1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt -passes="ipsccp<func-spec>" -funcspec-for-literal-constant=false -funcspec-max-clones=0 -funcspec-min-function-size=14 -S < %s | FileCheck %s --check-prefix=NONE 3; RUN: opt -passes="ipsccp<func-spec>" -funcspec-for-literal-constant=false -funcspec-max-clones=1 -funcspec-min-function-size=14 -S < %s | FileCheck %s --check-prefix=ONE 4; RUN: opt -passes="ipsccp<func-spec>" -funcspec-for-literal-constant=false -funcspec-max-clones=2 -funcspec-min-function-size=14 -S < %s | FileCheck %s --check-prefix=TWO 5; RUN: opt -passes="ipsccp<func-spec>" -funcspec-for-literal-constant=false -funcspec-max-clones=3 -funcspec-min-function-size=14 -S < %s | FileCheck %s --check-prefix=THREE 6 7; Make sure that we iterate correctly after sorting the specializations: 8; FnSpecialization: Specializations for function compute 9; FnSpecialization: Gain = 608 10; FnSpecialization: FormalArg = binop1, ActualArg = power 11; FnSpecialization: FormalArg = binop2, ActualArg = mul 12; FnSpecialization: Gain = 982 13; FnSpecialization: FormalArg = binop1, ActualArg = plus 14; FnSpecialization: FormalArg = binop2, ActualArg = minus 15; FnSpecialization: Gain = 795 16; FnSpecialization: FormalArg = binop1, ActualArg = minus 17; FnSpecialization: FormalArg = binop2, ActualArg = power 18 19define i64 @main(i64 %x, i64 %y, i1 %flag) { 20; NONE-LABEL: @main( 21; NONE-NEXT: entry: 22; NONE-NEXT: br i1 [[FLAG:%.*]], label [[PLUS:%.*]], label [[MINUS:%.*]] 23; NONE: plus: 24; NONE-NEXT: [[TMP0:%.*]] = call i64 @compute(i64 [[X:%.*]], i64 [[Y:%.*]], ptr @power, ptr @mul) 25; NONE-NEXT: br label [[MERGE:%.*]] 26; NONE: minus: 27; NONE-NEXT: [[TMP1:%.*]] = call i64 @compute(i64 [[X]], i64 [[Y]], ptr @plus, ptr @minus) 28; NONE-NEXT: br label [[MERGE]] 29; NONE: merge: 30; NONE-NEXT: [[TMP2:%.*]] = phi i64 [ [[TMP0]], [[PLUS]] ], [ [[TMP1]], [[MINUS]] ] 31; NONE-NEXT: [[TMP3:%.*]] = call i64 @compute(i64 [[TMP2]], i64 42, ptr @minus, ptr @power) 32; NONE-NEXT: ret i64 [[TMP3]] 33; 34; ONE-LABEL: @main( 35; ONE-NEXT: entry: 36; ONE-NEXT: br i1 [[FLAG:%.*]], label [[PLUS:%.*]], label [[MINUS:%.*]] 37; ONE: plus: 38; ONE-NEXT: [[TMP0:%.*]] = call i64 @compute(i64 [[X:%.*]], i64 [[Y:%.*]], ptr @power, ptr @mul) 39; ONE-NEXT: br label [[MERGE:%.*]] 40; ONE: minus: 41; ONE-NEXT: [[TMP1:%.*]] = call i64 @compute.specialized.1(i64 [[X]], i64 [[Y]], ptr @plus, ptr @minus) 42; ONE-NEXT: br label [[MERGE]] 43; ONE: merge: 44; ONE-NEXT: [[TMP2:%.*]] = phi i64 [ [[TMP0]], [[PLUS]] ], [ [[TMP1]], [[MINUS]] ] 45; ONE-NEXT: [[TMP3:%.*]] = call i64 @compute(i64 [[TMP2]], i64 42, ptr @minus, ptr @power) 46; ONE-NEXT: ret i64 [[TMP3]] 47; 48; TWO-LABEL: @main( 49; TWO-NEXT: entry: 50; TWO-NEXT: br i1 [[FLAG:%.*]], label [[PLUS:%.*]], label [[MINUS:%.*]] 51; TWO: plus: 52; TWO-NEXT: [[TMP0:%.*]] = call i64 @compute(i64 [[X:%.*]], i64 [[Y:%.*]], ptr @power, ptr @mul) 53; TWO-NEXT: br label [[MERGE:%.*]] 54; TWO: minus: 55; TWO-NEXT: [[TMP1:%.*]] = call i64 @compute.specialized.2(i64 [[X]], i64 [[Y]], ptr @plus, ptr @minus) 56; TWO-NEXT: br label [[MERGE]] 57; TWO: merge: 58; TWO-NEXT: [[TMP2:%.*]] = phi i64 [ [[TMP0]], [[PLUS]] ], [ [[TMP1]], [[MINUS]] ] 59; TWO-NEXT: [[TMP3:%.*]] = call i64 @compute.specialized.1(i64 [[TMP2]], i64 42, ptr @minus, ptr @power) 60; TWO-NEXT: ret i64 [[TMP3]] 61; 62; THREE-LABEL: @main( 63; THREE-NEXT: entry: 64; THREE-NEXT: br i1 [[FLAG:%.*]], label [[PLUS:%.*]], label [[MINUS:%.*]] 65; THREE: plus: 66; THREE-NEXT: [[TMP0:%.*]] = call i64 @compute.specialized.1(i64 [[X:%.*]], i64 [[Y:%.*]], ptr @power, ptr @mul) 67; THREE-NEXT: br label [[MERGE:%.*]] 68; THREE: minus: 69; THREE-NEXT: [[TMP1:%.*]] = call i64 @compute.specialized.2(i64 [[X]], i64 [[Y]], ptr @plus, ptr @minus) 70; THREE-NEXT: br label [[MERGE]] 71; THREE: merge: 72; THREE-NEXT: [[TMP2:%.*]] = phi i64 [ [[TMP0]], [[PLUS]] ], [ [[TMP1]], [[MINUS]] ] 73; THREE-NEXT: [[TMP3:%.*]] = call i64 @compute.specialized.3(i64 [[TMP2]], i64 42, ptr @minus, ptr @power) 74; THREE-NEXT: ret i64 [[TMP3]] 75; 76entry: 77 br i1 %flag, label %plus, label %minus 78 79plus: 80 %tmp0 = call i64 @compute(i64 %x, i64 %y, ptr @power, ptr @mul) 81 br label %merge 82 83minus: 84 %tmp1 = call i64 @compute(i64 %x, i64 %y, ptr @plus, ptr @minus) 85 br label %merge 86 87merge: 88 %tmp2 = phi i64 [ %tmp0, %plus ], [ %tmp1, %minus] 89 %tmp3 = call i64 @compute(i64 %tmp2, i64 42, ptr @minus, ptr @power) 90 ret i64 %tmp3 91} 92 93; THREE-NOT: define internal i64 @compute 94; 95; THREE-LABEL: define internal i64 @compute.specialized.1(i64 %x, i64 %y, ptr %binop1, ptr %binop2) { 96; THREE-NEXT: entry: 97; THREE-NEXT: [[TMP0:%.+]] = call i64 @power(i64 %x, i64 %y) 98; THREE-NEXT: [[TMP1:%.+]] = call i64 @mul(i64 %x, i64 %y) 99; THREE-NEXT: [[TMP2:%.+]] = add i64 [[TMP0]], [[TMP1]] 100; THREE-NEXT: [[TMP3:%.+]] = sdiv i64 [[TMP2]], %x 101; THREE-NEXT: [[TMP4:%.+]] = sub i64 [[TMP3]], %y 102; THREE-NEXT: [[TMP5:%.+]] = mul i64 [[TMP4]], 2 103; THREE-NEXT: ret i64 [[TMP5]] 104; THREE-NEXT: } 105; 106; THREE-LABEL: define internal i64 @compute.specialized.2(i64 %x, i64 %y, ptr %binop1, ptr %binop2) { 107; THREE-NEXT: entry: 108; THREE-NEXT: [[TMP0:%.+]] = call i64 @plus(i64 %x, i64 %y) 109; THREE-NEXT: [[TMP1:%.+]] = call i64 @minus(i64 %x, i64 %y) 110; THREE-NEXT: [[TMP2:%.+]] = add i64 [[TMP0]], [[TMP1]] 111; THREE-NEXT: [[TMP3:%.+]] = sdiv i64 [[TMP2]], %x 112; THREE-NEXT: [[TMP4:%.+]] = sub i64 [[TMP3]], %y 113; THREE-NEXT: [[TMP5:%.+]] = mul i64 [[TMP4]], 2 114; THREE-NEXT: ret i64 [[TMP5]] 115; THREE-NEXT: } 116; 117; THREE-LABEL: define internal i64 @compute.specialized.3(i64 %x, i64 %y, ptr %binop1, ptr %binop2) { 118; THREE-NEXT: entry: 119; THREE-NEXT: [[TMP0:%.+]] = call i64 @minus(i64 %x, i64 %y) 120; THREE-NEXT: [[TMP1:%.+]] = call i64 @power(i64 %x, i64 %y) 121; THREE-NEXT: [[TMP2:%.+]] = add i64 [[TMP0]], [[TMP1]] 122; THREE-NEXT: [[TMP3:%.+]] = sdiv i64 [[TMP2]], %x 123; THREE-NEXT: [[TMP4:%.+]] = sub i64 [[TMP3]], %y 124; THREE-NEXT: [[TMP5:%.+]] = mul i64 [[TMP4]], 2 125; THREE-NEXT: ret i64 [[TMP5]] 126; THREE-NEXT: } 127; 128define internal i64 @compute(i64 %x, i64 %y, ptr %binop1, ptr %binop2) { 129entry: 130 %tmp0 = call i64 %binop1(i64 %x, i64 %y) 131 %tmp1 = call i64 %binop2(i64 %x, i64 %y) 132 %add = add i64 %tmp0, %tmp1 133 %div = sdiv i64 %add, %x 134 %sub = sub i64 %div, %y 135 %mul = mul i64 %sub, 2 136 ret i64 %mul 137} 138 139define internal i64 @plus(i64 %x, i64 %y) { 140entry: 141 %tmp0 = add i64 %x, %y 142 ret i64 %tmp0 143} 144 145define internal i64 @minus(i64 %x, i64 %y) { 146entry: 147 %tmp0 = sub i64 %x, %y 148 ret i64 %tmp0 149} 150 151define internal i64 @mul(i64 %x, i64 %n) { 152entry: 153 %cmp6 = icmp sgt i64 %n, 1 154 br i1 %cmp6, label %for.body, label %for.cond.cleanup 155 156for.cond.cleanup: ; preds = %for.body, %entry 157 %x.addr.0.lcssa = phi i64 [ %x, %entry ], [ %add, %for.body ] 158 ret i64 %x.addr.0.lcssa 159 160for.body: ; preds = %entry, %for.body 161 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 1, %entry ] 162 %x.addr.07 = phi i64 [ %add, %for.body ], [ %x, %entry ] 163 %add = shl nsw i64 %x.addr.07, 1 164 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 165 %exitcond.not = icmp eq i64 %indvars.iv.next, %n 166 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body 167} 168 169define internal i64 @power(i64 %x, i64 %n) { 170entry: 171 %cmp6 = icmp sgt i64 %n, 1 172 br i1 %cmp6, label %for.body, label %for.cond.cleanup 173 174for.cond.cleanup: ; preds = %for.body, %entry 175 %x.addr.0.lcssa = phi i64 [ %x, %entry ], [ %mul, %for.body ] 176 ret i64 %x.addr.0.lcssa 177 178for.body: ; preds = %entry, %for.body 179 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 1, %entry ] 180 %x.addr.07 = phi i64 [ %mul, %for.body ], [ %x, %entry ] 181 %mul = mul nsw i64 %x.addr.07, %x.addr.07 182 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1 183 %exitcond.not = icmp eq i64 %indvars.iv.next, %n 184 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body 185} 186