From 1f0b92d3b23ea1b1e592aa4feee2ab49404b42d2 Mon Sep 17 00:00:00 2001 From: Nikita Popov Date: Thu, 30 May 2024 09:57:21 +0200 Subject: [PATCH] [ConstantFold] Remove non-trivial gep-of-gep fold This fold is subtly incorrect, because DL-unaware constant folding does not know the correct index type to use, and just performs the addition in the type that happens to already be there. This is incorrect, since sext(X)+sext(Y) is generally not the same as sext(X+Y). See the `@constexpr_gep_of_gep_with_narrow_type()` for a miscompile with the current implementation. One could try to restrict the fold to cases where no overflow occurs, but I'm not bothering with that here, because the DL-aware constant folding will take care of this anyway. I've only kept the straightforward zero-index case, where we just concatenate two GEPs. --- .../CodeGenCXX/2011-12-19-init-list-ctor.cpp | 2 +- ...initializer-stdinitializerlist-pr12086.cpp | 8 +- .../cxx0x-initializer-stdinitializerlist.cpp | 4 +- clang/test/OpenMP/threadprivate_codegen.cpp | 96 +++++++++---------- llvm/lib/IR/ConstantFold.cpp | 56 +---------- llvm/test/Other/constant-fold-gep.ll | 2 +- llvm/test/Transforms/InstCombine/gepgep.ll | 2 +- .../Transforms/InstCombine/getelementptr.ll | 3 +- .../InstCombine/ptrtoint-nullgep.ll | 16 ++-- 9 files changed, 70 insertions(+), 119 deletions(-) diff --git a/clang/test/CodeGenCXX/2011-12-19-init-list-ctor.cpp b/clang/test/CodeGenCXX/2011-12-19-init-list-ctor.cpp index 4033adc8f03909..6495fdebacf02f 100644 --- a/clang/test/CodeGenCXX/2011-12-19-init-list-ctor.cpp +++ b/clang/test/CodeGenCXX/2011-12-19-init-list-ctor.cpp @@ -22,5 +22,5 @@ struct S { // CHECK: call void @_ZN1AC1EPKc(ptr {{[^,]*}} getelementptr inbounds (%struct.S, ptr @arr, i32 0, i32 1), ptr noundef @.str) // CHECK: store i32 1, ptr getelementptr inbounds (%struct.S, ptr @arr, i64 1) // CHECK: call void @_ZN1AC1EPKc(ptr {{[^,]*}} getelementptr inbounds (%struct.S, ptr @arr, i64 1, i32 1), ptr noundef @.str.1) -// CHECK: store i32 2, ptr getelementptr inbounds (%struct.S, ptr @arr, i64 2) +// CHECK: store i32 2, ptr getelementptr inbounds (%struct.S, ptr getelementptr inbounds (%struct.S, ptr @arr, i64 1), i64 1) // CHECK: call void @_ZN1AC1EPKc(ptr {{[^,]*}} getelementptr inbounds (%struct.S, ptr @arr, i64 2, i32 1), ptr noundef @.str.2) diff --git a/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist-pr12086.cpp b/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist-pr12086.cpp index 6fbe4c7fd17a76..4f4c88d9e96a5c 100644 --- a/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist-pr12086.cpp +++ b/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist-pr12086.cpp @@ -83,8 +83,8 @@ std::initializer_list> nested = { // CHECK-DYNAMIC-BL: store i32 5, ptr @_ZGR6nested2_ // CHECK-DYNAMIC-BL: store i32 {{.*}}, ptr getelementptr inbounds (i32, ptr @_ZGR6nested2_, i64 1) // CHECK-DYNAMIC-BL: store ptr @_ZGR6nested2_, -// CHECK-DYNAMIC-BL: ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 2), align 8 -// CHECK-DYNAMIC-BL: store i64 2, ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 2, i32 1), align 8 +// CHECK-DYNAMIC-BL: ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 1), i64 1), align 8 +// CHECK-DYNAMIC-BL: store i64 2, ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 1), i64 1, i32 1), align 8 // CHECK-DYNAMIC-BL: store ptr @_ZGR6nested_, // CHECK-DYNAMIC-BL: ptr @nested, align 8 // CHECK-DYNAMIC-BL: store i64 3, ptr getelementptr inbounds ({{.*}}, ptr @nested, i32 0, i32 1), align 8 @@ -123,9 +123,9 @@ std::initializer_list> nested = { // CHECK-DYNAMIC-BE: store i32 5, ptr @_ZGR6nested2_ // CHECK-DYNAMIC-BE: store i32 {{.*}}, ptr getelementptr inbounds (i32, ptr @_ZGR6nested2_, i64 1) // CHECK-DYNAMIC-BE: store ptr @_ZGR6nested2_, -// CHECK-DYNAMIC-BE: ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 2), align 8 +// CHECK-DYNAMIC-BE: ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 1), i64 1), align 8 // CHECK-DYNAMIC-BE: store ptr getelementptr inbounds ([2 x i32], ptr @_ZGR6nested2_, i64 0, i64 2), -// CHECK-DYNAMIC-BE: ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 2, i32 1), align 8 +// CHECK-DYNAMIC-BE: ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr @_ZGR6nested_, i64 1), i64 1, i32 1), align 8 // CHECK-DYNAMIC-BE: store ptr @_ZGR6nested_, // CHECK-DYNAMIC-BE: ptr @nested, align 8 // CHECK-DYNAMIC-BE: store ptr getelementptr inbounds ([3 x {{.*}}], ptr @_ZGR6nested_, i64 0, i64 3), diff --git a/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist.cpp b/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist.cpp index aa2f078a5fb0ca..ac05910151a462 100644 --- a/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist.cpp +++ b/clang/test/CodeGenCXX/cxx0x-initializer-stdinitializerlist.cpp @@ -370,8 +370,8 @@ namespace partly_constant { // // Third init list. // CHECK-NOT: @[[PARTLY_CONSTANT_THIRD]], - // CHECK: store ptr {{.*}}@[[PARTLY_CONSTANT_THIRD]]{{.*}}, ptr getelementptr inbounds ({{.*}}, ptr {{.*}}@[[PARTLY_CONSTANT_INNER]]{{.*}}, i64 2) - // CHECK: store i64 4, ptr getelementptr inbounds ({{.*}}, ptr {{.*}}@[[PARTLY_CONSTANT_INNER]]{{.*}}, i64 2, i32 1) + // CHECK: store ptr {{.*}}@[[PARTLY_CONSTANT_THIRD]]{{.*}}, ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr {{.*}}@[[PARTLY_CONSTANT_INNER]]{{.*}}, i64 1), i64 1) + // CHECK: store i64 4, ptr getelementptr inbounds ({{.*}}, ptr getelementptr inbounds ({{.*}}, ptr {{.*}}@[[PARTLY_CONSTANT_INNER]]{{.*}}, i64 1), i64 1, i32 1) // CHECK-NOT: @[[PARTLY_CONSTANT_THIRD]], // // Outer init list. diff --git a/clang/test/OpenMP/threadprivate_codegen.cpp b/clang/test/OpenMP/threadprivate_codegen.cpp index b27783be829d66..7094decd410411 100644 --- a/clang/test/OpenMP/threadprivate_codegen.cpp +++ b/clang/test/OpenMP/threadprivate_codegen.cpp @@ -1260,8 +1260,8 @@ int foobar() { // CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK1-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // CHECK1: invoke.cont2: -// CHECK1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK1-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // CHECK1: invoke.cont3: // CHECK1-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -1273,8 +1273,8 @@ int foobar() { // CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK1-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // CHECK1: invoke.cont8: -// CHECK1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// CHECK1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK1-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // CHECK1: invoke.cont9: // CHECK1-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]] @@ -1783,8 +1783,8 @@ int foobar() { // CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK2-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // CHECK2: invoke.cont2: -// CHECK2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK2-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // CHECK2: invoke.cont3: // CHECK2-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -1796,8 +1796,8 @@ int foobar() { // CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK2-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // CHECK2: invoke.cont8: -// CHECK2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// CHECK2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK2-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // CHECK2: invoke.cont9: // CHECK2-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]] @@ -2458,8 +2458,8 @@ int foobar() { // SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // SIMD1-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // SIMD1: invoke.cont2: -// SIMD1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// SIMD1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // SIMD1-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // SIMD1: invoke.cont3: // SIMD1-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -2471,8 +2471,8 @@ int foobar() { // SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // SIMD1-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // SIMD1: invoke.cont8: -// SIMD1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// SIMD1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// SIMD1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // SIMD1-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // SIMD1: invoke.cont9: // SIMD1-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]] @@ -2922,8 +2922,8 @@ int foobar() { // SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // SIMD2-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG158:![0-9]+]] // SIMD2: invoke.cont2: -// SIMD2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG156]] -// SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// SIMD2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG156]] +// SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // SIMD2-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG159:![0-9]+]] // SIMD2: invoke.cont3: // SIMD2-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG154]] @@ -2935,8 +2935,8 @@ int foobar() { // SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // SIMD2-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG162:![0-9]+]] // SIMD2: invoke.cont8: -// SIMD2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG160]] -// SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// SIMD2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG160]] +// SIMD2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // SIMD2-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG163:![0-9]+]] // SIMD2: invoke.cont9: // SIMD2-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]], !dbg [[DBG164:![0-9]+]] @@ -3451,8 +3451,8 @@ int foobar() { // CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK-TLS1-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // CHECK-TLS1: invoke.cont2: -// CHECK-TLS1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK-TLS1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK-TLS1-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // CHECK-TLS1: invoke.cont3: // CHECK-TLS1-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -3464,8 +3464,8 @@ int foobar() { // CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK-TLS1-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // CHECK-TLS1: invoke.cont8: -// CHECK-TLS1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK-TLS1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// CHECK-TLS1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK-TLS1-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // CHECK-TLS1: invoke.cont9: // CHECK-TLS1-NEXT: [[TMP0:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]] @@ -4178,8 +4178,8 @@ int foobar() { // CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK-TLS2-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // CHECK-TLS2: invoke.cont2: -// CHECK-TLS2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK-TLS2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK-TLS2-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // CHECK-TLS2: invoke.cont3: // CHECK-TLS2-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -4191,8 +4191,8 @@ int foobar() { // CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK-TLS2-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // CHECK-TLS2: invoke.cont8: -// CHECK-TLS2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK-TLS2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// CHECK-TLS2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK-TLS2-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // CHECK-TLS2: invoke.cont9: // CHECK-TLS2-NEXT: [[TMP0:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR4]] @@ -4521,8 +4521,8 @@ int foobar() { // CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK-TLS3-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG189:![0-9]+]] // CHECK-TLS3: invoke.cont2: -// CHECK-TLS3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG187]] -// CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK-TLS3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG187]] +// CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK-TLS3-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG190:![0-9]+]] // CHECK-TLS3: invoke.cont3: // CHECK-TLS3-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG185]] @@ -4534,8 +4534,8 @@ int foobar() { // CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK-TLS3-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG193:![0-9]+]] // CHECK-TLS3: invoke.cont8: -// CHECK-TLS3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG191]] -// CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK-TLS3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG191]] +// CHECK-TLS3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK-TLS3-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG194:![0-9]+]] // CHECK-TLS3: invoke.cont9: // CHECK-TLS3-NEXT: [[TMP0:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]], !dbg [[DBG195:![0-9]+]] @@ -5280,8 +5280,8 @@ int foobar() { // CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // CHECK-TLS4-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG261:![0-9]+]] // CHECK-TLS4: invoke.cont2: -// CHECK-TLS4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG259]] -// CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// CHECK-TLS4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG259]] +// CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // CHECK-TLS4-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG262:![0-9]+]] // CHECK-TLS4: invoke.cont3: // CHECK-TLS4-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG257]] @@ -5293,8 +5293,8 @@ int foobar() { // CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // CHECK-TLS4-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG265:![0-9]+]] // CHECK-TLS4: invoke.cont8: -// CHECK-TLS4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG263]] -// CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// CHECK-TLS4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG263]] +// CHECK-TLS4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // CHECK-TLS4-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG266:![0-9]+]] // CHECK-TLS4: invoke.cont9: // CHECK-TLS4-NEXT: [[TMP0:%.*]] = call i32 @__cxa_thread_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR5]], !dbg [[DBG267:![0-9]+]] @@ -5571,8 +5571,8 @@ int foobar() { // SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // SIMD3-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]] // SIMD3: invoke.cont2: -// SIMD3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 -// SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// SIMD3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8 +// SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // SIMD3-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]] // SIMD3: invoke.cont3: // SIMD3-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8 @@ -5584,8 +5584,8 @@ int foobar() { // SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // SIMD3-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]] // SIMD3: invoke.cont8: -// SIMD3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 -// SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// SIMD3-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8 +// SIMD3-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // SIMD3-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]] // SIMD3: invoke.cont9: // SIMD3-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]] @@ -6035,8 +6035,8 @@ int foobar() { // SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // SIMD4-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG158:![0-9]+]] // SIMD4: invoke.cont2: -// SIMD4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG156]] -// SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// SIMD4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG156]] +// SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // SIMD4-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG159:![0-9]+]] // SIMD4: invoke.cont3: // SIMD4-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG154]] @@ -6048,8 +6048,8 @@ int foobar() { // SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // SIMD4-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG162:![0-9]+]] // SIMD4: invoke.cont8: -// SIMD4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG160]] -// SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// SIMD4-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG160]] +// SIMD4-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // SIMD4-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG163:![0-9]+]] // SIMD4: invoke.cont9: // SIMD4-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR3]], !dbg [[DBG164:![0-9]+]] @@ -6740,8 +6740,8 @@ int foobar() { // DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // DEBUG1-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG221:![0-9]+]] // DEBUG1: invoke.cont2: -// DEBUG1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG219]] -// DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// DEBUG1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG219]] +// DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // DEBUG1-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG222:![0-9]+]] // DEBUG1: invoke.cont3: // DEBUG1-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG217]] @@ -6753,8 +6753,8 @@ int foobar() { // DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // DEBUG1-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG225:![0-9]+]] // DEBUG1: invoke.cont8: -// DEBUG1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG223]] -// DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// DEBUG1-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG223]] +// DEBUG1-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // DEBUG1-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG226:![0-9]+]] // DEBUG1: invoke.cont9: // DEBUG1-NEXT: [[TMP0:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR4]], !dbg [[DBG227:![0-9]+]] @@ -7286,8 +7286,8 @@ int foobar() { // DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i32 noundef 2) // DEBUG2-NEXT: to label [[INVOKE_CONT2:%.*]] unwind label [[LPAD]], !dbg [[DBG171:![0-9]+]] // DEBUG2: invoke.cont2: -// DEBUG2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG169]] -// DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 2), i32 noundef 3) +// DEBUG2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT1]], align 8, !dbg [[DBG169]] +// DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr @arr_x, i64 1), i64 1), i32 noundef 3) // DEBUG2-NEXT: to label [[INVOKE_CONT3:%.*]] unwind label [[LPAD]], !dbg [[DBG172:![0-9]+]] // DEBUG2: invoke.cont3: // DEBUG2-NEXT: store ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), ptr [[ARRAYINIT_ENDOFINIT]], align 8, !dbg [[DBG167]] @@ -7299,8 +7299,8 @@ int foobar() { // DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i32 noundef 5) // DEBUG2-NEXT: to label [[INVOKE_CONT8:%.*]] unwind label [[LPAD6]], !dbg [[DBG175:![0-9]+]] // DEBUG2: invoke.cont8: -// DEBUG2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG173]] -// DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 2), i32 noundef 6) +// DEBUG2-NEXT: store ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), ptr [[ARRAYINIT_ENDOFINIT5]], align 8, !dbg [[DBG173]] +// DEBUG2-NEXT: invoke void @_ZN2S1C1Ei(ptr noundef nonnull align 4 dereferenceable(4) getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([[STRUCT_S1]], ptr getelementptr inbounds ([3 x %struct.S1], ptr @arr_x, i64 1), i64 1), i64 1), i32 noundef 6) // DEBUG2-NEXT: to label [[INVOKE_CONT9:%.*]] unwind label [[LPAD6]], !dbg [[DBG176:![0-9]+]] // DEBUG2: invoke.cont9: // DEBUG2-NEXT: [[TMP1:%.*]] = call i32 @__cxa_atexit(ptr @__cxx_global_array_dtor, ptr null, ptr @__dso_handle) #[[ATTR4]], !dbg [[DBG166]] diff --git a/llvm/lib/IR/ConstantFold.cpp b/llvm/lib/IR/ConstantFold.cpp index 9a7d437e6da6b3..a18ca19a9dccba 100644 --- a/llvm/lib/IR/ConstantFold.cpp +++ b/llvm/lib/IR/ConstantFold.cpp @@ -1429,64 +1429,16 @@ static Constant *foldGEPOfGEP(GEPOperator *GEP, Type *PointeeTy, bool InBounds, if (GEP->getInRange()) return nullptr; + // Only handle simple case with leading zero index. We cannot perform an + // actual addition as we don't know the correct index type size to use. Constant *Idx0 = cast(Idxs[0]); - if (Idx0->isNullValue()) { - // Handle the simple case of a zero index. - SmallVector NewIndices; - NewIndices.reserve(Idxs.size() + GEP->getNumIndices()); - NewIndices.append(GEP->idx_begin(), GEP->idx_end()); - NewIndices.append(Idxs.begin() + 1, Idxs.end()); - return ConstantExpr::getGetElementPtr( - GEP->getSourceElementType(), cast(GEP->getPointerOperand()), - NewIndices, InBounds && GEP->isInBounds()); - } - - gep_type_iterator LastI = gep_type_end(GEP); - for (gep_type_iterator I = gep_type_begin(GEP), E = gep_type_end(GEP); - I != E; ++I) - LastI = I; - - // We can't combine GEPs if the last index is a struct type. - if (!LastI.isSequential()) - return nullptr; - // We could perform the transform with non-constant index, but prefer leaving - // it as GEP of GEP rather than GEP of add for now. - ConstantInt *CI = dyn_cast(Idx0); - if (!CI) - return nullptr; - - // TODO: This code may be extended to handle vectors as well. - auto *LastIdx = cast(GEP->getOperand(GEP->getNumOperands()-1)); - Type *LastIdxTy = LastIdx->getType(); - if (LastIdxTy->isVectorTy()) + if (!Idx0->isNullValue()) return nullptr; SmallVector NewIndices; NewIndices.reserve(Idxs.size() + GEP->getNumIndices()); - NewIndices.append(GEP->idx_begin(), GEP->idx_end() - 1); - - // Add the last index of the source with the first index of the new GEP. - // Make sure to handle the case when they are actually different types. - if (LastIdxTy != Idx0->getType()) { - unsigned CommonExtendedWidth = - std::max(LastIdxTy->getIntegerBitWidth(), - Idx0->getType()->getIntegerBitWidth()); - CommonExtendedWidth = std::max(CommonExtendedWidth, 64U); - - Type *CommonTy = - Type::getIntNTy(LastIdxTy->getContext(), CommonExtendedWidth); - if (Idx0->getType() != CommonTy) - Idx0 = ConstantFoldCastInstruction(Instruction::SExt, Idx0, CommonTy); - if (LastIdx->getType() != CommonTy) - LastIdx = - ConstantFoldCastInstruction(Instruction::SExt, LastIdx, CommonTy); - if (!Idx0 || !LastIdx) - return nullptr; - } - - NewIndices.push_back(ConstantExpr::get(Instruction::Add, Idx0, LastIdx)); + NewIndices.append(GEP->idx_begin(), GEP->idx_end()); NewIndices.append(Idxs.begin() + 1, Idxs.end()); - return ConstantExpr::getGetElementPtr( GEP->getSourceElementType(), cast(GEP->getPointerOperand()), NewIndices, InBounds && GEP->isInBounds()); diff --git a/llvm/test/Other/constant-fold-gep.ll b/llvm/test/Other/constant-fold-gep.ll index 9af300ac9907f0..0090327f5e1e3d 100644 --- a/llvm/test/Other/constant-fold-gep.ll +++ b/llvm/test/Other/constant-fold-gep.ll @@ -104,7 +104,7 @@ ; Fold GEP of a GEP. Very simple cases are folded without targetdata. -; PLAIN: @Y = global ptr getelementptr inbounds ([3 x { i32, i32 }], ptr @ext, i64 2) +; PLAIN: @Y = global ptr getelementptr inbounds ([3 x { i32, i32 }], ptr getelementptr inbounds ([3 x { i32, i32 }], ptr @ext, i64 1), i64 1) ; PLAIN: @Z = global ptr getelementptr inbounds (i32, ptr getelementptr inbounds ([3 x { i32, i32 }], ptr @ext, i64 0, i64 1, i32 0), i64 1) ; OPT: @Y = local_unnamed_addr global ptr getelementptr inbounds (i8, ptr @ext, i64 48) ; OPT: @Z = local_unnamed_addr global ptr getelementptr inbounds (i8, ptr @ext, i64 12) diff --git a/llvm/test/Transforms/InstCombine/gepgep.ll b/llvm/test/Transforms/InstCombine/gepgep.ll index d2a0e1db392ba0..6b6a1597d560be 100644 --- a/llvm/test/Transforms/InstCombine/gepgep.ll +++ b/llvm/test/Transforms/InstCombine/gepgep.ll @@ -10,7 +10,7 @@ declare void @use(ptr) define void @f() { ; CHECK-LABEL: define void @f() { -; CHECK-NEXT: call void @use(ptr getelementptr (i8, ptr @buffer, i64 add (i64 sub (i64 0, i64 ptrtoint (ptr @buffer to i64)), i64 127))) +; CHECK-NEXT: call void @use(ptr getelementptr (i8, ptr getelementptr (i8, ptr @buffer, i64 add (i64 sub (i64 0, i64 ptrtoint (ptr @buffer to i64)), i64 63)), i64 64)) ; CHECK-NEXT: ret void ; call void @use(ptr getelementptr (i8, ptr getelementptr (i8, ptr @buffer, i64 add (i64 sub (i64 0, i64 ptrtoint (ptr @buffer to i64)), i64 63)), i64 64)) diff --git a/llvm/test/Transforms/InstCombine/getelementptr.ll b/llvm/test/Transforms/InstCombine/getelementptr.ll index 33ab1c37a2f43b..c91f307e818626 100644 --- a/llvm/test/Transforms/InstCombine/getelementptr.ll +++ b/llvm/test/Transforms/InstCombine/getelementptr.ll @@ -1683,10 +1683,9 @@ if.else: @g = external global i8 -; FIXME: This is a miscompile define ptr @constexpr_gep_of_gep_with_narrow_type() { ; CHECK-LABEL: @constexpr_gep_of_gep_with_narrow_type( -; CHECK-NEXT: ret ptr getelementptr (i8, ptr @g, i64 -2) +; CHECK-NEXT: ret ptr getelementptr (i8, ptr @g, i64 254) ; ret ptr getelementptr (i8, ptr getelementptr (i8, ptr @g, i8 127), i8 127) } diff --git a/llvm/test/Transforms/InstCombine/ptrtoint-nullgep.ll b/llvm/test/Transforms/InstCombine/ptrtoint-nullgep.ll index e6e63778d82f00..9d6f0abce84c7c 100644 --- a/llvm/test/Transforms/InstCombine/ptrtoint-nullgep.ll +++ b/llvm/test/Transforms/InstCombine/ptrtoint-nullgep.ll @@ -68,10 +68,10 @@ define void @constant_fold_ptrtoint_of_gep_of_nullgep() { ; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 1234) to i64)) ; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 1234) to i64)) ; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 1234) to i64)) -; LLPARSER-NEXT: call void @use_i64(i64 0) -; LLPARSER-NEXT: call void @use_i64(i64 0) -; LLPARSER-NEXT: call void @use_i64(i64 0) -; LLPARSER-NEXT: call void @use_i64(i64 0) +; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; LLPARSER-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) ; LLPARSER-NEXT: ret void ; ; INSTSIMPLIFY-LABEL: define {{[^@]+}}@constant_fold_ptrtoint_of_gep_of_nullgep() { @@ -83,10 +83,10 @@ define void @constant_fold_ptrtoint_of_gep_of_nullgep() { ; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 1234) to i64)) ; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 1234) to i64)) ; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 1234) to i64)) -; INSTSIMPLIFY-NEXT: call void @use_i64(i64 0) -; INSTSIMPLIFY-NEXT: call void @use_i64(i64 0) -; INSTSIMPLIFY-NEXT: call void @use_i64(i64 0) -; INSTSIMPLIFY-NEXT: call void @use_i64(i64 0) +; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr inbounds (i8, ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) +; INSTSIMPLIFY-NEXT: call void @use_i64(i64 ptrtoint (ptr addrspace(1) getelementptr (i8, ptr addrspace(1) getelementptr (i8, ptr addrspace(1) null, i64 -1), i64 1) to i64)) ; INSTSIMPLIFY-NEXT: ret void ; ; INSTCOMBINE-LABEL: define {{[^@]+}}@constant_fold_ptrtoint_of_gep_of_nullgep() {