[mono] Use SROA-friendly struct type declarations (dotnet#59007)

LLVM's SROA can decompose loads and stores of aggregate type into a
sequence of aggregate-element-typed loads and stores. Before this
change, Mono translated .NET-level value types into LLVM IR-level
structs containing nothing but `i8` elements.

When a value type field has reference type, and a value of this value
type is copied using a `memcpy` intrinsic or an LLVM IR load followed by
a store, LLVM will emit code that loads managed references in multiple
byte-sized fragments before reconstructing the original pointer using a
sequence of ALU ops. This causes sgen to fail to pin the referent.

This change works around this by translating value types to LLVM IR
structs with pointer-sized fields. Packed value types with non-standard
alignment will be translated into LLVM IR structs containing
alignment-sized fields.

Note that this does not completely guarantee that the code we generate
will respect sgen's requirements. No specific guarantees are provided
about the translation of non-atomic LLVM IR loads and stores to machine
code. And we'll need some alternative means (perhaps a special
`gc_copy_unaligned` runtime call or similar) to copy packed or
misaligned value types that contain managed references. For stronger
LLVM IR-level guarantees, we'll want to make use of unordered atomic
loads and stores and unordered atomic memcpy, but that work is out of
scope for this change.

Fixes dotnet#58062, but see the previous paragraph for caveats.

See:
- https://github.com/dotnet/llvm-project/blob/release/11.x/llvm/lib/Transforms/Scalar/SROA.cpp#L3371-L3388
- https://github.com/dotnet/llvm-project/blob/release/11.x/llvm/lib/Transforms/Scalar/SROA.cpp#L3327-L3340

Author: imhameed

src/mono/mono/mini/mini-llvm.c

@@ -530,23 +530,27 @@ ThisType (void)
 	return TARGET_SIZEOF_VOID_P == 8 ? LLVMPointerType (LLVMInt64Type (), 0) : LLVMPointerType (LLVMInt32Type (), 0);
 }
 
+typedef struct {
+	int32_t size;
+	uint32_t align;
+} MonoSizeAlign;
+
 /*
  * get_vtype_size:
  *
  *   Return the size of the LLVM representation of the vtype T.
  */
-static guint32
-get_vtype_size (MonoType *t)
+static MonoSizeAlign
+get_vtype_size_align (MonoType *t)
 {
-	int size;
-
-	size = mono_class_value_size (mono_class_from_mono_type_internal (t), NULL);
+	uint32_t align = 0;
+	int32_t size = mono_class_value_size (mono_class_from_mono_type_internal (t), &align);
 
 	/* LLVMArgAsIArgs depends on this since it stores whole words */
 	while (size < 2 * TARGET_SIZEOF_VOID_P && mono_is_power_of_two (size) == -1)
 		size ++;
-
-	return size;
+	MonoSizeAlign ret = { size, align };
+	return ret;
 }
 
 /*
@@ -681,11 +685,17 @@ create_llvm_type_for_type (MonoLLVMModule *module, MonoClass *klass)
 		for (i = 0; i < size; ++i)
 			eltypes [i] = esize == 4 ? LLVMFloatType () : LLVMDoubleType ();
 	} else {
-		size = get_vtype_size (t);
-
-		eltypes = g_new (LLVMTypeRef, size);
-		for (i = 0; i < size; ++i)
-			eltypes [i] = LLVMInt8Type ();
+		MonoSizeAlign size_align = get_vtype_size_align (t);
+		eltypes = g_new (LLVMTypeRef, size_align.size);
+		size = 0;
+		uint32_t bytes = 0;
+		uint32_t chunk = size_align.align < TARGET_SIZEOF_VOID_P ? size_align.align : TARGET_SIZEOF_VOID_P;
+		for (; chunk > 0; chunk = chunk >> 1) {
+			for (; (bytes + chunk) <= size_align.size; bytes += chunk) {
+				eltypes [size] = LLVMIntType (chunk * 8);
+				++size;
+			}
+		}
 	}
 
 	name = mono_type_full_name (m_class_get_byval_arg (klass));
@@ -2672,11 +2682,11 @@ static void
 emit_vtype_to_args (EmitContext *ctx, LLVMBuilderRef builder, MonoType *t, LLVMValueRef address, LLVMArgInfo *ainfo, LLVMValueRef *args, guint32 *nargs)
 {
 	int pindex = 0;
-	int j, size, nslots;
+	int j, nslots;
 	LLVMTypeRef arg_type;
 
 	t = mini_get_underlying_type (t);
-	size = get_vtype_size (t);
+	int32_t size = get_vtype_size_align (t).size;
 
 	if (MONO_CLASS_IS_SIMD (ctx->cfg, mono_class_from_mono_type_internal (t)))
 		address = LLVMBuildBitCast (ctx->builder, address, LLVMPointerType (LLVMInt8Type (), 0), "");
@@ -7342,7 +7352,7 @@ process_bb (EmitContext *ctx, MonoBasicBlock *bb)
 				} else if (ainfo->storage == LLVMArgVtypeAddr || values [ins->sreg1] == addresses [ins->sreg1]) {
 					/* LLVMArgVtypeByRef/LLVMArgVtypeAddr, have to make a copy */
 					addresses [ins->dreg] = build_alloca (ctx, t);
-					LLVMValueRef v = LLVMBuildLoad (builder, addresses [ins->sreg1], "");
+					LLVMValueRef v = LLVMBuildLoad (builder, addresses [ins->sreg1], "llvm_outarg_vt_copy");
 					LLVMBuildStore (builder, convert (ctx, v, type_to_llvm_type (ctx, t)), addresses [ins->dreg]);
 				} else {
 					addresses [ins->dreg] = addresses [ins->sreg1];

src/tests/issues.targets

@@ -2317,9 +2317,6 @@
         <ExcludeList Include="$(XunitTestBinBase)/JIT/Regression/CLR-x86-JIT/V1.1-M1-Beta1/b143840/b143840/*">
             <Issue>https://github.com/dotnet/runtime/issues/48914</Issue>
         </ExcludeList>
-        <ExcludeList Include = "$(XunitTestBinBase)/JIT/Performance/CodeQuality/Roslyn/CscBench/**">
-            <Issue>https://github.com/dotnet/runtime/issues/58062</Issue>
-        </ExcludeList>
         <ExcludeList Include = "$(XunitTestBinBase)/tracing/eventpipe/eventsourceerror/**">
             <Issue> needs triage </Issue>
         </ExcludeList>