Add inline non-transcendental cache version of log to lithium.

src/ia32/lithium-codegen-ia32.cc

Index: src/ia32/lithium-codegen-ia32.cc
===================================================================
--- src/ia32/lithium-codegen-ia32.cc	(revision 7492)
+++ src/ia32/lithium-codegen-ia32.cc	(working copy)
@@ -2777,9 +2777,31 @@
 
 void LCodeGen::DoMathLog(LUnaryMathOperation* instr) {
   ASSERT(ToDoubleRegister(instr->result()).is(xmm1));
-  TranscendentalCacheStub stub(TranscendentalCache::LOG,
-                               TranscendentalCacheStub::UNTAGGED);
-  CallCode(stub.GetCode(), RelocInfo::CODE_TARGET, instr, false);
+  XMMRegister input_reg = ToDoubleRegister(instr->InputAt(0));

[fschneider, 2011/04/04 13:05:54]

Drive-by: for optimizal performance make sure to also update the register
constraints in  LChunkBuilder::DoUnaryMathOperation. e.g. it seems MathLog does
not need MarkAsCall and fixed input/output registers anymore.

[Rico, 2011/04/05 07:12:45]

Done.

+  NearLabel positive, done, zero, negative;
+  __ xorpd(xmm0, xmm0);
+  __ ucomisd(input_reg, xmm0);
+  __ j(above, &positive);
+  __ j(equal, &zero);
+  ExternalReference nan = ExternalReference::address_of_nan();
+  __ movdbl(input_reg, Operand::StaticVariable(nan));
+  __ jmp(&done);
+  __ bind(&zero);
+  __ push(Immediate(0xFFF00000));
+  __ push(Immediate(0));
+  __ movdbl(input_reg, Operand(esp, 0));

[Lasse Reichstein, 2011/04/04 13:38:23]

You can load -Infinity into an XMM register without using memory by loading it
as float and converting to double (if you have a temp register).
I.e., something like
  mov(temp, Immediate(kFloatNegInfinityBits)); // whatever the bits are
  movd(input_reg, temp);  
  cvtss2sd(input_reg, input_reg);

(If you don't have a temp register, the code is fine. It's not like it's the
common path for real code :).

[Rico, 2011/04/05 07:12:45]

No temp register

+  __ add(Operand(esp), Immediate(kDoubleSize));
+  __ jmp(&done);
+  __ bind(&positive);
+  __ fldln2();
+  __ sub(Operand(esp), Immediate(kDoubleSize));
+  __ movdbl(Operand(esp, 0), input_reg);
+  __ fld_d(Operand(esp, 0));
+  __ fyl2x();
+  __ fstp_d(Operand(esp, 0));
+  __ movdbl(input_reg, Operand(esp,0));
+  __ add(Operand(esp), Immediate(kDoubleSize));
+  __ bind(&done);
 }