8024344: PPC64 (part 112): C argument in register AND stack slot.

Summary: On PPC, the first 13 floating point arguments to C calls are passed in floating point registers. Also, all but the first 8 arguments are passed on the stack. So there can be floating point arguments that are passed on the stack and in a register. We duplicate the regs datastructure in c_calling_convention() to represent this.
Reviewed-by: kvn, cjplummer

src/cpu/sparc/vm/sharedRuntime_sparc.cpp

@@ -1118,7 +1118,9 @@ static VMReg int_stk_helper( int i ) {
 
 int SharedRuntime::c_calling_convention(const BasicType *sig_bt,
                                          VMRegPair *regs,
+                                         VMRegPair *regs2,
                                          int total_args_passed) {
+    assert(regs2 == NULL, "not needed on sparc");
 
     // Return the number of VMReg stack_slots needed for the args.
     // This value does not include an abi space (like register window
@@ -2096,7 +2098,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
   // the 1st six register arguments). It's weird see int_stk_helper.
   //
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   if (is_critical_native) {
     // Critical natives may have to call out so they need a save area
@@ -2843,7 +2845,7 @@ nmethod *SharedRuntime::generate_dtrace_nmethod(
   // the 1st six register arguments). It's weird see int_stk_helper.
   //
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   // Calculate the total number of stack slots we will need.
 

src/cpu/sparc/vm/sparc.ad

@@ -3218,7 +3218,7 @@ frame %{
   // C.
   c_calling_convention %{
     // This is obviously always outgoing
-    (void) SharedRuntime::c_calling_convention(sig_bt, regs, length);
+    (void) SharedRuntime::c_calling_convention(sig_bt, regs, /*regs2=*/NULL, length);
   %}
 
   // Location of native (C/C++) and interpreter return values.  This is specified to
@@ -9118,7 +9118,7 @@ instruct branch_short(label labl) %{
   size(4);
   ins_cost(BRANCH_COST);
   format %{ "BA     $labl\t! short branch" %}
-  ins_encode %{ 
+  ins_encode %{
     Label* L = $labl$$label;
     assert(__ use_cbcond(*L), "back to back cbcond");
     __ ba_short(*L);

src/cpu/x86/vm/sharedRuntime_x86_32.cpp

@@ -977,7 +977,9 @@ AdapterHandlerEntry* SharedRuntime::generate_i2c2i_adapters(MacroAssembler *masm
 
 int SharedRuntime::c_calling_convention(const BasicType *sig_bt,
                                          VMRegPair *regs,
+                                         VMRegPair *regs2,
                                          int total_args_passed) {
+  assert(regs2 == NULL, "not needed on x86");
 // We return the amount of VMRegImpl stack slots we need to reserve for all
 // the arguments NOT counting out_preserve_stack_slots.
 
@@ -1624,7 +1626,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
   // Now figure out where the args must be stored and how much stack space
   // they require.
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   // Compute framesize for the wrapper.  We need to handlize all oops in
   // registers a max of 2 on x86.
@@ -2495,7 +2497,7 @@ nmethod *SharedRuntime::generate_dtrace_nmethod(
   // they require (neglecting out_preserve_stack_slots).
 
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   // Calculate the total number of stack slots we will need.
 

src/cpu/x86/vm/sharedRuntime_x86_64.cpp

@@ -889,7 +889,9 @@ AdapterHandlerEntry* SharedRuntime::generate_i2c2i_adapters(MacroAssembler *masm
 
 int SharedRuntime::c_calling_convention(const BasicType *sig_bt,
                                          VMRegPair *regs,
+                                         VMRegPair *regs2,
                                          int total_args_passed) {
+  assert(regs2 == NULL, "not needed on x86");
 // We return the amount of VMRegImpl stack slots we need to reserve for all
 // the arguments NOT counting out_preserve_stack_slots.
 
@@ -1857,7 +1859,7 @@ nmethod* SharedRuntime::generate_native_wrapper(MacroAssembler* masm,
   // Now figure out where the args must be stored and how much stack space
   // they require.
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   // Compute framesize for the wrapper.  We need to handlize all oops in
   // incoming registers
@@ -2761,7 +2763,7 @@ nmethod *SharedRuntime::generate_dtrace_nmethod(MacroAssembler *masm,
   // the 1st six register arguments). It's weird see int_stk_helper.
 
   int out_arg_slots;
-  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, total_c_args);
+  out_arg_slots = c_calling_convention(out_sig_bt, out_regs, NULL, total_c_args);
 
   // Calculate the total number of stack slots we will need.
 

src/cpu/x86/vm/x86_32.ad

@@ -3755,7 +3755,7 @@ frame %{
   // automatically biased by the preserve_stack_slots field above.
   c_calling_convention %{
     // This is obviously always outgoing
-    (void) SharedRuntime::c_calling_convention(sig_bt, regs, length);
+    (void) SharedRuntime::c_calling_convention(sig_bt, regs, /*regs2=*/NULL, length);
   %}
 
   // Location of C & interpreter return values

src/cpu/x86/vm/x86_64.ad

@@ -2941,7 +2941,7 @@ frame
   c_calling_convention
   %{
     // This is obviously always outgoing
-    (void) SharedRuntime::c_calling_convention(sig_bt, regs, length);
+    (void) SharedRuntime::c_calling_convention(sig_bt, regs, /*regs2=*/NULL, length);
   %}
 
   // Location of compiled Java return values.  Same as C for now.

src/cpu/zero/vm/globals_zero.hpp

@@ -43,6 +43,7 @@ define_pd_global(bool,  UncommonNullCast,     true);
 define_pd_global(intx,  CodeEntryAlignment,   32);
 define_pd_global(intx,  OptoLoopAlignment,    16);
 define_pd_global(intx,  InlineFrequencyCount, 100);
+define_pd_global(intx,  InlineSmallCode,      1000 );
 define_pd_global(intx,  PreInflateSpin,       10);
 
 define_pd_global(intx,  StackYellowPages,     2);

src/cpu/zero/vm/sharedRuntime_zero.cpp

@@ -135,6 +135,7 @@ RuntimeStub* SharedRuntime::generate_resolve_blob(address destination, const cha
 
 int SharedRuntime::c_calling_convention(const BasicType *sig_bt,
                                          VMRegPair *regs,
+                                         VMRegPair *regs2,
                                          int total_args_passed) {
   ShouldNotCallThis();
   return 0;

src/cpu/zero/vm/shark_globals_zero.hpp

@@ -50,7 +50,6 @@ define_pd_global(intx,     Tier4BackEdgeThreshold,       100000);
 
 define_pd_global(intx,     OnStackReplacePercentage,     933  );
 define_pd_global(intx,     FreqInlineSize,               325  );
-define_pd_global(intx,     InlineSmallCode,              1000 );
 define_pd_global(uintx,    NewRatio,                     12   );
 define_pd_global(intx,     NewSizeThreadIncrease,        4*K  );
 define_pd_global(intx,     InitialCodeCacheSize,         160*K);

src/share/vm/c1/c1_FrameMap.cpp

@@ -133,7 +133,7 @@ CallingConvention* FrameMap::c_calling_convention(const BasicTypeArray* signatur
     }
   }
 
-  intptr_t out_preserve = SharedRuntime::c_calling_convention(sig_bt, regs, sizeargs);
+  intptr_t out_preserve = SharedRuntime::c_calling_convention(sig_bt, regs, NULL, sizeargs);
   LIR_OprList* args = new LIR_OprList(signature->length());
   for (i = 0; i < sizeargs;) {
     BasicType t = sig_bt[i];

src/share/vm/runtime/sharedRuntime.hpp

@@ -356,7 +356,15 @@ class SharedRuntime: AllStatic {
                                                           const VMRegPair* regs) NOT_DEBUG_RETURN;
 
   // Ditto except for calling C
-  static int c_calling_convention(const BasicType *sig_bt, VMRegPair *regs, int total_args_passed);
+  //
+  // C argument in register AND stack slot.
+  // Some architectures require that an argument must be passed in a register
+  // AND in a stack slot. These architectures provide a second VMRegPair array
+  // to be filled by the c_calling_convention method. On other architectures,
+  // NULL is being passed as the second VMRegPair array, so arguments are either
+  // passed in a register OR in a stack slot.
+  static int c_calling_convention(const BasicType *sig_bt, VMRegPair *regs, VMRegPair *regs2,
+                                  int total_args_passed);
 
   // Generate I2C and C2I adapters. These adapters are simple argument marshalling
   // blobs. Unlike adapters in the tiger and earlier releases the code in these