1 ; RUN: llc < %s -march=x86-64 -mcpu=corei7 | FileCheck %s 2 3 define <2 x double> @ld(<2 x double> %p) nounwind optsize ssp { 4 ; CHECK: unpcklpd 5 %shuffle = shufflevector <2 x double> %p, <2 x double> undef, <2 x i32> zeroinitializer 6 ret <2 x double> %shuffle 7 } 8 9 define <2 x double> @hd(<2 x double> %p) nounwind optsize ssp { 10 ; CHECK: unpckhpd 11 %shuffle = shufflevector <2 x double> %p, <2 x double> undef, <2 x i32> <i32 1, i32 1> 12 ret <2 x double> %shuffle 13 } 14 15 define <2 x i64> @ldi(<2 x i64> %p) nounwind optsize ssp { 16 ; CHECK: punpcklqdq 17 %shuffle = shufflevector <2 x i64> %p, <2 x i64> undef, <2 x i32> zeroinitializer 18 ret <2 x i64> %shuffle 19 } 20 21 define <2 x i64> @hdi(<2 x i64> %p) nounwind optsize ssp { 22 ; CHECK: punpckhqdq 23 %shuffle = shufflevector <2 x i64> %p, <2 x i64> undef, <2 x i32> <i32 1, i32 1> 24 ret <2 x i64> %shuffle 25 } 26 27 ; rdar://10050549 28 %struct.Float2 = type { float, float } 29 30 define <4 x float> @loadhpi(%struct.Float2* %vPtr, <4 x float> %vecin1) nounwind readonly ssp { 31 entry: 32 ; CHECK: loadhpi 33 ; CHECK-NOT: movq 34 ; CHECK: movhps ( 35 %tmp1 = bitcast %struct.Float2* %vPtr to <1 x i64>* 36 %addptr7 = getelementptr inbounds <1 x i64>* %tmp1, i64 0 37 %tmp2 = bitcast <1 x i64>* %addptr7 to float* 38 %tmp3 = load float* %tmp2, align 4 39 %vec = insertelement <4 x float> undef, float %tmp3, i32 0 40 %addptr.i12 = getelementptr inbounds float* %tmp2, i64 1 41 %tmp4 = load float* %addptr.i12, align 4 42 %vecin2 = insertelement <4 x float> %vec, float %tmp4, i32 1 43 %shuffle = shufflevector <4 x float> %vecin1, <4 x float> %vecin2, <4 x i32> <i32 0, i32 1, i32 4, i32 5> 44 ret <4 x float> %shuffle 45 } 46 47 ; rdar://10119696 48 ; CHECK: f 49 define <4 x float> @f(<4 x float> %x, double* nocapture %y) nounwind readonly ssp { 50 entry: 51 ; CHECK: movlps (%{{rdi|rdx}}), %xmm0 52 %u110.i = load double* %y, align 1 53 %tmp8.i = insertelement <2 x double> undef, double %u110.i, i32 0 54 %tmp9.i = bitcast <2 x double> %tmp8.i to <4 x float> 55 %shuffle.i = shufflevector <4 x float> %x, <4 x float> %tmp9.i, <4 x i32> <i32 4, i32 5, i32 2, i32 3> 56 ret <4 x float> %shuffle.i 57 } 58 59 define <4 x float> @loadhpi2(%struct.Float2* nocapture %vHiCoefPtr_0, %struct.Float2* nocapture %vLoCoefPtr_0, i32 %s) nounwind readonly ssp { 60 entry: 61 ; CHECK: loadhpi2 62 ; CHECK: movhps ( 63 ; CHECK-NOT: movlhps 64 %0 = bitcast %struct.Float2* %vHiCoefPtr_0 to <1 x i64>* 65 %idx.ext = sext i32 %s to i64 66 %add.ptr = getelementptr inbounds <1 x i64>* %0, i64 %idx.ext 67 %add.ptr.val = load <1 x i64>* %add.ptr, align 1 68 %1 = bitcast <1 x i64> %add.ptr.val to <2 x float> 69 %shuffle.i = shufflevector <2 x float> %1, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef> 70 %2 = bitcast %struct.Float2* %vLoCoefPtr_0 to <1 x i64>* 71 %add.ptr2 = getelementptr inbounds <1 x i64>* %2, i64 %idx.ext 72 %add.ptr2.val = load <1 x i64>* %add.ptr2, align 1 73 %3 = bitcast <1 x i64> %add.ptr2.val to <2 x float> 74 %shuffle.i4 = shufflevector <2 x float> %3, <2 x float> undef, <4 x i32> <i32 0, i32 1, i32 undef, i32 undef> 75 %shuffle1.i5 = shufflevector <4 x float> %shuffle.i, <4 x float> %shuffle.i4, <4 x i32> <i32 0, i32 1, i32 4, i32 5> 76 ret <4 x float> %shuffle1.i5 77 } 78
