[x64] Introduce FMA3 instructions on scalar data elements.

test/cctest/test-disasm-x64.cc

 // Copyright 2011 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 33 matching lines @@
 
 
 static void DummyStaticFunction(Object* result) {
 }
 
 
 TEST(DisasmX64) {
   CcTest::InitializeVM();
   Isolate* isolate = CcTest::i_isolate();
   HandleScope scope(isolate);
-  v8::internal::byte buffer[2048];
+  v8::internal::byte buffer[4096];
   Assembler assm(isolate, buffer, sizeof buffer);
   DummyStaticFunction(NULL);  // just bloody use it (DELETE; debugging)
 
   // Short immediate instructions
   __ addq(rax, Immediate(12345678));
   __ orq(rax, Immediate(12345678));
   __ subq(rax, Immediate(12345678));
   __ xorq(rax, Immediate(12345678));
   __ andq(rax, Immediate(12345678));
 
@@ skipping 322 matching lines @@
 
     // logic operation
     __ andps(xmm0, xmm1);
     __ andps(xmm0, Operand(rbx, rcx, times_4, 10000));
     __ orps(xmm0, xmm1);
     __ orps(xmm0, Operand(rbx, rcx, times_4, 10000));
     __ xorps(xmm0, xmm1);
     __ xorps(xmm0, Operand(rbx, rcx, times_4, 10000));
 
     // Arithmetic operation
+    __ addss(xmm1, xmm0);
+    __ addss(xmm1, Operand(rbx, rcx, times_4, 10000));
+    __ mulss(xmm1, xmm0);
+    __ mulss(xmm1, Operand(rbx, rcx, times_4, 10000));
+    __ subss(xmm1, xmm0);
+    __ subss(xmm1, Operand(rbx, rcx, times_4, 10000));
+    __ divss(xmm1, xmm0);
+    __ divss(xmm1, Operand(rbx, rcx, times_4, 10000));
     __ addps(xmm1, xmm0);
     __ addps(xmm1, Operand(rbx, rcx, times_4, 10000));
     __ subps(xmm1, xmm0);
     __ subps(xmm1, Operand(rbx, rcx, times_4, 10000));
     __ mulps(xmm1, xmm0);
     __ mulps(xmm1, Operand(rbx, rcx, times_4, 10000));
     __ divps(xmm1, xmm0);
     __ divps(xmm1, Operand(rbx, rcx, times_4, 10000));
+
+    __ ucomiss(xmm0, xmm1);
+    __ ucomiss(xmm0, Operand(rbx, rcx, times_4, 10000));
   }
   // SSE 2 instructions
   {
     __ cvttsd2si(rdx, Operand(rbx, rcx, times_4, 10000));
     __ cvttsd2si(rdx, xmm1);
     __ cvttsd2siq(rdx, xmm1);
     __ cvttsd2siq(rdx, Operand(rbx, rcx, times_4, 10000));
     __ cvtqsi2sd(xmm1, Operand(rbx, rcx, times_4, 10000));
     __ cvtqsi2sd(xmm1, rdx);
     __ movsd(xmm1, Operand(rbx, rcx, times_4, 10000));
@@ skipping 42 matching lines @@
     __ cmovq(greater, rax, Operand(rdx, 3));
   }
 
   {
     if (CpuFeatures::IsSupported(SSE4_1)) {
       CpuFeatureScope scope(&assm, SSE4_1);
       __ extractps(rax, xmm1, 0);
     }
   }
 
+  // FMA3 instruction
+  {
+    if (CpuFeatures::IsSupported(FMA3)) {
+      CpuFeatureScope scope(&assm, FMA3);
+      __ vfmadd132sd(xmm0, xmm1, xmm2);
+      __ vfmadd132sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmadd213sd(xmm0, xmm1, xmm2);
+      __ vfmadd213sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmadd231sd(xmm0, xmm1, xmm2);
+      __ vfmadd231sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfmadd132sd(xmm9, xmm10, xmm11);
+      __ vfmadd132sd(xmm9, xmm10, Operand(r9, r11, times_4, 10000));
+      __ vfmadd213sd(xmm9, xmm10, xmm11);
+      __ vfmadd213sd(xmm9, xmm10, Operand(r9, r11, times_4, 10000));
+      __ vfmadd231sd(xmm9, xmm10, xmm11);
+      __ vfmadd231sd(xmm9, xmm10, Operand(r9, r11, times_4, 10000));
+
+      __ vfmsub132sd(xmm0, xmm1, xmm2);
+      __ vfmsub132sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmsub213sd(xmm0, xmm1, xmm2);
+      __ vfmsub213sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmsub231sd(xmm0, xmm1, xmm2);
+      __ vfmsub231sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfnmadd132sd(xmm0, xmm1, xmm2);
+      __ vfnmadd132sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmadd213sd(xmm0, xmm1, xmm2);
+      __ vfnmadd213sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmadd231sd(xmm0, xmm1, xmm2);
+      __ vfnmadd231sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfnmsub132sd(xmm0, xmm1, xmm2);
+      __ vfnmsub132sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmsub213sd(xmm0, xmm1, xmm2);
+      __ vfnmsub213sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmsub231sd(xmm0, xmm1, xmm2);
+      __ vfnmsub231sd(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfmadd132ss(xmm0, xmm1, xmm2);
+      __ vfmadd132ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmadd213ss(xmm0, xmm1, xmm2);
+      __ vfmadd213ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmadd231ss(xmm0, xmm1, xmm2);
+      __ vfmadd231ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfmsub132ss(xmm0, xmm1, xmm2);
+      __ vfmsub132ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmsub213ss(xmm0, xmm1, xmm2);
+      __ vfmsub213ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfmsub231ss(xmm0, xmm1, xmm2);
+      __ vfmsub231ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfnmadd132ss(xmm0, xmm1, xmm2);
+      __ vfnmadd132ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmadd213ss(xmm0, xmm1, xmm2);
+      __ vfnmadd213ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmadd231ss(xmm0, xmm1, xmm2);
+      __ vfnmadd231ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+
+      __ vfnmsub132ss(xmm0, xmm1, xmm2);
+      __ vfnmsub132ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmsub213ss(xmm0, xmm1, xmm2);
+      __ vfnmsub213ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+      __ vfnmsub231ss(xmm0, xmm1, xmm2);
+      __ vfnmsub231ss(xmm0, xmm1, Operand(rbx, rcx, times_4, 10000));
+    }
+  }
   // xchg.
   {
     __ xchgq(rax, rax);
     __ xchgq(rax, rbx);
     __ xchgq(rbx, rbx);
     __ xchgq(rbx, Operand(rsp, 12));
   }
 
   // Nop instructions
   for (int i = 0; i < 16; i++) {
@@ skipping 10 matching lines @@
 #ifdef OBJECT_PRINT
   OFStream os(stdout);
   code->Print(os);
   byte* begin = code->instruction_start();
   byte* end = begin + code->instruction_size();
   disasm::Disassembler::Disassemble(stdout, begin, end);
 #endif
 }
 
 #undef __