From e19460cbd7c847bd5452b2c7bf96b38d06b2a182 Mon Sep 17 00:00:00 2001
From: Jeroen van Rijn <Kelimion@users.noreply.github.com>
Date: Fri, 10 Nov 2023 19:37:08 +0100
Subject: Add -microarch:?

---
 src/llvm_backend.cpp | 7 +++++++
 1 file changed, 7 insertions(+)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index 276abc2d4..707ef2969 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -2518,6 +2518,13 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 		}
 	}
 
+	// 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);
 	}
-- 
cgit v1.2.3


From f9039510167c50e0a86e4c25fad5aa08aca24bf0 Mon Sep 17 00:00:00 2001
From: Jeroen van Rijn <Kelimion@users.noreply.github.com>
Date: Fri, 10 Nov 2023 20:14:00 +0100
Subject: Facored out `get_default_microarchitecture`

Moved `generic` -> `x86-64-v2` selection into its own procedure so that `llvm_backend.cpp` and `main.cpp` can share the same logic.
---
 src/build_settings.cpp | 13 ++++++-------
 src/llvm_backend.cpp   | 35 ++++++++++++++++++++---------------
 src/main.cpp           | 15 +++++++++------
 3 files changed, 35 insertions(+), 28 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/build_settings.cpp b/src/build_settings.cpp
index b9b591a7f..94807a852 100644
--- a/src/build_settings.cpp
+++ b/src/build_settings.cpp
@@ -86,17 +86,17 @@ 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 (default),x86-64-v3,x86-64-v4,znver1,znver2,znver3,znver4"),
+	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 (default),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"),
+	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 (default),iwmmxt,krait,kryo,mpcore,mpcorenovfp,native,neoverse-n1,neoverse-n2,neoverse-v1,sc000,sc300,strongarm,strongarm110,strongarm1100,strongarm1110,swift,xscale"),
+	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 (default),kryo,native,neoverse-512tvb,neoverse-e1,neoverse-n1,neoverse-n2,neoverse-v1,neoverse-v2,saphira,thunderx,thunderx2t99,thunderx3t110,thunderxt81,thunderxt83,thunderxt88,tsv110"),
+	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 (default)"),
+	str_lit("generic"),
 	// TargetArch_wasm64p32,
-	str_lit("generic (default)"),
+	str_lit("generic"),
 };
 
 gb_global String target_endian_names[TargetEndian_COUNT] = {
@@ -638,7 +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/llvm_backend.cpp b/src/llvm_backend.cpp
index 707ef2969..effd9d28e 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"
 
+char *get_default_microarchitecture() {
+	char * default_march = "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 = "x86-64";
+			} else {
+				default_march = "x86-64-v2";
+			}
+		}
+	}
+	return default_march;
+}
 
 gb_internal void lb_add_foreign_library_path(lbModule *m, Entity *e) {
 	if (e == nullptr) {
@@ -2491,7 +2510,7 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 	}
 
 	char const *host_cpu_name = LLVMGetHostCPUName();
-	char const *llvm_cpu = "generic";
+	char const *llvm_cpu = get_default_microarchitecture();
 	char const *llvm_features = "";
 	if (build_context.microarch.len != 0) {
 		if (build_context.microarch == "native") {
@@ -2502,20 +2521,6 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 		if (gb_strcmp(llvm_cpu, host_cpu_name) == 0) {
 			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.
diff --git a/src/main.cpp b/src/main.cpp
index 8a917090b..bf91a0889 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -2529,10 +2529,6 @@ int main(int arg_count, char const **arg_ptr) {
 		for (;;) {
 			String str = string_split_iterator(&it, ',');
 			if (str == "") break;
-			// If it's the entry in the list marked (default), we strip off the suffix before the match.
-			if (string_ends_with(str, str_lit(" (default)"))) {
-				str = substring(str, 0, str.len - 10);
-			}
 			if (str == build_context.microarch) {
 				// Found matching microarch
 				print_microarch_list = false;
@@ -2547,12 +2543,19 @@ int main(int arg_count, char const **arg_ptr) {
 		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 march_list  = target_microarch_list[build_context.metrics.arch];
 		String_Iterator it = {march_list, 0};
+
+		String default_march = make_string_c(get_default_microarchitecture());
+
 		for (;;) {
 			String str = string_split_iterator(&it, ',');
 			if (str == "") break;
-			gb_printf("\t%.*s\n", LIT(str));
+			if (str == default_march) {
+				gb_printf("\t%.*s (default)\n", LIT(str));
+			} else {
+				gb_printf("\t%.*s\n", LIT(str));
+			}
 		}
 		return 0;
 	}
-- 
cgit v1.2.3


From c5c46c507370e99a1c3459aaa3c6990ca6f6c549 Mon Sep 17 00:00:00 2001
From: Jeroen van Rijn <Kelimion@users.noreply.github.com>
Date: Sun, 3 Dec 2023 18:08:18 +0100
Subject: Silence writable string warnings when compiling Odin on Linux.

---
 src/llvm_backend.cpp | 18 +++++++++---------
 src/main.cpp         |  4 ++--
 2 files changed, 11 insertions(+), 11 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index effd9d28e..f61c297bd 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -21,8 +21,8 @@
 #include "llvm_backend_stmt.cpp"
 #include "llvm_backend_proc.cpp"
 
-char *get_default_microarchitecture() {
-	char * default_march = "generic";
+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
 		//
@@ -32,9 +32,9 @@ char *get_default_microarchitecture() {
 		// x86-64-v4: AVX512F, AVX512BW, AVX512CD, AVX512DQ, AVX512VL
 		if (ODIN_LLVM_MINIMUM_VERSION_12) {
 			if (build_context.metrics.os == TargetOs_freestanding) {
-				default_march = "x86-64";
+				default_march = str_lit("x86-64");
 			} else {
-				default_march = "x86-64-v2";
+				default_march = str_lit("x86-64-v2");
 			}
 		}
 	}
@@ -2509,16 +2509,16 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 		code_mode = LLVMCodeModelKernel;
 	}
 
-	char const *host_cpu_name = LLVMGetHostCPUName();
-	char const *llvm_cpu = get_default_microarchitecture();
+	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();
 		}
 	}
@@ -2578,7 +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/main.cpp b/src/main.cpp
index e3c44fff4..14f7e84ec 100644
--- a/src/main.cpp
+++ b/src/main.cpp
@@ -2544,6 +2544,7 @@ int main(int arg_count, char const **arg_ptr) {
 		}
 	}
 
+	String default_march = get_default_microarchitecture();
 	if (print_microarch_list) {
 		if (build_context.microarch != "?") {
 			gb_printf("Unknown microarchitecture '%.*s'.\n", LIT(build_context.microarch));
@@ -2554,8 +2555,6 @@ int main(int arg_count, char const **arg_ptr) {
 		String march_list  = target_microarch_list[build_context.metrics.arch];
 		String_Iterator it = {march_list, 0};
 
-		String default_march = make_string_c(get_default_microarchitecture());
-
 		for (;;) {
 			String str = string_split_iterator(&it, ',');
 			if (str == "") break;
@@ -2574,6 +2573,7 @@ int main(int arg_count, char const **arg_ptr) {
 	}
 
 	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));
-- 
cgit v1.2.3


From feba52002e15d3f5b808762bc7cec3f01a173835 Mon Sep 17 00:00:00 2001
From: gingerBill <bill@gingerbill.org>
Date: Wed, 13 Dec 2023 17:46:00 +0000
Subject: Reduce repetition on initializing global type info member arrays

---
 src/llvm_backend.cpp | 62 +++++++++-------------------------------------------
 1 file changed, 10 insertions(+), 52 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index f61c297bd..ca71a0f45 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -2697,64 +2697,22 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 				}
 			}
 
-			{
-				char const *name = LB_TYPE_INFO_TYPES_NAME;
-				Type *t = alloc_type_array(t_type_info_ptr, count);
+			auto const global_type_info_make = [](lbModule *m, char const *name, Type *elem_type, i64 count) -> lbAddr {
+				Type *t = alloc_type_array(elem_type, count);
 				LLVMValueRef g = LLVMAddGlobal(m->mod, lb_type(m, t), name);
 				LLVMSetInitializer(g, LLVMConstNull(lb_type(m, t)));
 				LLVMSetLinkage(g, LLVMInternalLinkage);
 				if (LB_USE_GIANT_PACKED_STRUCT) {
 					lb_make_global_private_const(g);
 				}
-				lb_global_type_info_member_types = lb_addr({g, alloc_type_pointer(t)});
-
-			}
-			{
-				char const *name = LB_TYPE_INFO_NAMES_NAME;
-				Type *t = alloc_type_array(t_string, count);
-				LLVMValueRef g = LLVMAddGlobal(m->mod, lb_type(m, t), name);
-				LLVMSetInitializer(g, LLVMConstNull(lb_type(m, t)));
-				LLVMSetLinkage(g, LLVMInternalLinkage);
-				if (LB_USE_GIANT_PACKED_STRUCT) {
-					lb_make_global_private_const(g);
-				}
-				lb_global_type_info_member_names = lb_addr({g, alloc_type_pointer(t)});
-			}
-			{
-				char const *name = LB_TYPE_INFO_OFFSETS_NAME;
-				Type *t = alloc_type_array(t_uintptr, count);
-				LLVMValueRef g = LLVMAddGlobal(m->mod, lb_type(m, t), name);
-				LLVMSetInitializer(g, LLVMConstNull(lb_type(m, t)));
-				LLVMSetLinkage(g, LLVMInternalLinkage);
-				if (LB_USE_GIANT_PACKED_STRUCT) {
-					lb_make_global_private_const(g);
-				}
-				lb_global_type_info_member_offsets = lb_addr({g, alloc_type_pointer(t)});
-			}
-
-			{
-				char const *name = LB_TYPE_INFO_USINGS_NAME;
-				Type *t = alloc_type_array(t_bool, count);
-				LLVMValueRef g = LLVMAddGlobal(m->mod, lb_type(m, t), name);
-				LLVMSetInitializer(g, LLVMConstNull(lb_type(m, t)));
-				LLVMSetLinkage(g, LLVMInternalLinkage);
-				if (LB_USE_GIANT_PACKED_STRUCT) {
-					lb_make_global_private_const(g);
-				}
-				lb_global_type_info_member_usings = lb_addr({g, alloc_type_pointer(t)});
-			}
-
-			{
-				char const *name = LB_TYPE_INFO_TAGS_NAME;
-				Type *t = alloc_type_array(t_string, count);
-				LLVMValueRef g = LLVMAddGlobal(m->mod, lb_type(m, t), name);
-				LLVMSetInitializer(g, LLVMConstNull(lb_type(m, t)));
-				LLVMSetLinkage(g, LLVMInternalLinkage);
-				if (LB_USE_GIANT_PACKED_STRUCT) {
-					lb_make_global_private_const(g);
-				}
-				lb_global_type_info_member_tags = lb_addr({g, alloc_type_pointer(t)});
-			}
+				return lb_addr({g, alloc_type_pointer(t)});
+			};
+
+			lb_global_type_info_member_types   = global_type_info_make(m, LB_TYPE_INFO_TYPES_NAME,   t_type_info_ptr, count);
+			lb_global_type_info_member_names   = global_type_info_make(m, LB_TYPE_INFO_NAMES_NAME,   t_string,        count);
+			lb_global_type_info_member_offsets = global_type_info_make(m, LB_TYPE_INFO_OFFSETS_NAME, t_uintptr,       count);
+			lb_global_type_info_member_usings  = global_type_info_make(m, LB_TYPE_INFO_USINGS_NAME,  t_bool,          count);
+			lb_global_type_info_member_tags    = global_type_info_make(m, LB_TYPE_INFO_TAGS_NAME,    t_string,        count);
 		}
 	}
 
-- 
cgit v1.2.3


From cd65a15d81b32636c5097200446cc6d6afc7199b Mon Sep 17 00:00:00 2001
From: Yawning Angel <yawning@schwanenlied.me>
Date: Fri, 15 Dec 2023 17:29:59 +0900
Subject: src: `enable_target_feature` should add features, not overwrite

`llvm_features` being empty is the default state, and implies the
presence of certain features.

Previously if any target features were explicitly enabled by the
`enable_target_feature` attribute, they were added comma separated
to `llvm_features`.

For example: `lzcnt,popcnt,...,sse4.2,sse`

This was causing LLVM to try to target a CPU that *ONLY* has the
explicitly enabled features.  This now will prefix explicitly enabled
features with a `+`, and preserve the existing `llvm_features` string
by appending to it if it is set.
---
 src/build_settings.cpp |  4 +++-
 src/llvm_backend.cpp   | 41 ++++++++++++++++++++++++++++++++++++++++-
 2 files changed, 43 insertions(+), 2 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/build_settings.cpp b/src/build_settings.cpp
index 18ad8ac0d..9d909fcae 100644
--- a/src/build_settings.cpp
+++ b/src/build_settings.cpp
@@ -1493,7 +1493,7 @@ gb_internal void enable_target_feature(TokenPos pos, String const &target_featur
 }
 
 
-gb_internal char const *target_features_set_to_cstring(gbAllocator allocator, bool with_quotes) {
+gb_internal char const *target_features_set_to_cstring(gbAllocator allocator, bool with_quotes, bool with_plus) {
 	isize len = 0;
 	isize i = 0;
 	for (String const &feature : build_context.target_features_set) {
@@ -1502,6 +1502,7 @@ gb_internal char const *target_features_set_to_cstring(gbAllocator allocator, bo
 		}
 		len += feature.len;
 		if (with_quotes) len += 2;
+		if (with_plus) len += 1;
 		i += 1;
 	}
 	char *features = gb_alloc_array(allocator, char, len+1);
@@ -1513,6 +1514,7 @@ gb_internal char const *target_features_set_to_cstring(gbAllocator allocator, bo
 		}
 
 		if (with_quotes) features[len++] = '"';
+		if (with_plus) features[len++] = '+';
 		gb_memmove(features + len, feature.text, feature.len);
 		len += feature.len;
 		if (with_quotes) features[len++] = '"';
diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index ca71a0f45..b90fd8495 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -2531,7 +2531,46 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 	*/
 
 	if (build_context.target_features_set.entries.count != 0) {
-		llvm_features = target_features_set_to_cstring(permanent_allocator(), false);
+		// Prefix all of the features with a `+`, because we are
+		// enabling additional features.
+		char const *additional_features = target_features_set_to_cstring(permanent_allocator(), false, true);
+
+		String f_string = make_string_c(llvm_features);
+		String a_string = make_string_c(additional_features);
+		isize f_len = f_string.len;
+
+		if (f_len == 0) {
+			// The common case is that llvm_features is empty, so
+			// the target_features_set additions can be used as is.
+			llvm_features = additional_features;
+		} else {
+			// The user probably specified `-microarch:native`, so
+			// llvm_features is populated by LLVM's idea of what
+			// the host CPU supports.
+			//
+			// As far as I can tell, (which is barely better than
+			// wild guessing), a bitset is formed by parsing the
+			// string left to right.
+			//
+			// So, llvm_features + ',' + additonal_features, will
+			// makes the target_features_set override llvm_features.
+
+			char *tmp = gb_alloc_array(permanent_allocator(), char, f_len + 1 + a_string.len + 1);
+			isize len = 0;
+
+			// tmp = f_string
+			gb_memmove(tmp, f_string.text, f_string.len);
+			len += f_string.len;
+			// tmp += ','
+			tmp[len++] = ',';
+			// tmp += a_string
+			gb_memmove(tmp + len, a_string.text, a_string.len);
+			len += a_string.len;
+			// tmp += NUL
+			tmp[len++] = 0;
+
+			llvm_features = tmp;
+		}
 	}
 
 	// GB_ASSERT_MSG(LLVMTargetHasAsmBackend(target));
-- 
cgit v1.2.3


From f4782157d3b586a88983d3b770c2c0eeb17946d1 Mon Sep 17 00:00:00 2001
From: gingerBill <bill@gingerbill.org>
Date: Sun, 7 Jan 2024 21:34:44 +0000
Subject: Implement instrumentation pass

---
 core/runtime/core.odin                |  1 +
 core/runtime/entry_unix.odin          |  1 +
 core/runtime/entry_wasm.odin          |  1 +
 core/runtime/entry_windows.odin       |  1 +
 core/runtime/error_checks.odin        | 24 ++++++-----
 core/runtime/procs_windows_amd64.odin |  1 +
 core/runtime/procs_windows_i386.odin  |  1 +
 src/check_decl.cpp                    | 26 ++++++++----
 src/entity.cpp                        |  2 +-
 src/llvm_backend.cpp                  |  2 -
 src/llvm_backend.hpp                  |  2 +-
 src/llvm_backend_general.cpp          | 13 ++++++
 src/llvm_backend_opt.cpp              | 77 +++++++++++++++++++++++++++++++++++
 src/llvm_backend_proc.cpp             | 12 ++++++
 14 files changed, 143 insertions(+), 21 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/core/runtime/core.odin b/core/runtime/core.odin
index 2d176f909..e12d9a43d 100644
--- a/core/runtime/core.odin
+++ b/core/runtime/core.odin
@@ -18,6 +18,7 @@
 // This could change at a later date if the all these data structures are
 // implemented within the compiler rather than in this "preload" file
 //
+//+no-instrumentation
 package runtime
 
 import "core:intrinsics"
diff --git a/core/runtime/entry_unix.odin b/core/runtime/entry_unix.odin
index 78e545c22..f494a509e 100644
--- a/core/runtime/entry_unix.odin
+++ b/core/runtime/entry_unix.odin
@@ -1,5 +1,6 @@
 //+private
 //+build linux, darwin, freebsd, openbsd
+//+no-instrumentation
 package runtime
 
 import "core:intrinsics"
diff --git a/core/runtime/entry_wasm.odin b/core/runtime/entry_wasm.odin
index 235d5611b..e7f3f156f 100644
--- a/core/runtime/entry_wasm.odin
+++ b/core/runtime/entry_wasm.odin
@@ -1,5 +1,6 @@
 //+private
 //+build wasm32, wasm64p32
+//+no-instrumentation
 package runtime
 
 import "core:intrinsics"
diff --git a/core/runtime/entry_windows.odin b/core/runtime/entry_windows.odin
index a315c1209..277daecca 100644
--- a/core/runtime/entry_windows.odin
+++ b/core/runtime/entry_windows.odin
@@ -1,5 +1,6 @@
 //+private
 //+build windows
+//+no-instrumentation
 package runtime
 
 import "core:intrinsics"
diff --git a/core/runtime/error_checks.odin b/core/runtime/error_checks.odin
index 9d484979a..ea6333c29 100644
--- a/core/runtime/error_checks.odin
+++ b/core/runtime/error_checks.odin
@@ -1,5 +1,6 @@
 package runtime
 
+@(no_instrumentation)
 bounds_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_array_bounds()
@@ -8,6 +9,7 @@ bounds_trap :: proc "contextless" () -> ! {
 	}
 }
 
+@(no_instrumentation)
 type_assertion_trap :: proc "contextless" () -> ! {
 	when ODIN_OS == .Windows {
 		windows_trap_type_assertion()
@@ -21,7 +23,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	if uint(index) < uint(count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, index, count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Index ")
@@ -34,6 +36,7 @@ bounds_check_error :: proc "contextless" (file: string, line, column: i32, index
 	handle_error(file, line, column, index, count)
 }
 
+@(no_instrumentation)
 slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int, len: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -46,6 +49,7 @@ slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, h
 	bounds_trap()
 }
 
+@(no_instrumentation)
 multi_pointer_slice_handle_error :: proc "contextless" (file: string, line, column: i32, lo, hi: int) -> ! {
 	print_caller_location(Source_Code_Location{file, line, column, ""})
 	print_string(" Invalid slice indices ")
@@ -82,7 +86,7 @@ dynamic_array_expr_error :: proc "contextless" (file: string, line, column: i32,
 	if 0 <= low && low <= high && high <= max {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, low, high, max: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Invalid dynamic array indices ")
@@ -103,7 +107,7 @@ matrix_bounds_check_error :: proc "contextless" (file: string, line, column: i32
 	   uint(column_index) < uint(column_count) {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (file: string, line, column: i32, row_index, column_index, row_count, column_count: int) -> ! {
 		print_caller_location(Source_Code_Location{file, line, column, ""})
 		print_string(" Matrix indices [")
@@ -127,7 +131,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -140,7 +144,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion\n")
@@ -153,7 +157,7 @@ when ODIN_NO_RTTI {
 		if ok {
 			return
 		}
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid) -> ! {
 			print_caller_location(Source_Code_Location{file, line, column, ""})
 			print_string(" Invalid type assertion from ")
@@ -198,7 +202,7 @@ when ODIN_NO_RTTI {
 			return id
 		}
 
-		@(cold)
+		@(cold, no_instrumentation)
 		handle_error :: proc "contextless" (file: string, line, column: i32, from, to: typeid, from_data: rawptr) -> ! {
 
 			actual := variant_type(from, from_data)
@@ -224,7 +228,7 @@ make_slice_error_loc :: #force_inline proc "contextless" (loc := #caller_locatio
 	if 0 <= len {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len: int) -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid slice length for make: ")
@@ -239,7 +243,7 @@ make_dynamic_array_error_loc :: #force_inline proc "contextless" (loc := #caller
 	if 0 <= len && len <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, len, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid dynamic array parameters for make: ")
@@ -256,7 +260,7 @@ make_map_expr_error_loc :: #force_inline proc "contextless" (loc := #caller_loca
 	if 0 <= cap {
 		return
 	}
-	@(cold)
+	@(cold, no_instrumentation)
 	handle_error :: proc "contextless" (loc: Source_Code_Location, cap: int)  -> ! {
 		print_caller_location(loc)
 		print_string(" Invalid map capacity for make: ")
diff --git a/core/runtime/procs_windows_amd64.odin b/core/runtime/procs_windows_amd64.odin
index e430357be..a30985d5c 100644
--- a/core/runtime/procs_windows_amd64.odin
+++ b/core/runtime/procs_windows_amd64.odin
@@ -1,4 +1,5 @@
 //+private
+//+no-instrumentation
 package runtime
 
 foreign import kernel32 "system:Kernel32.lib"
diff --git a/core/runtime/procs_windows_i386.odin b/core/runtime/procs_windows_i386.odin
index f810197f1..4f606da8f 100644
--- a/core/runtime/procs_windows_i386.odin
+++ b/core/runtime/procs_windows_i386.odin
@@ -1,4 +1,5 @@
 //+private
+//+no-instrumentation
 package runtime
 
 @require foreign import "system:int64.lib"
diff --git a/src/check_decl.cpp b/src/check_decl.cpp
index c69d8185e..a530945f9 100644
--- a/src/check_decl.cpp
+++ b/src/check_decl.cpp
@@ -910,24 +910,24 @@ gb_internal void check_proc_decl(CheckerContext *ctx, Entity *e, DeclInfo *d) {
 	e->Procedure.entry_point_only = ac.entry_point_only;
 	e->Procedure.is_export = ac.is_export;
 
-	bool no_instrumentation = false;
+	bool has_instrumentation = false;
 	if (pl->body == nullptr) {
-		no_instrumentation = true;
+		has_instrumentation = false;
 		if (ac.no_instrumentation != Instrumentation_Default) {
 			error(e->token, "@(no_instrumentation) is not allowed on foreign procedures");
 		}
 	} else {
-		if (e->file) {
-			no_instrumentation = (e->file->flags & AstFile_NoInstrumentation) != 0;
+		AstFile *file = e->token.pos.file_id ? global_files[e->token.pos.file_id] : nullptr;
+		if (file) {
+			has_instrumentation = (file->flags & AstFile_NoInstrumentation) == 0;
 		}
 
 		switch (ac.no_instrumentation) {
-		case Instrumentation_Enabled:  no_instrumentation = false; break;
+		case Instrumentation_Enabled:  has_instrumentation = true; break;
 		case Instrumentation_Default:  break;
-		case Instrumentation_Disabled: no_instrumentation = true;  break;
+		case Instrumentation_Disabled: has_instrumentation = false;  break;
 		}
 	}
-	e->Procedure.no_instrumentation = no_instrumentation;
 
 	auto const is_valid_instrumentation_call = [](Type *type) -> bool {
 		if (type == nullptr || type->kind != Type_Proc) {
@@ -949,6 +949,9 @@ gb_internal void check_proc_decl(CheckerContext *ctx, Entity *e, DeclInfo *d) {
 
 	if (ac.instrumentation_enter && ac.instrumentation_exit) {
 		error(e->token, "A procedure cannot be marked with both @(instrumentation_enter) and @(instrumentation_exit)");
+
+		has_instrumentation = false;
+		e->flags |= EntityFlag_Require;
 	} else if (ac.instrumentation_enter) {
 		if (!is_valid_instrumentation_call(e->type)) {
 			gbString s = type_to_string(e->type);
@@ -961,6 +964,9 @@ gb_internal void check_proc_decl(CheckerContext *ctx, Entity *e, DeclInfo *d) {
 		} else {
 			ctx->info->instrumentation_enter_entity = e;
 		}
+
+		has_instrumentation = false;
+		e->flags |= EntityFlag_Require;
 	} else if (ac.instrumentation_exit) {
 		if (!is_valid_instrumentation_call(e->type)) {
 			gbString s = type_to_string(e->type);
@@ -973,8 +979,14 @@ gb_internal void check_proc_decl(CheckerContext *ctx, Entity *e, DeclInfo *d) {
 		} else {
 			ctx->info->instrumentation_exit_entity = e;
 		}
+
+		has_instrumentation = false;
+		e->flags |= EntityFlag_Require;
 	}
 
+	e->Procedure.has_instrumentation = has_instrumentation;
+
+
 	e->deprecated_message = ac.deprecated_message;
 	e->warning_message = ac.warning_message;
 	ac.link_name = handle_link_name(ctx, e->token, ac.link_name, ac.link_prefix);
diff --git a/src/entity.cpp b/src/entity.cpp
index 0539386d0..e6c46d37e 100644
--- a/src/entity.cpp
+++ b/src/entity.cpp
@@ -251,7 +251,7 @@ struct Entity {
 			bool    generated_from_polymorphic : 1;
 			bool    target_feature_disabled    : 1;
 			bool    entry_point_only           : 1;
-			bool    no_instrumentation         : 1;
+			bool    has_instrumentation        : 1;
 			String  target_feature;
 		} Procedure;
 		struct {
diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index b90fd8495..0175d039e 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -1497,8 +1497,6 @@ gb_internal WORKER_TASK_PROC(lb_llvm_module_pass_worker_proc) {
 	auto passes = array_make<char const *>(heap_allocator(), 0, 64);
 	defer (array_free(&passes));
 
-
-
 	LLVMPassBuilderOptionsRef pb_options = LLVMCreatePassBuilderOptions();
 	defer (LLVMDisposePassBuilderOptions(pb_options));
 
diff --git a/src/llvm_backend.hpp b/src/llvm_backend.hpp
index 4e193bcea..b645d66d6 100644
--- a/src/llvm_backend.hpp
+++ b/src/llvm_backend.hpp
@@ -563,7 +563,7 @@ gb_internal LLVMTypeRef OdinLLVMGetVectorElementType(LLVMTypeRef type);
 
 gb_internal String lb_filepath_ll_for_module(lbModule *m);
 
-
+gb_internal LLVMTypeRef lb_type_internal_for_procedures_raw(lbModule *m, Type *type);
 
 gb_internal LLVMTypeRef llvm_array_type(LLVMTypeRef ElementType, uint64_t ElementCount) {
 #if LB_USE_NEW_PASS_SYSTEM
diff --git a/src/llvm_backend_general.cpp b/src/llvm_backend_general.cpp
index 54327cc54..f0f5327c6 100644
--- a/src/llvm_backend_general.cpp
+++ b/src/llvm_backend_general.cpp
@@ -2348,6 +2348,15 @@ gb_internal LLVMAttributeRef lb_create_enum_attribute(LLVMContextRef ctx, char c
 	return LLVMCreateEnumAttribute(ctx, kind, value);
 }
 
+gb_internal LLVMAttributeRef lb_create_string_attribute(LLVMContextRef ctx, String const &key, String const &value) {
+	LLVMAttributeRef attr = LLVMCreateStringAttribute(
+		ctx,
+		cast(char const *)key.text,   cast(unsigned)key.len,
+		cast(char const *)value.text, cast(unsigned)value.len);
+	return attr;
+}
+
+
 gb_internal void lb_add_proc_attribute_at_index(lbProcedure *p, isize index, char const *name, u64 value) {
 	LLVMAttributeRef attr = lb_create_enum_attribute(p->module->ctx, name, value);
 	GB_ASSERT(attr != nullptr);
@@ -2361,6 +2370,10 @@ gb_internal void lb_add_proc_attribute_at_index(lbProcedure *p, isize index, cha
 gb_internal void lb_add_attribute_to_proc(lbModule *m, LLVMValueRef proc_value, char const *name, u64 value=0) {
 	LLVMAddAttributeAtIndex(proc_value, LLVMAttributeIndex_FunctionIndex, lb_create_enum_attribute(m->ctx, name, value));
 }
+gb_internal void lb_add_attribute_to_proc_with_string(lbModule *m, LLVMValueRef proc_value, String const &name, String const &value) {
+	LLVMAttributeRef attr = lb_create_string_attribute(m->ctx, name, value);
+	LLVMAddAttributeAtIndex(proc_value, LLVMAttributeIndex_FunctionIndex, attr);
+}
 
 
 
diff --git a/src/llvm_backend_opt.cpp b/src/llvm_backend_opt.cpp
index 2e03b7974..d5487d259 100644
--- a/src/llvm_backend_opt.cpp
+++ b/src/llvm_backend_opt.cpp
@@ -380,6 +380,80 @@ gb_internal void lb_run_remove_dead_instruction_pass(lbProcedure *p) {
 	}
 }
 
+gb_internal LLVMValueRef lb_run_instrumentation_pass_insert_call(lbProcedure *p, Entity *entity, LLVMBuilderRef dummy_builder) {
+	lbModule *m = p->module;
+
+	lbValue cc = lb_find_procedure_value_from_entity(m, entity);
+
+	LLVMValueRef args[2] = {};
+	args[0] = p->value;
+
+	LLVMValueRef returnaddress_args[1] = {};
+
+	returnaddress_args[0] = LLVMConstInt(LLVMInt32TypeInContext(m->ctx), 0, false);
+
+	char const *instrinsic_name = "llvm.returnaddress";
+	unsigned id = LLVMLookupIntrinsicID(instrinsic_name, gb_strlen(instrinsic_name));
+	GB_ASSERT_MSG(id != 0, "Unable to find %s", instrinsic_name);
+	LLVMValueRef ip = LLVMGetIntrinsicDeclaration(m->mod, id, nullptr, 0);
+	LLVMTypeRef call_type = LLVMIntrinsicGetType(m->ctx, id, nullptr, 0);
+	args[1] = LLVMBuildCall2(dummy_builder, call_type, ip, returnaddress_args, gb_count_of(returnaddress_args), "");
+
+	LLVMTypeRef fnp = lb_type_internal_for_procedures_raw(p->module, entity->type);
+	return LLVMBuildCall2(dummy_builder, fnp, cc.value, args, 2, "");
+}
+
+
+gb_internal void lb_run_instrumentation_pass(lbProcedure *p) {
+	lbModule *m = p->module;
+	Entity *enter = m->info->instrumentation_enter_entity;
+	Entity *exit  = m->info->instrumentation_exit_entity;
+	if (enter == nullptr || exit == nullptr) {
+		return;
+	}
+	if (!(p->entity &&
+	      p->entity->kind == Entity_Procedure &&
+	      p->entity->Procedure.has_instrumentation)) {
+		return;
+	}
+
+#define LLVM_V_NAME(x) x, cast(unsigned)(gb_count_of(x)-1)
+
+	LLVMBuilderRef dummy_builder = LLVMCreateBuilderInContext(m->ctx);
+	defer (LLVMDisposeBuilder(dummy_builder));
+
+	LLVMBasicBlockRef entry_bb = p->entry_block->block;
+	LLVMPositionBuilder(dummy_builder, entry_bb, LLVMGetFirstInstruction(entry_bb));
+	lb_run_instrumentation_pass_insert_call(p, enter, dummy_builder);
+	LLVMRemoveStringAttributeAtIndex(p->value, LLVMAttributeIndex_FunctionIndex, LLVM_V_NAME("instrument-function-entry"));
+
+	unsigned bb_count = LLVMCountBasicBlocks(p->value);
+	LLVMBasicBlockRef *bbs = gb_alloc_array(temporary_allocator(), LLVMBasicBlockRef, bb_count);
+	LLVMGetBasicBlocks(p->value, bbs);
+	for (unsigned i = 0; i < bb_count; i++) {
+		LLVMBasicBlockRef bb = bbs[i];
+		LLVMValueRef terminator = LLVMGetBasicBlockTerminator(bb);
+		if (terminator == nullptr ||
+		    !LLVMIsAReturnInst(terminator)) {
+			continue;
+		}
+
+		// TODO(bill): getTerminatingMustTailCall()
+		// If T is preceded by a musttail call, that's the real terminator.
+		// if (CallInst *CI = BB.getTerminatingMustTailCall())
+		// 	T = CI;
+
+
+		LLVMPositionBuilderBefore(dummy_builder, terminator);
+		lb_run_instrumentation_pass_insert_call(p, exit, dummy_builder);
+	}
+
+	LLVMRemoveStringAttributeAtIndex(p->value, LLVMAttributeIndex_FunctionIndex, LLVM_V_NAME("instrument-function-exit"));
+
+#undef LLVM_V_NAME
+}
+
+
 
 gb_internal void lb_run_function_pass_manager(LLVMPassManagerRef fpm, lbProcedure *p, lbFunctionPassManagerKind pass_manager_kind) {
 	if (p == nullptr) {
@@ -401,6 +475,7 @@ gb_internal void lb_run_function_pass_manager(LLVMPassManagerRef fpm, lbProcedur
 	    }
 	    break;
 	}
+	lb_run_instrumentation_pass(p);
 
 	LLVMRunFunctionPassManager(fpm, p->value);
 }
@@ -552,3 +627,5 @@ gb_internal void lb_run_remove_unused_globals_pass(lbModule *m) {
 		}
 	}
 }
+
+
diff --git a/src/llvm_backend_proc.cpp b/src/llvm_backend_proc.cpp
index d4eae84bc..09bebd0cf 100644
--- a/src/llvm_backend_proc.cpp
+++ b/src/llvm_backend_proc.cpp
@@ -329,6 +329,18 @@ gb_internal lbProcedure *lb_create_procedure(lbModule *m, Entity *entity, bool i
 		}
 	}
 
+	if (p->body && entity->Procedure.has_instrumentation) {
+		Entity *instrumentation_enter = m->info->instrumentation_enter_entity;
+		Entity *instrumentation_exit  = m->info->instrumentation_exit_entity;
+		if (instrumentation_enter && instrumentation_exit) {
+			String enter = lb_get_entity_name(m, instrumentation_enter);
+			String exit  = lb_get_entity_name(m, instrumentation_exit);
+
+			lb_add_attribute_to_proc_with_string(m, p->value, make_string_c("instrument-function-entry"), enter);
+			lb_add_attribute_to_proc_with_string(m, p->value, make_string_c("instrument-function-exit"),  exit);
+		}
+	}
+
 	lbValue proc_value = {p->value, p->type};
 	lb_add_entity(m, entity,  proc_value);
 	lb_add_member(m, p->name, proc_value);
-- 
cgit v1.2.3


From 76f52dd6c9d37683e8e1ef85755b77e7e9b71c7b Mon Sep 17 00:00:00 2001
From: codename-irvin <brian.budge@protonmail.com>
Date: Mon, 15 Jan 2024 19:49:34 -0500
Subject: Add freestanding aarch64 target

---
 src/build_settings.cpp | 10 +++++++++-
 src/llvm_backend.cpp   |  2 +-
 2 files changed, 10 insertions(+), 2 deletions(-)

(limited to 'src/llvm_backend.cpp')

diff --git a/src/build_settings.cpp b/src/build_settings.cpp
index 9d909fcae..db09eabcf 100644
--- a/src/build_settings.cpp
+++ b/src/build_settings.cpp
@@ -582,7 +582,14 @@ gb_global TargetMetrics target_freestanding_amd64_sysv = {
 	TargetABI_SysV,
 };
 
-
+gb_global TargetMetrics target_freestanding_arm64_sysv = {
+	TargetOs_freestanding,
+	TargetArch_arm64,
+	8, 8, 8, 16,
+	str_lit("aarch64-none-elf"),
+	str_lit("e-m:o-p:32:32-Fi8-i64:64-v128:64:128-a:0:32-n32-S64"),
+	TargetABI_SysV,
+};
 
 struct NamedTargetMetrics {
 	String name;
@@ -617,6 +624,7 @@ gb_global NamedTargetMetrics named_targets[] = {
 	{ str_lit("wasi_wasm64p32"),         &target_wasi_wasm64p32 },
 
 	{ str_lit("freestanding_amd64_sysv"), &target_freestanding_amd64_sysv },
+	{ str_lit("freestanding_arm64_sysv"), &target_freestanding_arm64_sysv },
 };
 
 gb_global NamedTargetMetrics *selected_target_metrics;
diff --git a/src/llvm_backend.cpp b/src/llvm_backend.cpp
index 0175d039e..003424e0a 100644
--- a/src/llvm_backend.cpp
+++ b/src/llvm_backend.cpp
@@ -2503,7 +2503,7 @@ gb_internal bool lb_generate_code(lbGenerator *gen) {
 	LLVMCodeModel code_mode = LLVMCodeModelDefault;
 	if (is_arch_wasm()) {
 		code_mode = LLVMCodeModelJITDefault;
-	} else if (build_context.metrics.os == TargetOs_freestanding) {
+	} else if (is_arch_x86() && build_context.metrics.os == TargetOs_freestanding) {
 		code_mode = LLVMCodeModelKernel;
 	}
 
-- 
cgit v1.2.3