1 ; Test extensions of f32 to f64. 2 ; 3 ; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z10 \ 4 ; RUN: | FileCheck -check-prefix=CHECK -check-prefix=CHECK-SCALAR %s 5 ; RUN: llc < %s -mtriple=s390x-linux-gnu -mcpu=z13 | FileCheck %s 6 7 ; Check register extension. 8 define double @f1(float %val) { 9 ; CHECK-LABEL: f1: 10 ; CHECK: ldebr %f0, %f0 11 ; CHECK: br %r14 12 %res = fpext float %val to double 13 ret double %res 14 } 15 16 ; Check the low end of the LDEB range. 17 define double @f2(float *%ptr) { 18 ; CHECK-LABEL: f2: 19 ; CHECK: ldeb %f0, 0(%r2) 20 ; CHECK: br %r14 21 %val = load float , float *%ptr 22 %res = fpext float %val to double 23 ret double %res 24 } 25 26 ; Check the high end of the aligned LDEB range. 27 define double @f3(float *%base) { 28 ; CHECK-LABEL: f3: 29 ; CHECK: ldeb %f0, 4092(%r2) 30 ; CHECK: br %r14 31 %ptr = getelementptr float, float *%base, i64 1023 32 %val = load float , float *%ptr 33 %res = fpext float %val to double 34 ret double %res 35 } 36 37 ; Check the next word up, which needs separate address logic. 38 ; Other sequences besides this one would be OK. 39 define double @f4(float *%base) { 40 ; CHECK-LABEL: f4: 41 ; CHECK: aghi %r2, 4096 42 ; CHECK: ldeb %f0, 0(%r2) 43 ; CHECK: br %r14 44 %ptr = getelementptr float, float *%base, i64 1024 45 %val = load float , float *%ptr 46 %res = fpext float %val to double 47 ret double %res 48 } 49 50 ; Check negative displacements, which also need separate address logic. 51 define double @f5(float *%base) { 52 ; CHECK-LABEL: f5: 53 ; CHECK: aghi %r2, -4 54 ; CHECK: ldeb %f0, 0(%r2) 55 ; CHECK: br %r14 56 %ptr = getelementptr float, float *%base, i64 -1 57 %val = load float , float *%ptr 58 %res = fpext float %val to double 59 ret double %res 60 } 61 62 ; Check that LDEB allows indices. 63 define double @f6(float *%base, i64 %index) { 64 ; CHECK-LABEL: f6: 65 ; CHECK: sllg %r1, %r3, 2 66 ; CHECK: ldeb %f0, 400(%r1,%r2) 67 ; CHECK: br %r14 68 %ptr1 = getelementptr float, float *%base, i64 %index 69 %ptr2 = getelementptr float, float *%ptr1, i64 100 70 %val = load float , float *%ptr2 71 %res = fpext float %val to double 72 ret double %res 73 } 74 75 ; Test a case where we spill the source of at least one LDEBR. We want 76 ; to use LDEB if possible. 77 define void @f7(double *%ptr1, float *%ptr2) { 78 ; CHECK-LABEL: f7: 79 ; CHECK-SCALAR: ldeb {{%f[0-9]+}}, 16{{[04]}}(%r15) 80 ; CHECK: br %r14 81 %val0 = load volatile float , float *%ptr2 82 %val1 = load volatile float , float *%ptr2 83 %val2 = load volatile float , float *%ptr2 84 %val3 = load volatile float , float *%ptr2 85 %val4 = load volatile float , float *%ptr2 86 %val5 = load volatile float , float *%ptr2 87 %val6 = load volatile float , float *%ptr2 88 %val7 = load volatile float , float *%ptr2 89 %val8 = load volatile float , float *%ptr2 90 %val9 = load volatile float , float *%ptr2 91 %val10 = load volatile float , float *%ptr2 92 %val11 = load volatile float , float *%ptr2 93 %val12 = load volatile float , float *%ptr2 94 %val13 = load volatile float , float *%ptr2 95 %val14 = load volatile float , float *%ptr2 96 %val15 = load volatile float , float *%ptr2 97 %val16 = load volatile float , float *%ptr2 98 99 %ext0 = fpext float %val0 to double 100 %ext1 = fpext float %val1 to double 101 %ext2 = fpext float %val2 to double 102 %ext3 = fpext float %val3 to double 103 %ext4 = fpext float %val4 to double 104 %ext5 = fpext float %val5 to double 105 %ext6 = fpext float %val6 to double 106 %ext7 = fpext float %val7 to double 107 %ext8 = fpext float %val8 to double 108 %ext9 = fpext float %val9 to double 109 %ext10 = fpext float %val10 to double 110 %ext11 = fpext float %val11 to double 111 %ext12 = fpext float %val12 to double 112 %ext13 = fpext float %val13 to double 113 %ext14 = fpext float %val14 to double 114 %ext15 = fpext float %val15 to double 115 %ext16 = fpext float %val16 to double 116 117 store volatile float %val0, float *%ptr2 118 store volatile float %val1, float *%ptr2 119 store volatile float %val2, float *%ptr2 120 store volatile float %val3, float *%ptr2 121 store volatile float %val4, float *%ptr2 122 store volatile float %val5, float *%ptr2 123 store volatile float %val6, float *%ptr2 124 store volatile float %val7, float *%ptr2 125 store volatile float %val8, float *%ptr2 126 store volatile float %val9, float *%ptr2 127 store volatile float %val10, float *%ptr2 128 store volatile float %val11, float *%ptr2 129 store volatile float %val12, float *%ptr2 130 store volatile float %val13, float *%ptr2 131 store volatile float %val14, float *%ptr2 132 store volatile float %val15, float *%ptr2 133 store volatile float %val16, float *%ptr2 134 135 store volatile double %ext0, double *%ptr1 136 store volatile double %ext1, double *%ptr1 137 store volatile double %ext2, double *%ptr1 138 store volatile double %ext3, double *%ptr1 139 store volatile double %ext4, double *%ptr1 140 store volatile double %ext5, double *%ptr1 141 store volatile double %ext6, double *%ptr1 142 store volatile double %ext7, double *%ptr1 143 store volatile double %ext8, double *%ptr1 144 store volatile double %ext9, double *%ptr1 145 store volatile double %ext10, double *%ptr1 146 store volatile double %ext11, double *%ptr1 147 store volatile double %ext12, double *%ptr1 148 store volatile double %ext13, double *%ptr1 149 store volatile double %ext14, double *%ptr1 150 store volatile double %ext15, double *%ptr1 151 store volatile double %ext16, double *%ptr1 152 153 ret void 154 } 155
