Port r11517 (not deopt Math.floor on negative input) to x64, sse2.

BUG=v8:873
TEST=math-floor-negative.js

Review URL: https://chromiumcodereview.appspot.com/10636057

git-svn-id: http://v8.googlecode.com/svn/branches/bleeding_edge@11935 ce2b1a6d-e550-0410-aec6-3dcde31c8c00

src/ia32/lithium-codegen-ia32.cc

@@ -3169,8 +3169,7 @@ void LCodeGen::DoMathFloor(LUnaryMathOperation* instr) {
     __ cmp(output_reg, 0x80000000u);
     DeoptimizeIf(equal, instr->environment());
   } else {
-    Label negative_sign;
-    Label done;
+    Label negative_sign, done;
     // Deoptimize on unordered.
     __ xorps(xmm_scratch, xmm_scratch);  // Zero the register.
     __ ucomisd(input_reg, xmm_scratch);
@@ -3196,9 +3195,9 @@ void LCodeGen::DoMathFloor(LUnaryMathOperation* instr) {
     DeoptimizeIf(equal, instr->environment());
     __ jmp(&done, Label::kNear);
 
-    // Non-zero negative reaches here
+    // Non-zero negative reaches here.
     __ bind(&negative_sign);
-    // Truncate, then compare and compensate
+    // Truncate, then compare and compensate.
     __ cvttsd2si(output_reg, Operand(input_reg));
     __ cvtsi2sd(xmm_scratch, output_reg);
     __ ucomisd(input_reg, xmm_scratch);

src/x64/assembler-x64.h

@@ -455,6 +455,7 @@ class CpuFeatures : public AllStatic {
     ASSERT(initialized_);
     if (f == SSE2 && !FLAG_enable_sse2) return false;
     if (f == SSE3 && !FLAG_enable_sse3) return false;
+    if (f == SSE4_1 && !FLAG_enable_sse4_1) return false;
     if (f == CMOV && !FLAG_enable_cmov) return false;
     if (f == RDTSC && !FLAG_enable_rdtsc) return false;
     if (f == SAHF && !FLAG_enable_sahf) return false;

src/x64/lithium-codegen-x64.cc

@@ -3031,7 +3031,6 @@ void LCodeGen::DoMathFloor(LUnaryMathOperation* instr) {
   XMMRegister xmm_scratch = xmm0;
   Register output_reg = ToRegister(instr->result());
   XMMRegister input_reg = ToDoubleRegister(instr->InputAt(0));
-  Label done;
 
   if (CpuFeatures::IsSupported(SSE4_1)) {
     CpuFeatures::Scope scope(SSE4_1);
@@ -3046,10 +3045,13 @@ void LCodeGen::DoMathFloor(LUnaryMathOperation* instr) {
     __ cmpl(output_reg, Immediate(0x80000000));
     DeoptimizeIf(equal, instr->environment());
   } else {
+    Label negative_sign, done;
     // Deoptimize on negative inputs.
     __ xorps(xmm_scratch, xmm_scratch);  // Zero the register.
     __ ucomisd(input_reg, xmm_scratch);
-    DeoptimizeIf(below, instr->environment());
+    DeoptimizeIf(parity_even, instr->environment());
+    __ j(below, &negative_sign, Label::kNear);
+
     if (instr->hydrogen()->CheckFlag(HValue::kBailoutOnMinusZero)) {
       // Check for negative zero.
       Label positive_sign;
@@ -3064,12 +3066,23 @@ void LCodeGen::DoMathFloor(LUnaryMathOperation* instr) {
 
     // Use truncating instruction (OK because input is positive).
     __ cvttsd2si(output_reg, input_reg);
-
     // Overflow is signalled with minint.
     __ cmpl(output_reg, Immediate(0x80000000));
     DeoptimizeIf(equal, instr->environment());
+    __ jmp(&done, Label::kNear);
+
+    // Non-zero negative reaches here.
+    __ bind(&negative_sign);
+    // Truncate, then compare and compensate.
+    __ cvttsd2si(output_reg, input_reg);
+    __ cvtlsi2sd(xmm_scratch, output_reg);
+    __ ucomisd(input_reg, xmm_scratch);
+    __ j(equal, &done, Label::kNear);
+    __ subl(output_reg, Immediate(1));
+    DeoptimizeIf(overflow, instr->environment());
+
+    __ bind(&done);
   }
-  __ bind(&done);
 }
 
 

test/mjsunit/math-floor-negative.js

+// Copyright 2012 the V8 project authors. All rights reserved.
+// Redistribution and use in source and binary forms, with or without
+// modification, are permitted provided that the following conditions are
+// met:
+//
+//     * Redistributions of source code must retain the above copyright
+//       notice, this list of conditions and the following disclaimer.
+//     * Redistributions in binary form must reproduce the above
+//       copyright notice, this list of conditions and the following
+//       disclaimer in the documentation and/or other materials provided
+//       with the distribution.
+//     * Neither the name of Google Inc. nor the names of its
+//       contributors may be used to endorse or promote products derived
+//       from this software without specific prior written permission.
+//
+// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+// Flags: --noenable_sse4_1 --allow-natives-syntax
+
+function test1() {
+  // Trigger overflow when converting/truncating double to integer.
+  // Divide by 10 to avoid overflow when smi-tagging at the end.
+  return Math.floor(-100000000000.5) / 10;
+}
+
+function test2() {
+  // Trigger no overflow.
+  return Math.floor(-100.2);
+}
+
+function test3() {
+  // Trigger overflow when compensating by subtracting after compare.
+  // Divide by 10 to avoid overflow when smi-tagging at the end.
+  return Math.floor(-2147483648.1) / 10;
+}
+
+test1();
+test1();
+%OptimizeFunctionOnNextCall(test1);
+test2();
+test2();
+%OptimizeFunctionOnNextCall(test2);
+test3();
+test3();
+%OptimizeFunctionOnNextCall(test3);
+
+assertEquals(-10000000000.1, test1());
+assertEquals(-101, test2());
+assertEquals(-214748364.9, test3());