Index: llvm/lib/Target/X86/X86ISelLowering.cpp
===================================================================
--- llvm/lib/Target/X86/X86ISelLowering.cpp
+++ llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -1936,7 +1936,8 @@
       if (Subtarget.hasSSE2())
         return MVT::v16i8;
       // TODO: Can SSE1 handle a byte vector?
-      if (Subtarget.hasSSE1())
+
+      if (Subtarget.hasSSE1() && Subtarget.hasX87())
         return MVT::v4f32;
     } else if ((!IsMemset || ZeroMemset) && !MemcpyStrSrc && Size >= 8 &&
                !Subtarget.is64Bit() && Subtarget.hasSSE2()) {
Index: llvm/test/CodeGen/X86/pr38738.ll
===================================================================
--- /dev/null
+++ llvm/test/CodeGen/X86/pr38738.ll
@@ -0,0 +1,41 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -mtriple=x86_64-unknown-linux-gnu -o - -mattr=-x87,+sse,-sse2 %s | FileCheck --check-prefix X64 %s
+; RUN: llc -mtriple=i686-unknown-linux-gnu -o - -mattr=-x87,+sse,-sse2 %s | FileCheck   --check-prefix X86 %s
+
+%struct.params = type { double, double }
+
+define dso_local i32 @pr38738() {
+; X64-LABEL: pr38738:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movq $0, -{{[0-9]+}}(%rsp)
+; X64-NEXT:    movq $0, -{{[0-9]+}}(%rsp)
+; X64-NEXT:    movl $0, -{{[0-9]+}}(%rsp)
+; X64-NEXT:    movl -{{[0-9]+}}(%rsp), %eax
+; X64-NEXT:    retq
+;
+; X86-LABEL: pr38738:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    subl $28, %esp
+; X86-NEXT:    .cfi_def_cfa_offset 32
+; X86-NEXT:    movl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl $0, (%esp)
+; X86-NEXT:    movl $0, {{[0-9]+}}(%esp)
+; X86-NEXT:    movl {{[0-9]+}}(%esp), %eax
+; X86-NEXT:    addl $28, %esp
+; X86-NEXT:    .cfi_def_cfa_offset 4
+; X86-NEXT:    retl
+entry:
+  %retval = alloca i32, align 4
+  %dlg_sys_param = alloca %struct.params, align 8
+  %total_active_bw = alloca float, align 4
+  %0 = bitcast %struct.params* %dlg_sys_param to i8*
+  call void @llvm.memset.p0i8.i64(i8* align 8 %0, i8 0, i64 16, i1 false)
+  store float 0.000000e+00, float* %total_active_bw, align 4
+  %1 = load i32, i32* %retval, align 4
+  ret i32 %1
+}
+
+declare void @llvm.memset.p0i8.i64(i8* nocapture writeonly, i8, i64, i1) argmemonly nounwind
+