Reland "Calculate the number of out vregs."

This reverts commit 434a327234f74eed3ef4072314d2e2bdb73e4dda.

Reason for revert: Relanding with no change. The regressions
that were the reason for the revert may reappear. However,
these regressions are probably caused by subtle effects that
are not directly related to this change. For example, a code
size improvement can regress performance simply by moving
the start of a loop from an aligned address to an unaligned
address, or by splitting a loop across two cache lines.

Bug: 358519867
Bug: 359722268
Change-Id: I997b8a4219418f79b3a5fc4e7e50817911f0a737

1 files changed, 17 insertions, 3 deletions

diff --git a/compiler/optimizing/instruction_simplifier.cc b/compiler/optimizing/instruction_simplifier.cc
index a09a9f0ead..cd2371d90c 100644
--- a/compiler/optimizing/instruction_simplifier.cc
+++ b/compiler/optimizing/instruction_simplifier.cc
@@ -2996,7 +2996,7 @@ static bool NoEscapeForStringBufferReference(HInstruction* reference, HInstructi
   return false;
 }
 
-static bool TryReplaceStringBuilderAppend(HInvoke* invoke) {
+static bool TryReplaceStringBuilderAppend(CodeGenerator* codegen, HInvoke* invoke) {
   DCHECK_EQ(invoke->GetIntrinsic(), Intrinsics::kStringBuilderToString);
   if (invoke->CanThrowIntoCatchBlock()) {
     return false;
@@ -3153,11 +3153,25 @@ static bool TryReplaceStringBuilderAppend(HInvoke* invoke) {
     }
   }
 
+  // Calculate outgoing vregs, including padding for 64-bit arg alignment.
+  const PointerSize pointer_size = InstructionSetPointerSize(codegen->GetInstructionSet());
+  const size_t method_vregs = static_cast<size_t>(pointer_size) / kVRegSize;
+  uint32_t number_of_out_vregs = method_vregs;  // For correct alignment padding; subtracted below.
+  for (uint32_t f = format; f != 0u; f >>= StringBuilderAppend::kBitsPerArg) {
+    auto a = enum_cast<StringBuilderAppend::Argument>(f & StringBuilderAppend::kArgMask);
+    if (a == StringBuilderAppend::Argument::kLong || a == StringBuilderAppend::Argument::kDouble) {
+      number_of_out_vregs += /* alignment */ ((number_of_out_vregs) & 1u) + /* vregs */ 2u;
+    } else {
+      number_of_out_vregs += /* vregs */ 1u;
+    }
+  }
+  number_of_out_vregs -= method_vregs;
+
   // Create replacement instruction.
   HIntConstant* fmt = block->GetGraph()->GetIntConstant(static_cast<int32_t>(format));
   ArenaAllocator* allocator = block->GetGraph()->GetAllocator();
   HStringBuilderAppend* append = new (allocator) HStringBuilderAppend(
-      fmt, num_args, has_fp_args, allocator, invoke->GetDexPc());
+      fmt, num_args, number_of_out_vregs, has_fp_args, allocator, invoke->GetDexPc());
   append->SetReferenceTypeInfoIfValid(invoke->GetReferenceTypeInfo());
   for (size_t i = 0; i != num_args; ++i) {
     append->SetArgumentAt(i, args[num_args - 1u - i]);
@@ -3202,7 +3216,7 @@ void InstructionSimplifierVisitor::SimplifyAllocationIntrinsic(HInvoke* invoke)
       RecordSimplification();
     }
   } else if (invoke->GetIntrinsic() == Intrinsics::kStringBuilderToString &&
-             TryReplaceStringBuilderAppend(invoke)) {
+             TryReplaceStringBuilderAppend(codegen_, invoke)) {
     RecordSimplification();
   }
 }