Move Assertions #306

tgymnich · 2021-08-19T07:03:23Z

invertPointerM uses lookupM...

wsmoses · 2021-09-02T01:00:46Z

enzyme/Enzyme/GradientUtils.cpp

    // analysis honeypot no longer exists
-    return lookupM(getNewFromOriginal(oval), BuilderM);
+
+    switch (mode) {


I think the better solution here would be to have invertPointerM not also call lookup, but any code outside that calls invertpointerM that requires it explicitly call lookup. Thus invertpointer wouldn't need this.

tgymnich · 2021-09-13T23:18:42Z

I managed to move lookupM out of invertPointerM but it now appears at call sites of invertPointerM like setPtrDiffe, addToInvertedPtrDiffe and unwrapM

wsmoses · 2021-09-14T01:59:02Z

enzyme/Enzyme/GradientUtils.cpp

+
+        switch (mode) {
+        case DerivativeMode::ForwardMode:
+          pidx = invertPointerM(dli->getOperand(0), BuilderM);


This should never need a lookup since the only time this inverted phi happens is adjacent to the original value in the fwd pass.

Are you sure?
Removing this leads to 7 failed integration tests:

Failed Tests (7): Enzyme :: Integration/ReverseMode/eigensumsq.cpp Enzyme :: Integration/ReverseMode/eigensumsqdyn-notmp.cpp Enzyme :: Integration/ReverseMode/eigensumsqdyn.cpp Enzyme :: Integration/ReverseMode/mixedstruct1-simplefda.c Enzyme :: Integration/ReverseMode/mixedstruct1-simpleps.c Enzyme :: Integration/ReverseMode/simpleeigen.cpp Enzyme :: Integration/ReverseMode/vecmax.cpp

Instruction does not dominate all uses! %"'ipg18" = getelementptr inbounds %"class.std::vector", %"class.std::vector"* %"'", i64 0, i32 0, i32 0, i32 0, i32 0 %"'il_phi8_unwrap" = load double*, double** %"'ipg18", align 8, !tbaa !12

Also debugging Integration/ReverseMode/mixedstruct1-simplefda.c, I found that it's using this lookup: (GradientUtils:2739 on main)

{ auto ifound = invertedPointers.find(oval); if (ifound != invertedPointers.end()) { return lookupM(&*ifound->second, BuilderM); } }

Yeah that lookup should also be removed from invertpointer.

Can you post the full log for one of those cases? Specifically seeing how the domination errs might reveal what's going awry

FAIL: Enzyme :: Integration/ReverseMode/eigensumsqdyn.cpp (9 of 65) ******************** TEST 'Enzyme :: Integration/ReverseMode/eigensumsqdyn.cpp' FAILED ******************** Script: -- : 'RUN: at line 1'; /usr/lib/llvm-12/bin/clang++ -mllvm -force-vector-width=1 -ffast-math -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 2'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 3'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 5'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 -Xclang -disable-llvm-optzns /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 6'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 7'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 8'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - -- Exit Code: 255 Command Output (stderr): -- freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !17 freeing without malloc %41 = load i8*, i8** %40, align 8, !tbaa !12 freeing without malloc %51 = load i8*, i8** %50, align 8, !tbaa !22 freeing without malloc %48 = load i8*, i8** %47, align 8, !tbaa !21 freeing without malloc %76 = select i1 %75, double* %74, double* null freeing without malloc %54 = select i1 %53, double* %52, double* null freeing without malloc %38 = load i8*, i8** %37, align 8, !tbaa !14 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !17 freeing without malloc %41 = load i8*, i8** %40, align 8, !tbaa !12 freeing without malloc %51 = load i8*, i8** %50, align 8, !tbaa !22 freeing without malloc %48 = load i8*, i8** %47, align 8, !tbaa !21 freeing without malloc %76 = select i1 %75, double* %74, double* null freeing without malloc %54 = select i1 %53, double* %52, double* null freeing without malloc %38 = load i8*, i8** %37, align 8, !tbaa !14 freeing without malloc %119 = load i8*, i8** %118, align 8, !tbaa !9 freeing without malloc %123 = phi i8* [ %121, %115 ], [ %45, %61 ] Instruction does not dominate all uses! %"'ipg67" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'il_phi25_unwrap" = load double*, double** %"'ipg67", align 8, !tbaa !44 ; Function Attrs: noinline nounwind uwtable define internal double @preprocess__ZL6matvecPKN5Eigen6MatrixIdLin1ELin1ELi0ELin1ELin1EEES3_(%"class.Eigen::Matrix"* noalias %0, %"class.Eigen::Matrix"* noalias %1) #4 { %3 = alloca %"class.Eigen::internal::redux_evaluator", align 8 %4 = alloca %"class.Eigen::Matrix", align 8 %5 = alloca %"class.Eigen::Product", align 8 %6 = bitcast %"class.Eigen::Matrix"* %4 to i8* call void @llvm.lifetime.start.p0i8(i64 24, i8* nonnull %6) #19 %7 = bitcast %"class.Eigen::Matrix"* %4 to i64* store i64 0, i64* %7, align 8 %8 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 1 %9 = load i64, i64* %8, align 8, !tbaa !36 %10 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 2 %11 = load i64, i64* %10, align 8, !tbaa !42 %12 = icmp eq i64 %9, 0 %13 = icmp eq i64 %11, 0 %14 = or i1 %12, %13 br i1 %14, label %25, label %15 15: ; preds = %2 %16 = sdiv i64 9223372036854775807, %11 %17 = icmp slt i64 %16, %9 br i1 %17, label %18, label %20 18: ; preds = %15 %19 = tail call noalias nonnull i8* @_Znwm(i64 -1) #20 br label %20 20: ; preds = %18, %15 %21 = sdiv i64 9223372036854775807, %11 %22 = icmp slt i64 %21, %9 br i1 %22, label %23, label %25 23: ; preds = %20 %24 = tail call noalias nonnull i8* @_Znwm(i64 -1) #20 br label %25 25: ; preds = %23, %20, %2 %26 = mul i64 %11, %9 %27 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %4, i64 0, i32 0, i32 0, i32 1 %28 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %4, i64 0, i32 0, i32 0, i32 2 %29 = icmp eq i64 %26, 0 br i1 %29, label %44, label %30 30: ; preds = %25 %31 = bitcast %"class.Eigen::Matrix"* %4 to i8** %32 = icmp ugt i64 %26, 2305843009213693951 br i1 %32, label %33, label %35 33: ; preds = %30 %34 = tail call noalias nonnull i8* @_Znwm(i64 -1) #20 br label %35 35: ; preds = %33, %30 %36 = shl i64 %26, 3 %37 = tail call noalias i8* @malloc(i64 %36) #19 %38 = icmp ne i64 %36, 0 %39 = and i1 %38, false br i1 %39, label %40, label %42 40: ; preds = %35 %41 = tail call noalias nonnull i8* @_Znwm(i64 -1) #20 br label %42 42: ; preds = %40, %35 store i8* %37, i8** %31, align 8, !tbaa !33 %43 = bitcast i8* %37 to double* br label %44 44: ; preds = %42, %25 %45 = phi i8* [ null, %25 ], [ %37, %42 ] %46 = phi double* [ null, %25 ], [ %43, %42 ] store i64 %9, i64* %27, align 8, !tbaa !36 store i64 %11, i64* %28, align 8, !tbaa !42 %47 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %0, i64 0, i32 0, i32 0, i32 0 %48 = load double*, double** %47, align 8, !tbaa !33 %49 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 0 %50 = load double*, double** %49, align 8, !tbaa !33 %51 = icmp sgt i64 %26, 0 br i1 %51, label %.preheader2, label %61 .preheader2: ; preds = %44 br label %52 52: ; preds = %.preheader2, %52 %tiv6 = phi i64 [ 0, %.preheader2 ], [ %tiv.next7, %52 ] %tiv.next7 = add nuw nsw i64 %tiv6, 1 %53 = getelementptr inbounds double, double* %46, i64 %tiv6 %54 = getelementptr inbounds double, double* %48, i64 %tiv6 %55 = getelementptr inbounds double, double* %50, i64 %tiv6 %56 = load double, double* %54, align 8, !tbaa !37 %57 = load double, double* %55, align 8, !tbaa !37 %58 = fsub fast double %56, %57 store double %58, double* %53, align 8, !tbaa !37 %59 = add nuw nsw i64 %tiv6, 1 %60 = icmp eq i64 %59, %26 br i1 %60, label %.loopexit3, label %52, !llvm.loop !43 .loopexit3: ; preds = %52 br label %61 61: ; preds = %.loopexit3, %44 %62 = bitcast %"class.Eigen::Product"* %5 to i8* call void @llvm.lifetime.start.p0i8(i64 16, i8* nonnull %62) #19 %63 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %5, i64 0, i32 0 store %"class.Eigen::Matrix"* %4, %"class.Eigen::Matrix"** %63, align 8 %64 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %5, i64 0, i32 1 store %"class.Eigen::Matrix"* %4, %"class.Eigen::Matrix"** %64, align 8 %65 = bitcast %"class.Eigen::Matrix"** %64 to %"class.Eigen::PlainObjectBase"** br i1 %29, label %122, label %66 66: ; preds = %61 %67 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %68 = bitcast %"class.Eigen::internal::redux_evaluator"* %3 to i8* call void @llvm.lifetime.start.p0i8(i64 48, i8* nonnull %68) #19 %69 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 call void @_ZN5Eigen8internal17product_evaluatorINS_7ProductINS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES4_Li0EEELi8ENS_10DenseShapeES6_ddEC2ERKS5_(%"struct.Eigen::internal::product_evaluator"* nonnull dereferenceable(40) %69, %"class.Eigen::Product"* nonnull align 8 dereferenceable(16) %5) #19 %70 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 1 store %"class.Eigen::Product"* %5, %"class.Eigen::Product"** %70, align 8, !tbaa !38 %71 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %72 = load double*, double** %71, align 8, !tbaa !44 %73 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %74 = load double, double* %72, align 8, !tbaa !37 %75 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %67, align 8, !tbaa !46 %76 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %75, i64 0, i32 0, i32 1 %77 = load i64, i64* %76, align 8, !tbaa !36 %78 = icmp sgt i64 %77, 1 br i1 %78, label %.preheader, label %79 .preheader: ; preds = %66 br label %107 .loopexit1: ; preds = %107 br label %79 79: ; preds = %.loopexit1, %66 %80 = phi double [ %74, %66 ], [ %112, %.loopexit1 ] %81 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %65, align 8, !tbaa !48 %82 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %81, i64 0, i32 0, i32 2 %83 = load i64, i64* %82, align 8, !tbaa !42 %84 = icmp sgt i64 %83, 1 %85 = icmp sgt i64 %77, 0 %86 = load i64, i64* %73, align 8 %87 = and i1 %85, %84 br i1 %87, label %88, label %115 88: ; preds = %79 %89 = icmp sgt i64 %83, 2 %90 = select i1 %89, i64 %83, i64 2 br label %91 91: ; preds = %104, %88 %tiv = phi i64 [ %tiv.next, %104 ], [ 0, %88 ] %92 = phi double [ %101, %104 ], [ %80, %88 ] %93 = add i64 %tiv, 1 %tiv.next = add nuw nsw i64 %tiv, 1 %94 = mul nsw i64 %93, %86 br label %95 95: ; preds = %95, %91 %96 = phi i64 [ 0, %91 ], [ %102, %95 ] %97 = phi double [ %92, %91 ], [ %101, %95 ] %98 = add nsw i64 %96, %94 %99 = getelementptr inbounds double, double* %72, i64 %98 %100 = load double, double* %99, align 8, !tbaa !37 %101 = fadd fast double %100, %97 %102 = add nuw nsw i64 %96, 1 %103 = icmp eq i64 %102, %77 br i1 %103, label %104, label %95, !llvm.loop !49 104: ; preds = %95 %105 = add nuw nsw i64 %93, 1 %106 = icmp eq i64 %105, %90 br i1 %106, label %.loopexit, label %91, !llvm.loop !50 107: ; preds = %.preheader, %107 %tiv4 = phi i64 [ 0, %.preheader ], [ %tiv.next5, %107 ] %108 = phi double [ %112, %107 ], [ %74, %.preheader ] %109 = add i64 %tiv4, 1 %tiv.next5 = add nuw nsw i64 %tiv4, 1 %110 = getelementptr inbounds double, double* %72, i64 %109 %111 = load double, double* %110, align 8, !tbaa !37 %112 = fadd fast double %111, %108 %113 = add nuw nsw i64 %109, 1 %114 = icmp eq i64 %113, %77 br i1 %114, label %.loopexit1, label %107, !llvm.loop !51 .loopexit: ; preds = %104 br label %115 115: ; preds = %.loopexit, %79 %116 = phi double [ %80, %79 ], [ %101, %.loopexit ] %117 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 1, i32 0 %118 = bitcast %"class.Eigen::PlainObjectBase"* %117 to i8** %119 = load i8*, i8** %118, align 8, !tbaa !33 call void @free(i8* %119) #19 call void @llvm.lifetime.end.p0i8(i64 48, i8* nonnull %68) #19 %120 = bitcast %"class.Eigen::Matrix"* %4 to i8** %121 = load i8*, i8** %120, align 8, !tbaa !33 br label %122 122: ; preds = %115, %61 %123 = phi i8* [ %121, %115 ], [ %45, %61 ] %124 = phi double [ %116, %115 ], [ 0.000000e+00, %61 ] call void @llvm.lifetime.end.p0i8(i64 16, i8* nonnull %62) #19 call void @free(i8* %123) #19 call void @llvm.lifetime.end.p0i8(i64 24, i8* nonnull %6) #19 ret double %124 } ; Function Attrs: noinline nounwind uwtable define internal void @diffe_ZL6matvecPKN5Eigen6MatrixIdLin1ELin1ELi0ELin1ELin1EEES3_(%"class.Eigen::Matrix"* noalias %0, %"class.Eigen::Matrix"* %"'", %"class.Eigen::Matrix"* noalias %1, %"class.Eigen::Matrix"* %"'1", double %differeturn) #4 { %"iv'ac" = alloca i64, align 8 %"iv28'ac" = alloca i64, align 8 %"iv30'ac" = alloca i64, align 8 %"iv32'ac" = alloca i64, align 8 %"'de" = alloca double, align 8 store double 0.000000e+00, double* %"'de", align 8 %_cache = alloca i8*, align 8 %_cache37 = alloca i8*, align 8 %_cache40 = alloca i8*, align 8 %"'mi_cache" = alloca i8*, align 8 %_cache41 = alloca i8*, align 8 %_cache43 = alloca i8*, align 8 %"'de55" = alloca double, align 8 store double 0.000000e+00, double* %"'de55", align 8 %"'de56" = alloca double, align 8 store double 0.000000e+00, double* %"'de56", align 8 %"'de57" = alloca double, align 8 store double 0.000000e+00, double* %"'de57", align 8 %"'de65" = alloca double, align 8 store double 0.000000e+00, double* %"'de65", align 8 %_augmented_cache = alloca { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 }, align 8 %"'de82" = alloca double, align 8 store double 0.000000e+00, double* %"'de82", align 8 %"'de83" = alloca double, align 8 store double 0.000000e+00, double* %"'de83", align 8 %"'de91" = alloca double, align 8 store double 0.000000e+00, double* %"'de91", align 8 %"'de92" = alloca double, align 8 store double 0.000000e+00, double* %"'de92", align 8 %"'de95" = alloca double, align 8 store double 0.000000e+00, double* %"'de95", align 8 %"'de96" = alloca double, align 8 store double 0.000000e+00, double* %"'de96", align 8 %"'de112" = alloca double, align 8 store double 0.000000e+00, double* %"'de112", align 8 %"'de113" = alloca double, align 8 store double 0.000000e+00, double* %"'de113", align 8 %"'de133" = alloca double, align 8 store double 0.000000e+00, double* %"'de133", align 8 %"'ipa66" = alloca %"class.Eigen::internal::redux_evaluator", align 8 store %"class.Eigen::internal::redux_evaluator" zeroinitializer, %"class.Eigen::internal::redux_evaluator"* %"'ipa66", align 8 %3 = alloca %"class.Eigen::internal::redux_evaluator", align 8 %"'ipa" = alloca %"class.Eigen::Matrix", align 8 store %"class.Eigen::Matrix" zeroinitializer, %"class.Eigen::Matrix"* %"'ipa", align 8 %4 = alloca %"class.Eigen::Matrix", align 8 %"'ipa61" = alloca %"class.Eigen::Product", align 8 store %"class.Eigen::Product" zeroinitializer, %"class.Eigen::Product"* %"'ipa61", align 8 %5 = alloca %"class.Eigen::Product", align 8 %"'ipc" = bitcast %"class.Eigen::Matrix"* %"'ipa" to i64* %6 = bitcast %"class.Eigen::Matrix"* %4 to i64* store i64 0, i64* %"'ipc", align 8 store i64 0, i64* %6, align 8 %7 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 1 %8 = load i64, i64* %7, align 8, !tbaa !36 %9 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 2 %10 = load i64, i64* %9, align 8, !tbaa !42 %11 = icmp eq i64 %8, 0 %12 = icmp eq i64 %10, 0 %13 = or i1 %11, %12 br i1 %13, label %24, label %14 14: ; preds = %2 %15 = sdiv i64 9223372036854775807, %10 %16 = icmp slt i64 %15, %8 br i1 %16, label %17, label %19 17: ; preds = %14 %18 = tail call noalias nonnull i8* @_Znwm(i64 -1) #19 store i8* %18, i8** %_cache, align 8, !invariant.group !178 br label %19 19: ; preds = %17, %14 %20 = sdiv i64 9223372036854775807, %10 %21 = icmp slt i64 %20, %8 br i1 %21, label %22, label %24 22: ; preds = %19 %23 = tail call noalias nonnull i8* @_Znwm(i64 -1) #19 store i8* %23, i8** %_cache37, align 8, !invariant.group !179 br label %24 24: ; preds = %22, %19, %2 %25 = mul i64 %10, %8 %"'ipg51" = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %"'ipa", i64 0, i32 0, i32 0, i32 1 %26 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %4, i64 0, i32 0, i32 0, i32 1 %"'ipg50" = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %"'ipa", i64 0, i32 0, i32 0, i32 2 %27 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %4, i64 0, i32 0, i32 0, i32 2 %28 = icmp eq i64 %25, 0 br i1 %28, label %43, label %29 29: ; preds = %24 %"'ipc44" = bitcast %"class.Eigen::Matrix"* %"'ipa" to i8** %30 = bitcast %"class.Eigen::Matrix"* %4 to i8** %31 = icmp ugt i64 %25, 2305843009213693951 br i1 %31, label %32, label %34 32: ; preds = %29 %33 = tail call noalias nonnull i8* @_Znwm(i64 -1) #19 store i8* %33, i8** %_cache40, align 8, !invariant.group !180 br label %34 34: ; preds = %32, %29 %35 = shl i64 %25, 3 %36 = tail call noalias i8* @malloc(i64 %35) #20 %"'mi" = tail call noalias nonnull i8* @malloc(i64 %35) #20 store i8* %"'mi", i8** %"'mi_cache", align 8, !invariant.group !181 store i8* %36, i8** %_cache41, align 8, !invariant.group !182 call void @llvm.memset.p0i8.i64(i8* nonnull %"'mi", i8 0, i64 %35, i1 false) %37 = icmp ne i64 %35, 0 %38 = and i1 %37, false br i1 %38, label %39, label %41 39: ; preds = %34 %40 = tail call noalias nonnull i8* @_Znwm(i64 -1) #19 store i8* %40, i8** %_cache43, align 8, !invariant.group !183 br label %41 41: ; preds = %39, %34 store i8* %"'mi", i8** %"'ipc44", align 8 store i8* %36, i8** %30, align 8, !tbaa !33 %"'ipc53" = bitcast i8* %"'mi" to double* %42 = bitcast i8* %36 to double* br label %43 43: ; preds = %41, %24 %44 = phi double* [ null, %24 ], [ %"'ipc53", %41 ] %45 = phi double* [ null, %24 ], [ %42, %41 ] store i64 %8, i64* %"'ipg51", align 8 store i64 %8, i64* %26, align 8, !tbaa !36 store i64 %10, i64* %"'ipg50", align 8 store i64 %10, i64* %27, align 8, !tbaa !42 %"'ipg48" = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %"'", i64 0, i32 0, i32 0, i32 0 %46 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %0, i64 0, i32 0, i32 0, i32 0 %"'ipl49" = load double*, double** %"'ipg48", align 8 %47 = load double*, double** %46, align 8, !tbaa !33 %"'ipg" = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %"'1", i64 0, i32 0, i32 0, i32 0 %48 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 0 %"'ipl" = load double*, double** %"'ipg", align 8 %49 = load double*, double** %48, align 8, !tbaa !33 %50 = icmp sgt i64 %25, 0 br i1 %50, label %.preheader2, label %60 .preheader2: ; preds = %43 %51 = add i64 %25, -1 br label %52 52: ; preds = %52, %.preheader2 %iv = phi i64 [ %iv.next, %52 ], [ 0, %.preheader2 ] %iv.next = add nuw nsw i64 %iv, 1 %"'ipg54" = getelementptr inbounds double, double* %44, i64 %iv %53 = getelementptr inbounds double, double* %45, i64 %iv %"'ipg59" = getelementptr inbounds double, double* %"'ipl49", i64 %iv %54 = getelementptr inbounds double, double* %47, i64 %iv %"'ipg58" = getelementptr inbounds double, double* %"'ipl", i64 %iv %55 = getelementptr inbounds double, double* %49, i64 %iv %56 = load double, double* %54, align 8, !tbaa !37 %57 = load double, double* %55, align 8, !tbaa !37 %58 = fsub fast double %56, %57 store double %58, double* %53, align 8, !tbaa !37 %59 = icmp eq i64 %iv.next, %25 br i1 %59, label %.loopexit3, label %52, !llvm.loop !43 .loopexit3: ; preds = %52 br label %60 60: ; preds = %.loopexit3, %43 %"'ipg63" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa61", i64 0, i32 0 %61 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %5, i64 0, i32 0 store %"class.Eigen::Matrix"* %"'ipa", %"class.Eigen::Matrix"** %"'ipg63", align 8 store %"class.Eigen::Matrix"* %4, %"class.Eigen::Matrix"** %61, align 8 %"'ipg62" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa61", i64 0, i32 1 %62 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %5, i64 0, i32 1 store %"class.Eigen::Matrix"* %"'ipa", %"class.Eigen::Matrix"** %"'ipg62", align 8 store %"class.Eigen::Matrix"* %4, %"class.Eigen::Matrix"** %62, align 8 %63 = bitcast %"class.Eigen::Matrix"** %62 to %"class.Eigen::PlainObjectBase"** br i1 %28, label %99, label %64 64: ; preds = %60 %65 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %"'ipg74" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0 %66 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 %_augmented = call { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 } @augmented__ZN5Eigen8internal17product_evaluatorINS_7ProductINS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES4_Li0EEELi8ENS_10DenseShapeES6_ddEC2ERKS5_(%"struct.Eigen::internal::product_evaluator"* %66, %"struct.Eigen::internal::product_evaluator"* %"'ipg74", %"class.Eigen::Product"* %5, %"class.Eigen::Product"* %"'ipa61") %"'ipg72" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 1 %67 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 1 store %"class.Eigen::Product"* %"'ipa61", %"class.Eigen::Product"** %"'ipg72", align 8 store { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 } %_augmented, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 }* %_augmented_cache, align 8, !invariant.group !184 store %"class.Eigen::Product"* %5, %"class.Eigen::Product"** %67, align 8, !tbaa !38 %"'ipg67" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl69" = load double*, double** %"'ipg67", align 8, !invariant.group !185 %68 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %69 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %65, align 8, !tbaa !46, !invariant.group !186 %70 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %69, i64 0, i32 0, i32 1 %71 = load i64, i64* %70, align 8, !tbaa !36, !invariant.group !187 %72 = icmp sgt i64 %71, 1 br i1 %72, label %.preheader, label %74 .preheader: ; preds = %64 %73 = add i64 %71, -2 br label %95 .loopexit1: ; preds = %95 br label %74 74: ; preds = %.loopexit1, %64 %75 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %63, align 8, !tbaa !48, !invariant.group !188 %76 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %75, i64 0, i32 0, i32 2 %77 = load i64, i64* %76, align 8, !tbaa !42, !invariant.group !189 %78 = icmp sgt i64 %77, 1 %79 = icmp sgt i64 %71, 0 %80 = load i64, i64* %68, align 8, !invariant.group !190 %81 = and i1 %79, %78 br i1 %81, label %82, label %98 82: ; preds = %74 %83 = icmp sgt i64 %77, 2 %84 = select i1 %83, i64 %77, i64 2 %85 = add nsw i64 %84, -2 %86 = add i64 %71, -1 br label %87 87: ; preds = %92, %82 %iv28 = phi i64 [ %iv.next29, %92 ], [ 0, %82 ] %iv.next29 = add nuw nsw i64 %iv28, 1 %88 = mul nsw i64 %iv.next29, %80 br label %89 89: ; preds = %89, %87 %iv30 = phi i64 [ %iv.next31, %89 ], [ 0, %87 ] %iv.next31 = add nuw nsw i64 %iv30, 1 %90 = add nsw i64 %iv30, %88 %"'ipg97" = getelementptr inbounds double, double* %"'ipl69", i64 %90 %91 = icmp eq i64 %iv.next31, %71 br i1 %91, label %92, label %89, !llvm.loop !49 92: ; preds = %89 %93 = add nuw nsw i64 %iv.next29, 1 %94 = icmp eq i64 %93, %84 br i1 %94, label %.loopexit, label %87, !llvm.loop !50 95: ; preds = %95, %.preheader %iv32 = phi i64 [ %iv.next33, %95 ], [ 0, %.preheader ] %iv.next33 = add nuw nsw i64 %iv32, 1 %"'ipg114" = getelementptr inbounds double, double* %"'ipl69", i64 %iv.next33 %96 = add nuw nsw i64 %iv.next33, 1 %97 = icmp eq i64 %96, %71 br i1 %97, label %.loopexit1, label %95, !llvm.loop !51 .loopexit: ; preds = %92 br label %98 98: ; preds = %.loopexit, %74 br label %99 99: ; preds = %98, %60 br label %invert23 invert: ; preds = %invert6, %invert2 ret void invert2: ; preds = %invert4, %invert3 br label %invert invert3: ; preds = %invert4 %100 = load i8*, i8** %_cache, align 8, !invariant.group !178 tail call void @_ZdlPv(i8* %100) br label %invert2 invert4: ; preds = %staging, %invert5 %_unwrap = sdiv i64 9223372036854775807, %10 %_unwrap36 = icmp slt i64 %_unwrap, %8 br i1 %_unwrap36, label %invert3, label %invert2 invert5: ; preds = %staging %101 = load i8*, i8** %_cache37, align 8, !invariant.group !179 tail call void @_ZdlPv(i8* %101) br label %invert4 invert6: ; preds = %invert12, %invert7 %_unwrap38 = sdiv i64 9223372036854775807, %10 %_unwrap39 = icmp slt i64 %_unwrap38, %8 br i1 %13, label %invert, label %staging invert7: ; preds = %invert9, %invert8 br label %invert6 invert8: ; preds = %invert9 %102 = load i8*, i8** %_cache40, align 8, !invariant.group !180 tail call void @_ZdlPv(i8* %102) br label %invert7 invert9: ; preds = %invert11, %invert10 %103 = load i8*, i8** %"'mi_cache", align 8, !invariant.group !181 tail call void @free(i8* nonnull %103) %104 = load i8*, i8** %_cache41, align 8, !invariant.group !182 tail call void @free(i8* %104) %_unwrap42 = icmp ugt i64 %25, 2305843009213693951 br i1 %_unwrap42, label %invert8, label %invert7 invert10: ; preds = %invert11 %105 = load i8*, i8** %_cache43, align 8, !invariant.group !183 tail call void @_ZdlPv(i8* %105) br label %invert9 invert11: ; preds = %invert12 %_unwrap45 = shl i64 %25, 3 %_unwrap46 = icmp ne i64 %_unwrap45, 0 %_unwrap47 = and i1 %_unwrap46, false br i1 %_unwrap47, label %invert10, label %invert9 invert12: ; preds = %invert14, %invert.preheader2 br i1 %28, label %invert6, label %invert11 invert.preheader2: ; preds = %invert13 br label %invert12 invert13: ; preds = %mergeinvert13_.loopexit3, %incinvert13 %106 = load i64, i64* %"iv'ac", align 8 %"'ipg54_unwrap" = getelementptr inbounds double, double* %44, i64 %106 %107 = load double, double* %"'ipg54_unwrap", align 8 store double 0.000000e+00, double* %"'ipg54_unwrap", align 8 %108 = load double, double* %"'de55", align 8 %109 = fadd fast double %108, %107 store double %109, double* %"'de55", align 8 %110 = load double, double* %"'de55", align 8 %111 = fneg fast double %110 store double 0.000000e+00, double* %"'de55", align 8 %112 = load double, double* %"'de56", align 8 %113 = fadd fast double %112, %110 store double %113, double* %"'de56", align 8 %114 = load double, double* %"'de57", align 8 %115 = fadd fast double %114, %111 store double %115, double* %"'de57", align 8 %116 = load double, double* %"'de57", align 8 store double 0.000000e+00, double* %"'de57", align 8 %117 = load i64, i64* %"iv'ac", align 8 %"'ipg58_unwrap" = getelementptr inbounds double, double* %"'ipl", i64 %117 %118 = load double, double* %"'ipg58_unwrap", align 8 %119 = fadd fast double %118, %116 store double %119, double* %"'ipg58_unwrap", align 8 %120 = load double, double* %"'de56", align 8 store double 0.000000e+00, double* %"'de56", align 8 %121 = load i64, i64* %"iv'ac", align 8 %"'ipg59_unwrap" = getelementptr inbounds double, double* %"'ipl49", i64 %121 %122 = load double, double* %"'ipg59_unwrap", align 8 %123 = fadd fast double %122, %120 store double %123, double* %"'ipg59_unwrap", align 8 %124 = load i64, i64* %"iv'ac", align 8 %125 = icmp eq i64 %124, 0 %126 = xor i1 %125, true br i1 %125, label %invert.preheader2, label %incinvert13 incinvert13: ; preds = %invert13 %127 = load i64, i64* %"iv'ac", align 8 %128 = add nsw i64 %127, -1 store i64 %128, i64* %"iv'ac", align 8 br label %invert13 invert.loopexit3: ; preds = %invert14 %_unwrap60 = add i64 %25, -1 br label %mergeinvert13_.loopexit3 mergeinvert13_.loopexit3: ; preds = %invert.loopexit3 store i64 %_unwrap60, i64* %"iv'ac", align 8 br label %invert13 invert14: ; preds = %invert23, %invert15 br i1 %50, label %invert.loopexit3, label %invert12 invert15: ; preds = %invert16, %invert.preheader %129 = load double, double* %"'de65", align 8 store double 0.000000e+00, double* %"'de65", align 8 %"'il_phi25_unwrap" = load double*, double** %"'ipg67", align 8, !tbaa !44 %130 = load double, double* %"'il_phi25_unwrap", align 8 %131 = fadd fast double %130, %129 store double %131, double* %"'il_phi25_unwrap", align 8 %_unwrap73 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 %"'ipg74_unwrap" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0 %132 = load { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 }* %_augmented_cache, align 8, !invariant.group !184 call void @diffe_ZN5Eigen8internal17product_evaluatorINS_7ProductINS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES4_Li0EEELi8ENS_10DenseShapeES6_ddEC2ERKS5_(%"struct.Eigen::internal::product_evaluator"* %_unwrap73, %"struct.Eigen::internal::product_evaluator"* %"'ipg74_unwrap", %"class.Eigen::Product"* %5, %"class.Eigen::Product"* %"'ipa61", { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, { %"class.Eigen::DenseBase"*, %"class.Eigen::DenseBase"*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, i1, i1, i1, double, i64, i64, i64, double*, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, %"class.Eigen::Matrix"*, %"class.Eigen::Matrix"*, i64 } %132) br label %invert14 invert.preheader: ; preds = %invert21 br label %invert15 invert.loopexit1: ; preds = %invert16 %_unwrap76 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %_unwrap77 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap76, align 8, !tbaa !46, !invariant.group !186 %_unwrap78 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap77, i64 0, i32 0, i32 1 %_unwrap79 = load i64, i64* %_unwrap78, align 8, !tbaa !36, !invariant.group !187 %_unwrap80 = add i64 %_unwrap79, -2 br label %mergeinvert21_.loopexit1 mergeinvert21_.loopexit1: ; preds = %invert.loopexit1 store i64 %_unwrap80, i64* %"iv32'ac", align 8 br label %invert21 invert16: ; preds = %invert22, %invert17 %133 = load double, double* %"'de82", align 8 store double 0.000000e+00, double* %"'de82", align 8 %_unwrap85 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %_unwrap86 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap85, align 8, !tbaa !46, !invariant.group !186 %_unwrap87 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap86, i64 0, i32 0, i32 1 %_unwrap88 = load i64, i64* %_unwrap87, align 8, !tbaa !36, !invariant.group !187 %_unwrap89 = icmp sgt i64 %_unwrap88, 1 %134 = xor i1 %_unwrap89, true %135 = select fast i1 %_unwrap89, double %133, double 0.000000e+00 %136 = load double, double* %"'de83", align 8 %137 = fadd fast double %136, %133 %138 = select fast i1 %_unwrap89, double %137, double %136 store double %138, double* %"'de83", align 8 %139 = select fast i1 %134, double %133, double 0.000000e+00 %140 = load double, double* %"'de65", align 8 %141 = fadd fast double %140, %133 %142 = select fast i1 %_unwrap89, double %140, double %141 store double %142, double* %"'de65", align 8 br i1 %_unwrap89, label %invert.loopexit1, label %invert15 invert17: ; preds = %invert18 br label %invert16 invert18: ; preds = %invert19 %143 = load double, double* %"'de91", align 8 store double 0.000000e+00, double* %"'de91", align 8 %144 = load i64, i64* %"iv28'ac", align 8 %145 = icmp eq i64 %144, 0 %146 = xor i1 %145, true %147 = select fast i1 %146, double %143, double 0.000000e+00 %148 = load double, double* %"'de92", align 8 %149 = fadd fast double %148, %143 %150 = select fast i1 %145, double %148, double %149 store double %150, double* %"'de92", align 8 %151 = select fast i1 %145, double %143, double 0.000000e+00 %152 = load double, double* %"'de82", align 8 %153 = fadd fast double %152, %143 %154 = select fast i1 %145, double %153, double %152 store double %154, double* %"'de82", align 8 br i1 %145, label %invert17, label %incinvert18 incinvert18: ; preds = %invert18 %155 = load i64, i64* %"iv28'ac", align 8 %156 = add nsw i64 %155, -1 store i64 %156, i64* %"iv28'ac", align 8 br label %invert20 invert19: ; preds = %mergeinvert19_, %incinvert19 %157 = load double, double* %"'de92", align 8 store double 0.000000e+00, double* %"'de92", align 8 %158 = load double, double* %"'de95", align 8 %159 = fadd fast double %158, %157 store double %159, double* %"'de95", align 8 %160 = load double, double* %"'de96", align 8 %161 = fadd fast double %160, %157 store double %161, double* %"'de96", align 8 %162 = load double, double* %"'de95", align 8 store double 0.000000e+00, double* %"'de95", align 8 %163 = load i64, i64* %"iv30'ac", align 8 %164 = load i64, i64* %"iv28'ac", align 8 %"'ipg67_unwrap" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl69_unwrap" = load double*, double** %"'ipg67_unwrap", align 8, !invariant.group !185 %iv.next29_unwrap = add nuw nsw i64 %164, 1 %_unwrap98 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %_unwrap99 = load i64, i64* %_unwrap98, align 8, !invariant.group !190 %_unwrap100 = mul nsw i64 %iv.next29_unwrap, %_unwrap99 %_unwrap101 = add nsw i64 %163, %_unwrap100 %"'ipg97_unwrap" = getelementptr inbounds double, double* %"'ipl69_unwrap", i64 %_unwrap101 %165 = load double, double* %"'ipg97_unwrap", align 8 %166 = fadd fast double %165, %162 store double %166, double* %"'ipg97_unwrap", align 8 %167 = load double, double* %"'de96", align 8 store double 0.000000e+00, double* %"'de96", align 8 %168 = load i64, i64* %"iv30'ac", align 8 %169 = icmp eq i64 %168, 0 %170 = xor i1 %169, true %171 = select fast i1 %170, double %167, double 0.000000e+00 %172 = load double, double* %"'de92", align 8 %173 = fadd fast double %172, %167 %174 = select fast i1 %169, double %172, double %173 store double %174, double* %"'de92", align 8 %175 = select fast i1 %169, double %167, double 0.000000e+00 %176 = load double, double* %"'de91", align 8 %177 = fadd fast double %176, %167 %178 = select fast i1 %169, double %177, double %176 store double %178, double* %"'de91", align 8 br i1 %169, label %invert18, label %incinvert19 incinvert19: ; preds = %invert19 %179 = load i64, i64* %"iv30'ac", align 8 %180 = add nsw i64 %179, -1 store i64 %180, i64* %"iv30'ac", align 8 br label %invert19 invert20: ; preds = %mergeinvert18_.loopexit, %incinvert18 %181 = load i64, i64* %"iv28'ac", align 8 %_unwrap106 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %_unwrap107 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap106, align 8, !tbaa !46, !invariant.group !186 %_unwrap108 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap107, i64 0, i32 0, i32 1 %_unwrap109 = load i64, i64* %_unwrap108, align 8, !tbaa !36, !invariant.group !187 %_unwrap110 = add i64 %_unwrap109, -1 br label %mergeinvert19_ mergeinvert19_: ; preds = %invert20 store i64 %_unwrap110, i64* %"iv30'ac", align 8 br label %invert19 invert21: ; preds = %incinvert21, %mergeinvert21_.loopexit1 %182 = load double, double* %"'de83", align 8 store double 0.000000e+00, double* %"'de83", align 8 %183 = load double, double* %"'de112", align 8 %184 = fadd fast double %183, %182 store double %184, double* %"'de112", align 8 %185 = load double, double* %"'de113", align 8 %186 = fadd fast double %185, %182 store double %186, double* %"'de113", align 8 %187 = load double, double* %"'de112", align 8 store double 0.000000e+00, double* %"'de112", align 8 %188 = load i64, i64* %"iv32'ac", align 8 %"'ipg67_unwrap115" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa66", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl69_unwrap116" = load double*, double** %"'ipg67_unwrap115", align 8, !invariant.group !185 %iv.next33_unwrap = add nuw nsw i64 %188, 1 %"'ipg114_unwrap" = getelementptr inbounds double, double* %"'ipl69_unwrap116", i64 %iv.next33_unwrap %189 = load double, double* %"'ipg114_unwrap", align 8 %190 = fadd fast double %189, %187 store double %190, double* %"'ipg114_unwrap", align 8 %191 = load double, double* %"'de113", align 8 store double 0.000000e+00, double* %"'de113", align 8 %192 = load i64, i64* %"iv32'ac", align 8 %193 = icmp eq i64 %192, 0 %194 = xor i1 %193, true %195 = select fast i1 %193, double %191, double 0.000000e+00 %196 = load double, double* %"'de65", align 8 %197 = fadd fast double %196, %191 %198 = select fast i1 %193, double %197, double %196 store double %198, double* %"'de65", align 8 %199 = select fast i1 %194, double %191, double 0.000000e+00 %200 = load double, double* %"'de83", align 8 %201 = fadd fast double %200, %191 %202 = select fast i1 %193, double %200, double %201 store double %202, double* %"'de83", align 8 br i1 %193, label %invert.preheader, label %incinvert21 incinvert21: ; preds = %invert21 %203 = load i64, i64* %"iv32'ac", align 8 %204 = add nsw i64 %203, -1 store i64 %204, i64* %"iv32'ac", align 8 br label %invert21 invert.loopexit: ; preds = %invert22 %_unwrap121 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %63, align 8, !tbaa !48, !invariant.group !188 %_unwrap122 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap121, i64 0, i32 0, i32 2 %_unwrap123 = load i64, i64* %_unwrap122, align 8, !tbaa !42, !invariant.group !189 %_unwrap124 = icmp sgt i64 %_unwrap123, 2 %_unwrap125 = select i1 %_unwrap124, i64 %_unwrap123, i64 2 %_unwrap126 = add nsw i64 %_unwrap125, -2 br label %mergeinvert18_.loopexit mergeinvert18_.loopexit: ; preds = %invert.loopexit store i64 %_unwrap126, i64* %"iv28'ac", align 8 br label %invert20 invert22: ; preds = %invert23 %205 = load double, double* %"'de133", align 8 store double 0.000000e+00, double* %"'de133", align 8 %_unwrap135 = bitcast %"class.Eigen::Product"* %5 to %"class.Eigen::PlainObjectBase"** %_unwrap136 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap135, align 8, !tbaa !46, !invariant.group !186 %_unwrap137 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap136, i64 0, i32 0, i32 1 %_unwrap138 = load i64, i64* %_unwrap137, align 8, !tbaa !36, !invariant.group !187 %_unwrap139 = icmp sgt i64 %_unwrap138, 0 %_unwrap140 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %63, align 8, !tbaa !48, !invariant.group !188 %_unwrap141 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap140, i64 0, i32 0, i32 2 %_unwrap142 = load i64, i64* %_unwrap141, align 8, !tbaa !42, !invariant.group !189 %_unwrap143 = icmp sgt i64 %_unwrap142, 1 %_unwrap144 = and i1 %_unwrap139, %_unwrap143 %206 = xor i1 %_unwrap144, true %207 = select fast i1 %_unwrap144, double %205, double 0.000000e+00 %208 = load double, double* %"'de92", align 8 %209 = fadd fast double %208, %205 %210 = select fast i1 %_unwrap144, double %209, double %208 store double %210, double* %"'de92", align 8 %211 = select fast i1 %206, double %205, double 0.000000e+00 %212 = load double, double* %"'de82", align 8 %213 = fadd fast double %212, %205 %214 = select fast i1 %_unwrap144, double %212, double %213 store double %214, double* %"'de82", align 8 br i1 %_unwrap144, label %invert.loopexit, label %invert16 invert23: ; preds = %99 store double %differeturn, double* %"'de", align 8 %215 = load double, double* %"'de", align 8 store double 0.000000e+00, double* %"'de", align 8 %216 = xor i1 %28, true %217 = select fast i1 %216, double %215, double 0.000000e+00 %218 = load double, double* %"'de133", align 8 %219 = fadd fast double %218, %215 %220 = select fast i1 %28, double %218, double %219 store double %220, double* %"'de133", align 8 br i1 %28, label %invert14, label %invert22 staging: ; preds = %invert6 br i1 %_unwrap39, label %invert5, label %invert4 } LLVM ERROR: function failed verification (4) PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace. Stack dump: 0. Program arguments: /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S 1. Running pass 'Enzyme Pass' on module '<stdin>'. #0 0x00007f085e1d49a3 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbda9a3) #1 0x00007f085e1d2cc0 llvm::sys::RunSignalHandlers() (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbd8cc0) #2 0x00007f085e1d500f (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbdb00f) #3 0x00007f085d5e73c0 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x153c0) #4 0x00007f085d0da18b raise (/lib/x86_64-linux-gnu/libc.so.6+0x4618b) #5 0x00007f085d0b9859 abort (/lib/x86_64-linux-gnu/libc.so.6+0x25859) #6 0x00007f085e12fc09 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35c09) #7 0x00007f085e12fa26 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35a26) #8 0x00007f085ad88a2a EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x58aa2a) #9 0x00007f085ad57ddf (anonymous namespace)::Enzyme::HandleAutoDiff(llvm::CallInst*, llvm::TargetLibraryInfo&, bool, DerivativeMode) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x559ddf) #10 0x00007f085ad5c593 (anonymous namespace)::Enzyme::lowerEnzymeCalls(llvm::Function&, bool, bool&, std::set<llvm::Function*, std::less<llvm::Function*>, std::allocator<llvm::Function*> >&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x55e593) #11 0x00007f085ad5e6d6 (anonymous namespace)::Enzyme::runOnModule(llvm::Module&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5606d6) #12 0x00007f085e2fc4ff llvm::legacy::PassManagerImpl::run(llvm::Module&) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xd024ff) #13 0x0000000000431dfb main (/usr/lib/llvm-12/bin/opt+0x431dfb) #14 0x00007f085d0bb0b3 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b3) #15 0x000000000041e68e _start (/usr/lib/llvm-12/bin/opt+0x41e68e) /usr/lib/llvm-12/bin/lli: error: 'main' function not found in module. --

******************** FAIL: Enzyme :: Integration/ReverseMode/eigensumsqdyn-notmp.cpp (11 of 65) ******************** TEST 'Enzyme :: Integration/ReverseMode/eigensumsqdyn-notmp.cpp' FAILED ******************** Script: -- : 'RUN: at line 1'; /usr/lib/llvm-12/bin/clang++ -mllvm -force-vector-width=1 -ffast-math -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 2'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 3'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 5'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 -Xclang -disable-llvm-optzns /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 6'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 7'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 8'; /usr/lib/llvm-12/bin/clang++ -fno-unroll-loops -fno-vectorize -fno-slp-vectorize -fno-exceptions -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/eigensumsqdyn-notmp.cpp -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - -- Exit Code: 255 Command Output (stderr): -- freeing without malloc %45 = load i8*, i8** %44, align 8, !tbaa !17 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %66 = load i8*, i8** %65, align 8, !tbaa !3 freeing without malloc %156 = load i8*, i8** %155, align 8, !tbaa !3 freeing without malloc %154 = load i8*, i8** %153, align 8, !tbaa !3 freeing without malloc %79 = load i8*, i8** %78, align 8, !tbaa !23 freeing without malloc %77 = load i8*, i8** %76, align 8, !tbaa !23 freeing without malloc %75 = load i8*, i8** %74, align 8, !tbaa !25 freeing without malloc %72 = load i8*, i8** %71, align 8, !tbaa !24 freeing without malloc %76 = select i1 %75, double* %74, double* null freeing without malloc %54 = select i1 %53, double* %52, double* null freeing without malloc %45 = load i8*, i8** %44, align 8, !tbaa !12 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %38 = load i8*, i8** %37, align 8, !tbaa !14 freeing without malloc %45 = load i8*, i8** %44, align 8, !tbaa !17 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %66 = load i8*, i8** %65, align 8, !tbaa !3 freeing without malloc %156 = load i8*, i8** %155, align 8, !tbaa !3 freeing without malloc %154 = load i8*, i8** %153, align 8, !tbaa !3 freeing without malloc %79 = load i8*, i8** %78, align 8, !tbaa !23 freeing without malloc %77 = load i8*, i8** %76, align 8, !tbaa !23 freeing without malloc %75 = load i8*, i8** %74, align 8, !tbaa !25 freeing without malloc %72 = load i8*, i8** %71, align 8, !tbaa !24 freeing without malloc %76 = select i1 %75, double* %74, double* null freeing without malloc %54 = select i1 %53, double* %52, double* null freeing without malloc %45 = load i8*, i8** %44, align 8, !tbaa !12 freeing without malloc %44 = load i8*, i8** %43, align 8, !tbaa !8 freeing without malloc %38 = load i8*, i8** %37, align 8, !tbaa !14 freeing without malloc %70 = load i8*, i8** %69, align 8, !tbaa !23 Instruction does not dominate all uses! %"'ipg23" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'il_phi_unwrap" = load double*, double** %"'ipg23", align 8, !tbaa !43 ; Function Attrs: noinline nounwind uwtable define internal double @preprocess__ZL6matvecPKN5Eigen6MatrixIdLin1ELin1ELi0ELin1ELin1EEES3_(%"class.Eigen::Matrix"* noalias %0, %"class.Eigen::Matrix"* noalias %1) #4 { %3 = alloca %"class.Eigen::internal::redux_evaluator", align 8 %4 = alloca %"class.Eigen::Product", align 8 %5 = bitcast %"class.Eigen::Product"* %4 to i8* call void @llvm.lifetime.start.p0i8(i64 48, i8* nonnull %5) #19 %6 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 0, i32 0 store %"class.Eigen::Matrix"* %0, %"class.Eigen::Matrix"** %6, align 8 %7 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 0, i32 1 store %"class.Eigen::Matrix"* %1, %"class.Eigen::Matrix"** %7, align 8 %8 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 1, i32 0 store %"class.Eigen::Matrix"* %0, %"class.Eigen::Matrix"** %8, align 8 %9 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 1, i32 1 store %"class.Eigen::Matrix"* %1, %"class.Eigen::Matrix"** %9, align 8 %10 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 1 %11 = load i64, i64* %10, align 8, !tbaa !36 %12 = bitcast %"class.Eigen::Matrix"** %9 to %"class.Eigen::PlainObjectBase"** %13 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 2 %14 = load i64, i64* %13, align 8, !tbaa !42 %15 = mul nsw i64 %14, %11 %16 = icmp eq i64 %15, 0 br i1 %16, label %71, label %17 17: ; preds = %2 %18 = bitcast %"class.Eigen::Matrix"** %7 to %"class.Eigen::PlainObjectBase"** %19 = bitcast %"class.Eigen::internal::redux_evaluator"* %3 to i8* call void @llvm.lifetime.start.p0i8(i64 48, i8* nonnull %19) #19 %20 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 call void @_ZN5Eigen8internal17product_evaluatorINS_7ProductINS_13CwiseBinaryOpINS0_20scalar_difference_opIddEEKNS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES8_EES9_Li0EEELi8ENS_10DenseShapeESB_ddEC2ERKSA_(%"struct.Eigen::internal::product_evaluator"* nonnull dereferenceable(40) %20, %"class.Eigen::Product"* nonnull align 8 dereferenceable(48) %4) #19 %21 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 1 store %"class.Eigen::Product"* %4, %"class.Eigen::Product"** %21, align 8, !tbaa !38 %22 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %23 = load double*, double** %22, align 8, !tbaa !43 %24 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %25 = load double, double* %23, align 8, !tbaa !37 %26 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %18, align 8, !tbaa !45 %27 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %26, i64 0, i32 0, i32 1 %28 = load i64, i64* %27, align 8, !tbaa !36 %29 = icmp sgt i64 %28, 1 br i1 %29, label %.preheader, label %30 .preheader: ; preds = %17 br label %58 .loopexit1: ; preds = %58 br label %30 30: ; preds = %.loopexit1, %17 %31 = phi double [ %25, %17 ], [ %63, %.loopexit1 ] %32 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %12, align 8, !tbaa !45 %33 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %32, i64 0, i32 0, i32 2 %34 = load i64, i64* %33, align 8, !tbaa !42 %35 = icmp sgt i64 %34, 1 %36 = icmp sgt i64 %28, 0 %37 = load i64, i64* %24, align 8 %38 = and i1 %36, %35 br i1 %38, label %39, label %66 39: ; preds = %30 %40 = icmp sgt i64 %34, 2 %41 = select i1 %40, i64 %34, i64 2 br label %42 42: ; preds = %55, %39 %tiv = phi i64 [ %tiv.next, %55 ], [ 0, %39 ] %43 = phi double [ %52, %55 ], [ %31, %39 ] %44 = add i64 %tiv, 1 %tiv.next = add nuw nsw i64 %tiv, 1 %45 = mul nsw i64 %44, %37 br label %46 46: ; preds = %46, %42 %47 = phi i64 [ 0, %42 ], [ %53, %46 ] %48 = phi double [ %43, %42 ], [ %52, %46 ] %49 = add nsw i64 %47, %45 %50 = getelementptr inbounds double, double* %23, i64 %49 %51 = load double, double* %50, align 8, !tbaa !37 %52 = fadd fast double %51, %48 %53 = add nuw nsw i64 %47, 1 %54 = icmp eq i64 %53, %28 br i1 %54, label %55, label %46, !llvm.loop !48 55: ; preds = %46 %56 = add nuw nsw i64 %44, 1 %57 = icmp eq i64 %56, %41 br i1 %57, label %.loopexit, label %42, !llvm.loop !49 58: ; preds = %.preheader, %58 %tiv2 = phi i64 [ 0, %.preheader ], [ %tiv.next3, %58 ] %59 = phi double [ %63, %58 ], [ %25, %.preheader ] %60 = add i64 %tiv2, 1 %tiv.next3 = add nuw nsw i64 %tiv2, 1 %61 = getelementptr inbounds double, double* %23, i64 %60 %62 = load double, double* %61, align 8, !tbaa !37 %63 = fadd fast double %62, %59 %64 = add nuw nsw i64 %60, 1 %65 = icmp eq i64 %64, %28 br i1 %65, label %.loopexit1, label %58, !llvm.loop !50 .loopexit: ; preds = %55 br label %66 66: ; preds = %.loopexit, %30 %67 = phi double [ %31, %30 ], [ %52, %.loopexit ] %68 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 1, i32 0 %69 = bitcast %"class.Eigen::PlainObjectBase"* %68 to i8** %70 = load i8*, i8** %69, align 8, !tbaa !33 call void @free(i8* %70) #19 call void @llvm.lifetime.end.p0i8(i64 48, i8* nonnull %19) #19 br label %71 71: ; preds = %66, %2 %72 = phi double [ %67, %66 ], [ 0.000000e+00, %2 ] call void @llvm.lifetime.end.p0i8(i64 48, i8* nonnull %5) #19 ret double %72 } ; Function Attrs: noinline nounwind uwtable define internal void @diffe_ZL6matvecPKN5Eigen6MatrixIdLin1ELin1ELi0ELin1ELin1EEES3_(%"class.Eigen::Matrix"* noalias %0, %"class.Eigen::Matrix"* %"'", %"class.Eigen::Matrix"* noalias %1, %"class.Eigen::Matrix"* %"'1", double %differeturn) #4 { %"iv'ac" = alloca i64, align 8 %"iv13'ac" = alloca i64, align 8 %"iv15'ac" = alloca i64, align 8 %"'de" = alloca double, align 8 store double 0.000000e+00, double* %"'de", align 8 %"'de21" = alloca double, align 8 store double 0.000000e+00, double* %"'de21", align 8 %_augmented_cache = alloca { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 }, align 8 %"'de36" = alloca double, align 8 store double 0.000000e+00, double* %"'de36", align 8 %"'de37" = alloca double, align 8 store double 0.000000e+00, double* %"'de37", align 8 %"'de45" = alloca double, align 8 store double 0.000000e+00, double* %"'de45", align 8 %"'de46" = alloca double, align 8 store double 0.000000e+00, double* %"'de46", align 8 %"'de49" = alloca double, align 8 store double 0.000000e+00, double* %"'de49", align 8 %"'de50" = alloca double, align 8 store double 0.000000e+00, double* %"'de50", align 8 %"'de66" = alloca double, align 8 store double 0.000000e+00, double* %"'de66", align 8 %"'de67" = alloca double, align 8 store double 0.000000e+00, double* %"'de67", align 8 %"'de85" = alloca double, align 8 store double 0.000000e+00, double* %"'de85", align 8 %"'ipa22" = alloca %"class.Eigen::internal::redux_evaluator", align 8 store %"class.Eigen::internal::redux_evaluator" zeroinitializer, %"class.Eigen::internal::redux_evaluator"* %"'ipa22", align 8 %3 = alloca %"class.Eigen::internal::redux_evaluator", align 8 %"'ipa" = alloca %"class.Eigen::Product", align 8 store %"class.Eigen::Product" zeroinitializer, %"class.Eigen::Product"* %"'ipa", align 8 %4 = alloca %"class.Eigen::Product", align 8 %"'ipg20" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa", i64 0, i32 0, i32 0 %5 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 0, i32 0 store %"class.Eigen::Matrix"* %"'", %"class.Eigen::Matrix"** %"'ipg20", align 8 store %"class.Eigen::Matrix"* %0, %"class.Eigen::Matrix"** %5, align 8 %"'ipg19" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa", i64 0, i32 0, i32 1 %6 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 0, i32 1 store %"class.Eigen::Matrix"* %"'1", %"class.Eigen::Matrix"** %"'ipg19", align 8 store %"class.Eigen::Matrix"* %1, %"class.Eigen::Matrix"** %6, align 8 %"'ipg18" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa", i64 0, i32 1, i32 0 %7 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 1, i32 0 store %"class.Eigen::Matrix"* %"'", %"class.Eigen::Matrix"** %"'ipg18", align 8 store %"class.Eigen::Matrix"* %0, %"class.Eigen::Matrix"** %7, align 8 %"'ipg" = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %"'ipa", i64 0, i32 1, i32 1 %8 = getelementptr inbounds %"class.Eigen::Product", %"class.Eigen::Product"* %4, i64 0, i32 1, i32 1 store %"class.Eigen::Matrix"* %"'1", %"class.Eigen::Matrix"** %"'ipg", align 8 store %"class.Eigen::Matrix"* %1, %"class.Eigen::Matrix"** %8, align 8 %9 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 1 %10 = load i64, i64* %9, align 8, !tbaa !36 %11 = bitcast %"class.Eigen::Matrix"** %8 to %"class.Eigen::PlainObjectBase"** %12 = getelementptr inbounds %"class.Eigen::Matrix", %"class.Eigen::Matrix"* %1, i64 0, i32 0, i32 0, i32 2 %13 = load i64, i64* %12, align 8, !tbaa !42 %14 = mul nsw i64 %13, %10 %15 = icmp eq i64 %14, 0 br i1 %15, label %51, label %16 16: ; preds = %2 %17 = bitcast %"class.Eigen::Matrix"** %6 to %"class.Eigen::PlainObjectBase"** %"'ipg28" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0 %18 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 %_augmented = call { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 } @augmented__ZN5Eigen8internal17product_evaluatorINS_7ProductINS_13CwiseBinaryOpINS0_20scalar_difference_opIddEEKNS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES8_EES9_Li0EEELi8ENS_10DenseShapeESB_ddEC2ERKSA_(%"struct.Eigen::internal::product_evaluator"* %18, %"struct.Eigen::internal::product_evaluator"* %"'ipg28", %"class.Eigen::Product"* %4, %"class.Eigen::Product"* %"'ipa") %"'ipg27" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 1 %19 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 1 store %"class.Eigen::Product"* %"'ipa", %"class.Eigen::Product"** %"'ipg27", align 8 store { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 } %_augmented, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 }* %_augmented_cache, align 8, !invariant.group !180 store %"class.Eigen::Product"* %4, %"class.Eigen::Product"** %19, align 8, !tbaa !38 %"'ipg23" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl" = load double*, double** %"'ipg23", align 8, !invariant.group !181 %20 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %21 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %17, align 8, !tbaa !45, !invariant.group !182 %22 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %21, i64 0, i32 0, i32 1 %23 = load i64, i64* %22, align 8, !tbaa !36, !invariant.group !183 %24 = icmp sgt i64 %23, 1 br i1 %24, label %.preheader, label %26 .preheader: ; preds = %16 %25 = add i64 %23, -2 br label %47 .loopexit1: ; preds = %47 br label %26 26: ; preds = %.loopexit1, %16 %27 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %11, align 8, !tbaa !45, !invariant.group !184 %28 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %27, i64 0, i32 0, i32 2 %29 = load i64, i64* %28, align 8, !tbaa !42, !invariant.group !185 %30 = icmp sgt i64 %29, 1 %31 = icmp sgt i64 %23, 0 %32 = load i64, i64* %20, align 8, !invariant.group !186 %33 = and i1 %31, %30 br i1 %33, label %34, label %50 34: ; preds = %26 %35 = icmp sgt i64 %29, 2 %36 = select i1 %35, i64 %29, i64 2 %37 = add nsw i64 %36, -2 %38 = add i64 %23, -1 br label %39 39: ; preds = %44, %34 %iv = phi i64 [ %iv.next, %44 ], [ 0, %34 ] %iv.next = add nuw nsw i64 %iv, 1 %40 = mul nsw i64 %iv.next, %32 br label %41 41: ; preds = %41, %39 %iv13 = phi i64 [ %iv.next14, %41 ], [ 0, %39 ] %iv.next14 = add nuw nsw i64 %iv13, 1 %42 = add nsw i64 %iv13, %40 %"'ipg51" = getelementptr inbounds double, double* %"'ipl", i64 %42 %43 = icmp eq i64 %iv.next14, %23 br i1 %43, label %44, label %41, !llvm.loop !48 44: ; preds = %41 %45 = add nuw nsw i64 %iv.next, 1 %46 = icmp eq i64 %45, %36 br i1 %46, label %.loopexit, label %39, !llvm.loop !49 47: ; preds = %47, %.preheader %iv15 = phi i64 [ %iv.next16, %47 ], [ 0, %.preheader ] %iv.next16 = add nuw nsw i64 %iv15, 1 %"'ipg68" = getelementptr inbounds double, double* %"'ipl", i64 %iv.next16 %48 = add nuw nsw i64 %iv.next16, 1 %49 = icmp eq i64 %48, %23 br i1 %49, label %.loopexit1, label %47, !llvm.loop !50 .loopexit: ; preds = %44 br label %50 50: ; preds = %.loopexit, %26 br label %51 51: ; preds = %50, %2 br label %invert10 invert: ; preds = %invert10, %invert2 ret void invert2: ; preds = %invert3, %invert.preheader %52 = load double, double* %"'de21", align 8 store double 0.000000e+00, double* %"'de21", align 8 %"'il_phi_unwrap" = load double*, double** %"'ipg23", align 8, !tbaa !43 %53 = load double, double* %"'il_phi_unwrap", align 8 %54 = fadd fast double %53, %52 store double %54, double* %"'il_phi_unwrap", align 8 %_unwrap = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0 %"'ipg28_unwrap" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0 %55 = load { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 }* %_augmented_cache, align 8, !invariant.group !180 call void @diffe_ZN5Eigen8internal17product_evaluatorINS_7ProductINS_13CwiseBinaryOpINS0_20scalar_difference_opIddEEKNS_6MatrixIdLin1ELin1ELi0ELin1ELin1EEES8_EES9_Li0EEELi8ENS_10DenseShapeESB_ddEC2ERKSA_(%"struct.Eigen::internal::product_evaluator"* %_unwrap, %"struct.Eigen::internal::product_evaluator"* %"'ipg28_unwrap", %"class.Eigen::Product"* %4, %"class.Eigen::Product"* %"'ipa", { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double*, i8*, i8*, i8*, i8*, { double*, double*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, double**, double**, i64, i64, i64, i1, i1, i64*, i64*, double**, double**, i64*, double**, double**, double**, double** }, { i8*, i8*, i8*, i8*, i8*, i8*, { i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, i8*, double*, i8*, i8*, i8*, i8*, { double*, i64, i64* }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double**, double**, double**, i64*, double*, double*, double*, double**, double**, double**, double**, <8 x double>*, <8 x double>*, i64*, double**, double**, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, double*, <4 x double>*, <4 x double>*, double**, double**, double*, <4 x double>*, i64*, double*, double*, double*, double**, double** }*, { double*, double*, i64, i64*, i64, i64* }*, { double*, i64, i64* }*, i64, i64, i64, double*, double* }, double*, double*, double*, { i8*, i8*, i8*, i8*, i8*, i8*, { { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i32, i32, i32, i32 }, { i8*, i1*, i8*, i1 }, i32*, i32* }, i1, i1, i64, i64, i64, i64, i1, i1, i64, i64, i64, i64 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, { i8*, i8*, i8*, i8*, i8*, i8*, { double*, %"class.Eigen::Matrix"*, double*, %"class.Eigen::Matrix"*, i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i64, i64, i1 }, i1, i1, i1, double, i64, i64, i64, double*, i64, double*, i64, double*, i64 }, { i8*, i8*, i8*, i8*, i8*, double*, i64, i64, i64, i1, i1 }, i64, i64, i1, i64 } %55) br label %invert invert.preheader: ; preds = %invert8 br label %invert2 invert.loopexit1: ; preds = %invert3 %_unwrap30 = bitcast %"class.Eigen::Matrix"** %6 to %"class.Eigen::PlainObjectBase"** %_unwrap31 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap30, align 8, !tbaa !45, !invariant.group !182 %_unwrap32 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap31, i64 0, i32 0, i32 1 %_unwrap33 = load i64, i64* %_unwrap32, align 8, !tbaa !36, !invariant.group !183 %_unwrap34 = add i64 %_unwrap33, -2 br label %mergeinvert8_.loopexit1 mergeinvert8_.loopexit1: ; preds = %invert.loopexit1 store i64 %_unwrap34, i64* %"iv15'ac", align 8 br label %invert8 invert3: ; preds = %invert9, %invert4 %56 = load double, double* %"'de36", align 8 store double 0.000000e+00, double* %"'de36", align 8 %_unwrap39 = bitcast %"class.Eigen::Matrix"** %6 to %"class.Eigen::PlainObjectBase"** %_unwrap40 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap39, align 8, !tbaa !45, !invariant.group !182 %_unwrap41 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap40, i64 0, i32 0, i32 1 %_unwrap42 = load i64, i64* %_unwrap41, align 8, !tbaa !36, !invariant.group !183 %_unwrap43 = icmp sgt i64 %_unwrap42, 1 %57 = xor i1 %_unwrap43, true %58 = select fast i1 %_unwrap43, double %56, double 0.000000e+00 %59 = load double, double* %"'de37", align 8 %60 = fadd fast double %59, %56 %61 = select fast i1 %_unwrap43, double %60, double %59 store double %61, double* %"'de37", align 8 %62 = select fast i1 %57, double %56, double 0.000000e+00 %63 = load double, double* %"'de21", align 8 %64 = fadd fast double %63, %56 %65 = select fast i1 %_unwrap43, double %63, double %64 store double %65, double* %"'de21", align 8 br i1 %_unwrap43, label %invert.loopexit1, label %invert2 invert4: ; preds = %invert5 br label %invert3 invert5: ; preds = %invert6 %66 = load double, double* %"'de45", align 8 store double 0.000000e+00, double* %"'de45", align 8 %67 = load i64, i64* %"iv'ac", align 8 %68 = icmp eq i64 %67, 0 %69 = xor i1 %68, true %70 = select fast i1 %69, double %66, double 0.000000e+00 %71 = load double, double* %"'de46", align 8 %72 = fadd fast double %71, %66 %73 = select fast i1 %68, double %71, double %72 store double %73, double* %"'de46", align 8 %74 = select fast i1 %68, double %66, double 0.000000e+00 %75 = load double, double* %"'de36", align 8 %76 = fadd fast double %75, %66 %77 = select fast i1 %68, double %76, double %75 store double %77, double* %"'de36", align 8 br i1 %68, label %invert4, label %incinvert5 incinvert5: ; preds = %invert5 %78 = load i64, i64* %"iv'ac", align 8 %79 = add nsw i64 %78, -1 store i64 %79, i64* %"iv'ac", align 8 br label %invert7 invert6: ; preds = %mergeinvert6_, %incinvert6 %80 = load double, double* %"'de46", align 8 store double 0.000000e+00, double* %"'de46", align 8 %81 = load double, double* %"'de49", align 8 %82 = fadd fast double %81, %80 store double %82, double* %"'de49", align 8 %83 = load double, double* %"'de50", align 8 %84 = fadd fast double %83, %80 store double %84, double* %"'de50", align 8 %85 = load double, double* %"'de49", align 8 store double 0.000000e+00, double* %"'de49", align 8 %86 = load i64, i64* %"iv13'ac", align 8 %87 = load i64, i64* %"iv'ac", align 8 %"'ipg23_unwrap" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl_unwrap" = load double*, double** %"'ipg23_unwrap", align 8, !invariant.group !181 %iv.next_unwrap = add nuw nsw i64 %87, 1 %_unwrap52 = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %3, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1, i32 0 %_unwrap53 = load i64, i64* %_unwrap52, align 8, !invariant.group !186 %_unwrap54 = mul nsw i64 %iv.next_unwrap, %_unwrap53 %_unwrap55 = add nsw i64 %86, %_unwrap54 %"'ipg51_unwrap" = getelementptr inbounds double, double* %"'ipl_unwrap", i64 %_unwrap55 %88 = load double, double* %"'ipg51_unwrap", align 8 %89 = fadd fast double %88, %85 store double %89, double* %"'ipg51_unwrap", align 8 %90 = load double, double* %"'de50", align 8 store double 0.000000e+00, double* %"'de50", align 8 %91 = load i64, i64* %"iv13'ac", align 8 %92 = icmp eq i64 %91, 0 %93 = xor i1 %92, true %94 = select fast i1 %93, double %90, double 0.000000e+00 %95 = load double, double* %"'de46", align 8 %96 = fadd fast double %95, %90 %97 = select fast i1 %92, double %95, double %96 store double %97, double* %"'de46", align 8 %98 = select fast i1 %92, double %90, double 0.000000e+00 %99 = load double, double* %"'de45", align 8 %100 = fadd fast double %99, %90 %101 = select fast i1 %92, double %100, double %99 store double %101, double* %"'de45", align 8 br i1 %92, label %invert5, label %incinvert6 incinvert6: ; preds = %invert6 %102 = load i64, i64* %"iv13'ac", align 8 %103 = add nsw i64 %102, -1 store i64 %103, i64* %"iv13'ac", align 8 br label %invert6 invert7: ; preds = %mergeinvert5_.loopexit, %incinvert5 %104 = load i64, i64* %"iv'ac", align 8 %_unwrap60 = bitcast %"class.Eigen::Matrix"** %6 to %"class.Eigen::PlainObjectBase"** %_unwrap61 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap60, align 8, !tbaa !45, !invariant.group !182 %_unwrap62 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap61, i64 0, i32 0, i32 1 %_unwrap63 = load i64, i64* %_unwrap62, align 8, !tbaa !36, !invariant.group !183 %_unwrap64 = add i64 %_unwrap63, -1 br label %mergeinvert6_ mergeinvert6_: ; preds = %invert7 store i64 %_unwrap64, i64* %"iv13'ac", align 8 br label %invert6 invert8: ; preds = %incinvert8, %mergeinvert8_.loopexit1 %105 = load double, double* %"'de37", align 8 store double 0.000000e+00, double* %"'de37", align 8 %106 = load double, double* %"'de66", align 8 %107 = fadd fast double %106, %105 store double %107, double* %"'de66", align 8 %108 = load double, double* %"'de67", align 8 %109 = fadd fast double %108, %105 store double %109, double* %"'de67", align 8 %110 = load double, double* %"'de66", align 8 store double 0.000000e+00, double* %"'de66", align 8 %111 = load i64, i64* %"iv15'ac", align 8 %"'ipg23_unwrap69" = getelementptr inbounds %"class.Eigen::internal::redux_evaluator", %"class.Eigen::internal::redux_evaluator"* %"'ipa22", i64 0, i32 0, i32 0, i32 0, i32 0, i32 0 %"'ipl_unwrap70" = load double*, double** %"'ipg23_unwrap69", align 8, !invariant.group !181 %iv.next16_unwrap = add nuw nsw i64 %111, 1 %"'ipg68_unwrap" = getelementptr inbounds double, double* %"'ipl_unwrap70", i64 %iv.next16_unwrap %112 = load double, double* %"'ipg68_unwrap", align 8 %113 = fadd fast double %112, %110 store double %113, double* %"'ipg68_unwrap", align 8 %114 = load double, double* %"'de67", align 8 store double 0.000000e+00, double* %"'de67", align 8 %115 = load i64, i64* %"iv15'ac", align 8 %116 = icmp eq i64 %115, 0 %117 = xor i1 %116, true %118 = select fast i1 %116, double %114, double 0.000000e+00 %119 = load double, double* %"'de21", align 8 %120 = fadd fast double %119, %114 %121 = select fast i1 %116, double %120, double %119 store double %121, double* %"'de21", align 8 %122 = select fast i1 %117, double %114, double 0.000000e+00 %123 = load double, double* %"'de37", align 8 %124 = fadd fast double %123, %114 %125 = select fast i1 %116, double %123, double %124 store double %125, double* %"'de37", align 8 br i1 %116, label %invert.preheader, label %incinvert8 incinvert8: ; preds = %invert8 %126 = load i64, i64* %"iv15'ac", align 8 %127 = add nsw i64 %126, -1 store i64 %127, i64* %"iv15'ac", align 8 br label %invert8 invert.loopexit: ; preds = %invert9 %_unwrap75 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %11, align 8, !tbaa !45, !invariant.group !184 %_unwrap76 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap75, i64 0, i32 0, i32 2 %_unwrap77 = load i64, i64* %_unwrap76, align 8, !tbaa !42, !invariant.group !185 %_unwrap78 = icmp sgt i64 %_unwrap77, 2 %_unwrap79 = select i1 %_unwrap78, i64 %_unwrap77, i64 2 %_unwrap80 = add nsw i64 %_unwrap79, -2 br label %mergeinvert5_.loopexit mergeinvert5_.loopexit: ; preds = %invert.loopexit store i64 %_unwrap80, i64* %"iv'ac", align 8 br label %invert7 invert9: ; preds = %invert10 %128 = load double, double* %"'de85", align 8 store double 0.000000e+00, double* %"'de85", align 8 %_unwrap87 = bitcast %"class.Eigen::Matrix"** %6 to %"class.Eigen::PlainObjectBase"** %_unwrap88 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %_unwrap87, align 8, !tbaa !45, !invariant.group !182 %_unwrap89 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap88, i64 0, i32 0, i32 1 %_unwrap90 = load i64, i64* %_unwrap89, align 8, !tbaa !36, !invariant.group !183 %_unwrap91 = icmp sgt i64 %_unwrap90, 0 %_unwrap92 = load %"class.Eigen::PlainObjectBase"*, %"class.Eigen::PlainObjectBase"** %11, align 8, !tbaa !45, !invariant.group !184 %_unwrap93 = getelementptr inbounds %"class.Eigen::PlainObjectBase", %"class.Eigen::PlainObjectBase"* %_unwrap92, i64 0, i32 0, i32 2 %_unwrap94 = load i64, i64* %_unwrap93, align 8, !tbaa !42, !invariant.group !185 %_unwrap95 = icmp sgt i64 %_unwrap94, 1 %_unwrap96 = and i1 %_unwrap91, %_unwrap95 %129 = xor i1 %_unwrap96, true %130 = select fast i1 %_unwrap96, double %128, double 0.000000e+00 %131 = load double, double* %"'de46", align 8 %132 = fadd fast double %131, %128 %133 = select fast i1 %_unwrap96, double %132, double %131 store double %133, double* %"'de46", align 8 %134 = select fast i1 %129, double %128, double 0.000000e+00 %135 = load double, double* %"'de36", align 8 %136 = fadd fast double %135, %128 %137 = select fast i1 %_unwrap96, double %135, double %136 store double %137, double* %"'de36", align 8 br i1 %_unwrap96, label %invert.loopexit, label %invert3 invert10: ; preds = %51 store double %differeturn, double* %"'de", align 8 %138 = load double, double* %"'de", align 8 store double 0.000000e+00, double* %"'de", align 8 %139 = xor i1 %15, true %140 = select fast i1 %139, double %138, double 0.000000e+00 %141 = load double, double* %"'de85", align 8 %142 = fadd fast double %141, %138 %143 = select fast i1 %15, double %141, double %142 store double %143, double* %"'de85", align 8 br i1 %15, label %invert, label %invert9 } LLVM ERROR: function failed verification (4) PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace. Stack dump: 0. Program arguments: /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S 1. Running pass 'Enzyme Pass' on module '<stdin>'. #0 0x00007fef5f57b9a3 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbda9a3) #1 0x00007fef5f579cc0 llvm::sys::RunSignalHandlers() (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbd8cc0) #2 0x00007fef5f57c00f (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbdb00f) #3 0x00007fef5e98e3c0 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x153c0) #4 0x00007fef5e48118b raise (/lib/x86_64-linux-gnu/libc.so.6+0x4618b) #5 0x00007fef5e460859 abort (/lib/x86_64-linux-gnu/libc.so.6+0x25859) #6 0x00007fef5f4d6c09 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35c09) #7 0x00007fef5f4d6a26 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35a26) #8 0x00007fef5c12fa2a EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x58aa2a) #9 0x00007fef5c0feddf (anonymous namespace)::Enzyme::HandleAutoDiff(llvm::CallInst*, llvm::TargetLibraryInfo&, bool, DerivativeMode) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x559ddf) #10 0x00007fef5c103593 (anonymous namespace)::Enzyme::lowerEnzymeCalls(llvm::Function&, bool, bool&, std::set<llvm::Function*, std::less<llvm::Function*>, std::allocator<llvm::Function*> >&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x55e593) #11 0x00007fef5c1056d6 (anonymous namespace)::Enzyme::runOnModule(llvm::Module&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5606d6) #12 0x00007fef5f6a34ff llvm::legacy::PassManagerImpl::run(llvm::Module&) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xd024ff) #13 0x0000000000431dfb main (/usr/lib/llvm-12/bin/opt+0x431dfb) #14 0x00007fef5e4620b3 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b3) #15 0x000000000041e68e _start (/usr/lib/llvm-12/bin/opt+0x41e68e) /usr/lib/llvm-12/bin/lli: error: 'main' function not found in module. -- ******************** FAIL: Enzyme :: Integration/ReverseMode/mixedstruct1-simplefda.c (19 of 65) ******************** TEST 'Enzyme :: Integration/ReverseMode/mixedstruct1-simplefda.c' FAILED ******************** Script: -- : 'RUN: at line 1'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 -disable-llvm-passes /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 2'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 3'; /usr/lib/llvm-12/bin/clang -std=c11 -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 4'; /usr/lib/llvm-12/bin/clang -std=c11 -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 5'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 -disable-llvm-passes /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 6'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 7'; /usr/lib/llvm-12/bin/clang -std=c11 -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 8'; /usr/lib/llvm-12/bin/clang -std=c11 -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - -- Exit Code: 255 Command Output (stderr): -- clang: warning: argument unused during compilation: '-disable-llvm-passes' [-Wunused-command-line-argument] /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c:46:38: warning: format specifies type 'int' but the argument has type 'long long' [-Wformat] printf("inner considering %d\n", frame->R_start); ~~ ^~~~~~~~~~~~~~ %lld /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simplefda.c:54:32: warning: format specifies type 'int' but the argument has type 'long long' [-Wformat] printf("considering %d\n", frame->R_start); ~~ ^~~~~~~~~~~~~~ %lld 2 warnings generated. Instruction does not dominate all uses! %"'ipg25" = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %"'", i64 0, i32 2 %"'il_phi5_unwrap" = load float*, float** %"'ipg25", align 8, !tbaa !9 ; Function Attrs: nounwind uwtable define dso_local float @preprocess_tile_multiply(%struct.WINDOW_FRAME* nocapture readonly %0) local_unnamed_addr #0 { %2 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 1 %3 = load i64, i64* %2, align 8, !tbaa !2 %4 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 0 %5 = load i64, i64* %4, align 8, !tbaa !8 %6 = sub nsw i64 %3, %5 %7 = icmp sgt i64 %6, 1 br i1 %7, label %8, label %26 8: ; preds = %1 %9 = call noalias dereferenceable_or_null(24) i8* @malloc(i64 24) #8 %10 = bitcast i8* %9 to %struct.WINDOW_FRAME* %11 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 2 %12 = load float*, float** %11, align 8, !tbaa !9 %13 = getelementptr inbounds i8, i8* %9, i64 16 %14 = bitcast i8* %13 to float** store float* %12, float** %14, align 8, !tbaa !9 %15 = add nsw i64 %5, 1 %16 = bitcast i8* %9 to i64* store i64 %15, i64* %16, align 8, !tbaa !8 %17 = getelementptr inbounds i8, i8* %9, i64 8 %18 = bitcast i8* %17 to i64* store i64 %3, i64* %18, align 8, !tbaa !2 %19 = call i32 (i8*, ...) @printf(i8* nonnull dereferenceable(1) getelementptr inbounds ([22 x i8], [22 x i8]* @.str, i64 0, i64 0), i64 %5) %20 = load float*, float** %11, align 8, !tbaa !9 %21 = load i64, i64* %4, align 8, !tbaa !8 %22 = getelementptr inbounds float, float* %20, i64 %21 %23 = load float, float* %22, align 4, !tbaa !10 %24 = call float @tile_multiply(%struct.WINDOW_FRAME* %10) call void @free(i8* %9) #8 %25 = fadd float %23, %24 br label %34 26: ; preds = %1 %27 = call i32 (i8*, ...) @printf(i8* nonnull dereferenceable(1) getelementptr inbounds ([16 x i8], [16 x i8]* @.str.1, i64 0, i64 0), i64 %5) %28 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 2 %29 = load float*, float** %28, align 8, !tbaa !9 %30 = load i64, i64* %4, align 8, !tbaa !8 %31 = getelementptr inbounds float, float* %29, i64 %30 %32 = load float, float* %31, align 4, !tbaa !10 %33 = fadd float %32, 0.000000e+00 br label %34 34: ; preds = %26, %8 %35 = phi float [ %25, %8 ], [ %33, %26 ] ret float %35 } ; Function Attrs: nounwind uwtable define internal void @diffetile_multiply(%struct.WINDOW_FRAME* nocapture readonly %0, %struct.WINDOW_FRAME* nocapture %"'", float %differeturn) local_unnamed_addr #0 { %"'de" = alloca float, align 4 store float 0.000000e+00, float* %"'de", align 4 %"'de6" = alloca float, align 4 store float 0.000000e+00, float* %"'de6", align 4 %"'de7" = alloca float, align 4 store float 0.000000e+00, float* %"'de7", align 4 %"'de8" = alloca float, align 4 store float 0.000000e+00, float* %"'de8", align 4 %_cache = alloca i8*, align 8 %"'mi_cache" = alloca i8*, align 8 %"'il_phi4_cache" = alloca float*, align 8 %_cache10 = alloca i64, align 8 %"'de22" = alloca float, align 4 store float 0.000000e+00, float* %"'de22", align 4 %"'de23" = alloca float, align 4 store float 0.000000e+00, float* %"'de23", align 4 %2 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 1 %3 = load i64, i64* %2, align 8, !tbaa !2 %4 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 0 %5 = load i64, i64* %4, align 8, !tbaa !8 %6 = sub nsw i64 %3, %5 %7 = icmp sgt i64 %6, 1 br i1 %7, label %8, label %21 8: ; preds = %1 %9 = call noalias nonnull dereferenceable(24) dereferenceable_or_null(24) i8* @malloc(i64 24) #8 %"'mi" = call noalias nonnull dereferenceable(24) dereferenceable_or_null(24) i8* @malloc(i64 24) #8 call void @llvm.memset.p0i8.i64(i8* nonnull dereferenceable(24) dereferenceable_or_null(24) %"'mi", i8 0, i64 24, i1 false) %"'ipc" = bitcast i8* %"'mi" to %struct.WINDOW_FRAME* %10 = bitcast i8* %9 to %struct.WINDOW_FRAME* %"'ipg13" = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %"'", i64 0, i32 2 %11 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 2 %"'ipl20" = load float*, float** %"'ipg13", align 8 %12 = load float*, float** %11, align 8, !tbaa !9 %"'ipg18" = getelementptr inbounds i8, i8* %"'mi", i64 16 %13 = getelementptr inbounds i8, i8* %9, i64 16 %"'ipc19" = bitcast i8* %"'ipg18" to float** %14 = bitcast i8* %13 to float** store float* %"'ipl20", float** %"'ipc19", align 8 store float* %12, float** %14, align 8, !tbaa !9 store i8* %"'mi", i8** %"'mi_cache", align 8, !invariant.group !17 store i8* %9, i8** %_cache, align 8, !invariant.group !18 %15 = add nsw i64 %5, 1 %"'ipc17" = bitcast i8* %"'mi" to i64* %16 = bitcast i8* %9 to i64* store i64 %15, i64* %"'ipc17", align 8 store i64 %15, i64* %16, align 8, !tbaa !8 %"'ipg15" = getelementptr inbounds i8, i8* %"'mi", i64 8 %17 = getelementptr inbounds i8, i8* %9, i64 8 %"'ipc16" = bitcast i8* %"'ipg15" to i64* %18 = bitcast i8* %17 to i64* store i64 %3, i64* %"'ipc16", align 8 store i64 %3, i64* %18, align 8, !tbaa !2 %19 = call i32 (i8*, ...) @printf(i8* nonnull dereferenceable(1) getelementptr inbounds ([22 x i8], [22 x i8]* @.str, i64 0, i64 0), i64 %5) %"'ipl" = load float*, float** %"'ipg13", align 8 store float* %"'ipl", float** %"'il_phi4_cache", align 8, !invariant.group !19 %20 = load i64, i64* %4, align 8, !tbaa !8 store i64 %20, i64* %_cache10, align 8, !invariant.group !20 %"'ipg" = getelementptr inbounds float, float* %"'ipl", i64 %20 br label %24 21: ; preds = %1 %22 = call i32 (i8*, ...) @printf(i8* nonnull dereferenceable(1) getelementptr inbounds ([16 x i8], [16 x i8]* @.str.1, i64 0, i64 0), i64 %5) %"'ipg25" = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %"'", i64 0, i32 2 %"'ipl29" = load float*, float** %"'ipg25", align 8 %23 = load i64, i64* %4, align 8, !tbaa !8, !invariant.group !21 %"'ipg24" = getelementptr inbounds float, float* %"'ipl29", i64 %23 br label %24 24: ; preds = %21, %8 br label %invert3 invert: ; preds = %invert2, %invert1 ret void invert1: ; preds = %invert3 %25 = load float, float* %"'de6", align 4 store float 0.000000e+00, float* %"'de6", align 4 %26 = load float, float* %"'de7", align 4 %27 = fadd fast float %26, %25 store float %27, float* %"'de7", align 4 %28 = load float, float* %"'de8", align 4 %29 = fadd fast float %28, %25 store float %29, float* %"'de8", align 4 %30 = load i8*, i8** %_cache, align 8, !invariant.group !18 %_unwrap = bitcast i8* %30 to %struct.WINDOW_FRAME* %31 = load i8*, i8** %"'mi_cache", align 8, !invariant.group !17 %"'ipc_unwrap" = bitcast i8* %31 to %struct.WINDOW_FRAME* %32 = load float, float* %"'de8", align 4 call void @diffetile_multiply(%struct.WINDOW_FRAME* %_unwrap, %struct.WINDOW_FRAME* %"'ipc_unwrap", float %32) store float 0.000000e+00, float* %"'de8", align 4 %33 = load float, float* %"'de7", align 4 store float 0.000000e+00, float* %"'de7", align 4 %34 = load float*, float** %"'il_phi4_cache", align 8, !invariant.group !19 %35 = load i64, i64* %_cache10, align 8, !invariant.group !20 %"'ipg_unwrap" = getelementptr inbounds float, float* %34, i64 %35 %36 = load float, float* %"'ipg_unwrap", align 4 %37 = fadd fast float %36, %33 store float %37, float* %"'ipg_unwrap", align 4 %38 = load i8*, i8** %"'mi_cache", align 8, !invariant.group !17 tail call void @free(i8* nonnull %38) tail call void @free(i8* %30) br label %invert invert2: ; preds = %invert3 %39 = load float, float* %"'de22", align 4 store float 0.000000e+00, float* %"'de22", align 4 %40 = load float, float* %"'de23", align 4 %41 = fadd fast float %40, %39 store float %41, float* %"'de23", align 4 %42 = load float, float* %"'de23", align 4 store float 0.000000e+00, float* %"'de23", align 4 %"'il_phi5_unwrap" = load float*, float** %"'ipg25", align 8, !tbaa !9 %_unwrap26 = load i64, i64* %4, align 8, !tbaa !8, !invariant.group !21 %"'ipg24_unwrap" = getelementptr inbounds float, float* %"'il_phi5_unwrap", i64 %_unwrap26 %43 = load float, float* %"'ipg24_unwrap", align 4 %44 = fadd fast float %43, %42 store float %44, float* %"'ipg24_unwrap", align 4 br label %invert invert3: ; preds = %24 store float %differeturn, float* %"'de", align 4 %45 = load float, float* %"'de", align 4 store float 0.000000e+00, float* %"'de", align 4 %46 = xor i1 %7, true %47 = select fast i1 %46, float %45, float 0.000000e+00 %48 = load float, float* %"'de22", align 4 %49 = fadd fast float %48, %45 %50 = select fast i1 %7, float %48, float %49 store float %50, float* %"'de22", align 4 %51 = select fast i1 %7, float %45, float 0.000000e+00 %52 = load float, float* %"'de6", align 4 %53 = fadd fast float %52, %45 %54 = select fast i1 %7, float %53, float %52 store float %54, float* %"'de6", align 4 br i1 %7, label %invert1, label %invert2 } LLVM ERROR: function failed verification (4) PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace. Stack dump: 0. Program arguments: /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S 1. Running pass 'Enzyme Pass' on module '<stdin>'. #0 0x00007f2f2a2149a3 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbda9a3) #1 0x00007f2f2a212cc0 llvm::sys::RunSignalHandlers() (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbd8cc0) #2 0x00007f2f2a21500f (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbdb00f) #3 0x00007f2f296273c0 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x153c0) #4 0x00007f2f2911a18b raise (/lib/x86_64-linux-gnu/libc.so.6+0x4618b) #5 0x00007f2f290f9859 abort (/lib/x86_64-linux-gnu/libc.so.6+0x25859) #6 0x00007f2f2a16fc09 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35c09) #7 0x00007f2f2a16fa26 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35a26) #8 0x00007f2f26dc8a2a EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x58aa2a) #9 0x00007f2f26e4a44a AdjointGenerator<AugmentedReturn const*>::visitCallInst(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x60c44a) #10 0x00007f2f26e35014 llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::delegateCallInst(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5f7014) #11 0x00007f2f26e0bdad llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visitCall(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5cddad) #12 0x00007f2f26df929f llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visit(llvm::Instruction&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5bb29f) #13 0x00007f2f26deada7 llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visit(llvm::Instruction*) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5acda7) #14 0x00007f2f26dc62ed EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5882ed) #15 0x00007f2f26d97ddf (anonymous namespace)::Enzyme::HandleAutoDiff(llvm::CallInst*, llvm::TargetLibraryInfo&, bool, DerivativeMode) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x559ddf) #16 0x00007f2f26d9c593 (anonymous namespace)::Enzyme::lowerEnzymeCalls(llvm::Function&, bool, bool&, std::set<llvm::Function*, std::less<llvm::Function*>, std::allocator<llvm::Function*> >&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x55e593) #17 0x00007f2f26d9e6d6 (anonymous namespace)::Enzyme::runOnModule(llvm::Module&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5606d6) #18 0x00007f2f2a33c4ff llvm::legacy::PassManagerImpl::run(llvm::Module&) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xd024ff) #19 0x0000000000431dfb main (/usr/lib/llvm-12/bin/opt+0x431dfb) #20 0x00007f2f290fb0b3 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b3) #21 0x000000000041e68e _start (/usr/lib/llvm-12/bin/opt+0x41e68e) /usr/lib/llvm-12/bin/lli: error: 'main' function not found in module. --

******************** FAIL: Enzyme :: Integration/ReverseMode/mixedstruct1-simpleps.c (20 of 65) ******************** TEST 'Enzyme :: Integration/ReverseMode/mixedstruct1-simpleps.c' FAILED ******************** Script: -- : 'RUN: at line 1'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 -disable-llvm-passes /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 2'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 3'; /usr/lib/llvm-12/bin/clang -std=c11 -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 4'; /usr/lib/llvm-12/bin/clang -std=c11 -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 5'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 -disable-llvm-passes /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 6'; /usr/lib/llvm-12/bin/clang -std=c11 -O1 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 7'; /usr/lib/llvm-12/bin/clang -std=c11 -O2 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - : 'RUN: at line 8'; /usr/lib/llvm-12/bin/clang -std=c11 -O3 /workspaces/Enzyme/enzyme/test/Integration/ReverseMode/mixedstruct1-simpleps.c -S -emit-llvm -o - | /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -enzyme-inline=1 -S | /usr/lib/llvm-12/bin/lli - -- Exit Code: 255 Command Output (stderr): -- clang: warning: argument unused during compilation: '-disable-llvm-passes' [-Wunused-command-line-argument] Instruction does not dominate all uses! %"'ipg10" = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %"'", i64 0, i32 1 %"'il_phi_unwrap" = load float*, float** %"'ipg10", align 8, !tbaa !8 ; Function Attrs: nounwind uwtable define dso_local float @preprocess_tile_multiply(%struct.WINDOW_FRAME* noalias nocapture readonly %0) local_unnamed_addr #0 { %2 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 0 %3 = load i64, i64* %2, align 8, !tbaa !2 %4 = icmp slt i64 %3, 8 br i1 %4, label %5, label %18 5: ; preds = %1 %6 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 1 %7 = load float*, float** %6, align 8, !tbaa !8 %8 = getelementptr inbounds float, float* %7, i64 %3 %9 = load float, float* %8, align 4, !tbaa !9 %10 = call noalias dereferenceable_or_null(16) i8* @malloc(i64 16) #8 %11 = bitcast i8* %10 to %struct.WINDOW_FRAME* %12 = getelementptr inbounds i8, i8* %10, i64 8 %13 = bitcast i8* %12 to float** store float* %7, float** %13, align 8, !tbaa !8 %14 = add nsw i64 %3, 1 %15 = bitcast i8* %10 to i64* store i64 %14, i64* %15, align 8, !tbaa !2 %16 = call float @tile_multiply(%struct.WINDOW_FRAME* %11) call void @free(i8* %10) #8 %17 = fadd float %9, %16 br label %18 18: ; preds = %5, %1 %19 = phi float [ %17, %5 ], [ 0.000000e+00, %1 ] ret float %19 } ; Function Attrs: nounwind uwtable define internal void @diffetile_multiply(%struct.WINDOW_FRAME* noalias nocapture readonly %0, %struct.WINDOW_FRAME* nocapture %"'", float %differeturn) local_unnamed_addr #0 { %"'de" = alloca float, align 4 store float 0.000000e+00, float* %"'de", align 4 %"'de3" = alloca float, align 4 store float 0.000000e+00, float* %"'de3", align 4 %"'de4" = alloca float, align 4 store float 0.000000e+00, float* %"'de4", align 4 %"'de5" = alloca float, align 4 store float 0.000000e+00, float* %"'de5", align 4 %_cache = alloca i8*, align 8 %"'mi_cache" = alloca i8*, align 8 %2 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 0 %3 = load i64, i64* %2, align 8, !tbaa !2 %4 = icmp slt i64 %3, 8 br i1 %4, label %5, label %14 5: ; preds = %1 %"'ipg10" = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %"'", i64 0, i32 1 %6 = getelementptr inbounds %struct.WINDOW_FRAME, %struct.WINDOW_FRAME* %0, i64 0, i32 1 %"'ipl" = load float*, float** %"'ipg10", align 8 %7 = load float*, float** %6, align 8, !tbaa !8 %"'ipg9" = getelementptr inbounds float, float* %"'ipl", i64 %3 %8 = call noalias nonnull dereferenceable(16) dereferenceable_or_null(16) i8* @malloc(i64 16) #8 %"'mi" = call noalias nonnull dereferenceable(16) dereferenceable_or_null(16) i8* @malloc(i64 16) #8 call void @llvm.memset.p0i8.i64(i8* nonnull dereferenceable(16) dereferenceable_or_null(16) %"'mi", i8 0, i64 16, i1 false) %"'ipc" = bitcast i8* %"'mi" to %struct.WINDOW_FRAME* %9 = bitcast i8* %8 to %struct.WINDOW_FRAME* %"'ipg" = getelementptr inbounds i8, i8* %"'mi", i64 8 %10 = getelementptr inbounds i8, i8* %8, i64 8 %"'ipc8" = bitcast i8* %"'ipg" to float** %11 = bitcast i8* %10 to float** store float* %"'ipl", float** %"'ipc8", align 8 store float* %7, float** %11, align 8, !tbaa !8 store i8* %"'mi", i8** %"'mi_cache", align 8, !invariant.group !16 store i8* %8, i8** %_cache, align 8, !invariant.group !17 %12 = add nsw i64 %3, 1 %"'ipc7" = bitcast i8* %"'mi" to i64* %13 = bitcast i8* %8 to i64* store i64 %12, i64* %"'ipc7", align 8 store i64 %12, i64* %13, align 8, !tbaa !2 br label %14 14: ; preds = %5, %1 br label %invert2 invert: ; preds = %invert2, %invert1 ret void invert1: ; preds = %invert2 %15 = load float, float* %"'de3", align 4 store float 0.000000e+00, float* %"'de3", align 4 %16 = load float, float* %"'de4", align 4 %17 = fadd fast float %16, %15 store float %17, float* %"'de4", align 4 %18 = load float, float* %"'de5", align 4 %19 = fadd fast float %18, %15 store float %19, float* %"'de5", align 4 %20 = load i8*, i8** %_cache, align 8, !invariant.group !17 %_unwrap = bitcast i8* %20 to %struct.WINDOW_FRAME* %21 = load i8*, i8** %"'mi_cache", align 8, !invariant.group !16 %"'ipc_unwrap" = bitcast i8* %21 to %struct.WINDOW_FRAME* %22 = load float, float* %"'de5", align 4 call void @diffetile_multiply(%struct.WINDOW_FRAME* %_unwrap, %struct.WINDOW_FRAME* %"'ipc_unwrap", float %22) store float 0.000000e+00, float* %"'de5", align 4 %23 = load i8*, i8** %"'mi_cache", align 8, !invariant.group !16 tail call void @free(i8* nonnull %23) tail call void @free(i8* %20) %24 = load float, float* %"'de4", align 4 store float 0.000000e+00, float* %"'de4", align 4 %"'il_phi_unwrap" = load float*, float** %"'ipg10", align 8, !tbaa !8 %"'ipg9_unwrap" = getelementptr inbounds float, float* %"'il_phi_unwrap", i64 %3 %25 = load float, float* %"'ipg9_unwrap", align 4 %26 = fadd fast float %25, %24 store float %26, float* %"'ipg9_unwrap", align 4 br label %invert invert2: ; preds = %14 store float %differeturn, float* %"'de", align 4 %27 = load float, float* %"'de", align 4 store float 0.000000e+00, float* %"'de", align 4 %28 = select fast i1 %4, float %27, float 0.000000e+00 %29 = load float, float* %"'de3", align 4 %30 = fadd fast float %29, %27 %31 = select fast i1 %4, float %30, float %29 store float %31, float* %"'de3", align 4 br i1 %4, label %invert1, label %invert } LLVM ERROR: function failed verification (4) PLEASE submit a bug report to https://bugs.llvm.org/ and include the crash backtrace. Stack dump: 0. Program arguments: /usr/lib/llvm-12/bin/opt - -load=/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so -enzyme -S 1. Running pass 'Enzyme Pass' on module '<stdin>'. #0 0x00007f2662a7c9a3 llvm::sys::PrintStackTrace(llvm::raw_ostream&, int) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbda9a3) #1 0x00007f2662a7acc0 llvm::sys::RunSignalHandlers() (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbd8cc0) #2 0x00007f2662a7d00f (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xbdb00f) #3 0x00007f2661e8f3c0 __restore_rt (/lib/x86_64-linux-gnu/libpthread.so.0+0x153c0) #4 0x00007f266198218b raise (/lib/x86_64-linux-gnu/libc.so.6+0x4618b) #5 0x00007f2661961859 abort (/lib/x86_64-linux-gnu/libc.so.6+0x25859) #6 0x00007f26629d7c09 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35c09) #7 0x00007f26629d7a26 (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xb35a26) #8 0x00007f265f630a2a EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x58aa2a) #9 0x00007f265f6b244a AdjointGenerator<AugmentedReturn const*>::visitCallInst(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x60c44a) #10 0x00007f265f69d014 llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::delegateCallInst(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5f7014) #11 0x00007f265f673dad llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visitCall(llvm::CallInst&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5cddad) #12 0x00007f265f66129f llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visit(llvm::Instruction&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5bb29f) #13 0x00007f265f652da7 llvm::InstVisitor<AdjointGenerator<AugmentedReturn const*>, void>::visit(llvm::Instruction*) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5acda7) #14 0x00007f265f62e2ed EnzymeLogic::CreatePrimalAndGradient(llvm::Function*, DIFFE_TYPE, std::vector<DIFFE_TYPE, std::allocator<DIFFE_TYPE> > const&, llvm::TargetLibraryInfo&, TypeAnalysis&, bool, bool, DerivativeMode, llvm::Type*, FnTypeInfo const&, std::map<llvm::Argument*, bool, std::less<llvm::Argument*>, std::allocator<std::pair<llvm::Argument* const, bool> > >, AugmentedReturn const*, bool, bool, bool) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5882ed) #15 0x00007f265f5ffddf (anonymous namespace)::Enzyme::HandleAutoDiff(llvm::CallInst*, llvm::TargetLibraryInfo&, bool, DerivativeMode) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x559ddf) #16 0x00007f265f604593 (anonymous namespace)::Enzyme::lowerEnzymeCalls(llvm::Function&, bool, bool&, std::set<llvm::Function*, std::less<llvm::Function*>, std::allocator<llvm::Function*> >&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x55e593) #17 0x00007f265f6066d6 (anonymous namespace)::Enzyme::runOnModule(llvm::Module&) (/workspaces/Enzyme/enzyme/build/Enzyme/LLVMEnzyme-12.so+0x5606d6) #18 0x00007f2662ba44ff llvm::legacy::PassManagerImpl::run(llvm::Module&) (/usr/lib/llvm-12/lib/libLLVM-12.so.1+0xd024ff) #19 0x0000000000431dfb main (/usr/lib/llvm-12/bin/opt+0x431dfb) #20 0x00007f26619630b3 __libc_start_main (/lib/x86_64-linux-gnu/libc.so.6+0x270b3) #21 0x000000000041e68e _start (/usr/lib/llvm-12/bin/opt+0x41e68e) /usr/lib/llvm-12/bin/lli: error: 'main' function not found in module. --

Ah I see, yeah this should just be changed to check if the insertion point is in a reverse block and only then lookup.

https://gist.github.com/tgymnich/02d10c83ce9c018ec5e27d35478d8202
https://gist.github.com/tgymnich/257bcc1a2dfb6d4844cafc7227d4c0a6

wsmoses · 2021-09-14T02:00:42Z

enzyme/Enzyme/GradientUtils.h

    if (auto arg = dyn_cast<Argument>(ptr)) {
      assert(arg->getParent() == oldFunc);
    }
-    ptr = invertPointerM(ptr, BuilderM);


Can you change this to instead check if BuilderM is a forward pass block or not to choose whether to do the lookup?

wsmoses · 2021-09-14T02:01:26Z

enzyme/Enzyme/GradientUtils.h

+    case DerivativeMode::ForwardMode:
+      ptr = invertPointerM(origptr, BuilderM);
+      break;
+    case DerivativeMode::ReverseModePrimal:


Add an assert for reverse mode primal since this method should never be called in that case.

This reverts commit 864fd54.

tgymnich self-assigned this Aug 19, 2021

tgymnich mentioned this pull request Aug 19, 2021

Forward Mode CallInst #297

Merged

3 tasks

tgymnich force-pushed the move-assertions branch from 1ba72bc to d578d00 Compare September 1, 2021 15:23

tgymnich marked this pull request as ready for review September 1, 2021 18:51

wsmoses reviewed Sep 2, 2021

View reviewed changes

tgymnich force-pushed the move-assertions branch from f97e663 to 864fd54 Compare September 7, 2021 09:37

wsmoses reviewed Sep 14, 2021

View reviewed changes

wsmoses approved these changes Sep 14, 2021

View reviewed changes

tgymnich added 7 commits September 18, 2021 23:22

move assertions

e66ef80

don't use lookup in invertPointerM when running in forward mode

1d19c16

Revert "don't use lookup in invertPointerM when running in forward mode"

fe2328a

This reverts commit 864fd54.

stop using lookup inside invertPointer

15f9658

handle lookup in unwrap

6bf3154

add assertion

00d8030

only perform lookup for reverse builders

85c522d

tgymnich force-pushed the move-assertions branch from b03a851 to 85c522d Compare September 18, 2021 21:23

tgymnich merged commit 51988b0 into main Sep 19, 2021

tgymnich deleted the move-assertions branch September 20, 2021 21:53

Move Assertions #306

Move Assertions #306

Uh oh!

Conversation

tgymnich commented Aug 19, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tgymnich commented Sep 13, 2021

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tgymnich Sep 16, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tgymnich Sep 16, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tgymnich Sep 16, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

wsmoses Sep 14, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

tgymnich commented Aug 19, 2021 •

edited

Loading

tgymnich Sep 16, 2021 •

edited

Loading

tgymnich Sep 16, 2021 •

edited

Loading

tgymnich Sep 16, 2021 •

edited

Loading

wsmoses Sep 14, 2021 •

edited

Loading