8320379: C2: Sort spilling/unspilling sequence for better ld/st mergi…

…ng into ldp/stp on AArch64

Macro-assembler on aarch64 can merge adjacent loads or stores
into ldp/stp[1]. For example, it can merge:
```
str     w20, [sp, openjdk#16]
str     w10, [sp, openjdk#20]
```
into
```
stp     w20, w10, [sp, openjdk#16]
```

But C2 may generate a sequence like:
```
str     x21, [sp, openjdk#8]
str     w20, [sp, openjdk#16]
str     x19, [sp, openjdk#24] <---
str     w10, [sp, openjdk#20] <--- Before sorting
str     x11, [sp, openjdk#40]
str     w13, [sp, openjdk#48]
str     x16, [sp, openjdk#56]
```
We can't do any merging for non-adjacent loads or stores.

The patch is to sort the spilling or unspilling sequence in
the order of offset during instruction scheduling and bundling
phase. After that, we can get a new sequence:
```
str     x21, [sp, openjdk#8]
str     w20, [sp, openjdk#16]
str     w10, [sp, openjdk#20] <---
str     x19, [sp, openjdk#24] <--- After sorting
str     x11, [sp, openjdk#40]
str     w13, [sp, openjdk#48]
str     x16, [sp, openjdk#56]
```

Then macro-assembler can do ld/st merging:
```
str     x21, [sp, openjdk#8]
stp     w20, w10, [sp, openjdk#16] <--- Merged
str     x19, [sp, openjdk#24]
str     x11, [sp, openjdk#40]
str     w13, [sp, openjdk#48]
str     x16, [sp, openjdk#56]
```

To justify the patch, we run `HelloWorld.java`
```
public class HelloWorld {
    public static void main(String [] args) {
        System.out.println("Hello World!");
    }
}
```
with `java -Xcomp -XX:-TieredCompilation HelloWorld`.

Before the patch, macro-assembler can do ld/st merging for
3688 times. After the patch, the number of ld/st merging
increases to 3871 times, by ~5 %.

Tested tier1~3 on x86 and AArch64.

[1] https://github.com/openjdk/jdk/blob/a95062b39a431b4937ab6e9e73de4d2b8ea1ac49/src/hotspot/cpu/aarch64/macroAssembler_aarch64.cpp#L2079

src/hotspot/share/opto/output.cpp

@@ -169,6 +169,12 @@ class Scheduling {
   // Add a node to the current bundle
   void AddNodeToBundle(Node *n, const Block *bb);
 
+  // Return true only when the stack offset of the first spill node is
+  // greater than the stack offset of the second one. Otherwise, return false.
+  // When compare_two_spill_nodes(first, second) returns true, we think that
+  // "second" should be scheduled before "first" in the final basic block.
+  bool compare_two_spill_nodes(Node* first, Node* second);
+
   // Add a node to the list of available nodes
   void AddNodeToAvailableList(Node *n);
 
@@ -2271,6 +2277,29 @@ Node * Scheduling::ChooseNodeToBundle() {
   return _available[0];
 }
 
+bool Scheduling::compare_two_spill_nodes(Node* first, Node* second) {
+  assert(first->is_MachSpillCopy() && second->is_MachSpillCopy(), "");
+
+  OptoReg::Name first_src_lo = _regalloc->get_reg_first(first->in(1));
+  OptoReg::Name first_dst_lo = _regalloc->get_reg_first(first);
+  OptoReg::Name second_src_lo = _regalloc->get_reg_first(second->in(1));
+  OptoReg::Name second_dst_lo = _regalloc->get_reg_first(second);
+
+  // Comparison between stack -> reg and stack -> reg
+  if (OptoReg::is_stack(first_src_lo) && OptoReg::is_stack(second_src_lo) &&
+      OptoReg::is_reg(first_dst_lo) && OptoReg::is_reg(second_dst_lo)) {
+    return _regalloc->reg2offset(first_src_lo) > _regalloc->reg2offset(second_src_lo);
+  }
+
+  // Comparison between reg -> stack and reg -> stack
+  if (OptoReg::is_stack(first_dst_lo) && OptoReg::is_stack(second_dst_lo) &&
+      OptoReg::is_reg(first_src_lo) && OptoReg::is_reg(second_src_lo)) {
+    return _regalloc->reg2offset(first_dst_lo) > _regalloc->reg2offset(second_dst_lo);
+  }
+
+  return false;
+}
+
 void Scheduling::AddNodeToAvailableList(Node *n) {
   assert( !n->is_Proj(), "projections never directly made available" );
 #ifndef PRODUCT
@@ -2282,11 +2311,20 @@ void Scheduling::AddNodeToAvailableList(Node *n) {
 
   int latency = _current_latency[n->_idx];
 
-  // Insert in latency order (insertion sort)
+  // Insert in latency order (insertion sort). If two MachSpillCopyNodes
+  // for stack spilling or unspilling have the same latency, we sort
+  // them in the order of stack offset. Some backends (aarch64) may also
+  // have more opportunities to do ld/st merging
   uint i;
-  for ( i=0; i < _available.size(); i++ )
-    if (_current_latency[_available[i]->_idx] > latency)
+  for (i = 0; i < _available.size(); i++) {
+    if (_current_latency[_available[i]->_idx] > latency) {
       break;
+    } else if (_current_latency[_available[i]->_idx] == latency &&
+               n->is_MachSpillCopy() && _available[i]->is_MachSpillCopy() &&
+               compare_two_spill_nodes(n, _available[i])) {
+      break;
+    }
+  }
 
   // Special Check for compares following branches
   if( n->is_Mach() && _scheduled.size() > 0 ) {