PPC: Fix "Fix receiver when calling eval() bound by with scope"

PPC: Fix "Fix receiver when calling eval() bound by with scope"

R=wingo@igalia.com, dstence@us.ibm.com, michael_dawson@ca.ibm.com
BUG=

Committed: https://crrev.com/01009640b7590b2bd68a14171f96c5d42a2ab75f
Cr-Commit-Position: refs/heads/master@{#29300}

[dstence, 2015-06-24 18:10:13]

lgtm

[MTBrandyberry, 2015-06-24 18:20:43]

The CQ bit was checked by mbrandy@us.ibm.com

[commit-bot: I haz the power, 2015-06-24 18:20:58]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1209703002/1

[commit-bot: I haz the power, 2015-06-24 19:12:20]

The CQ bit was unchecked by commit-bot@chromium.org

[commit-bot: I haz the power, 2015-06-24 19:12:21]

Failed to apply patch for src/ppc/full-codegen-ppc.cc:
While running git apply --index -3 -p1;
  error: patch failed: src/ppc/full-codegen-ppc.cc:3148
  Falling back to three-way merge...
  Applied patch to 'src/ppc/full-codegen-ppc.cc' with conflicts.
  U src/ppc/full-codegen-ppc.cc

Patch:       src/ppc/full-codegen-ppc.cc
Index: src/ppc/full-codegen-ppc.cc
diff --git a/src/ppc/full-codegen-ppc.cc b/src/ppc/full-codegen-ppc.cc
index
e89be857cfd4c70a83fba91986b99ccb50f2d3e6..83526dd7d87920f196023f244aa68ce08c49f792
100644
--- a/src/ppc/full-codegen-ppc.cc
+++ b/src/ppc/full-codegen-ppc.cc
@@ -3148,34 +3148,35 @@ void
FullCodeGenerator::PushCalleeAndWithBaseObject(Call* expr) {
       EmitDynamicLookupFastCase(callee, NOT_INSIDE_TYPEOF, &slow, &done);
     }
     __ bind(&slow);
-    // Call the runtime to find the function to call (returned in rax) and
-    // the object holding it (returned in rdx).
-    __ Push(context_register());
-    __ Push(callee->name());
+    // Call the runtime to find the function to call (returned in r3) and
+    // the object holding it (returned in r4).
+    DCHECK(!context_register().is(r5));
+    __ mov(r5, Operand(callee->name()));
+    __ Push(context_register(), r5);
     __ CallRuntime(Runtime::kLoadLookupSlot, 2);
-    __ Push(rax);  // Function.
-    __ Push(rdx);  // Receiver.
+    __ Push(r3, r4);  // Function, receiver.
     PrepareForBailoutForId(expr->LookupId(), NO_REGISTERS);
 
     // If fast case code has been generated, emit code to push the function
     // and receiver and have the slow path jump around this code.
     if (done.is_linked()) {
       Label call;
-      __ jmp(&call, Label::kNear);
+      __ b(&call);
       __ bind(&done);
       // Push function.
-      __ Push(rax);
+      __ push(r3);
       // Pass undefined as the receiver, which is the WithBaseObject of a
       // non-object environment record.  If the callee is sloppy, it will patch
       // it up to be the global receiver.
-      __ PushRoot(Heap::kUndefinedValueRootIndex);
+      __ LoadRoot(r4, Heap::kUndefinedValueRootIndex);
+      __ push(r4);
       __ bind(&call);
     }
   } else {
     VisitForStackValue(callee);
     // refEnv.WithBaseObject()
     __ LoadRoot(r5, Heap::kUndefinedValueRootIndex);
-    __ push(r5);
+    __ push(r5);  // Reserved receiver slot.
   }
 }
 
@@ -3796,7 +3797,7 @@ void FullCodeGenerator::EmitArguments(CallRuntime* expr) {
   ZoneList<Expression*>* args = expr->arguments();
   DCHECK(args->length() == 1);
 
-  // ArgumentsAccessStub expects the key in edx and the formal
+  // ArgumentsAccessStub expects the key in r4 and the formal
   // parameter count in r3.
   VisitForAccumulatorValue(args->at(0));
   __ mr(r4, r3);

[MTBrandyberry, 2015-06-24 20:00:48]

On 2015/06/24 19:12:21, commit-bot: I haz the power wrote:
> Failed to apply patch for src/ppc/full-codegen-ppc.cc:
> While running git apply --index -3 -p1;
>   error: patch failed: src/ppc/full-codegen-ppc.cc:3148
>   Falling back to three-way merge...
>   Applied patch to 'src/ppc/full-codegen-ppc.cc' with conflicts.
>   U src/ppc/full-codegen-ppc.cc
> 
> Patch:       src/ppc/full-codegen-ppc.cc
> Index: src/ppc/full-codegen-ppc.cc
> diff --git a/src/ppc/full-codegen-ppc.cc b/src/ppc/full-codegen-ppc.cc
> index
>
e89be857cfd4c70a83fba91986b99ccb50f2d3e6..83526dd7d87920f196023f244aa68ce08c49f792
> 100644
> --- a/src/ppc/full-codegen-ppc.cc
> +++ b/src/ppc/full-codegen-ppc.cc
> @@ -3148,34 +3148,35 @@ void
> FullCodeGenerator::PushCalleeAndWithBaseObject(Call* expr) {
>        EmitDynamicLookupFastCase(callee, NOT_INSIDE_TYPEOF, &slow, &done);
>      }
>      __ bind(&slow);
> -    // Call the runtime to find the function to call (returned in rax) and
> -    // the object holding it (returned in rdx).
> -    __ Push(context_register());
> -    __ Push(callee->name());
> +    // Call the runtime to find the function to call (returned in r3) and
> +    // the object holding it (returned in r4).
> +    DCHECK(!context_register().is(r5));
> +    __ mov(r5, Operand(callee->name()));
> +    __ Push(context_register(), r5);
>      __ CallRuntime(Runtime::kLoadLookupSlot, 2);
> -    __ Push(rax);  // Function.
> -    __ Push(rdx);  // Receiver.
> +    __ Push(r3, r4);  // Function, receiver.
>      PrepareForBailoutForId(expr->LookupId(), NO_REGISTERS);
>  
>      // If fast case code has been generated, emit code to push the function
>      // and receiver and have the slow path jump around this code.
>      if (done.is_linked()) {
>        Label call;
> -      __ jmp(&call, Label::kNear);
> +      __ b(&call);
>        __ bind(&done);
>        // Push function.
> -      __ Push(rax);
> +      __ push(r3);
>        // Pass undefined as the receiver, which is the WithBaseObject of a
>        // non-object environment record.  If the callee is sloppy, it will
patch
>        // it up to be the global receiver.
> -      __ PushRoot(Heap::kUndefinedValueRootIndex);
> +      __ LoadRoot(r4, Heap::kUndefinedValueRootIndex);
> +      __ push(r4);
>        __ bind(&call);
>      }
>    } else {
>      VisitForStackValue(callee);
>      // refEnv.WithBaseObject()
>      __ LoadRoot(r5, Heap::kUndefinedValueRootIndex);
> -    __ push(r5);
> +    __ push(r5);  // Reserved receiver slot.
>    }
>  }
>  
> @@ -3796,7 +3797,7 @@ void FullCodeGenerator::EmitArguments(CallRuntime* expr)
{
>    ZoneList<Expression*>* args = expr->arguments();
>    DCHECK(args->length() == 1);
>  
> -  // ArgumentsAccessStub expects the key in edx and the formal
> +  // ArgumentsAccessStub expects the key in r4 and the formal
>    // parameter count in r3.
>    VisitForAccumulatorValue(args->at(0));
>    __ mr(r4, r3);

Holding off on this until the revert of the original commit is sorted.

[wingo, 2015-06-25 12:53:33]

On 2015/06/24 20:00:48, mtbrandyberry wrote:
> On 2015/06/24 19:12:21, commit-bot: I haz the power wrote:
> > Failed to apply patch for src/ppc/full-codegen-ppc.cc:
> > While running git apply --index -3 -p1;
> >   error: patch failed: src/ppc/full-codegen-ppc.cc:3148
> >   Falling back to three-way merge...
> >   Applied patch to 'src/ppc/full-codegen-ppc.cc' with conflicts.
> >   U src/ppc/full-codegen-ppc.cc
> > 
> > Patch:       src/ppc/full-codegen-ppc.cc
> > Index: src/ppc/full-codegen-ppc.cc
> > diff --git a/src/ppc/full-codegen-ppc.cc b/src/ppc/full-codegen-ppc.cc
> > index
> >
>
e89be857cfd4c70a83fba91986b99ccb50f2d3e6..83526dd7d87920f196023f244aa68ce08c49f792
> > 100644
> > --- a/src/ppc/full-codegen-ppc.cc
> > +++ b/src/ppc/full-codegen-ppc.cc
> > @@ -3148,34 +3148,35 @@ void
> > FullCodeGenerator::PushCalleeAndWithBaseObject(Call* expr) {
> >        EmitDynamicLookupFastCase(callee, NOT_INSIDE_TYPEOF, &slow, &done);
> >      }
> >      __ bind(&slow);
> > -    // Call the runtime to find the function to call (returned in rax) and
> > -    // the object holding it (returned in rdx).
> > -    __ Push(context_register());
> > -    __ Push(callee->name());
> > +    // Call the runtime to find the function to call (returned in r3) and
> > +    // the object holding it (returned in r4).
> > +    DCHECK(!context_register().is(r5));
> > +    __ mov(r5, Operand(callee->name()));
> > +    __ Push(context_register(), r5);
> >      __ CallRuntime(Runtime::kLoadLookupSlot, 2);
> > -    __ Push(rax);  // Function.
> > -    __ Push(rdx);  // Receiver.
> > +    __ Push(r3, r4);  // Function, receiver.
> >      PrepareForBailoutForId(expr->LookupId(), NO_REGISTERS);
> >  
> >      // If fast case code has been generated, emit code to push the function
> >      // and receiver and have the slow path jump around this code.
> >      if (done.is_linked()) {
> >        Label call;
> > -      __ jmp(&call, Label::kNear);
> > +      __ b(&call);
> >        __ bind(&done);
> >        // Push function.
> > -      __ Push(rax);
> > +      __ push(r3);
> >        // Pass undefined as the receiver, which is the WithBaseObject of a
> >        // non-object environment record.  If the callee is sloppy, it will
> patch
> >        // it up to be the global receiver.
> > -      __ PushRoot(Heap::kUndefinedValueRootIndex);
> > +      __ LoadRoot(r4, Heap::kUndefinedValueRootIndex);
> > +      __ push(r4);
> >        __ bind(&call);
> >      }
> >    } else {
> >      VisitForStackValue(callee);
> >      // refEnv.WithBaseObject()
> >      __ LoadRoot(r5, Heap::kUndefinedValueRootIndex);
> > -    __ push(r5);
> > +    __ push(r5);  // Reserved receiver slot.
> >    }
> >  }
> >  
> > @@ -3796,7 +3797,7 @@ void FullCodeGenerator::EmitArguments(CallRuntime*
expr)
> {
> >    ZoneList<Expression*>* args = expr->arguments();
> >    DCHECK(args->length() == 1);
> >  
> > -  // ArgumentsAccessStub expects the key in edx and the formal
> > +  // ArgumentsAccessStub expects the key in r4 and the formal
> >    // parameter count in r3.
> >    VisitForAccumulatorValue(args->at(0));
> >    __ mr(r4, r3);
> 
> Holding off on this until the revert of the original commit is sorted.

Did I just paste x64 assembly into this thing?  My bad :P  I meant to cut and
paste from the code below but I obviously flubbed it.  Sorry about that!

[MTBrandyberry, 2015-06-25 16:16:25]

The CQ bit was checked by mbrandy@us.ibm.com

[MTBrandyberry, 2015-06-25 16:16:26]

The patchset sent to the CQ was uploaded after l-g-t-m from dstence@us.ibm.com
Link to the patchset: https://codereview.chromium.org/1209703002/#ps20001
(title: "Rebase and reapply")

[commit-bot: I haz the power, 2015-06-25 16:16:34]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1209703002/20001

[wingo, 2015-06-25 16:20:05]

lgtm

[commit-bot: I haz the power, 2015-06-25 16:44:28]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-06-25 16:44:43]

Patchset 2 (id:??) landed as
https://crrev.com/01009640b7590b2bd68a14171f96c5d42a2ab75f
Cr-Commit-Position: refs/heads/master@{#29300}