Fix BinaryOpIC implementation on ARM.

On a pair of smis HEAP_NUMBERS stub is significantly slower than GENERIC. This slows
down some tests dramatically (crypto-aes from SunSpider).
With this change HEAP_NUMBERS stub switches to GENERIC stub the first time it sees 2 smis
as its operands.

Review URL: http://codereview.chromium.org/1687005

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@4483 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

diff --git a/src/arm/codegen-arm.cc b/src/arm/codegen-arm.cc
index 8145f5a..af46d81 100644 (file)
--- a/src/arm/codegen-arm.cc
+++ b/src/arm/codegen-arm.cc
@@ -6468,11 +6468,22 @@ void GenericBinaryOpStub::HandleBinaryOpSlowCases(
       __ pop(lr);
     }
 
+    // HEAP_NUMBERS stub is slower than GENERIC on a pair of smis.
+    // r0 is known to be a smi. If r1 is also a smi then switch to GENERIC.
+    Label r1_is_not_smi;
+    if (runtime_operands_type_ == BinaryOpIC::HEAP_NUMBERS) {
+      __ tst(r1, Operand(kSmiTagMask));
+      __ b(ne, &r1_is_not_smi);
+      GenerateTypeTransition(masm);
+      __ jmp(&r1_is_smi);
+    }
+
     __ bind(&finished_loading_r0);
 
     // Move r1 to a double in r0-r1.
     __ tst(r1, Operand(kSmiTagMask));
     __ b(eq, &r1_is_smi);  // It's a Smi so don't check it's a heap number.
+    __ bind(&r1_is_not_smi);
     __ CompareObjectType(r1, r4, r4, HEAP_NUMBER_TYPE);
     __ b(ne, &slow);
     if (mode_ == OVERWRITE_LEFT) {