diff --git a/src/codegen/x64/assembler-x64.h b/src/codegen/x64/assembler-x64.h
index 206c59acc4..6509094083 100644
--- a/src/codegen/x64/assembler-x64.h
+++ b/src/codegen/x64/assembler-x64.h
@@ -1433,9 +1433,8 @@ class V8_EXPORT_PRIVATE Assembler : public AssemblerBase {
     vinstr(0x0b, dst, src1, src2, k66, k0F3A, kWIG);
     emit(static_cast<byte>(mode) | 0x8);  // Mask precision exception.
   }
-  void vroundps(XMMRegister dst, XMMRegister src1, XMMRegister src2,
-                RoundingMode mode) {
-    vinstr(0x08, dst, src1, src2, k66, k0F3A, kWIG);
+  void vroundps(XMMRegister dst, XMMRegister src, RoundingMode mode) {
+    vinstr(0x08, dst, xmm0, src, k66, k0F3A, kWIG);
     emit(static_cast<byte>(mode) | 0x8);  // Mask precision exception.
   }
 
diff --git a/src/codegen/x64/macro-assembler-x64.h b/src/codegen/x64/macro-assembler-x64.h
index ab1745a760..18674f9f54 100644
--- a/src/codegen/x64/macro-assembler-x64.h
+++ b/src/codegen/x64/macro-assembler-x64.h
@@ -174,9 +174,6 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
   AVX_OP(Cmpneqpd, cmpneqpd)
   AVX_OP(Cmpnltpd, cmpnltpd)
   AVX_OP(Cmpnlepd, cmpnlepd)
-  AVX_OP(Roundss, roundss)
-  AVX_OP(Roundsd, roundsd)
-  AVX_OP(Roundps, roundps)
   AVX_OP(Sqrtss, sqrtss)
   AVX_OP(Sqrtsd, sqrtsd)
   AVX_OP(Sqrtps, sqrtps)
@@ -284,6 +281,9 @@ class V8_EXPORT_PRIVATE TurboAssembler : public TurboAssemblerBase {
   AVX_OP_SSE4_1(Pmovzxwd, pmovzxwd)
   AVX_OP_SSE4_1(Pmovzxdq, pmovzxdq)
   AVX_OP_SSE4_1(Pextrq, pextrq)
+  AVX_OP_SSE4_1(Roundps, roundps)
+  AVX_OP_SSE4_1(Roundss, roundss)
+  AVX_OP_SSE4_1(Roundsd, roundsd)
   AVX_OP_SSE4_2(Pcmpgtq, pcmpgtq)
 
 #undef AVX_OP
diff --git a/src/diagnostics/x64/disasm-x64.cc b/src/diagnostics/x64/disasm-x64.cc
index 4d0760b17c..2251e0d7b4 100644
--- a/src/diagnostics/x64/disasm-x64.cc
+++ b/src/diagnostics/x64/disasm-x64.cc
@@ -947,6 +947,11 @@ int DisassemblerX64::AVXInstruction(byte* data) {
     int mod, regop, rm, vvvv = vex_vreg();
     get_modrm(*current, &mod, &regop, &rm);
     switch (opcode) {
+      case 0x08:
+        AppendToBuffer("vroundps %s,", NameOfXMMRegister(regop));
+        current += PrintRightXMMOperand(current);
+        AppendToBuffer(",0x%x", *current++);
+        break;
       case 0x0A:
         AppendToBuffer("vroundss %s,%s,", NameOfXMMRegister(regop),
                        NameOfXMMRegister(vvvv));
@@ -1840,6 +1845,12 @@ int DisassemblerX64::TwoByteOpcodeInstruction(byte* data) {
         current += PrintRightOperand(current);
         AppendToBuffer(",%s,%d", NameOfXMMRegister(regop), (*current) & 3);
         current += 1;
+      } else if (third_byte == 0x08) {
+        get_modrm(*current, &mod, &regop, &rm);
+        AppendToBuffer("roundps %s,", NameOfXMMRegister(regop));
+        current += PrintRightXMMOperand(current);
+        AppendToBuffer(",0x%x", (*current) & 3);
+        current += 1;
       } else if (third_byte == 0x0A) {
         get_modrm(*current, &mod, &regop, &rm);
         AppendToBuffer("roundss %s,", NameOfXMMRegister(regop));
diff --git a/test/cctest/test-disasm-x64.cc b/test/cctest/test-disasm-x64.cc
index 8e9eadca25..0bd7605a98 100644
--- a/test/cctest/test-disasm-x64.cc
+++ b/test/cctest/test-disasm-x64.cc
@@ -588,6 +588,7 @@ TEST(DisasmX64) {
       __ blendvpd(xmm5, xmm1);
       __ blendvpd(xmm5, Operand(rdx, 4));
 
+      __ roundps(xmm8, xmm3, kRoundUp);
       __ roundss(xmm8, xmm3, kRoundDown);
       __ roundsd(xmm8, xmm3, kRoundDown);
 
@@ -648,6 +649,7 @@ TEST(DisasmX64) {
       __ vmovdqu(xmm9, Operand(rbx, rcx, times_4, 10000));
       __ vmovdqu(Operand(rbx, rcx, times_4, 10000), xmm0);
 
+      __ vroundps(xmm9, xmm2, kRoundUp);
       __ vroundss(xmm9, xmm1, xmm2, kRoundDown);
       __ vroundsd(xmm8, xmm3, xmm0, kRoundDown);
       __ vucomisd(xmm9, xmm1);