[X86][FastISel] Don't force Nearest-Even rounding for VCVTPS2PH, use MXCSR.

FastISel counterpart to r259448.

llvm-svn: 259449
diff --git a/llvm/test/CodeGen/X86/fast-isel-float-half-convertion.ll b/llvm/test/CodeGen/X86/fast-isel-float-half-convertion.ll
index 707a325..acb85fd 100644
--- a/llvm/test/CodeGen/X86/fast-isel-float-half-convertion.ll
+++ b/llvm/test/CodeGen/X86/fast-isel-float-half-convertion.ll
@@ -4,7 +4,7 @@
 
 define i16 @test_fp32_to_fp16(float %a) {
 ; CHECK-LABEL: test_fp32_to_fp16:
-; CHECK: vcvtps2ph $0, %xmm0, %xmm0
+; CHECK: vcvtps2ph $4, %xmm0, %xmm0
 ; CHECK-NEXT: vmovd %xmm0, %eax
 ; CHECK-NEXT: retq
 entry: