[turbofan] Add CalleeSavedFPRegisters to CallDescriptor.

src/compiler/arm64/code-generator-arm64.cc

 // Copyright 2014 the V8 project authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "src/compiler/code-generator.h"
 
 #include "src/arm64/macro-assembler-arm64.h"
 #include "src/compiler/code-generator-impl.h"
 #include "src/compiler/gap-resolver.h"
 #include "src/compiler/node-matchers.h"
@@ skipping 1059 matching lines @@
 }
 
 
 void CodeGenerator::AssemblePrologue() {
   CallDescriptor* descriptor = linkage()->GetIncomingDescriptor();
   int stack_slots = frame()->GetSpillSlotCount();
   if (descriptor->kind() == CallDescriptor::kCallAddress) {
     __ SetStackPointer(csp);
     __ Push(lr, fp);
     __ Mov(fp, csp);
-    // TODO(dcarney): correct callee saved registers.
-    __ PushCalleeSavedRegisters();
-    frame()->SetRegisterSaveAreaSize(20 * kPointerSize);
+
+    // Save registers.

[paul.l..., 2015/06/23 02:15:02]

This looks a bit messy to me, but luckily there are arm64 macro-asm function to
do most of this, similar to the way we do on mips with MultiPush/Pop. I'll send
a complete diff separately, but the basic save code can look like this:

    CPURegList callee_save_fp =
        CPURegList(CPURegister::kFPRegister, kDRegSizeInBits,
                   descriptor->CalleeSavedFPRegisters());
    DCHECK(callee_save_fp.list() == CPURegList::GetCalleeSavedFP().list());
    int saved_count = callee_save_fp.Count();
    __ PushCPURegList(callee_save_fp);

    CPURegList callee_save = CPURegList(CPURegister::kRegister, kXRegSizeInBits,
                                        descriptor->CalleeSavedRegisters());
    // TODO(): TF save list not in sync with CPURegList::GetCalleeSaved() - x30
    // DCHECK(callee_save.list() == CPURegList::GetCalleeSaved().list());
    saved_count += callee_save.Count();
    __ PushCPURegList(callee_save);

    frame()->SetRegisterSaveAreaSize(saved_count * kPointerSize);

the 2nd DCHECK is a problem since the original TF code saves link-reg (X30), but
the assembler CPURegList::GetCalleeSaved() does not. I have a workaround I'll
send separately, but when we publish this, the arm team can help get the best
approach.

I don't think there is any perf issue here, as their Push/Pop macros do convert
into stp/ldp.

[akos.palfi.imgtec, 2015/06/23 19:48:57]

Thanks, it's much cleaner now.

+    MemOperand tos(csp, -2 * static_cast<int>(kXRegSize), PreIndex);
+    int saved_count = 0;
+    const RegList saves_fp = descriptor->CalleeSavedFPRegisters();
+    if (saves_fp != 0) {
+      // Save callee-saved FP registers.
+      int prev_reg = -1;
+      for (int i = FPRegister::kMaxNumRegisters - 1; i >= 0; i--) {
+        if (((1 << i) & saves_fp)) {
+          if (prev_reg != -1) {
+            __ stp(FPRegister::from_code(i), FPRegister::from_code(prev_reg),
+                   tos);
+            saved_count += 2;
+            prev_reg = -1;
+          } else {
+            prev_reg = i;
+          }
+        }
+      }
+      CHECK(prev_reg == -1);
+    }
+    const RegList saves = descriptor->CalleeSavedRegisters();
+    if (saves != 0) {
+      // Save callee-saved registers.
+      int prev_reg = -1;
+      for (int i = Register::kNumRegisters - 1; i >= 0; i--) {
+        if (((1 << i) & saves)) {
+          if (prev_reg != -1) {
+            __ stp(Register::from_code(i), Register::from_code(prev_reg), tos);
+            saved_count += 2;
+            prev_reg = -1;
+          } else {
+            prev_reg = i;
+          }
+        }
+      }
+      CHECK(prev_reg == -1);
+    }
+    frame()->SetRegisterSaveAreaSize(saved_count * kPointerSize);
   } else if (descriptor->IsJSFunctionCall()) {
     CompilationInfo* info = this->info();
     __ SetStackPointer(jssp);
     __ Prologue(info->IsCodePreAgingActive());
     frame()->SetRegisterSaveAreaSize(
         StandardFrameConstants::kFixedFrameSizeFromFp);
   } else if (needs_frame_) {
     __ SetStackPointer(jssp);
     __ StubPrologue();
     frame()->SetRegisterSaveAreaSize(
@@ skipping 28 matching lines @@
 
 void CodeGenerator::AssembleReturn() {
   CallDescriptor* descriptor = linkage()->GetIncomingDescriptor();
   int stack_slots = frame()->GetSpillSlotCount();
   if (descriptor->kind() == CallDescriptor::kCallAddress) {
     if (frame()->GetRegisterSaveAreaSize() > 0) {
       // Remove this frame's spill slots first.
       if (stack_slots > 0) {
         __ Add(csp, csp, AlignedStackSlots(stack_slots) * kPointerSize);
       }
+
       // Restore registers.
-      // TODO(dcarney): correct callee saved registers.
-      __ PopCalleeSavedRegisters();
+      MemOperand tos(csp, 2 * kXRegSize, PostIndex);
+      const RegList saves = descriptor->CalleeSavedRegisters();
+      if (saves != 0) {
+        // Restore callee-saved registers.
+        int prev_reg = -1;
+        for (int i = 0; i < Register::kNumRegisters; i++) {
+          if (((1 << i) & saves)) {
+            if (prev_reg != -1) {
+              __ ldp(Register::from_code(prev_reg), Register::from_code(i),
+                     tos);
+              prev_reg = -1;
+            } else {
+              prev_reg = i;
+            }
+          }
+        }
+        CHECK(prev_reg == -1);
+      }
+      const RegList saves_fp = descriptor->CalleeSavedFPRegisters();
+      if (saves_fp != 0) {
+        // Restore callee-saved FP registers.
+        int prev_reg = -1;
+        for (int i = 0; i < FPRegister::kMaxNumRegisters; i++) {
+          if (((1 << i) & saves_fp)) {
+            if (prev_reg != -1) {
+              __ ldp(FPRegister::from_code(prev_reg), FPRegister::from_code(i),
+                     tos);
+              prev_reg = -1;
+            } else {
+              prev_reg = i;
+            }
+          }
+        }
+        CHECK(prev_reg == -1);
+      }
     }
+
     __ Mov(csp, fp);
     __ Pop(fp, lr);
     __ Ret();
   } else if (descriptor->IsJSFunctionCall() || needs_frame_) {
     __ Mov(jssp, fp);
     __ Pop(fp, lr);
     int pop_count = descriptor->IsJSFunctionCall()
                         ? static_cast<int>(descriptor->JSParameterCount())
                         : 0;
     __ Drop(pop_count);
@@ skipping 185 matching lines @@
     }
   }
   MarkLazyDeoptSite();
 }
 
 #undef __
 
 }  // namespace compiler
 }  // namespace internal
 }  // namespace v8