1 ; RUN: opt < %s -cost-model -analyze -mtriple=systemz-unknown -mcpu=z13 | FileCheck %s 2 3 ; CHECK: shuffle 4 define void @shuffle() { 5 6 ;; Reverse shuffles 7 shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> 8 shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0> 9 10 shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> 11 shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0> 12 13 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 14 shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0> 15 16 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 17 shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0> 18 19 shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0> 20 shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0> 21 22 ;; Alternate shuffles 23 shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31> 24 shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 16, i32 1, i32 18, i32 3, i32 20, i32 5, i32 22, i32 7, i32 24, i32 9, i32 26, i32 11, i32 28, i32 13, i32 30, i32 15> 25 26 shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15> 27 shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7> 28 29 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 0, i32 5, i32 2, i32 7> 30 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 4, i32 1, i32 6, i32 3> 31 32 shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 0, i32 3> 33 shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 2, i32 1> 34 35 shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 0, i32 3> 36 shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 2, i32 1> 37 38 ;; Broadcast shuffles 39 shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> zeroinitializer 40 shufflevector <32 x i8> undef, <32 x i8> undef, <32 x i32> zeroinitializer 41 42 shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> zeroinitializer 43 shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> zeroinitializer 44 45 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> zeroinitializer 46 shufflevector <8 x i32> undef, <8 x i32> undef, <8 x i32> zeroinitializer 47 48 shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> zeroinitializer 49 shufflevector <4 x i64> undef, <4 x i64> undef, <4 x i32> zeroinitializer 50 51 shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> zeroinitializer 52 shufflevector <4 x double> undef, <4 x double> undef, <4 x i32> zeroinitializer 53 54 ;; Random shuffles 55 shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 4, i32 17, i32 2, i32 19, i32 0, i32 21, i32 8, i32 23, i32 6, i32 10, i32 10, i32 27, i32 29, i32 29, i32 14, i32 31> 56 shufflevector <18 x i8> undef, <18 x i8> undef, <18 x i32> <i32 4, i32 17, i32 2, i32 19, i32 0, i32 21, i32 8, i32 23, i32 6, i32 10, i32 10, i32 27, i32 29, i32 29, i32 14, i32 31, i32 0, i32 1> 57 58 shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 9, i32 9, i32 2, i32 2, i32 4, i32 13, i32 15, i32 15> 59 shufflevector <12 x i16> undef, <12 x i16> undef, <12 x i32> <i32 9, i32 9, i32 2, i32 2, i32 4, i32 13, i32 15, i32 15, i32 9, i32 2, i32 2, i32 4> 60 61 shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 4, i32 7> 62 shufflevector <6 x i32> undef, <6 x i32> undef, <6 x i32> <i32 0, i32 0, i32 4, i32 7, i32 4, i32 7> 63 64 shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 2> 65 shufflevector <4 x i64> undef, <4 x i64> undef, <4 x i32> <i32 1, i32 2, i32 0, i32 2> 66 67 shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 2, i32 1> 68 shufflevector <4 x double> undef, <4 x double> undef, <4 x i32> <i32 2, i32 1, i32 0, i32 2> 69 70 ret void 71 72 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %1 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> 73 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %2 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 0> 74 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %3 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0> 75 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %4 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 0> 76 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %5 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 77 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %6 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0> 78 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %7 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 79 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %8 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 0> 80 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %9 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 0> 81 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %10 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 1, i32 0> 82 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %11 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 17, i32 2, i32 19, i32 4, i32 21, i32 6, i32 23, i32 8, i32 25, i32 10, i32 27, i32 12, i32 29, i32 14, i32 31> 83 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %12 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 16, i32 1, i32 18, i32 3, i32 20, i32 5, i32 22, i32 7, i32 24, i32 9, i32 26, i32 11, i32 28, i32 13, i32 30, i32 15> 84 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %13 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 9, i32 2, i32 11, i32 4, i32 13, i32 6, i32 15> 85 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %14 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 8, i32 1, i32 10, i32 3, i32 12, i32 5, i32 14, i32 7> 86 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %15 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 0, i32 5, i32 2, i32 7> 87 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %16 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 4, i32 1, i32 6, i32 3> 88 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %17 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 0, i32 3> 89 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %18 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 2, i32 1> 90 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %19 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 0, i32 3> 91 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %20 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 2, i32 1> 92 ; CHECK: Cost Model: Found an estimated cost of 0 for instruction: %21 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> zeroinitializer 93 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %22 = shufflevector <32 x i8> undef, <32 x i8> undef, <32 x i32> zeroinitializer 94 ; CHECK: Cost Model: Found an estimated cost of 0 for instruction: %23 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> zeroinitializer 95 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %24 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> zeroinitializer 96 ; CHECK: Cost Model: Found an estimated cost of 0 for instruction: %25 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> zeroinitializer 97 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %26 = shufflevector <8 x i32> undef, <8 x i32> undef, <8 x i32> zeroinitializer 98 ; CHECK: Cost Model: Found an estimated cost of 0 for instruction: %27 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> zeroinitializer 99 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %28 = shufflevector <4 x i64> undef, <4 x i64> undef, <4 x i32> zeroinitializer 100 ; CHECK: Cost Model: Found an estimated cost of 0 for instruction: %29 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> zeroinitializer 101 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %30 = shufflevector <4 x double> undef, <4 x double> undef, <4 x i32> zeroinitializer 102 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %31 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 4, i32 17, i32 2, i32 19, i32 0, i32 21, i32 8, i32 23, i32 6, i32 10, i32 10, i32 27, i32 29, i32 29, i32 14, i32 31> 103 ; CHECK: Cost Model: Found an estimated cost of 2 for instruction: %32 = shufflevector <18 x i8> undef, <18 x i8> undef, <18 x i32> <i32 4, i32 17, i32 2, i32 19, i32 0, i32 21, i32 8, i32 23, i32 6, i32 10, i32 10, i32 27, i32 29, i32 29, i32 14, i32 31, i32 0, i32 1> 104 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %33 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 9, i32 9, i32 2, i32 2, i32 4, i32 13, i32 15, i32 15> 105 ; CHECK: Cost Model: Found an estimated cost of 2 for instruction: %34 = shufflevector <12 x i16> undef, <12 x i16> undef, <12 x i32> <i32 9, i32 9, i32 2, i32 2, i32 4, i32 13, i32 15, i32 15, i32 9, i32 2, i32 2, i32 4> 106 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %35 = shufflevector <4 x i32> undef, <4 x i32> undef, <4 x i32> <i32 0, i32 0, i32 4, i32 7> 107 ; CHECK: Cost Model: Found an estimated cost of 2 for instruction: %36 = shufflevector <6 x i32> undef, <6 x i32> undef, <6 x i32> <i32 0, i32 0, i32 4, i32 7, i32 4, i32 7> 108 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %37 = shufflevector <2 x i64> undef, <2 x i64> undef, <2 x i32> <i32 1, i32 2> 109 ; CHECK: Cost Model: Found an estimated cost of 2 for instruction: %38 = shufflevector <4 x i64> undef, <4 x i64> undef, <4 x i32> <i32 1, i32 2, i32 0, i32 2> 110 ; CHECK: Cost Model: Found an estimated cost of 1 for instruction: %39 = shufflevector <2 x double> undef, <2 x double> undef, <2 x i32> <i32 2, i32 1> 111 ; CHECK: Cost Model: Found an estimated cost of 2 for instruction: %40 = shufflevector <4 x double> undef, <4 x double> undef, <4 x i32> <i32 2, i32 1, i32 0, i32 2> 112 } 113
