1 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:64:128-a0:0:64-n32-S64" 2 target triple = "armv7-none-linux-gnueabi" 3 4 define internal <4 x float> @smear_4f(float %in) nounwind readnone alwaysinline { 5 %1 = insertelement <4 x float> undef, float %in, i32 0 6 %2 = insertelement <4 x float> %1, float %in, i32 1 7 %3 = insertelement <4 x float> %2, float %in, i32 2 8 %4 = insertelement <4 x float> %3, float %in, i32 3 9 ret <4 x float> %4 10 } 11 12 define internal <2 x float> @smear_2f(float %in) nounwind readnone alwaysinline { 13 %1 = insertelement <2 x float> undef, float %in, i32 0 14 %2 = insertelement <2 x float> %1, float %in, i32 1 15 ret <2 x float> %2 16 } 17 18 declare <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float>, <2 x float>) nounwind readnone 19 declare <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float>, <4 x float>) nounwind readnone 20 declare <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float>, <2 x float>) nounwind readnone 21 declare <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float>, <4 x float>) nounwind readnone 22 23 define <4 x float> @_Z5clampDv4_fS_S_(<4 x float> %value, <4 x float> %low, <4 x float> %high) nounwind readonly { 24 %1 = tail call <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float> %value, <4 x float> %high) nounwind readnone 25 %2 = tail call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> %1, <4 x float> %low) nounwind readnone 26 ret <4 x float> %2 27 } 28 29 define <4 x float> @_Z5clampDv4_fff(<4 x float> %value, float %low, float %high) nounwind readonly { 30 %_high = tail call <4 x float> @smear_4f(float %high) nounwind readnone 31 %_low = tail call <4 x float> @smear_4f(float %low) nounwind readnone 32 %out = tail call <4 x float> @_Z5clampDv4_fS_S_(<4 x float> %value, <4 x float> %_low, <4 x float> %_high) nounwind readonly 33 ret <4 x float> %out 34 } 35 36 define <3 x float> @_Z5clampDv3_fS_S_(<3 x float> %value, <3 x float> %low, <3 x float> %high) nounwind readonly { 37 %_value = shufflevector <3 x float> %value, <3 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 38 %_low = shufflevector <3 x float> %low, <3 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 39 %_high = shufflevector <3 x float> %high, <3 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 40 %a = tail call <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float> %_value, <4 x float> %_high) nounwind readnone 41 %b = tail call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> %a, <4 x float> %_low) nounwind readnone 42 %c = shufflevector <4 x float> %b, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2> 43 ret <3 x float> %c 44 } 45 46 define <3 x float> @_Z5clampDv3_fff(<3 x float> %value, float %low, float %high) nounwind readonly { 47 %_value = shufflevector <3 x float> %value, <3 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3> 48 %_high = tail call <4 x float> @smear_4f(float %high) nounwind readnone 49 %_low = tail call <4 x float> @smear_4f(float %low) nounwind readnone 50 %a = tail call <4 x float> @llvm.arm.neon.vmins.v4f32(<4 x float> %_value, <4 x float> %_high) nounwind readnone 51 %b = tail call <4 x float> @llvm.arm.neon.vmaxs.v4f32(<4 x float> %a, <4 x float> %_low) nounwind readnone 52 %c = shufflevector <4 x float> %b, <4 x float> undef, <3 x i32> <i32 0, i32 1, i32 2> 53 ret <3 x float> %c 54 } 55 56 57 define <2 x float> @_Z5clampDv2_fS_S_(<2 x float> %value, <2 x float> %low, <2 x float> %high) nounwind readonly { 58 %1 = tail call <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float> %value, <2 x float> %high) nounwind readnone 59 %2 = tail call <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float> %1, <2 x float> %low) nounwind readnone 60 ret <2 x float> %2 61 } 62 63 define <2 x float> @_Z5clampDv2_fff(<2 x float> %value, float %low, float %high) nounwind readonly { 64 %_high = tail call <2 x float> @smear_2f(float %high) nounwind readnone 65 %_low = tail call <2 x float> @smear_2f(float %low) nounwind readnone 66 %a = tail call <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float> %value, <2 x float> %_high) nounwind readnone 67 %b = tail call <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float> %a, <2 x float> %_low) nounwind readnone 68 ret <2 x float> %b 69 } 70 71 72 define float @_Z5clampfff(float %value, float %low, float %high) nounwind readonly { 73 %_value = tail call <2 x float> @smear_2f(float %value) nounwind readnone 74 %_low = tail call <2 x float> @smear_2f(float %low) nounwind readnone 75 %_high = tail call <2 x float> @smear_2f(float %high) nounwind readnone 76 %a = tail call <2 x float> @llvm.arm.neon.vmins.v2f32(<2 x float> %_value, <2 x float> %_high) nounwind readnone 77 %b = tail call <2 x float> @llvm.arm.neon.vmaxs.v2f32(<2 x float> %a, <2 x float> %_low) nounwind readnone 78 %c = extractelement <2 x float> %b, i32 0 79 ret float %c 80 } 81 82
