[ConstantFolding] Canonicalize constexpr GEPs to i8 #89872
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
llvmbot
added
clang
|
@llvm/pr-subscribers-pgo @llvm/pr-subscribers-function-specialization Author: Nikita Popov (nikic) ChangesThis patch canonicalizes constant expression GEPs to use i8 source element type, aka ptradd. This is the ConstantFolding equivalent of the InstCombine canonicalization introduced in #68882. I believe all our optimizations working on constant expression GEPs (like GlobalOpt etc) have already been switched to work on offsets, so I don't expect any significant fallout from this change. This is part of: Patch is 258.02 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/89872.diff 96 Files Affected:
diff --git a/clang/test/CodeGen/RISCV/riscv-inline-asm.c b/clang/test/CodeGen/RISCV/riscv-inline-asm.c
index 3565705dea7130..ed97add95e7116 100644
--- a/clang/test/CodeGen/RISCV/riscv-inline-asm.c
+++ b/clang/test/CodeGen/RISCV/riscv-inline-asm.c
@@ -49,9 +49,9 @@ extern int var, arr[2][2];
struct Pair { int a, b; } pair;
// CHECK-LABEL: test_s(
-// CHECK: call void asm sideeffect "// $0 $1 $2", "s,s,s"(ptr nonnull @var, ptr nonnull getelementptr inbounds ([2 x [2 x i32]], ptr @arr, {{.*}}), ptr nonnull @test_s)
-// CHECK: call void asm sideeffect "// $0", "s"(ptr nonnull getelementptr inbounds (%struct.Pair, ptr @pair, {{.*}}))
-// CHECK: call void asm sideeffect "// $0 $1 $2", "S,S,S"(ptr nonnull @var, ptr nonnull getelementptr inbounds ([2 x [2 x i32]], ptr @arr, {{.*}}), ptr nonnull @test_s)
+// CHECK: call void asm sideeffect "// $0 $1 $2", "s,s,s"(ptr nonnull @var, ptr nonnull getelementptr inbounds (i8, ptr @arr, {{.*}}), ptr nonnull @test_s)
+// CHECK: call void asm sideeffect "// $0", "s"(ptr nonnull getelementptr inbounds (i8, ptr @pair, {{.*}}))
+// CHECK: call void asm sideeffect "// $0 $1 $2", "S,S,S"(ptr nonnull @var, ptr nonnull getelementptr inbounds (i8, ptr @arr, {{.*}}), ptr nonnull @test_s)
void test_s(void) {
asm("// %0 %1 %2" :: "s"(&var), "s"(&arr[1][1]), "s"(test_s));
asm("// %0" :: "s"(&pair.b));
diff --git a/clang/test/CodeGen/attr-counted-by.c b/clang/test/CodeGen/attr-counted-by.c
index 1fb39f9a346667..6fc563aba2ae22 100644
--- a/clang/test/CodeGen/attr-counted-by.c
+++ b/clang/test/CodeGen/attr-counted-by.c
@@ -66,7 +66,7 @@ struct anon_struct {
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = icmp ult i64 [[IDXPROM]], [[TMP0]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP1]], label [[CONT3:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3:![0-9]+]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB2:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10:[0-9]+]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB1:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10:[0-9]+]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont3:
// SANITIZE-WITH-ATTR-NEXT: [[ARRAY:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 12
@@ -114,7 +114,7 @@ void test1(struct annotated *p, int index, int val) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], [[INDEX]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP1]], label [[CONT3:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB4:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB3:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont3:
// SANITIZE-WITH-ATTR-NEXT: [[ARRAY:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 12
@@ -203,7 +203,7 @@ size_t test2_bdos(struct annotated *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = icmp ugt i64 [[TMP0]], [[INDEX]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP1]], label [[CONT3:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB5:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB4:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont3:
// SANITIZE-WITH-ATTR-NEXT: [[ARRAY:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 12
@@ -308,7 +308,7 @@ size_t test3_bdos(struct annotated *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = icmp ult i64 [[IDXPROM]], [[TMP0]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP1]], label [[CONT4:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB6:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB5:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont4:
// SANITIZE-WITH-ATTR-NEXT: [[TMP2:%.*]] = icmp sgt i32 [[DOT_COUNTED_BY_LOAD]], 2
@@ -325,7 +325,7 @@ size_t test3_bdos(struct annotated *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP7:%.*]] = icmp ult i64 [[IDXPROM13]], [[TMP6]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP7]], label [[CONT20:%.*]], label [[HANDLER_OUT_OF_BOUNDS16:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds16:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB7:[0-9]+]], i64 [[IDXPROM13]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB6:[0-9]+]], i64 [[IDXPROM13]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont20:
// SANITIZE-WITH-ATTR-NEXT: [[TMP8:%.*]] = icmp sgt i32 [[DOT_COUNTED_BY_LOAD7]], 3
@@ -342,7 +342,7 @@ size_t test3_bdos(struct annotated *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP13:%.*]] = icmp ult i64 [[IDXPROM30]], [[TMP12]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP13]], label [[CONT37:%.*]], label [[HANDLER_OUT_OF_BOUNDS33:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds33:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB8:[0-9]+]], i64 [[IDXPROM30]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB7:[0-9]+]], i64 [[IDXPROM30]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont37:
// SANITIZE-WITH-ATTR-NEXT: [[ARRAYIDX35:%.*]] = getelementptr inbounds [0 x i32], ptr [[ARRAY]], i64 0, i64 [[IDXPROM30]]
@@ -494,7 +494,7 @@ size_t test4_bdos(struct annotated *p, int index) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP0:%.*]] = icmp ugt i64 [[DOT_COUNTED_BY_LOAD]], [[IDXPROM]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP0]], label [[CONT3:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB9:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB8:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont3:
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 16
@@ -590,7 +590,7 @@ size_t test5_bdos(struct anon_struct *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP0:%.*]] = icmp ugt i64 [[DOT_COUNTED_BY_LOAD]], [[IDXPROM]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP0]], label [[CONT3:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB10:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB9:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont3:
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 16
@@ -683,7 +683,7 @@ size_t test6_bdos(struct anon_struct *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP2:%.*]] = icmp ult i64 [[IDXPROM]], [[TMP1]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP2]], label [[CONT7:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB12:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB11:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont7:
// SANITIZE-WITH-ATTR-NEXT: [[INTS:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 9
@@ -756,7 +756,7 @@ size_t test7_bdos(struct union_of_fams *p) {
// SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = icmp ult i64 [[IDXPROM]], [[TMP0]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP1]], label [[CONT9:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB13:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB12:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont9:
// SANITIZE-WITH-ATTR-NEXT: [[INTS:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 9
@@ -1095,10 +1095,10 @@ int test12_a, test12_b;
// SANITIZE-WITH-ATTR-NEXT: [[DOTNOT:%.*]] = icmp eq i32 [[DOTCOUNTED_BY_LOAD]], 0
// SANITIZE-WITH-ATTR-NEXT: br i1 [[DOTNOT]], label [[HANDLER_OUT_OF_BOUNDS4:%.*]], label [[HANDLER_TYPE_MISMATCH6:%.*]], !prof [[PROF10:![0-9]+]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds4:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB19:[0-9]+]], i64 0) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB20:[0-9]+]], i64 0) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.type_mismatch6:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_type_mismatch_v1_abort(ptr nonnull @[[GLOB20:[0-9]+]], i64 ptrtoint (ptr getelementptr inbounds ([[STRUCT_ANON_5:%.*]], ptr @test12_foo, i64 1, i32 0, i32 0, i32 0) to i64)) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_type_mismatch_v1_abort(ptr nonnull @[[GLOB21:[0-9]+]], i64 ptrtoint (ptr getelementptr inbounds (i8, ptr @test12_foo, i64 4) to i64)) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
//
// NO-SANITIZE-WITH-ATTR-LABEL: define dso_local noundef i32 @test12(
@@ -1111,7 +1111,7 @@ int test12_a, test12_b;
// NO-SANITIZE-WITH-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [6 x i32], ptr [[BAZ]], i64 0, i64 [[IDXPROM]]
// NO-SANITIZE-WITH-ATTR-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITH-ATTR-NEXT: store i32 [[TMP0]], ptr @test12_b, align 4, !tbaa [[TBAA2]]
-// NO-SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds ([[STRUCT_ANON_5:%.*]], ptr @test12_foo, i64 1, i32 0, i32 0, i32 0), align 4, !tbaa [[TBAA2]]
+// NO-SANITIZE-WITH-ATTR-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds (i8, ptr @test12_foo, i64 4), align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITH-ATTR-NEXT: store i32 [[TMP1]], ptr @test12_a, align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITH-ATTR-NEXT: br label [[FOR_COND:%.*]]
// NO-SANITIZE-WITH-ATTR: for.cond:
@@ -1140,7 +1140,7 @@ int test12_a, test12_b;
// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB4:[0-9]+]], i64 0) #[[ATTR8]], !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR-NEXT: unreachable, !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR: handler.type_mismatch6:
-// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @__ubsan_handle_type_mismatch_v1_abort(ptr nonnull @[[GLOB5:[0-9]+]], i64 ptrtoint (ptr getelementptr inbounds ([[STRUCT_ANON_5:%.*]], ptr @test12_foo, i64 1, i32 0, i32 0, i32 0) to i64)) #[[ATTR8]], !nosanitize [[META9]]
+// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @__ubsan_handle_type_mismatch_v1_abort(ptr nonnull @[[GLOB5:[0-9]+]], i64 ptrtoint (ptr getelementptr inbounds (i8, ptr @test12_foo, i64 4) to i64)) #[[ATTR8]], !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR-NEXT: unreachable, !nosanitize [[META9]]
//
// NO-SANITIZE-WITHOUT-ATTR-LABEL: define dso_local noundef i32 @test12(
@@ -1153,7 +1153,7 @@ int test12_a, test12_b;
// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [6 x i32], ptr [[BAZ]], i64 0, i64 [[IDXPROM]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: store i32 [[TMP0]], ptr @test12_b, align 4, !tbaa [[TBAA2]]
-// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds ([[STRUCT_ANON_5:%.*]], ptr @test12_foo, i64 1, i32 0, i32 0, i32 0), align 4, !tbaa [[TBAA2]]
+// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[TMP1:%.*]] = load i32, ptr getelementptr inbounds (i8, ptr @test12_foo, i64 4), align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: store i32 [[TMP1]], ptr @test12_a, align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: br label [[FOR_COND:%.*]]
// NO-SANITIZE-WITHOUT-ATTR: for.cond:
@@ -1188,7 +1188,7 @@ struct test13_bar {
// SANITIZE-WITH-ATTR-NEXT: [[TMP2:%.*]] = icmp ugt i64 [[TMP1]], [[INDEX]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP2]], label [[CONT5:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB23:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB24:[0-9]+]], i64 [[INDEX]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: cont5:
// SANITIZE-WITH-ATTR-NEXT: [[REVMAP:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 16
@@ -1249,7 +1249,7 @@ struct test14_foo {
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP0]], label [[TRAP:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
// SANITIZE-WITH-ATTR-NEXT: [[IDXPROM:%.*]] = sext i32 [[IDX]] to i64
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB24:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB25:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: trap:
// SANITIZE-WITH-ATTR-NEXT: tail call void @llvm.trap() #[[ATTR10]]
@@ -1305,7 +1305,7 @@ int test14(int idx) {
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP0]], label [[TRAP:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
// SANITIZE-WITH-ATTR-NEXT: [[IDXPROM:%.*]] = sext i32 [[IDX]] to i64
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB25:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB27:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR10]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR-NEXT: unreachable, !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: trap:
// SANITIZE-WITH-ATTR-NEXT: tail call void @llvm.trap() #[[ATTR10]]
@@ -1315,7 +1315,7 @@ int test14(int idx) {
// NO-SANITIZE-WITH-ATTR-SAME: i32 noundef [[IDX:%.*]]) local_unnamed_addr #[[ATTR4]] {
// NO-SANITIZE-WITH-ATTR-NEXT: entry:
// NO-SANITIZE-WITH-ATTR-NEXT: [[IDXPROM:%.*]] = sext i32 [[IDX]] to i64
-// NO-SANITIZE-WITH-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [0 x i32], ptr getelementptr inbounds ([[STRUCT_ANON_8:%.*]], ptr @__const.test15.foo, i64 1, i32 0), i64 0, i64 [[IDXPROM]]
+// NO-SANITIZE-WITH-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [0 x i32], ptr getelementptr inbounds (i8, ptr @__const.test15.foo, i64 8), i64 0, i64 [[IDXPROM]]
// NO-SANITIZE-WITH-ATTR-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITH-ATTR-NEXT: ret i32 [[TMP0]]
//
@@ -1326,7 +1326,7 @@ int test14(int idx) {
// SANITIZE-WITHOUT-ATTR-NEXT: br i1 [[TMP0]], label [[TRAP:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF8]], !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR: handler.out_of_bounds:
// SANITIZE-WITHOUT-ATTR-NEXT: [[IDXPROM:%.*]] = sext i32 [[IDX]] to i64
-// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB10:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR8]], !nosanitize [[META9]]
+// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB11:[0-9]+]], i64 [[IDXPROM]]) #[[ATTR8]], !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR-NEXT: unreachable, !nosanitize [[META9]]
// SANITIZE-WITHOUT-ATTR: trap:
// SANITIZE-WITHOUT-ATTR-NEXT: tail call void @llvm.trap() #[[ATTR8]]
@@ -1336,7 +1336,7 @@ int test14(int idx) {
// NO-SANITIZE-WITHOUT-ATTR-SAME: i32 noundef [[IDX:%.*]]) local_unnamed_addr #[[ATTR1]] {
// NO-SANITIZE-WITHOUT-ATTR-NEXT: entry:
// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[IDXPROM:%.*]] = sext i32 [[IDX]] to i64
-// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [0 x i32], ptr getelementptr inbounds ([[STRUCT_ANON_8:%.*]], ptr @__const.test15.foo, i64 1, i32 0), i64 0, i64 [[IDXPROM]]
+// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds [0 x i32], ptr getelementptr inbounds (i8, ptr @__const.test15.foo, i64 8), i64 0, i64 [[IDXPROM]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4, !tbaa [[TBAA2]]
// NO-SANITIZE-WITHOUT-ATTR-NEXT: ret i32 [[TMP0]]
//
@@ -1487,7 +1487,7 @@ struct tests_foo {
// SANITIZE-WITH-ATTR-NEXT: [[TMP0:%.*]] = icmp ugt i32 [[DOTCOUNTED_BY_LOAD]], 10
// SANITIZE-WITH-ATTR-NEXT: br i1 [[TMP0]], label [[CONT4:%.*]], label [[HANDLER_OUT_OF_BOUNDS:%.*]], !prof [[PROF3]], !nosanitize [[META2]]
// SANITIZE-WITH-ATTR: handler.out_of_bounds:
-// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB26:[0-9]+]], i64 10) #[[ATTR10]], !nosanitize [[META2]]
+// SANITIZE-WITH-ATTR-NEXT: tail call void @__ubsan_handle_out_of_bounds_abort(ptr nonnull @[[GLOB28:[0-9]+]], i64 10) #[[ATTR10]], !nosanitize [[M...
[truncated]
|
nikic
commented
Apr 24, 2024
| bool CanBeNull, CanBeFreed; | ||
| uint64_t DerefBytes = | ||
| Ptr->getPointerDereferenceableBytes(DL, CanBeNull, CanBeFreed); | ||
| InBounds = DerefBytes != 0 && !CanBeNull && Offset.sle(DerefBytes); |
There was a problem hiding this comment.
I've added this inference in this patch, because we lose out on this inference in the DataLayout-independent constant folding:
llvm-project/llvm/lib/IR/ConstantFold.cpp
Lines 1717 to 1724 in 88b6186
There was a problem hiding this comment.
can we remove the other constant folding?
There was a problem hiding this comment.
I don't think we'd want to remove it at this point as it's still used in some key places like initial IR construction in clang -- longer term, I'd like to get rid of the two separate constant folding implementations altogether.
|
Hi @nikic |
The major patches have been #68882, #84341 and this one. There have been some smaller changes in between to fix related optimization regressions. |
nikic
force-pushed
the
constexpr-gep-i8-canon
branch
from
e78a5eb to
2fa7299
Compare
You can test this locally with the following command:git-clang-format --diff f6527774569790b5a5236f6e84f3f839ce6c2fff 10ed97d2fd16475736b7d2f7b67a4e3e4e0406a1 -- clang/test/CodeGen/RISCV/riscv-inline-asm.c clang/test/CodeGen/attr-counted-by.c clang/test/CodeGenCXX/atomicinit.cpp clang/test/CodeGenCXX/auto-var-init.cpp clang/test/Profile/c-unreachable-after-switch.c compiler-rt/test/profile/Linux/counter_promo_for.c compiler-rt/test/profile/Linux/counter_promo_while.c llvm/lib/Analysis/ConstantFolding.cppView the diff from clang-format here.diff --git a/compiler-rt/test/profile/Linux/counter_promo_for.c b/compiler-rt/test/profile/Linux/counter_promo_for.c
index aa77e6084b..d843510697 100644
--- a/compiler-rt/test/profile/Linux/counter_promo_for.c
+++ b/compiler-rt/test/profile/Linux/counter_promo_for.c
@@ -18,30 +18,30 @@ int g;
__attribute__((noinline)) void bar(int i) { g += i; }
__attribute__((noinline)) void foo(int n, int N) {
-// PROMO-LABEL: @foo
-// PROMO: load{{.*}}@__profc_foo{{.*}} 24){{.*}}
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 24){{.*}}
-// PROMO: load{{.*}}@__profc_foo, align
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo, align
-// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// PROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
-//
-// NOPROMO-LABEL: @foo
-// NOPROMO: load{{.*}}@__profc_foo, align
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo, align
-// NOPROMO: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// NOPROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // PROMO-LABEL: @foo
+ // PROMO: load{{.*}}@__profc_foo{{.*}} 24){{.*}}
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 24){{.*}}
+ // PROMO: load{{.*}}@__profc_foo, align
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo, align
+ // PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // PROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ //
+ // NOPROMO-LABEL: @foo
+ // NOPROMO: load{{.*}}@__profc_foo, align
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo, align
+ // NOPROMO: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // NOPROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
int i;
for (i = 0; i < N; i++) {
if (i < n + 1)
diff --git a/compiler-rt/test/profile/Linux/counter_promo_while.c b/compiler-rt/test/profile/Linux/counter_promo_while.c
index c6ea3a7282..e0a9ee3c24 100644
--- a/compiler-rt/test/profile/Linux/counter_promo_while.c
+++ b/compiler-rt/test/profile/Linux/counter_promo_while.c
@@ -16,27 +16,27 @@
int g;
__attribute__((noinline)) void bar(int i) { g += i; }
__attribute__((noinline)) void foo(int n, int N) {
-// PROMO-LABEL: @foo
-// PROMO: load{{.*}}@__profc_foo, align
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo, align
-// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
-// PROMO-NEXT: add
-// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
-//
-// NOPROMO-LABEL: @foo
-// NOPROMO: load{{.*}}@__profc_foo, align
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo, align
-// NOPROMO: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
-// NOPROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
-// NOPROMO-NEXT: add
-// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // PROMO-LABEL: @foo
+ // PROMO: load{{.*}}@__profc_foo, align
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo, align
+ // PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // PROMO-NEXT: add
+ // PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ //
+ // NOPROMO-LABEL: @foo
+ // NOPROMO: load{{.*}}@__profc_foo, align
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo, align
+ // NOPROMO: load{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 8){{.*}}
+ // NOPROMO: load{{.*}}@__profc_foo{{.*}} 16){{.*}}
+ // NOPROMO-NEXT: add
+ // NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 16){{.*}}
int i = 0;
while (i < N) {
if (i < n + 1)
|
aeubanks
reviewed
Apr 29, 2024
| Indices.push_back(APInt::getZero(isa<StructType>(ElemTy) ? 32 : BitWidth)); | ||
| ElemTy = NextTy; | ||
| // Try to infer inbounds for GEPs of globals. | ||
| if (!InBounds && Offset.isNonNegative()) { |
There was a problem hiding this comment.
is there a test for the case where !Offset.isNonNegative()?
There was a problem hiding this comment.
https://github.com/llvm/llvm-project/blob/main/llvm/test/Transforms/GlobalOpt/large-element-size.ll gets an incorrect inbounds without the check.
|
btw we're still looking into a performance regression caused by #68882 that still repros with LLVM head, even after the SROA enhancements. this patch looks good, but can we hold off a bit on submitting this? |
Sure! |
aeubanks
approved these changes
May 16, 2024
This patch canonicalizes constant expression GEPs to use i8 source element type, aka ptradd. This is the ConstantFolding equivalent of the InstCombine canonicalization introduced in llvm#68882. I believe all our optimizations working on constant expression GEPs (like GlobalOpt etc) have already been switched to work on offsets, so I don't expect any significant fallout from this change. This is part of: https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699
nikic
force-pushed
the
constexpr-gep-i8-canon
branch
from
2fa7299 to
10ed97d
Compare
MaheshRavishankar
pushed a commit
to iree-org/llvm-project
that referenced
this pull request
Jun 4, 2024
…)" This reverts commit 8e8d259.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This patch canonicalizes constant expression GEPs to use i8 source element type, aka ptradd. This is the ConstantFolding equivalent of the InstCombine canonicalization introduced in #68882.
I believe all our optimizations working on constant expression GEPs (like GlobalOpt etc) have already been switched to work on offsets, so I don't expect any significant fallout from this change.
This is part of:
https://discourse.llvm.org/t/rfc-replacing-getelementptr-with-ptradd/68699