[ARM] Improve VFP register moves.

test/cctest/test-macro-assembler-arm.cc

 // Copyright 2013 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
@@ skipping 151 matching lines @@
   T t;
 
   __ stm(db_w, sp, r4.bit() | r5.bit() | lr.bit());
 
   for (int i = 0; i < 4; i++) {
     __ mov(r4, Operand(i));
     __ vdup(Neon32, q1, r4);
     __ ExtractLane(r5, q1, NeonS32, i);
     __ str(r5, MemOperand(r0, offsetof(T, i32x4_low) + 4 * i));
     SwVfpRegister si = SwVfpRegister::from_code(i);
-    __ ExtractLane(si, q1, r4, i);
+    __ ExtractLane(si, q1, i);
     __ vstr(si, r0, offsetof(T, f32x4_low) + 4 * i);
   }
 
   for (int i = 0; i < 8; i++) {
     __ mov(r4, Operand(i));
     __ vdup(Neon16, q1, r4);
     __ ExtractLane(r5, q1, NeonS16, i);
     __ str(r5, MemOperand(r0, offsetof(T, i16x8_low) + 4 * i));
   }
 
@@ skipping 13 matching lines @@
     __ str(r5, MemOperand(r0, offsetof(T, i8x16_low_d) + 4 * (i + 8)));
   }
 
   if (CpuFeatures::IsSupported(VFP32DREGS)) {
     for (int i = 0; i < 4; i++) {
       __ mov(r4, Operand(-i));
       __ vdup(Neon32, q15, r4);
       __ ExtractLane(r5, q15, NeonS32, i);
       __ str(r5, MemOperand(r0, offsetof(T, i32x4_high) + 4 * i));
       SwVfpRegister si = SwVfpRegister::from_code(i);
-      __ ExtractLane(si, q15, r4, i);
+      __ ExtractLane(si, q15, i);
       __ vstr(si, r0, offsetof(T, f32x4_high) + 4 * i);
     }
 
     for (int i = 0; i < 8; i++) {
       __ mov(r4, Operand(-i));
       __ vdup(Neon16, q15, r4);
       __ ExtractLane(r5, q15, NeonS16, i);
       __ str(r5, MemOperand(r0, offsetof(T, i16x8_high) + 4 * i));
     }
 
@@ skipping 80 matching lines @@
     int16_t i16x8_high[8];
     int8_t i8x16_low[16];
     int8_t i8x16_high[16];
     int32_t f32x4_low[4];
     int32_t f32x4_high[4];
   } T;
   T t;
 
   __ stm(db_w, sp, r4.bit() | r5.bit() | r6.bit() | r7.bit() | lr.bit());
 
-  const Register kScratch = r5;
-
   __ veor(q0, q0, q0);  // Zero
   __ veor(q1, q1, q1);  // Zero
   for (int i = 0; i < 4; i++) {
     __ mov(r4, Operand(i));
     __ ReplaceLane(q0, q0, r4, NeonS32, i);
     SwVfpRegister si = SwVfpRegister::from_code(i);
     __ vmov(si, r4);
-    __ ReplaceLane(q1, q1, si, kScratch, i);
+    __ ReplaceLane(q1, q1, si, i);
   }
   __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, i32x4_low))));
   __ vst1(Neon8, NeonListOperand(q0), NeonMemOperand(r4));
   __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, f32x4_low))));
   __ vst1(Neon8, NeonListOperand(q1), NeonMemOperand(r4));
 
   __ veor(q0, q0, q0);  // Zero
   for (int i = 0; i < 8; i++) {
     __ mov(r4, Operand(i));
     __ ReplaceLane(q0, q0, r4, NeonS16, i);
@@ skipping 10 matching lines @@
   __ vst1(Neon8, NeonListOperand(q0), NeonMemOperand(r4));
 
   if (CpuFeatures::IsSupported(VFP32DREGS)) {
     __ veor(q14, q14, q14);  // Zero
     __ veor(q15, q15, q15);  // Zero
     for (int i = 0; i < 4; i++) {
       __ mov(r4, Operand(-i));
       __ ReplaceLane(q14, q14, r4, NeonS32, i);
       SwVfpRegister si = SwVfpRegister::from_code(i);
       __ vmov(si, r4);
-      __ ReplaceLane(q15, q15, si, kScratch, i);
+      __ ReplaceLane(q15, q15, si, i);
     }
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, i32x4_high))));
     __ vst1(Neon8, NeonListOperand(q14), NeonMemOperand(r4));
     __ add(r4, r0, Operand(static_cast<int32_t>(offsetof(T, f32x4_high))));
     __ vst1(Neon8, NeonListOperand(q15), NeonMemOperand(r4));
 
     __ veor(q14, q14, q14);  // Zero
     for (int i = 0; i < 8; i++) {
       __ mov(r4, Operand(-i));
       __ ReplaceLane(q14, q14, r4, NeonS16, i);
@@ skipping 41 matching lines @@
     for (int i = 0; i < 8; i++) {
       CHECK_EQ(-i, t.i16x8_high[i]);
     }
     for (int i = 0; i < 16; i++) {
       CHECK_EQ(-i, t.i8x16_high[i]);
     }
   }
 }
 
 #undef __