A64: Remove two TODOs that will not be done.

- For the LDP/STP TODO: The arguments need dereferencing, so using LDP/STP would
  require some additional checks.
- For the constant pools TODO: Unless we need to, there is no benefit from
  trying to generate constant pools more often, as it would likely only pollute
  the I-cache.

R=ulan@chromium.org

Review URL: https://codereview.chromium.org/151363004

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@19316 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

diff --git a/src/a64/assembler-a64.cc b/src/a64/assembler-a64.cc
index f2cc107..c06fbe9 100644 (file)
--- a/src/a64/assembler-a64.cc
+++ b/src/a64/assembler-a64.cc
@@ -2354,9 +2354,6 @@ void Assembler::BlockConstPoolFor(int instructions) {
 }
 
 
-// TODO(all): We are never trying to emit constant pools after unconditional
-// branches, because we only call it from Assembler::Emit() (or manually).
-// We should try to enable that.
 void Assembler::CheckConstPool(bool force_emit, bool require_jump) {
   // Some short sequence of instruction mustn't be broken up by constant pool
   // emission, such sequences are protected by calls to BlockConstPoolFor and

diff --git a/src/a64/builtins-a64.cc b/src/a64/builtins-a64.cc
index ee6654a..34156b3 100644 (file)
--- a/src/a64/builtins-a64.cc
+++ b/src/a64/builtins-a64.cc
@@ -684,8 +684,6 @@ static void Generate_JSEntryTrampolineHelper(MacroAssembler* masm,
     // Compute the copy end address.
     __ Add(x10, argv, Operand(argc, LSL, kPointerSizeLog2));
 
-    // TODO(all): This can potentially be optimized with ldp/stp to speed up
-    // arguments passing from C++ to JS.
     __ B(&entry);
     __ Bind(&loop);
     __ Ldr(x11, MemOperand(argv, kPointerSize, PostIndex));