/llvm_backend_general.cpp
index fdcf94f29..c149ec853 100644
--- a/src/llvm_backend_general.cpp
+++ b/src/llvm_backend_general.cpp
@@ -1332,6 +1332,8 @@ gb_internal void lb_emit_store_union_variant(lbProcedure *p, lbValue parent, lbV
 	Type *pt = base_type(type_deref(parent.type));
 	GB_ASSERT(pt->kind == Type_Union);
 	if (pt->Union.kind == UnionType_shared_nil) {
+		GB_ASSERT(type_size_of(variant_type));
+
 		lbBlock *if_nil     = lb_create_block(p, "shared_nil.if_nil");
 		lbBlock *if_not_nil = lb_create_block(p, "shared_nil.if_not_nil");
 		lbBlock *done       = lb_create_block(p, "shared_nil.done");
@@ -1353,9 +1355,13 @@ gb_internal void lb_emit_store_union_variant(lbProcedure *p, lbValue parent, lbV
 
 
 	} else {
-		lbValue underlying = lb_emit_conv(p, parent, alloc_type_pointer(variant_type));
-
-		lb_emit_store(p, underlying, variant);
+		if (type_size_of(variant_type) == 0) {
+			unsigned alignment = 1;
+			lb_mem_zero_ptr_internal(p, parent.value, pt->Union.variant_block_size, alignment, false);
+		} else {
+			lbValue underlying = lb_emit_conv(p, parent, alloc_type_pointer(variant_type));
+			lb_emit_store(p, underlying, variant);
+		}
 		lb_emit_store_union_variant_tag(p, parent, variant_type);
 	}
 }
diff --git a/src/llvm_backend_stmt.cpp b/src/llvm_backend_stmt.cpp
index 9d688be6a..002fef881 100644
--- a/src/llvm_backend_stmt.cpp
+++ b/src/llvm_backend_stmt.cpp
@@ -1846,9 +1846,25 @@ gb_internal void lb_build_return_stmt_internal(lbProcedure *p, lbValue res) {
 		LLVMBuildRetVoid(p->builder);
 	} else {
 		LLVMValueRef ret_val = res.value;
-		ret_val = OdinLLVMBuildTransmute(p, ret_val, p->abi_function_type->ret.type);
-		if (p->abi_function_type->ret.cast_type != nullptr) {
-			ret_val = OdinLLVMBuildTransmute(p, ret_val, p->abi_function_type->ret.cast_type);
+		LLVMTypeRef ret_type = p->abi_function_type->ret.type;
+		if (LLVMTypeRef cast_type = p->abi_function_type->ret.cast_type) {
+			ret_type = cast_type;
+		}
+
+		if (LLVMGetTypeKind(ret_type) == LLVMStructTypeKind) {
+			LLVMTypeRef src_type = LLVMTypeOf(ret_val);
+
+			if (p->temp_callee_return_struct_memory == nullptr) {
+				i64 max_align = gb_max(lb_alignof(ret_type), lb_alignof(src_type));
+				p->temp_callee_return_struct_memory = llvm_alloca(p, ret_type, max_align);
+			}
+			// reuse the temp return value memory where possible
+			LLVMValueRef ptr = p->temp_callee_return_struct_memory;
+			LLVMValueRef nptr = LLVMBuildPointerCast(p->builder, ptr, LLVMPointerType(src_type, 0), "");
+			LLVMBuildStore(p->builder, ret_val, nptr);
+			ret_val = LLVMBuildLoad2(p->builder, ret_type, ptr, "");
+		} else {
+			ret_val = OdinLLVMBuildTransmute(p, ret_val, ret_type);
 		}
 
 		lb_emit_defer_stmts(p, lbDeferExit_Return, nullptr);
diff --git a/src/llvm_backend_type.cpp b/src/llvm_backend_type.cpp
index 02dad2a3a..e291e40a5 100644
--- a/src/llvm_backend_type.cpp
+++ b/src/llvm_backend_type.cpp
@@ -9,7 +9,12 @@ gb_internal isize lb_type_info_index(CheckerInfo *info, Type *type, bool err_on_
 		}
 	}
 	if (err_on_not_found) {
-		GB_PANIC("NOT FOUND lb_type_info_index %s @ index %td", type_to_string(type), index);
+		gb_printf_err("NOT FOUND lb_type_info_index:\n\t%s\n\t@ index %td\n\tmax count: %u\nFound:\n", type_to_string(type), index, set->count);
+		for (auto const &entry : *set) {
+			isize type_info_index = entry.key;
+			gb_printf_err("\t%s\n", type_to_string(info->type_info_types[type_info_index]));
+		}
+		GB_PANIC("NOT FOUND");
 	}
 	return -1;
 }
diff --git a/src/llvm_backend_utility.cpp b/src/llvm_backend_utility.cpp
index d8dbfd736..be3ae9c8a 100644
--- a/src/llvm_backend_utility.cpp
+++ b/src/llvm_backend_utility.cpp
@@ -57,6 +57,10 @@ gb_internal lbValue lb_correct_endianness(lbProcedure *p, lbValue value) {
 	return value;
 }
 
+gb_internal LLVMValueRef lb_mem_zero_ptr_internal(lbProcedure *p, LLVMValueRef ptr, usize len, unsigned alignment, bool is_volatile) {
+	return lb_mem_zero_ptr_internal(p, ptr, LLVMConstInt(lb_type(p->module, t_uint), len, false), alignment, is_volatile);
+}
+
 gb_internal LLVMValueRef lb_mem_zero_ptr_internal(lbProcedure *p, LLVMValueRef ptr, LLVMValueRef len, unsigned alignment, bool is_volatile) {
 	bool is_inlinable = false;
 
diff --git a/src/main.cpp b/src/main.cpp
index 79c2b3561..14f7e84ec 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -199,19 +199,19 @@ gb_internal void print_usage_line(i32 indent, char const *fmt, ...) {
 }
 
 gb_internal void usage(String argv0) {
-	print_usage_line(0, "%.*s is a tool for managing Odin source code", LIT(argv0));
+	print_usage_line(0, "%.*s is a tool for managing Odin source code.", LIT(argv0));
 	print_usage_line(0, "Usage:");
 	print_usage_line(1, "%.*s command [arguments]", LIT(argv0));
 	print_usage_line(0, "Commands:");
-	print_usage_line(1, "build             compile directory of .odin files, as an executable.");
-	print_usage_line(1, "                  one must contain the program's entry point, all must be in the same package.");
-	print_usage_line(1, "run               same as 'build', but also then runs the newly compiled executable.");
-	print_usage_line(1, "check             parse, and type check a directory of .odin files");
-	print_usage_line(1, "strip-semicolon   parse, type check, and remove unneeded semicolons from the entire program");
-	print_usage_line(1, "test              build and runs procedures with the attribute @(test) in the initial package");
-	print_usage_line(1, "doc               generate documentation on a directory of .odin files");
-	print_usage_line(1, "version           print version");
-	print_usage_line(1, "report            print information useful to reporting a bug");
+	print_usage_line(1, "build             Compiles directory of .odin files, as an executable.");
+	print_usage_line(1, "                  One must contain the program's entry point, all must be in the same package.");
+	print_usage_line(1, "run               Same as 'build', but also then runs the newly compiled executable.");
+	print_usage_line(1, "check             Parses, and type checks a directory of .odin files.");
+	print_usage_line(1, "strip-semicolon   Parses, type checks, and removes unneeded semicolons from the entire program.");
+	print_usage_line(1, "test              Builds and runs procedures with the attribute @(test) in the initial package.");
+	print_usage_line(1, "doc               Generates documentation on a directory of .odin files.");
+	print_usage_line(1, "version           Prints version.");
+	print_usage_line(1, "report            Prints information useful to reporting a bug.");
 	print_usage_line(0, "");
 	print_usage_line(0, "For further details on a command, invoke command help:");
 	print_usage_line(1, "e.g. `odin build -help` or `odin help build`");
@@ -1580,45 +1580,45 @@ gb_internal void remove_temp_files(lbGenerator *gen) {
 
 
 gb_internal void print_show_help(String const arg0, String const &command) {
-	print_usage_line(0, "%.*s is a tool for managing Odin source code", LIT(arg0));
+	print_usage_line(0, "%.*s is a tool for managing Odin source code.", LIT(arg0));
 	print_usage_line(0, "Usage:");
 	print_usage_line(1, "%.*s %.*s [arguments]", LIT(arg0), LIT(command));
 	print_usage_line(0, "");
 
 	if (command == "build") {
-		print_usage_line(1, "build   Compile directory of .odin files as an executable.");
+		print_usage_line(1, "build   Compiles directory of .odin files as an executable.");
 		print_usage_line(2, "One must contain the program's entry point, all must be in the same package.");
 		print_usage_line(2, "Use `-file` to build a single file instead.");
 		print_usage_line(2, "Examples:");
-		print_usage_line(3, "odin build .                    # Build package in current directory");
-		print_usage_line(3, "odin build <dir>                # Build package in <dir>");
-		print_usage_line(3, "odin build filename.odin -file  # Build single-file package, must contain entry point.");
+		print_usage_line(3, "odin build .                     Builds package in current directory.");
+		print_usage_line(3, "odin build <dir>                 Builds package in <dir>.");
+		print_usage_line(3, "odin build filename.odin -file   Builds single-file package, must contain entry point.");
 	} else if (command == "run") {
 		print_usage_line(1, "run     Same as 'build', but also then runs the newly compiled executable.");
 		print_usage_line(2, "Append an empty flag and then the args, '-- <args>', to specify args for the output.");
 		print_usage_line(2, "Examples:");
-		print_usage_line(3, "odin run .                    # Build and run package in current directory");
-		print_usage_line(3, "odin run <dir>                # Build and run package in <dir>");
-		print_usage_line(3, "odin run filename.odin -file  # Build and run single-file package, must contain entry point.");
+		print_usage_line(3, "odin run .                     Builds and runs package in current directory.");
+		print_usage_line(3, "odin run <dir>                 Builds and runs package in <dir>.");
+		print_usage_line(3, "odin run filename.odin -file   Builds and runs single-file package, must contain entry point.");
 	} else if (command == "check") {
-		print_usage_line(1, "check   Parse and type check directory of .odin files");
+		print_usage_line(1, "check   Parses and type checks directory of .odin files.");
 		print_usage_line(2, "Examples:");
-		print_usage_line(3, "odin check .                    # Type check package in current directory");
-		print_usage_line(3, "odin check <dir>                # Type check package in <dir>");
-		print_usage_line(3, "odin check filename.odin -file  # Type check single-file package, must contain entry point.");
+		print_usage_line(3, "odin check .                     Type checks package in current directory.");
+		print_usage_line(3, "odin check <dir>                 Type checks package in <dir>.");
+		print_usage_line(3, "odin check filename.odin -file   Type checks single-file package, must contain entry point.");
 	} else if (command == "test") {
-		print_usage_line(1, "test      Build and runs procedures with the attribute @(test) in the initial package");
+		print_usage_line(1, "test    Builds and runs procedures with the attribute @(test) in the initial package.");
 	} else if (command == "doc") {
-		print_usage_line(1, "doc       generate documentation from a directory of .odin files");
+		print_usage_line(1, "doc     Generates documentation from a directory of .odin files.");
 		print_usage_line(2, "Examples:");
-		print_usage_line(3, "odin doc .                    # Generate documentation on package in current directory");
-		print_usage_line(3, "odin doc <dir>                # Generate documentation on package in <dir>");
-		print_usage_line(3, "odin doc filename.odin -file  # Generate documentation on single-file package.");
+		print_usage_line(3, "odin doc .                     Generates documentation on package in current directory.");
+		print_usage_line(3, "odin doc <dir>                 Generates documentation on package in <dir>.");
+		print_usage_line(3, "odin doc filename.odin -file   Generates documentation on single-file package.");
 	} else if (command == "version") {
-		print_usage_line(1, "version   print version");
+		print_usage_line(1, "version   Prints version.");
 	} else if (command == "strip-semicolon") {
 		print_usage_line(1, "strip-semicolon");
-		print_usage_line(2, "Parse and type check .odin file(s) and then remove unneeded semicolons from the entire project");
+		print_usage_line(2, "Parses and type checks .odin file(s) and then removes unneeded semicolons from the entire project.");
 	}
 
 	bool doc             = command == "doc";
@@ -1642,311 +1642,318 @@ gb_internal void print_show_help(String const arg0, String const &command) {
 
 	if (doc) {
 		print_usage_line(1, "-short");
-		print_usage_line(2, "Show shortened documentation for the packages");
+		print_usage_line(2, "Shows shortened documentation for the packages.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-all-packages");
-		print_usage_line(2, "Generates documentation for all packages used in the current project");
+		print_usage_line(2, "Generates documentation for all packages used in the current project.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-doc-format");
-		print_usage_line(2, "Generates documentation as the .odin-doc format (useful for external tooling)");
+		print_usage_line(2, "Generates documentation as the .odin-doc format (useful for external tooling).");
 		print_usage_line(0, "");
 	}
 
 	if (run_or_build) {
 		print_usage_line(1, "-out:<filepath>");
-		print_usage_line(2, "Set the file name of the outputted executable");
+		print_usage_line(2, "Sets the file name of the outputted executable.");
 		print_usage_line(2, "Example: -out:foo.exe");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-o:<string>");
-		print_usage_line(2, "Set the optimization mode for compilation");
+		print_usage_line(2, "Sets the optimization mode for compilation.");
+		print_usage_line(2, "Available options:");
+		print_usage_line(3, "-o:none");
+		print_usage_line(3, "-o:minimal");
+		print_usage_line(3, "-o:size");
+		print_usage_line(3, "-o:speed");
 		if (LB_USE_NEW_PASS_SYSTEM) {
-			print_usage_line(2, "Accepted values: none, minimal, size, speed, aggressive");
-		} else {
-			print_usage_line(2, "Accepted values: none, minimal, size, speed");
+			print_usage_line(3, "-o:aggressive");
 		}
-		print_usage_line(2, "Example: -o:speed");
-		print_usage_line(2, "The default is -o:minimal");
+		print_usage_line(2, "The default is -o:minimal.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-show-timings");
-		print_usage_line(2, "Shows basic overview of the timings of different stages within the compiler in milliseconds");
+		print_usage_line(2, "Shows basic overview of the timings of different stages within the compiler in milliseconds.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-show-more-timings");
-		print_usage_line(2, "Shows an advanced overview of the timings of different stages within the compiler in milliseconds");
+		print_usage_line(2, "Shows an advanced overview of the timings of different stages within the compiler in milliseconds.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-show-system-calls");
-		print_usage_line(2, "Prints the whole command and arguments for calls to external tools like linker and assembler");
+		print_usage_line(2, "Prints the whole command and arguments for calls to external tools like linker and assembler.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-export-timings:<format>");
-		print_usage_line(2, "Export timings to one of a few formats. Requires `-show-timings` or `-show-more-timings`");
+		print_usage_line(2, "Exports timings to one of a few formats. Requires `-show-timings` or `-show-more-timings`.");
 		print_usage_line(2, "Available options:");
-		print_usage_line(3, "-export-timings:json        Export compile time stats to JSON");
-		print_usage_line(3, "-export-timings:csv         Export compile time stats to CSV");
+		print_usage_line(3, "-export-timings:json   Exports compile time stats to JSON.");
+		print_usage_line(3, "-export-timings:csv    Exports compile time stats to CSV.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-export-timings-file:<filename>");
-		print_usage_line(2, "Specify the filename for `-export-timings`");
+		print_usage_line(2, "Specifies the filename for `-export-timings`.");
 		print_usage_line(2, "Example: -export-timings-file:timings.json");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-thread-count:<integer>");
-		print_usage_line(2, "Override the number of threads the compiler will use to compile with");
+		print_usage_line(2, "Overrides the number of threads the compiler will use to compile with.");
 		print_usage_line(2, "Example: -thread-count:2");
 		print_usage_line(0, "");
 	}
 
 	if (check_only) {
 		print_usage_line(1, "-show-unused");
-		print_usage_line(2, "Shows unused package declarations within the current project");
+		print_usage_line(2, "Shows unused package declarations within the current project.");
 		print_usage_line(0, "");
 		print_usage_line(1, "-show-unused-with-location");
-		print_usage_line(2, "Shows unused package declarations within the current project with the declarations source location");
+		print_usage_line(2, "Shows unused package declarations within the current project with the declarations source location.");
 		print_usage_line(0, "");
 	}
 
 	if (run_or_build) {
 		print_usage_line(1, "-keep-temp-files");
-		print_usage_line(2, "Keeps the temporary files generated during compilation");
+		print_usage_line(2, "Keeps the temporary files generated during compilation.");
 		print_usage_line(0, "");
 	} else if (strip_semicolon) {
 		print_usage_line(1, "-keep-temp-files");
-		print_usage_line(2, "Keeps the temporary files generated during stripping the unneeded semicolons from files");
+		print_usage_line(2, "Keeps the temporary files generated during stripping the unneeded semicolons from files.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-collection:<name>=<filepath>");
-		print_usage_line(2, "Defines a library collection used for imports");
+		print_usage_line(2, "Defines a library collection used for imports.");
 		print_usage_line(2, "Example: -collection:shared=dir/to/shared");
 		print_usage_line(2, "Usage in Code:");
 		print_usage_line(3, "import \"shared:foo\"");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-define:<name>=<value>");
-		print_usage_line(2, "Defines a scalar boolean, integer or string as global constant");
+		print_usage_line(2, "Defines a scalar boolean, integer or string as global constant.");
 		print_usage_line(2, "Example: -define:SPAM=123");
-		print_usage_line(2, "To use:  #config(SPAM, default_value)");
+		print_usage_line(2, "Usage in code:");
+		print_usage_line(3, "#config(SPAM, default_value)");
 		print_usage_line(0, "");
 	}
 
 	if (build) {
 		print_usage_line(1, "-build-mode:<mode>");
-		print_usage_line(2, "Sets the build mode");
+		print_usage_line(2, "Sets the build mode.");
 		print_usage_line(2, "Available options:");
-		print_usage_line(3, "-build-mode:exe       Build as an executable");
-		print_usage_line(3, "-build-mode:dll       Build as a dynamically linked library");
-		print_usage_line(3, "-build-mode:shared    Build as a dynamically linked library");
-		print_usage_line(3, "-build-mode:obj       Build as an object file");
-		print_usage_line(3, "-build-mode:object    Build as an object file");
-		print_usage_line(3, "-build-mode:assembly  Build as an assembly file");
-		print_usage_line(3, "-build-mode:assembler Build as an assembly file");
-		print_usage_line(3, "-build-mode:asm       Build as an assembly file");
-		print_usage_line(3, "-build-mode:llvm-ir   Build as an LLVM IR file");
-		print_usage_line(3, "-build-mode:llvm      Build as an LLVM IR file");
+		print_usage_line(3, "-build-mode:exe         Builds as an executable.");
+		print_usage_line(3, "-build-mode:dll         Builds as a dynamically linked library.");
+		print_usage_line(3, "-build-mode:shared      Builds as a dynamically linked library.");
+		print_usage_line(3, "-build-mode:obj         Builds as an object file.");
+		print_usage_line(3, "-build-mode:object      Builds as an object file.");
+		print_usage_line(3, "-build-mode:assembly    Builds as an assembly file.");
+		print_usage_line(3, "-build-mode:assembler   Builds as an assembly file.");
+		print_usage_line(3, "-build-mode:asm         Builds as an assembly file.");
+		print_usage_line(3, "-build-mode:llvm-ir     Builds as an LLVM IR file.");
+		print_usage_line(3, "-build-mode:llvm        Builds as an LLVM IR file.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-target:<string>");
-		print_usage_line(2, "Sets the target for the executable to be built in");
+		print_usage_line(2, "Sets the target for the executable to be built in.");
 		print_usage_line(0, "");
 	}
 
 	if (run_or_build) {
 		print_usage_line(1, "-debug");
-		print_usage_line(2, "Enabled debug information, and defines the global constant ODIN_DEBUG to be 'true'");
+		print_usage_line(2, "Enables debug information, and defines the global constant ODIN_DEBUG to be 'true'.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-disable-assert");
-		print_usage_line(2, "Disable the code generation of the built-in run-time 'assert' procedure, and defines the global constant ODIN_DISABLE_ASSERT to be 'true'");
+		print_usage_line(2, "Disables the code generation of the built-in run-time 'assert' procedure, and defines the global constant ODIN_DISABLE_ASSERT to be 'true'.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-no-bounds-check");
-		print_usage_line(2, "Disables bounds checking program wide");
+		print_usage_line(2, "Disables bounds checking program wide.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-no-crt");
-		print_usage_line(2, "Disables automatic linking with the C Run Time");
+		print_usage_line(2, "Disables automatic linking with the C Run Time.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-no-thread-local");
-		print_usage_line(2, "Ignore @thread_local attribute, effectively treating the program as if it is single-threaded");
+		print_usage_line(2, "Ignores @thread_local attribute, effectively treating the program as if it is single-threaded.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-lld");
-		print_usage_line(2, "Use the LLD linker rather than the default");
+		print_usage_line(2, "Uses the LLD linker rather than the default.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-use-separate-modules");
 		print_usage_line(1, "[EXPERIMENTAL]");
-		print_usage_line(2, "The backend generates multiple build units which are then linked together");
-		print_usage_line(2, "Normally, a single build unit is generated for a standard project");
+		print_usage_line(2, "The backend generates multiple build units which are then linked together.");
+		print_usage_line(2, "Normally, a single build unit is generated for a standard project.");
 		print_usage_line(0, "");
 
 	}
 
 	if (check) {
 		print_usage_line(1, "-no-threaded-checker");
-		print_usage_line(2, "Disabled multithreading in the semantic checker stage");
+		print_usage_line(2, "Disables multithreading in the semantic checker stage.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-vet");
-		print_usage_line(2, "Do extra checks on the code");
+		print_usage_line(2, "Does extra checks on the code.");
 		print_usage_line(2, "Extra checks include:");
-		print_usage_line(2, "-vet-unused");
-		print_usage_line(2, "-vet-shadowing");
-		print_usage_line(2, "-vet-using-stmt");
+		print_usage_line(3, "-vet-unused");
+		print_usage_line(3, "-vet-shadowing");
+		print_usage_line(3, "-vet-using-stmt");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-unused");
-		print_usage_line(2, "Checks for unused declarations");
+		print_usage_line(2, "Checks for unused declarations.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-shadowing");
-		print_usage_line(2, "Checks for variable shadowing within procedures");
+		print_usage_line(2, "Checks for variable shadowing within procedures.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-using-stmt");
-		print_usage_line(2, "Checks for the use of 'using' as a statement");
-		print_usage_line(2, "'using' is considered bad practice outside of immediate refactoring");
+		print_usage_line(2, "Checks for the use of 'using' as a statement.");
+		print_usage_line(2, "'using' is considered bad practice outside of immediate refactoring.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-using-param");
-		print_usage_line(2, "Checks for the use of 'using' on procedure parameters");
-		print_usage_line(2, "'using' is considered bad practice outside of immediate refactoring");
+		print_usage_line(2, "Checks for the use of 'using' on procedure parameters.");
+		print_usage_line(2, "'using' is considered bad practice outside of immediate refactoring.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-style");
-		print_usage_line(2, "Errs on missing trailing commas followed by a newline");
-		print_usage_line(2, "Errs on deprecated syntax");
-		print_usage_line(2, "Does not err on unneeded tokens (unlike -strict-style)");
+		print_usage_line(2, "Errs on missing trailing commas followed by a newline.");
+		print_usage_line(2, "Errs on deprecated syntax.");
+		print_usage_line(2, "Does not err on unneeded tokens (unlike -strict-style).");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-vet-semicolon");
-		print_usage_line(2, "Errs on unneeded semicolons");
+		print_usage_line(2, "Errs on unneeded semicolons.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-ignore-unknown-attributes");
-		print_usage_line(2, "Ignores unknown attributes");
-		print_usage_line(2, "This can be used with metaprogramming tools");
+		print_usage_line(2, "Ignores unknown attributes.");
+		print_usage_line(2, "This can be used with metaprogramming tools.");
 		print_usage_line(0, "");
 
 		if (command != "test") {
 			print_usage_line(1, "-no-entry-point");
-			print_usage_line(2, "Removes default requirement of an entry point (e.g. main procedure)");
+			print_usage_line(2, "Removes default requirement of an entry point (e.g. main procedure).");
 			print_usage_line(0, "");
 		}
 	}
 
 	if (test_only) {
 		print_usage_line(1, "-test-name:<string>");
-		print_usage_line(2, "Run specific test only by name");
+		print_usage_line(2, "Runs specific test only by name.");
 		print_usage_line(0, "");
 	}
 
 	if (run_or_build) {
 		print_usage_line(1, "-minimum-os-version:<string>");
-		print_usage_line(2, "Sets the minimum OS version targeted by the application");
-		print_usage_line(2, "e.g. -minimum-os-version:12.0.0");
-		print_usage_line(2, "(Only used when target is Darwin)");
+		print_usage_line(2, "Sets the minimum OS version targeted by the application.");
+		print_usage_line(2, "Example: -minimum-os-version:12.0.0");
+		print_usage_line(2, "(Only used when target is Darwin.)");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-extra-linker-flags:<string>");
-		print_usage_line(2, "Adds extra linker specific flags in a string");
+		print_usage_line(2, "Adds extra linker specific flags in a string.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-extra-assembler-flags:<string>");
-		print_usage_line(2, "Adds extra assembler specific flags in a string");
+		print_usage_line(2, "Adds extra assembler specific flags in a string.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-microarch:<string>");
-		print_usage_line(2, "Specifies the specific micro-architecture for the build in a string");
+		print_usage_line(2, "Specifies the specific micro-architecture for the build in a string.");
 		print_usage_line(2, "Examples:");
 		print_usage_line(3, "-microarch:sandybridge");
 		print_usage_line(3, "-microarch:native");
+		print_usage_line(3, "-microarch:? for a list");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-reloc-mode:<string>");
-		print_usage_line(2, "Specifies the reloc mode");
-		print_usage_line(2, "Options:");
-		print_usage_line(3, "default");
-		print_usage_line(3, "static");
-		print_usage_line(3, "pic");
-		print_usage_line(3, "dynamic-no-pic");
+		print_usage_line(2, "Specifies the reloc mode.");
+		print_usage_line(2, "Available options:");
+		print_usage_line(3, "-reloc-mode:default");
+		print_usage_line(3, "-reloc-mode:static");
+		print_usage_line(3, "-reloc-mode:pic");
+		print_usage_line(3, "-reloc-mode:dynamic-no-pic");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-disable-red-zone");
-		print_usage_line(2, "Disable red zone on a supported freestanding target");
+		print_usage_line(2, "Disables red zone on a supported freestanding target.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-dynamic-map-calls");
-		print_usage_line(2, "Use dynamic map calls to minimize code generation at the cost of runtime execution");
+		print_usage_line(2, "Uses dynamic map calls to minimize code generation at the cost of runtime execution.");
 		print_usage_line(0, "");
 	}
 
 	if (check) {
 		print_usage_line(1, "-disallow-do");
-		print_usage_line(2, "Disallows the 'do' keyword in the project");
+		print_usage_line(2, "Disallows the 'do' keyword in the project.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-default-to-nil-allocator");
-		print_usage_line(2, "Sets the default allocator to be the nil_allocator, an allocator which does nothing");
+		print_usage_line(2, "Sets the default allocator to be the nil_allocator, an allocator which does nothing.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-strict-style");
-		print_usage_line(2, "Errs on unneeded tokens, such as unneeded semicolons");
-		print_usage_line(2, "Errs on missing trailing commas followed by a newline");
-		print_usage_line(2, "Errs on deprecated syntax");
+		print_usage_line(2, "Errs on unneeded tokens, such as unneeded semicolons.");
+		print_usage_line(2, "Errs on missing trailing commas followed by a newline.");
+		print_usage_line(2, "Errs on deprecated syntax.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-ignore-warnings");
-		print_usage_line(2, "Ignores warning messages");
+		print_usage_line(2, "Ignores warning messages.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-warnings-as-errors");
-		print_usage_line(2, "Treats warning messages as error messages");
+		print_usage_line(2, "Treats warning messages as error messages.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-terse-errors");
-		print_usage_line(2, "Prints a terse error message without showing the code on that line and the location in that line");
+		print_usage_line(2, "Prints a terse error message without showing the code on that line and the location in that line.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-error-pos-style:<string>");
-		print_usage_line(2, "Options are 'unix', 'odin' and 'default' (odin)");
-		print_usage_line(2, "'odin'    file/path(45:3)");
-		print_usage_line(2, "'unix'    file/path:45:3:");
+		print_usage_line(2, "Available options:");
+		print_usage_line(3, "-error-pos-style:unix      file/path:45:3:");
+		print_usage_line(3, "-error-pos-style:odin      file/path(45:3)");
+		print_usage_line(3, "-error-pos-style:default   (Defaults to 'odin'.)");
 		print_usage_line(0, "");
 
-
 		print_usage_line(1, "-max-error-count:<integer>");
-		print_usage_line(2, "Set the maximum number of errors that can be displayed before the compiler terminates");
-		print_usage_line(2, "Must be an integer >0");
-		print_usage_line(2, "If not set, the default max error count is %d", DEFAULT_MAX_ERROR_COLLECTOR_COUNT);
+		print_usage_line(2, "Sets the maximum number of errors that can be displayed before the compiler terminates.");
+		print_usage_line(2, "Must be an integer >0.");
+		print_usage_line(2, "If not set, the default max error count is %d.", DEFAULT_MAX_ERROR_COLLECTOR_COUNT);
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-foreign-error-procedures");
-		print_usage_line(2, "States that the error procedues used in the runtime are defined in a separate translation unit");
+		print_usage_line(2, "States that the error procedures used in the runtime are defined in a separate translation unit.");
 		print_usage_line(0, "");
 
 	}
 
 	if (run_or_build) {
 		print_usage_line(1, "-sanitize:<string>");
-		print_usage_line(1, "Enables sanitization analysis");
-		print_usage_line(1, "Options are 'address', 'memory', and 'thread'");
-		print_usage_line(1, "NOTE: This flag can be used multiple times");
+		print_usage_line(2, "Enables sanitization analysis.");
+		print_usage_line(2, "Available options:");
+		print_usage_line(3, "-sanitize:address");
+		print_usage_line(3, "-sanitize:memory");
+		print_usage_line(3, "-sanitize:thread");
+		print_usage_line(2, "NOTE: This flag can be used multiple times.");
 		print_usage_line(0, "");
 
 	}
@@ -1955,27 +1962,27 @@ gb_internal void print_show_help(String const arg0, String const &command) {
 		#if defined(GB_SYSTEM_WINDOWS)
 		print_usage_line(1, "-ignore-vs-search");
 		print_usage_line(2, "[Windows only]");
-		print_usage_line(2, "Ignores the Visual Studio search for library paths");
+		print_usage_line(2, "Ignores the Visual Studio search for library paths.");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-resource:<filepath>");
 		print_usage_line(2, "[Windows only]");
-		print_usage_line(2, "Defines the resource file for the executable");
+		print_usage_line(2, "Defines the resource file for the executable.");
 		print_usage_line(2, "Example: -resource:path/to/file.rc");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-pdb-name:<filepath>");
 		print_usage_line(2, "[Windows only]");
-		print_usage_line(2, "Defines the generated PDB name when -debug is enabled");
+		print_usage_line(2, "Defines the generated PDB name when -debug is enabled.");
 		print_usage_line(2, "Example: -pdb-name:different.pdb");
 		print_usage_line(0, "");
 
 		print_usage_line(1, "-subsystem:<option>");
 		print_usage_line(2, "[Windows only]");
-		print_usage_line(2, "Defines the subsystem for the application");
+		print_usage_line(2, "Defines the subsystem for the application.");
 		print_usage_line(2, "Available options:");
-		print_usage_line(3, "console");
-		print_usage_line(3, "windows");
+		print_usage_line(3, "-subsystem:console");
+		print_usage_line(3, "-subsystem:windows");
 		print_usage_line(0, "");
 
 		#endif
@@ -2518,12 +2525,55 @@ int main(int arg_count, char const **arg_ptr) {
 	// 	return 1;
 	// }
 
+	// Check chosen microarchitecture. If not found or ?, print list.
+	bool print_microarch_list = true;
+	if (build_context.microarch.len == 0) {
+		// Autodetect, no need to print list.
+		print_microarch_list = false;
+	} else {
+		String march_list = target_microarch_list[build_context.metrics.arch];
+		String_Iterator it = {march_list, 0};
+		for (;;) {
+			String str = string_split_iterator(&it, ',');
+			if (str == "") break;
+			if (str == build_context.microarch) {
+				// Found matching microarch
+				print_microarch_list = false;
+				break;
+			}
+		}
+	}
+
+	String default_march = get_default_microarchitecture();
+	if (print_microarch_list) {
+		if (build_context.microarch != "?") {
+			gb_printf("Unknown microarchitecture '%.*s'.\n", LIT(build_context.microarch));
+		}
+		gb_printf("Possible -microarch values for target %.*s are:\n", LIT(target_arch_names[build_context.metrics.arch]));
+		gb_printf("\n");
+
+		String march_list  = target_microarch_list[build_context.metrics.arch];
+		String_Iterator it = {march_list, 0};
+
+		for (;;) {
+			String str = string_split_iterator(&it, ',');
+			if (str == "") break;
+			if (str == default_march) {
+				gb_printf("\t%.*s (default)\n", LIT(str));
+			} else {
+				gb_printf("\t%.*s\n", LIT(str));
+			}
+		}
+		return 0;
+	}
+
 	// Set and check build paths...
 	if (!init_build_paths(init_filename)) {
 		return 1;
 	}
 
 	if (build_context.show_debug_messages) {
+		debugf("Selected microarch: %.*s\n", LIT(default_march));
 		for_array(i, build_context.build_paths) {
 			String build_path = path_to_string(heap_allocator(), build_context.build_paths[i]);
 			debugf("build_paths[%ld]: %.*s\n", i, LIT(build_path));
diff --git a/src/string.cpp b/src/string.cpp
index 6eac4f53b..9fb933b1b 100644
--- a/src/string.cpp
+++ b/src/string.cpp
@@ -10,6 +10,10 @@ struct String {
 		return text[i];
 	}
 };
+struct String_Iterator {
+	String const &str;
+	isize  pos;
+};
 // NOTE(bill): used for printf style arguments
 #define LIT(x) ((int)(x).len), (x).text
 #if defined(GB_COMPILER_MSVC) && _MSC_VER < 1700
@@ -201,11 +205,31 @@ gb_internal gb_inline String string_trim_starts_with(String const &s, String con
 }
 
 
+gb_internal String string_split_iterator(String_Iterator *it, const char sep) {
+	isize start = it->pos;
+	isize end   = it->str.len;
+
+	if (start == end) {
+		return str_lit("");
+	}
+
+	isize i = start;
+	for (; i < it->str.len; i++) {
+		if (it->str[i] == sep) {
+			String res = substring(it->str, start, i);
+			it->pos += res.len + 1;
+			return res;
+		}
+	}
+	it->pos = end;
+	return substring(it->str, start, end);
+}
+
 gb_internal gb_inline isize string_extension_position(String const &str) {
 	isize dot_pos = -1;
 	isize i = str.len;
 	while (i --> 0) {
-		if (str[i] == GB_PATH_SEPARATOR)
+		if (str[i] == '\\' || str[i] == '/')
 			break;
 		if (str[i] == '.') {
 			dot_pos = i;
diff --git a/src/threading.cpp b/src/threading.cpp
index 3ddc05b0a..74aa3eb7e 100644
--- a/src/threading.cpp
+++ b/src/threading.cpp
@@ -660,7 +660,7 @@ gb_internal void futex_broadcast(Futex *addr) {
 gb_internal void futex_wait(Futex *addr, Footex val) {
 	for (;;) {
 		int ret = _umtx_op(addr, UMTX_OP_WAIT_UINT, val, 0, NULL);
-		if (ret == 0) {
+		if (ret == -1) {
 			if (errno == ETIMEDOUT || errno == EINTR) {
 				continue;
 			}