diff options
Diffstat (limited to 'clang/test/CodeGenCXX')
4 files changed, 69 insertions, 25 deletions
diff --git a/clang/test/CodeGenCXX/pragma-loop-safety-imperfectly_nested.cpp b/clang/test/CodeGenCXX/pragma-loop-safety-imperfectly_nested.cpp new file mode 100644 index 00000000000..da060f7902e --- /dev/null +++ b/clang/test/CodeGenCXX/pragma-loop-safety-imperfectly_nested.cpp @@ -0,0 +1,32 @@ +// RUN: %clang_cc1 -triple x86_64-apple-darwin -std=c++11 -emit-llvm -o - %s | FileCheck %s + +// Verify that the outer loop has the llvm.access.group property for the +// accesses outside and inside the inner loop, even when the inner loop +// is not perfectly nested. +void vectorize_imperfectly_nested_test(int *List, int Length) { +#pragma clang loop vectorize(assume_safety) interleave(disable) unroll(disable) + for (int i = 0; i < Length; ++i) { + List[i * Length] = 42; +#pragma clang loop vectorize(assume_safety) interleave(disable) unroll(disable) + for (int j = 1; j < Length - 1; ++j) + List[i * Length + j] = (i + j) * 2; + List[(i + 1) * Length - 1] = 21; + } +} + + +// CHECK: load i32, i32* %Length.addr, align 4, !llvm.access.group ![[ACCESS_GROUP_2:[0-9]+]] + +// CHECK: %[[MUL:.+]] = mul nsw i32 %add, 2 +// CHECK: store i32 %[[MUL]], i32* %{{.+}}, !llvm.access.group ![[ACCESS_GROUP_3:[0-9]+]] +// CHECK: br label %{{.+}}, !llvm.loop ![[INNER_LOOPID:[0-9]+]] +// CHECK: store i32 21, i32* %{{.+}}, !llvm.access.group ![[ACCESS_GROUP_2]] +// CHECK: br label %{{.+}}, !llvm.loop ![[OUTER_LOOPID:[0-9]+]] + +// CHECK: ![[ACCESS_GROUP_2]] = distinct !{} +// CHECK: ![[ACCESS_GROUP_LIST_3:[0-9]+]] = !{![[ACCESS_GROUP_2]], ![[ACCESS_GROUP_4:[0-9]+]]} +// CHECK: ![[ACCESS_GROUP_4]] = distinct !{} +// CHECK: ![[INNER_LOOPID]] = distinct !{![[INNER_LOOPID]], {{.*}} ![[PARALLEL_ACCESSES_8:[0-9]+]]} +// CHECK: ![[PARALLEL_ACCESSES_8]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_4]]} +// CHECK: ![[OUTER_LOOPID]] = distinct !{![[OUTER_LOOPID]], {{.*}} ![[PARALLEL_ACCESSES_10:[0-9]+]]} +// CHECK: ![[PARALLEL_ACCESSES_10]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_2]]} diff --git a/clang/test/CodeGenCXX/pragma-loop-safety-nested.cpp b/clang/test/CodeGenCXX/pragma-loop-safety-nested.cpp index 4641c953ee6..deec06bbc82 100644 --- a/clang/test/CodeGenCXX/pragma-loop-safety-nested.cpp +++ b/clang/test/CodeGenCXX/pragma-loop-safety-nested.cpp @@ -1,7 +1,7 @@ // RUN: %clang_cc1 -triple x86_64-apple-darwin -std=c++11 -emit-llvm -o - %s | FileCheck %s -// Verify that the inner access is tagged with a parallel_loop_access -// for the inner and outer loop using a list. +// Verify that the outer loop has the llvm.access.group property for the +// accesses outside and inside the inner loop. void vectorize_nested_test(int *List, int Length) { #pragma clang loop vectorize(assume_safety) interleave(disable) unroll(disable) for (int i = 0; i < Length; ++i) { @@ -11,11 +11,17 @@ void vectorize_nested_test(int *List, int Length) { } } + +// CHECK: load i32, i32* %Length.addr, align 4, !llvm.access.group ![[ACCESS_GROUP_2:[0-9]+]] // CHECK: %[[MUL:.+]] = mul -// CHECK: store i32 %[[MUL]], i32* %{{.+}}, !llvm.mem.parallel_loop_access ![[PARALLEL_LIST:[0-9]+]] +// CHECK: store i32 %[[MUL]], i32* %{{.+}}, !llvm.access.group ![[ACCESS_GROUP_LIST_3:[0-9]+]] // CHECK: br label %{{.+}}, !llvm.loop ![[INNER_LOOPID:[0-9]+]] // CHECK: br label %{{.+}}, !llvm.loop ![[OUTER_LOOPID:[0-9]+]] -// CHECK: ![[OUTER_LOOPID]] = distinct !{![[OUTER_LOOPID]], -// CHECK: ![[PARALLEL_LIST]] = !{![[OUTER_LOOPID]], ![[INNER_LOOPID]]} -// CHECK: ![[INNER_LOOPID]] = distinct !{![[INNER_LOOPID]], +// CHECK: ![[ACCESS_GROUP_2]] = distinct !{} +// CHECK: ![[ACCESS_GROUP_LIST_3]] = !{![[ACCESS_GROUP_2]], ![[ACCESS_GROUP_4:[0-9]+]]} +// CHECK: ![[ACCESS_GROUP_4]] = distinct !{} +// CHECK: ![[INNER_LOOPID]] = distinct !{![[INNER_LOOPID]], {{.*}} ![[PARALLEL_ACCESSES_8:[0-9]+]]} +// CHECK: ![[PARALLEL_ACCESSES_8]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_4]]} +// CHECK: ![[OUTER_LOOPID]] = distinct !{![[OUTER_LOOPID]], {{.*}} ![[PARALLEL_ACCESSES_10:[0-9]+]]} +// CHECK: ![[PARALLEL_ACCESSES_10]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_2]]} diff --git a/clang/test/CodeGenCXX/pragma-loop-safety-outer.cpp b/clang/test/CodeGenCXX/pragma-loop-safety-outer.cpp index 83f534e6512..d99b86ffe23 100644 --- a/clang/test/CodeGenCXX/pragma-loop-safety-outer.cpp +++ b/clang/test/CodeGenCXX/pragma-loop-safety-outer.cpp @@ -1,7 +1,7 @@ // RUN: %clang_cc1 -triple x86_64-apple-darwin -std=c++11 -emit-llvm -o - %s | FileCheck %s -// Verify that the inner access is tagged with a parallel_loop_access -// for the outer loop. +// Verify that the outer loop has the inner loop's access in its +// llvm.loop.parallel_accesses property. void vectorize_outer_test(int *List, int Length) { #pragma clang loop vectorize(assume_safety) interleave(disable) unroll(disable) for (int i = 0; i < Length; i += 2) { @@ -12,9 +12,11 @@ void vectorize_outer_test(int *List, int Length) { } // CHECK: %[[MUL:.+]] = mul -// CHECK: store i32 %[[MUL]], i32* %{{.+}}, !llvm.mem.parallel_loop_access ![[OUTER_LOOPID:[0-9]+]] +// CHECK: store i32 %[[MUL]], i32* %{{.+}}, !llvm.access.group ![[ACCESS_GROUP_2:[0-9]+]] // CHECK: br label %{{.+}}, !llvm.loop ![[INNER_LOOPID:[0-9]+]] -// CHECK: br label %{{.+}}, !llvm.loop ![[OUTER_LOOPID]] +// CHECK: br label %{{.+}}, !llvm.loop ![[OUTER_LOOPID:[0-9]+]] -// CHECK: ![[OUTER_LOOPID]] = distinct !{![[OUTER_LOOPID]], +// CHECK: ![[ACCESS_GROUP_2]] = distinct !{} // CHECK: ![[INNER_LOOPID]] = distinct !{![[INNER_LOOPID]], +// CHECK: ![[OUTER_LOOPID]] = distinct !{![[OUTER_LOOPID]], {{.*}} ![[PARALLEL_ACCESSES_9:[0-9]+]]} +// CHECK: ![[PARALLEL_ACCESSES_9]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_2]]} diff --git a/clang/test/CodeGenCXX/pragma-loop-safety.cpp b/clang/test/CodeGenCXX/pragma-loop-safety.cpp index c6ce82737d0..c0b10b0a6bd 100644 --- a/clang/test/CodeGenCXX/pragma-loop-safety.cpp +++ b/clang/test/CodeGenCXX/pragma-loop-safety.cpp @@ -3,19 +3,19 @@ // Verify assume_safety vectorization is recognized. void vectorize_test(int *List, int Length) { // CHECK: define {{.*}} @_Z14vectorize_test -// CHECK: [[LOAD1_IV:.+]] = load i32, i32* [[IV1:[^,]+]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID:[0-9]+]] -// CHECK-NEXT: [[LOAD1_LEN:.+]] = load i32, i32* [[LEN1:.+]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID]] +// CHECK: [[LOAD1_IV:.+]] = load i32, i32* [[IV1:[^,]+]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_2:[0-9]+]] +// CHECK-NEXT: [[LOAD1_LEN:.+]] = load i32, i32* [[LEN1:.+]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_2]] // CHECK-NEXT: [[CMP1:.+]] = icmp slt i32[[LOAD1_IV]],[[LOAD1_LEN]] // CHECK-NEXT: br i1[[CMP1]], label %[[LOOP1_BODY:[^,]+]], label %[[LOOP1_END:[^,]+]] #pragma clang loop vectorize(assume_safety) interleave(disable) unroll(disable) for (int i = 0; i < Length; i++) { - // CHECK: [[RHIV1:.+]] = load i32, i32* [[IV1]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID]] + // CHECK: [[RHIV1:.+]] = load i32, i32* [[IV1]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_2]] // CHECK-DAG: [[CALC1:.+]] = mul nsw i32[[RHIV1]], 2 - // CHECK-DAG: [[SIV1:.+]] = load i32, i32* [[IV1]]{{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID]] + // CHECK-DAG: [[SIV1:.+]] = load i32, i32* [[IV1]]{{.*}}!llvm.access.group ![[ACCESS_GROUP_2]] // CHECK-DAG: [[INDEX1:.+]] = sext i32[[SIV1]] to i64 - // CHECK-DAG: [[ARRAY1:.+]] = load i32*, i32** [[LIST1:.*]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID]] + // CHECK-DAG: [[ARRAY1:.+]] = load i32*, i32** [[LIST1:.*]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_2]] // CHECK-DAG: [[PTR1:.+]] = getelementptr inbounds i32, i32*[[ARRAY1]], i64[[INDEX1]] - // CHECK: store i32[[CALC1]], i32*[[PTR1]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP1_ID]] + // CHECK: store i32[[CALC1]], i32*[[PTR1]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_2]] // CHECK-NEXT: br label [[LOOP1_INC:[^,]+]] List[i] = i * 2; @@ -26,19 +26,19 @@ void vectorize_test(int *List, int Length) { // Verify assume_safety interleaving is recognized. void interleave_test(int *List, int Length) { // CHECK: define {{.*}} @_Z15interleave_test -// CHECK: [[LOAD2_IV:.+]] = load i32, i32* [[IV2:[^,]+]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID:[0-9]+]] -// CHECK-NEXT: [[LOAD2_LEN:.+]] = load i32, i32* [[LEN2:.+]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID]] +// CHECK: [[LOAD2_IV:.+]] = load i32, i32* [[IV2:[^,]+]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_8:[0-9]+]] +// CHECK-NEXT: [[LOAD2_LEN:.+]] = load i32, i32* [[LEN2:.+]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_8]] // CHECK-NEXT: [[CMP2:.+]] = icmp slt i32[[LOAD2_IV]],[[LOAD2_LEN]] // CHECK-NEXT: br i1[[CMP2]], label %[[LOOP2_BODY:[^,]+]], label %[[LOOP2_END:[^,]+]] #pragma clang loop interleave(assume_safety) vectorize(disable) unroll(disable) for (int i = 0; i < Length; i++) { - // CHECK: [[RHIV2:.+]] = load i32, i32* [[IV2]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID]] + // CHECK: [[RHIV2:.+]] = load i32, i32* [[IV2]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_8]] // CHECK-DAG: [[CALC2:.+]] = mul nsw i32[[RHIV2]], 2 - // CHECK-DAG: [[SIV2:.+]] = load i32, i32* [[IV2]]{{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID]] + // CHECK-DAG: [[SIV2:.+]] = load i32, i32* [[IV2]]{{.*}}!llvm.access.group ![[ACCESS_GROUP_8]] // CHECK-DAG: [[INDEX2:.+]] = sext i32[[SIV2]] to i64 - // CHECK-DAG: [[ARRAY2:.+]] = load i32*, i32** [[LIST2:.*]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID]] + // CHECK-DAG: [[ARRAY2:.+]] = load i32*, i32** [[LIST2:.*]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_8]] // CHECK-DAG: [[PTR2:.+]] = getelementptr inbounds i32, i32*[[ARRAY2]], i64[[INDEX2]] - // CHECK: store i32[[CALC2]], i32*[[PTR2]], {{.*}}!llvm.mem.parallel_loop_access ![[LOOP2_ID]] + // CHECK: store i32[[CALC2]], i32*[[PTR2]], {{.*}}!llvm.access.group ![[ACCESS_GROUP_8]] // CHECK-NEXT: br label [[LOOP2_INC:[^,]+]] List[i] = i * 2; @@ -46,9 +46,13 @@ void interleave_test(int *List, int Length) { } } -// CHECK: ![[LOOP1_HINTS]] = distinct !{![[LOOP1_HINTS]], ![[INTERLEAVE_1:[0-9]+]], ![[INTENABLE_1:[0-9]+]], ![[UNROLL_DISABLE:[0-9]+]]} +// CHECK: ![[ACCESS_GROUP_2]] = distinct !{} +// CHECK: ![[LOOP1_HINTS]] = distinct !{![[LOOP1_HINTS]], ![[INTERLEAVE_1:[0-9]+]], ![[INTENABLE_1:[0-9]+]], ![[UNROLL_DISABLE:[0-9]+]], ![[PARALLEL_ACCESSES_7:[0-9]+]]} // CHECK: ![[INTERLEAVE_1]] = !{!"llvm.loop.interleave.count", i32 1} // CHCCK: ![[INTENABLE_1]] = !{!"llvm.loop.vectorize.enable", i1 true} // CHECK: ![[UNROLL_DISABLE]] = !{!"llvm.loop.unroll.disable"} -// CHECK: ![[LOOP2_HINTS]] = distinct !{![[LOOP2_HINTS]], ![[WIDTH_1:[0-9]+]], ![[INTENABLE_1]], ![[UNROLL_DISABLE]]} +// CHECK: ![[PARALLEL_ACCESSES_7]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_2]]} +// CHECK: ![[ACCESS_GROUP_8]] = distinct !{} +// CHECK: ![[LOOP2_HINTS]] = distinct !{![[LOOP2_HINTS]], ![[WIDTH_1:[0-9]+]], ![[INTENABLE_1]], ![[UNROLL_DISABLE]], ![[PARALLEL_ACCESSES_11:[0-9]+]]} // CHECK: ![[WIDTH_1]] = !{!"llvm.loop.vectorize.width", i32 1} +// CHECK: ![[PARALLEL_ACCESSES_11]] = !{!"llvm.loop.parallel_accesses", ![[ACCESS_GROUP_8]]} |
