case Intrinsic::x86_sse3_addsub_pd:
case Intrinsic::x86_sse3_addsub_ps:
case Intrinsic::x86_avx_addsub_pd_256:
- case Intrinsic::x86_avx_addsub_ps_256:
+ case Intrinsic::x86_avx_addsub_ps_256: {
+ APInt SubMask = APInt::getSplat(VWidth, APInt(2, 0x1));
+ if (DemandedElts.isSubsetOf(SubMask))
+ return IC.Builder.CreateFSub(II.getArgOperand(0), II.getArgOperand(1));
+
+ APInt AddMask = APInt::getSplat(VWidth, APInt(2, 0x2));
+ if (DemandedElts.isSubsetOf(AddMask))
+ return IC.Builder.CreateFAdd(II.getArgOperand(0), II.getArgOperand(1));
+
simplifyAndSetOp(&II, 0, DemandedElts, UndefElts);
simplifyAndSetOp(&II, 1, DemandedElts, UndefElts2);
UndefElts &= UndefElts2;
break;
+ }
case Intrinsic::x86_sse2_packssdw_128:
case Intrinsic::x86_sse2_packsswb_128:
define double @elts_addsub_v2f64(<2 x double> %0, <2 x double> %1) {
; CHECK-LABEL: @elts_addsub_v2f64(
; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <2 x double> [[TMP1:%.*]], <2 x double> undef, <2 x i32> <i32 1, i32 undef>
-; CHECK-NEXT: [[TMP4:%.*]] = tail call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> [[TMP0:%.*]], <2 x double> [[TMP3]])
+; CHECK-NEXT: [[TMP4:%.*]] = fsub <2 x double> [[TMP0:%.*]], [[TMP3]]
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <2 x double> [[TMP4]], i32 0
; CHECK-NEXT: ret double [[TMP5]]
;
define double @elts_addsub_v2f64_sub(<2 x double> %0, <2 x double> %1) {
; CHECK-LABEL: @elts_addsub_v2f64_sub(
-; CHECK-NEXT: [[TMP3:%.*]] = tail call <2 x double> @llvm.x86.sse3.addsub.pd(<2 x double> [[TMP0:%.*]], <2 x double> [[TMP1:%.*]])
+; CHECK-NEXT: [[TMP3:%.*]] = fsub <2 x double> [[TMP0:%.*]], [[TMP1:%.*]]
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <2 x double> [[TMP3]], i32 0
; CHECK-NEXT: ret double [[TMP4]]
;
define float @elts_addsub_v4f32_add(<4 x float> %0, <4 x float> %1) {
; CHECK-LABEL: @elts_addsub_v4f32_add(
-; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x float> [[TMP0:%.*]], <4 x float> undef, <4 x i32> <i32 undef, i32 1, i32 undef, i32 1>
-; CHECK-NEXT: [[TMP4:%.*]] = shufflevector <4 x float> [[TMP1:%.*]], <4 x float> undef, <4 x i32> <i32 undef, i32 1, i32 undef, i32 1>
-; CHECK-NEXT: [[TMP5:%.*]] = tail call <4 x float> @llvm.x86.sse3.addsub.ps(<4 x float> [[TMP3]], <4 x float> [[TMP4]])
-; CHECK-NEXT: [[TMP6:%.*]] = extractelement <4 x float> [[TMP5]], i32 1
-; CHECK-NEXT: [[TMP7:%.*]] = extractelement <4 x float> [[TMP5]], i32 3
-; CHECK-NEXT: [[TMP8:%.*]] = fadd float [[TMP6]], [[TMP7]]
-; CHECK-NEXT: ret float [[TMP8]]
+; CHECK-NEXT: [[TMP3:%.*]] = fadd <4 x float> [[TMP0:%.*]], [[TMP1:%.*]]
+; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x float> [[TMP3]], i32 1
+; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x float> [[TMP3]], i32 1
+; CHECK-NEXT: [[TMP6:%.*]] = fadd float [[TMP4]], [[TMP5]]
+; CHECK-NEXT: ret float [[TMP6]]
;
%3 = shufflevector <4 x float> %0, <4 x float> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
%4 = shufflevector <4 x float> %1, <4 x float> undef, <4 x i32> <i32 0, i32 1, i32 0, i32 1>
define double @elts_addsub_v4f64_add(<4 x double> %0, <4 x double> %1) {
; CHECK-LABEL: @elts_addsub_v4f64_add(
-; CHECK-NEXT: [[TMP3:%.*]] = tail call <4 x double> @llvm.x86.avx.addsub.pd.256(<4 x double> [[TMP0:%.*]], <4 x double> [[TMP1:%.*]])
+; CHECK-NEXT: [[TMP3:%.*]] = fadd <4 x double> [[TMP0:%.*]], [[TMP1:%.*]]
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <4 x double> [[TMP3]], i32 1
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <4 x double> [[TMP3]], i32 3
; CHECK-NEXT: [[TMP6:%.*]] = fadd double [[TMP4]], [[TMP5]]
define float @elts_addsub_v8f32_sub(<8 x float> %0, <8 x float> %1) {
; CHECK-LABEL: @elts_addsub_v8f32_sub(
-; CHECK-NEXT: [[TMP3:%.*]] = tail call <8 x float> @llvm.x86.avx.addsub.ps.256(<8 x float> [[TMP0:%.*]], <8 x float> [[TMP1:%.*]])
+; CHECK-NEXT: [[TMP3:%.*]] = fsub <8 x float> [[TMP0:%.*]], [[TMP1:%.*]]
; CHECK-NEXT: [[TMP4:%.*]] = extractelement <8 x float> [[TMP3]], i32 0
; CHECK-NEXT: [[TMP5:%.*]] = extractelement <8 x float> [[TMP3]], i32 4
; CHECK-NEXT: [[TMP6:%.*]] = fadd float [[TMP4]], [[TMP5]]