1; RUN: opt < %s -S -passes='print<scalar-evolution>,loop-unroll<peeling;full-unroll-max=0>,print<scalar-evolution>' 2>&1 | FileCheck %s 2; 3; This test ensures that (extractvalue 0 (with-overflow-inst op0, op1)) 4; is invalidated by LoopPeel when the operands of with-overflow-inst 5; are changed. 6; 7; In the following case, LoopPeel modifies the CFG into another one 8; with %bb7 not dominating %bb2 and %bb3 although %extractvalue is 9; still the step for the %bb3 loop. %call has been modified and uses 10; different operands but the SCEV value for %extractvalue has not been 11; invalidated and still refers to %load in its SCEV operands 12; (SCEV(%extractvalue) := -2 + -2 * %load). 13; 14; When LoopUnroll tries to compute the SCEV for the %bb3 Phi, the 15; stale data for %extractvalue is used whereas %load is not available 16; in %bb3 which is wrong. 17; 18; for more details and nice pictures: https://github.com/llvm/llvm-project/issues/97586 19; 20; Although the reason for the bug was in forgetValue, it is still relevant to 21; test if LoopPeel invalidates %extractvalue after changing %call. 22; 23; forgetValue only walks the users, so calling it on the IV Phis does not 24; invalidate %extractvalue (thus forgetLoop does not invalidate it too). 25; It has to be done by LoopPeel itself. 26 27 28define void @loop_peeling_smul_with_overflow() { 29; before loop-unroll 30; CHECK: Classifying expressions for: @loop_peeling_smul_with_overflow 31; CHECK: %extractvalue = extractvalue { i32, i1 } %call, 0 32; CHECK-NEXT: --> (-2 + (-2 * %load)) 33; CHECK: %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ] 34; CHECK-NEXT: --> {0,+,(-2 + (-2 * %load))}<nuw><nsw><%bb3> 35; after loop-unroll 36; CHECK: Classifying expressions for: @loop_peeling_smul_with_overflow 37; CHECK: %extractvalue = extractvalue { i32, i1 } %call, 0 38; CHECK-NEXT: --> (-2 * %add8.lcssa) 39; CHECK: %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ] 40; CHECK-NEXT: --> {0,+,(-2 * %add8.lcssa)}<nuw><nsw><%bb3> 41; 42bb: 43 br label %bb1 44 45bb1: ; preds = %bb3, %bb 46 %phi = phi i32 [ 0, %bb ], [ %phi4, %bb3 ] 47 br label %bb5 48 49bb2: ; preds = %bb7 50 %call = call { i32, i1 } @llvm.smul.with.overflow.i32(i32 %add8, i32 -2) 51 %extractvalue = extractvalue { i32, i1 } %call, 0 52 br label %bb3 53 54bb3: ; preds = %bb3, %bb2 55 %phi4 = phi i32 [ %add, %bb3 ], [ 0, %bb2 ] 56 %add = add i32 %extractvalue, %phi4 57 br i1 false, label %bb3, label %bb1 58 59bb5: ; preds = %bb7, %bb1 60 %phi6 = phi i32 [ 1, %bb1 ], [ 0, %bb7 ] 61 %icmp = icmp eq i32 %phi, 0 62 br i1 %icmp, label %bb9, label %bb7 63 64bb7: ; preds = %bb5 65 %load = load i32, ptr addrspace(1) null, align 4 66 %add8 = add i32 %load, 1 67 br i1 false, label %bb2, label %bb5 68 69bb9: ; preds = %bb5 70 ret void 71} 72