| ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 6 |
| ; RUN: opt -S < %s -p loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 | FileCheck %s |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size(ptr noalias %p1, ptr noalias %p2) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR0:[0-9]+]] { |
| ; CHECK-NEXT: [[ENTRY:.*:]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ] |
| ; CHECK-NEXT: br label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1 |
| ; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1 |
| ; CHECK-NEXT: [[TMP4:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]] |
| ; CHECK-NEXT: [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4 |
| ; CHECK-NEXT: [[TMP3:%.*]] = freeze <4 x i1> [[TMP4]] |
| ; CHECK-NEXT: [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]]) |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 1024 |
| ; CHECK-NEXT: [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]] |
| ; CHECK-NEXT: br i1 [[TMP7]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP5]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: br label %[[LOOP_END:.*]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP4]], i1 false) |
| ; CHECK-NEXT: [[TMP9:%.*]] = add i64 [[INDEX1]], [[TMP8]] |
| ; CHECK-NEXT: br label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[TMP9]], %[[VECTOR_EARLY_EXIT]] ], [ -1, %[[MIDDLE_BLOCK]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ] |
| br label %loop |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, 1024 |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| define i64 @early_exit_alignment_and_deref_via_assumption_with_constant_size_too_small(ptr noalias %p1, ptr noalias %p2) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_via_assumption_with_constant_size_too_small( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ 0, %[[ENTRY]] ] |
| ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 |
| ; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1 |
| ; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]] |
| ; CHECK-NEXT: br i1 [[CMP3]], label %[[LOOP_INC]], label %[[LOOP_END:.*]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], 1025 |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[INDEX]], %[[LOOP]] ], [ -1, %[[LOOP_INC]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ] |
| br label %loop |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, 1025 |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero(ptr noalias %p1, ptr noalias %p2, i64 %n) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 [[N]]) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 [[N]]) ] |
| ; CHECK-NEXT: [[C:%.*]] = icmp ne i64 [[N]], 0 |
| ; CHECK-NEXT: br i1 [[C]], label %[[LOOP_PREHEADER:.*]], label %[[LOOP_END:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[N]], 4 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1 |
| ; CHECK-NEXT: [[TMP2:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]] |
| ; CHECK-NEXT: [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4 |
| ; CHECK-NEXT: [[TMP3:%.*]] = freeze <4 x i1> [[TMP2]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]]) |
| ; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT3]], [[N_VEC]] |
| ; CHECK-NEXT: [[TMP6:%.*]] = or i1 [[TMP4]], [[TMP5]] |
| ; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[LOOP_END_LOOPEXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP7:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP2]], i1 false) |
| ; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX1]], [[TMP7]] |
| ; CHECK-NEXT: br label %[[LOOP_END_LOOPEXIT]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 |
| ; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1 |
| ; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]] |
| ; CHECK-NEXT: br i1 [[CMP3]], label %[[LOOP_INC]], label %[[LOOP_END_LOOPEXIT]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], [[N]] |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[LOOP_END_LOOPEXIT]], !llvm.loop [[LOOP4:![0-9]+]] |
| ; CHECK: [[LOOP_END_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RETVAL_PH:%.*]] = phi i64 [ -1, %[[LOOP_INC]] ], [ [[INDEX]], %[[LOOP]] ], [ -1, %[[MIDDLE_BLOCK]] ], [ [[TMP8]], %[[VECTOR_EARLY_EXIT]] ] |
| ; CHECK-NEXT: br label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[RETVAL_PH]], %[[LOOP_END_LOOPEXIT]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 %n) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 %n) ] |
| %c = icmp ne i64 %n, 0 |
| br i1 %c, label %loop, label %loop.end |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, %n |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ 0, %entry ], [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_n_may_be_zero(ptr noalias %p1, ptr noalias %p2, i64 %n) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_n_may_be_zero( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]], i64 [[N:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 [[N]]) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 [[N]]) ] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ 0, %[[ENTRY]] ] |
| ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 |
| ; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1 |
| ; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]] |
| ; CHECK-NEXT: br i1 [[CMP3]], label %[[LOOP_INC]], label %[[LOOP_END:.*]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], [[N]] |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[INDEX]], %[[LOOP]] ], [ -1, %[[LOOP_INC]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 %n) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 %n) ] |
| br label %loop |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, %n |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero_i16(ptr noalias %p1, ptr noalias %p2, i32 %n) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero_i16( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]], i32 [[N:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[N_EXT:%.*]] = zext i32 [[N]] to i64 |
| ; CHECK-NEXT: [[N_SCALED:%.*]] = shl nuw nsw i64 [[N_EXT]], 1 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 [[N_SCALED]]) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 [[N_SCALED]]) ] |
| ; CHECK-NEXT: [[C:%.*]] = icmp ne i64 [[N_EXT]], 0 |
| ; CHECK-NEXT: br i1 [[C]], label %[[LOOP_PREHEADER:.*]], label %[[LOOP_END:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ 0, %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[ARRAYIDX]], align 1 |
| ; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]] |
| ; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1 |
| ; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]] |
| ; CHECK-NEXT: br i1 [[CMP3]], label %[[LOOP_INC]], label %[[LOOP_END_LOOPEXIT:.*]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], [[N_EXT]] |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[LOOP_END_LOOPEXIT]] |
| ; CHECK: [[LOOP_END_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RETVAL_PH:%.*]] = phi i64 [ -1, %[[LOOP_INC]] ], [ [[INDEX]], %[[LOOP]] ] |
| ; CHECK-NEXT: br label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[RETVAL_PH]], %[[LOOP_END_LOOPEXIT]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| %n.ext = zext i32 %n to i64 |
| %n.scaled = shl nuw nsw i64 %n.ext, 1 |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 %n.scaled) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 %n.scaled) ] |
| %c = icmp ne i64 %n.ext, 0 |
| br i1 %c, label %loop, label %loop.end |
| |
| loop: |
| %index = phi i64 [ 0, %entry ], [ %index.next, %loop.inc ] |
| %gep.p1 = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %gep.p1, align 1 |
| %gep.p2 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %gep.p2, align 1 |
| %c.0 = icmp eq i8 %ld1, %ld2 |
| br i1 %c.0, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, %n.ext |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ 0, %entry ], [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero_i16_ptr_iv(ptr %A, i32 noundef %n) nofree nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_n_not_zero_i16_ptr_iv( |
| ; CHECK-SAME: ptr [[A:%.*]], i32 noundef [[N:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[A]], i64 2) ] |
| ; CHECK-NEXT: [[N_EXT:%.*]] = zext i32 [[N]] to i64 |
| ; CHECK-NEXT: [[MUL:%.*]] = shl i64 [[N_EXT]], 1 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[A]], i64 [[MUL]]) ] |
| ; CHECK-NEXT: [[A_END:%.*]] = getelementptr i8, ptr [[A]], i64 [[MUL]] |
| ; CHECK-NEXT: [[PRE:%.*]] = icmp eq i32 [[N]], 0 |
| ; CHECK-NEXT: br i1 [[PRE]], label %[[EXIT:.*]], label %[[LOOP_HEADER_PREHEADER:.*]] |
| ; CHECK: [[LOOP_HEADER_PREHEADER]]: |
| ; CHECK-NEXT: [[TMP0:%.*]] = add nsw i64 [[MUL]], -2 |
| ; CHECK-NEXT: [[TMP1:%.*]] = lshr i64 [[TMP0]], 1 |
| ; CHECK-NEXT: [[TMP2:%.*]] = add nuw i64 [[TMP1]], 1 |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP2]], 4 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP2]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[N_VEC]], 2 |
| ; CHECK-NEXT: [[IV_NEXT1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[OFFSET_IDX]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[NEXT_GEP]], align 2 |
| ; CHECK-NEXT: [[TMP5:%.*]] = icmp eq <4 x i16> [[WIDE_LOAD]], zeroinitializer |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 |
| ; CHECK-NEXT: [[TMP6:%.*]] = freeze <4 x i1> [[TMP5]] |
| ; CHECK-NEXT: [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]]) |
| ; CHECK-NEXT: [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: [[TMP9:%.*]] = or i1 [[TMP7]], [[TMP8]] |
| ; CHECK-NEXT: br i1 [[TMP9]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP7]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT_LOOPEXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP10:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 false) |
| ; CHECK-NEXT: [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]] |
| ; CHECK-NEXT: [[TMP12:%.*]] = mul i64 [[TMP11]], 2 |
| ; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP12]] |
| ; CHECK-NEXT: br label %[[EXIT_LOOPEXIT]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IV_NEXT1]], %[[MIDDLE_BLOCK]] ], [ [[A]], %[[LOOP_HEADER_PREHEADER]] ] |
| ; CHECK-NEXT: br label %[[LOOP_HEADER:.*]] |
| ; CHECK: [[LOOP_HEADER]]: |
| ; CHECK-NEXT: [[IV:%.*]] = phi ptr [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[L:%.*]] = load i16, ptr [[IV]], align 2 |
| ; CHECK-NEXT: [[C_0:%.*]] = icmp eq i16 [[L]], 0 |
| ; CHECK-NEXT: br i1 [[C_0]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_LATCH]] |
| ; CHECK: [[LOOP_LATCH]]: |
| ; CHECK-NEXT: [[IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[IV]], i64 2 |
| ; CHECK-NEXT: [[EC:%.*]] = icmp eq ptr [[IV_NEXT]], [[A_END]] |
| ; CHECK-NEXT: br i1 [[EC]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP6:![0-9]+]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[P_PH:%.*]] = phi ptr [ [[A_END]], %[[LOOP_LATCH]] ], [ [[IV]], %[[LOOP_HEADER]] ], [ [[A_END]], %[[MIDDLE_BLOCK]] ], [ [[TMP13]], %[[VECTOR_EARLY_EXIT]] ] |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[P:%.*]] = phi ptr [ [[A]], %[[ENTRY]] ], [ [[P_PH]], %[[EXIT_LOOPEXIT]] ] |
| ; CHECK-NEXT: [[RES:%.*]] = ptrtoint ptr [[P]] to i64 |
| ; CHECK-NEXT: ret i64 [[RES]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %A, i64 2) ] |
| %n.ext = zext i32 %n to i64 |
| %mul = shl nuw nsw i64 %n.ext, 1 |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %A, i64 %mul) ] |
| %A.end = getelementptr i8, ptr %A, i64 %mul |
| %pre = icmp eq i32 %n, 0 |
| br i1 %pre, label %exit, label %loop.header |
| |
| loop.header: |
| %iv = phi ptr [ %iv.next, %loop.latch ], [ %A, %entry ] |
| %l = load i16, ptr %iv, align 2 |
| %c.0 = icmp eq i16 %l, 0 |
| br i1 %c.0, label %exit, label %loop.latch |
| |
| loop.latch: |
| %iv.next = getelementptr inbounds nuw i8, ptr %iv, i64 2 |
| %ec = icmp eq ptr %iv.next, %A.end |
| br i1 %ec, label %exit, label %loop.header |
| |
| exit: |
| %p = phi ptr [ %A, %entry ], [ %iv, %loop.header ], [ %A.end, %loop.latch ] |
| %res = ptrtoint ptr %p to i64 |
| ret i64 %res |
| } |
| |
| ; The existing assumptions is strong enough to vectorize this. |
| define ptr @find_deref_pointer_distance_align_attribute_argument(ptr align 2 %first, ptr align 2 %last) nofree nosync { |
| ; CHECK-LABEL: define ptr @find_deref_pointer_distance_align_attribute_argument( |
| ; CHECK-SAME: ptr align 2 [[FIRST:%.*]], ptr align 2 [[LAST:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[FIRST]], i64 2) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[LAST]], i64 2) ] |
| ; CHECK-NEXT: [[LAST_I64:%.*]] = ptrtoint ptr [[LAST]] to i64 |
| ; CHECK-NEXT: [[FIRST_I64:%.*]] = ptrtoint ptr [[FIRST]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[LAST_I64]], [[FIRST_I64]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[FIRST]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[C_0:%.*]] = icmp eq ptr [[FIRST]], [[LAST]] |
| ; CHECK-NEXT: br i1 [[C_0]], label %[[EXIT:.*]], label %[[LOOP_HEADER_PREHEADER:.*]] |
| ; CHECK: [[LOOP_HEADER_PREHEADER]]: |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[LAST_I64]], -2 |
| ; CHECK-NEXT: [[TMP1:%.*]] = sub i64 [[TMP0]], [[FIRST_I64]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 1 |
| ; CHECK-NEXT: [[TMP3:%.*]] = add nuw i64 [[TMP2]], 1 |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = mul i64 [[N_VEC]], 2 |
| ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[TMP4]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[OFFSET_IDX]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[NEXT_GEP]], align 2 |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i16> [[WIDE_LOAD]], splat (i16 1) |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 |
| ; CHECK-NEXT: [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]]) |
| ; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]] |
| ; CHECK-NEXT: br i1 [[TMP10]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP8]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT_LOOPEXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP11:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 false) |
| ; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], [[TMP11]] |
| ; CHECK-NEXT: [[TMP13:%.*]] = mul i64 [[TMP12]], 2 |
| ; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[TMP13]] |
| ; CHECK-NEXT: br label %[[EXIT_LOOPEXIT]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ [[FIRST]], %[[LOOP_HEADER_PREHEADER]] ] |
| ; CHECK-NEXT: br label %[[LOOP_HEADER:.*]] |
| ; CHECK: [[LOOP_HEADER]]: |
| ; CHECK-NEXT: [[IV:%.*]] = phi ptr [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[L:%.*]] = load i16, ptr [[IV]], align 2 |
| ; CHECK-NEXT: [[C_1:%.*]] = icmp eq i16 [[L]], 1 |
| ; CHECK-NEXT: br i1 [[C_1]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_LATCH]] |
| ; CHECK: [[LOOP_LATCH]]: |
| ; CHECK-NEXT: [[IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[IV]], i64 2 |
| ; CHECK-NEXT: [[C_2:%.*]] = icmp eq ptr [[IV_NEXT]], [[LAST]] |
| ; CHECK-NEXT: br i1 [[C_2]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP8:![0-9]+]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[FIRST_ADDR_0_LCSSA_I_PH:%.*]] = phi ptr [ [[IV_NEXT]], %[[LOOP_LATCH]] ], [ [[IV]], %[[LOOP_HEADER]] ], [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ [[TMP14]], %[[VECTOR_EARLY_EXIT]] ] |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[FIRST_ADDR_0_LCSSA_I:%.*]] = phi ptr [ [[FIRST]], %[[ENTRY]] ], [ [[FIRST_ADDR_0_LCSSA_I_PH]], %[[EXIT_LOOPEXIT]] ] |
| ; CHECK-NEXT: ret ptr [[FIRST_ADDR_0_LCSSA_I]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %first, i64 2) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %last, i64 2) ] |
| %last.i64 = ptrtoint ptr %last to i64 |
| %first.i64 = ptrtoint ptr %first to i64 |
| %distance = sub i64 %last.i64, %first.i64 |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %first, i64 %distance) ] |
| %c.0 = icmp eq ptr %first, %last |
| br i1 %c.0, label %exit, label %loop.header |
| |
| loop.header: |
| %iv = phi ptr [ %first, %entry ], [ %iv.next, %loop.latch ] |
| %l = load i16, ptr %iv, align 2 |
| %c.1 = icmp eq i16 %l, 1 |
| br i1 %c.1, label %exit, label %loop.latch |
| |
| loop.latch: |
| %iv.next = getelementptr inbounds nuw i8, ptr %iv, i64 2 |
| %c.2 = icmp eq ptr %iv.next, %last |
| br i1 %c.2, label %exit, label %loop.header |
| |
| exit: |
| %first.addr.0.lcssa.i = phi ptr [ %first, %entry ], [ %iv, %loop.header ], [ %iv.next, %loop.latch ] |
| ret ptr %first.addr.0.lcssa.i |
| } |
| |
| ; The existing assumptions is strong enough to vectorize this. |
| define ptr @find_deref_pointer_distance_align_assumption(ptr %first, ptr %last) nofree nosync { |
| ; CHECK-LABEL: define ptr @find_deref_pointer_distance_align_assumption( |
| ; CHECK-SAME: ptr [[FIRST:%.*]], ptr [[LAST:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[LAST_I64:%.*]] = ptrtoint ptr [[LAST]] to i64 |
| ; CHECK-NEXT: [[FIRST_I64:%.*]] = ptrtoint ptr [[FIRST]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[LAST_I64]], [[FIRST_I64]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[FIRST]], i64 2) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[LAST]], i64 2) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[FIRST]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[C_0:%.*]] = icmp eq ptr [[FIRST]], [[LAST]] |
| ; CHECK-NEXT: br i1 [[C_0]], label %[[EXIT:.*]], label %[[LOOP_HEADER_PREHEADER:.*]] |
| ; CHECK: [[LOOP_HEADER_PREHEADER]]: |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[LAST_I64]], -2 |
| ; CHECK-NEXT: [[TMP1:%.*]] = sub i64 [[TMP0]], [[FIRST_I64]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 1 |
| ; CHECK-NEXT: [[TMP3:%.*]] = add nuw i64 [[TMP2]], 1 |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = mul i64 [[N_VEC]], 2 |
| ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[TMP4]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[OFFSET_IDX]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[NEXT_GEP]], align 2 |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i16> [[WIDE_LOAD]], splat (i16 1) |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 |
| ; CHECK-NEXT: [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]]) |
| ; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]] |
| ; CHECK-NEXT: br i1 [[TMP10]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP8]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT_LOOPEXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP11:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 false) |
| ; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], [[TMP11]] |
| ; CHECK-NEXT: [[TMP13:%.*]] = mul i64 [[TMP12]], 2 |
| ; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[FIRST]], i64 [[TMP13]] |
| ; CHECK-NEXT: br label %[[EXIT_LOOPEXIT]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ [[FIRST]], %[[LOOP_HEADER_PREHEADER]] ] |
| ; CHECK-NEXT: br label %[[LOOP_HEADER:.*]] |
| ; CHECK: [[LOOP_HEADER]]: |
| ; CHECK-NEXT: [[IV:%.*]] = phi ptr [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[L:%.*]] = load i16, ptr [[IV]], align 2 |
| ; CHECK-NEXT: [[C_1:%.*]] = icmp eq i16 [[L]], 1 |
| ; CHECK-NEXT: br i1 [[C_1]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_LATCH]] |
| ; CHECK: [[LOOP_LATCH]]: |
| ; CHECK-NEXT: [[IV_NEXT]] = getelementptr inbounds nuw i8, ptr [[IV]], i64 2 |
| ; CHECK-NEXT: [[C_2:%.*]] = icmp eq ptr [[IV_NEXT]], [[LAST]] |
| ; CHECK-NEXT: br i1 [[C_2]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP10:![0-9]+]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[FIRST_ADDR_0_LCSSA_I_PH:%.*]] = phi ptr [ [[IV_NEXT]], %[[LOOP_LATCH]] ], [ [[IV]], %[[LOOP_HEADER]] ], [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ [[TMP14]], %[[VECTOR_EARLY_EXIT]] ] |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[FIRST_ADDR_0_LCSSA_I:%.*]] = phi ptr [ [[FIRST]], %[[ENTRY]] ], [ [[FIRST_ADDR_0_LCSSA_I_PH]], %[[EXIT_LOOPEXIT]] ] |
| ; CHECK-NEXT: ret ptr [[FIRST_ADDR_0_LCSSA_I]] |
| ; |
| entry: |
| %last.i64 = ptrtoint ptr %last to i64 |
| %first.i64 = ptrtoint ptr %first to i64 |
| %distance = sub i64 %last.i64, %first.i64 |
| call void @llvm.assume(i1 true) [ "align"(ptr %first, i64 2) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %last, i64 2) ] |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %first, i64 %distance) ] |
| %c.0 = icmp eq ptr %first, %last |
| br i1 %c.0, label %exit, label %loop.header |
| |
| loop.header: |
| %iv = phi ptr [ %first, %entry ], [ %iv.next, %loop.latch ] |
| %l = load i16, ptr %iv, align 2 |
| %c.1 = icmp eq i16 %l, 1 |
| br i1 %c.1, label %exit, label %loop.latch |
| |
| loop.latch: |
| %iv.next = getelementptr inbounds nuw i8, ptr %iv, i64 2 |
| %c.2 = icmp eq ptr %iv.next, %last |
| br i1 %c.2, label %exit, label %loop.header |
| |
| exit: |
| %first.addr.0.lcssa.i = phi ptr [ %first, %entry ], [ %iv, %loop.header ], [ %iv.next, %loop.latch ] |
| ret ptr %first.addr.0.lcssa.i |
| } |
| |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size_nofree_via_context(ptr noalias %p1, ptr noalias %p2) nosync { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size_nofree_via_context( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR1:[0-9]+]] { |
| ; CHECK-NEXT: [[ENTRY:.*:]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ] |
| ; CHECK-NEXT: br label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[ARRAYIDX2:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[ARRAYIDX2]], align 1 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1 |
| ; CHECK-NEXT: [[TMP2:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]] |
| ; CHECK-NEXT: [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4 |
| ; CHECK-NEXT: [[TMP3:%.*]] = freeze <4 x i1> [[TMP2]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP3]]) |
| ; CHECK-NEXT: [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 1024 |
| ; CHECK-NEXT: [[TMP6:%.*]] = or i1 [[TMP4]], [[TMP5]] |
| ; CHECK-NEXT: br i1 [[TMP6]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: br label %[[LOOP_END:.*]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP7:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP2]], i1 false) |
| ; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[INDEX1]], [[TMP7]] |
| ; CHECK-NEXT: br label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[TMP8]], %[[VECTOR_EARLY_EXIT]] ], [ -1, %[[MIDDLE_BLOCK]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ] |
| br label %loop |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, 1024 |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| } |
| |
| ; Test with an invoke terminator between the block containing the assume and loop. |
| ; This should NOT vectorize because the invoke could free memory. |
| define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size_invoke_may_free(ptr noalias %p1, ptr noalias %p2) nosync personality ptr @__gxx_personality_v0 { |
| ; CHECK-LABEL: define i64 @early_exit_alignment_and_deref_known_via_assumption_with_constant_size_invoke_may_free( |
| ; CHECK-SAME: ptr noalias [[P1:%.*]], ptr noalias [[P2:%.*]]) #[[ATTR1]] personality ptr @__gxx_personality_v0 { |
| ; CHECK-NEXT: [[ENTRY:.*:]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P1]], i64 4), "dereferenceable"(ptr [[P1]], i64 1024) ] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[P2]], i64 4), "dereferenceable"(ptr [[P2]], i64 1024) ] |
| ; CHECK-NEXT: invoke void @may_free() |
| ; CHECK-NEXT: to label %[[LOOP_PREHEADER:.*]] unwind label %[[LPAD:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[INDEX1:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ 0, %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[LD1:%.*]] = load i8, ptr [[TMP0]], align 1 |
| ; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX1]] |
| ; CHECK-NEXT: [[LD2:%.*]] = load i8, ptr [[TMP1]], align 1 |
| ; CHECK-NEXT: [[CMP3:%.*]] = icmp eq i8 [[LD1]], [[LD2]] |
| ; CHECK-NEXT: br i1 [[CMP3]], label %[[LOOP_INC]], label %[[LOOP_END:.*]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX1]], 1 |
| ; CHECK-NEXT: [[EXITCOND:%.*]] = icmp ne i64 [[INDEX_NEXT]], 1024 |
| ; CHECK-NEXT: br i1 [[EXITCOND]], label %[[LOOP]], label %[[LOOP_END]] |
| ; CHECK: [[LOOP_END]]: |
| ; CHECK-NEXT: [[RETVAL:%.*]] = phi i64 [ [[INDEX1]], %[[LOOP]] ], [ -1, %[[LOOP_INC]] ] |
| ; CHECK-NEXT: ret i64 [[RETVAL]] |
| ; CHECK: [[LPAD]]: |
| ; CHECK-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } |
| ; CHECK-NEXT: cleanup |
| ; CHECK-NEXT: resume { ptr, i32 } [[LP]] |
| ; |
| entry: |
| call void @llvm.assume(i1 true) [ "align"(ptr %p1, i64 4), "dereferenceable"(ptr %p1, i64 1024) ] |
| call void @llvm.assume(i1 true) [ "align"(ptr %p2, i64 4), "dereferenceable"(ptr %p2, i64 1024) ] |
| invoke void @may_free() to label %loop unwind label %lpad |
| |
| loop: |
| %index = phi i64 [ %index.next, %loop.inc ], [ 0, %entry ] |
| %arrayidx = getelementptr inbounds i8, ptr %p1, i64 %index |
| %ld1 = load i8, ptr %arrayidx, align 1 |
| %arrayidx1 = getelementptr inbounds i8, ptr %p2, i64 %index |
| %ld2 = load i8, ptr %arrayidx1, align 1 |
| %cmp3 = icmp eq i8 %ld1, %ld2 |
| br i1 %cmp3, label %loop.inc, label %loop.end |
| |
| loop.inc: |
| %index.next = add i64 %index, 1 |
| %exitcond = icmp ne i64 %index.next, 1024 |
| br i1 %exitcond, label %loop, label %loop.end |
| |
| loop.end: |
| %retval = phi i64 [ %index, %loop ], [ -1, %loop.inc ] |
| ret i64 %retval |
| |
| lpad: |
| %lp = landingpad { ptr, i32 } cleanup |
| resume { ptr, i32 } %lp |
| } |
| |
| define i64 @find_if_pointer_distance_deref_via_assumption(ptr %vec) nofree nosync { |
| ; CHECK-LABEL: define i64 @find_if_pointer_distance_deref_via_assumption( |
| ; CHECK-SAME: ptr [[VEC:%.*]]) #[[ATTR0]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[BEGIN:%.*]] = load ptr, ptr [[VEC]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[BEGIN]], i64 2) ] |
| ; CHECK-NEXT: [[BEGIN_INT:%.*]] = ptrtoint ptr [[BEGIN]] to i64 |
| ; CHECK-NEXT: [[END_GEP:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8 |
| ; CHECK-NEXT: [[END:%.*]] = load ptr, ptr [[END_GEP]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 2) ] |
| ; CHECK-NEXT: [[END_INT:%.*]] = ptrtoint ptr [[END]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[END_INT]], [[BEGIN_INT]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[BEGIN]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[IS_EMPTY:%.*]] = icmp eq ptr [[BEGIN]], [[END]] |
| ; CHECK-NEXT: br i1 [[IS_EMPTY]], label %[[EXIT:.*]], label %[[LOOP_PREHEADER:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: [[END_PTR:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[DISTANCE]] |
| ; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[END_INT]], -2 |
| ; CHECK-NEXT: [[TMP1:%.*]] = sub i64 [[TMP0]], [[BEGIN_INT]] |
| ; CHECK-NEXT: [[TMP2:%.*]] = lshr i64 [[TMP1]], 1 |
| ; CHECK-NEXT: [[TMP3:%.*]] = add nuw i64 [[TMP2]], 1 |
| ; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 4 |
| ; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]] |
| ; CHECK: [[VECTOR_PH]]: |
| ; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 4 |
| ; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]] |
| ; CHECK-NEXT: [[TMP4:%.*]] = mul i64 [[N_VEC]], 2 |
| ; CHECK-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[TMP4]] |
| ; CHECK-NEXT: br label %[[VECTOR_BODY:.*]] |
| ; CHECK: [[VECTOR_BODY]]: |
| ; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ] |
| ; CHECK-NEXT: [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2 |
| ; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[OFFSET_IDX]] |
| ; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[NEXT_GEP]], align 2 |
| ; CHECK-NEXT: [[TMP6:%.*]] = icmp eq <4 x i16> [[WIDE_LOAD]], splat (i16 1) |
| ; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4 |
| ; CHECK-NEXT: [[TMP7:%.*]] = freeze <4 x i1> [[TMP6]] |
| ; CHECK-NEXT: [[TMP8:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP7]]) |
| ; CHECK-NEXT: [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]] |
| ; CHECK-NEXT: [[TMP10:%.*]] = or i1 [[TMP8]], [[TMP9]] |
| ; CHECK-NEXT: br i1 [[TMP10]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]] |
| ; CHECK: [[MIDDLE_SPLIT]]: |
| ; CHECK-NEXT: br i1 [[TMP8]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]] |
| ; CHECK: [[MIDDLE_BLOCK]]: |
| ; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]] |
| ; CHECK-NEXT: br i1 [[CMP_N]], label %[[EXIT_LOOPEXIT:.*]], label %[[SCALAR_PH]] |
| ; CHECK: [[VECTOR_EARLY_EXIT]]: |
| ; CHECK-NEXT: [[TMP11:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 false) |
| ; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], [[TMP11]] |
| ; CHECK-NEXT: [[TMP13:%.*]] = mul i64 [[TMP12]], 2 |
| ; CHECK-NEXT: [[TMP14:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[TMP13]] |
| ; CHECK-NEXT: br label %[[EXIT_LOOPEXIT]] |
| ; CHECK: [[SCALAR_PH]]: |
| ; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP5]], %[[MIDDLE_BLOCK]] ], [ [[BEGIN]], %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: br label %[[LOOP_HEADER:.*]] |
| ; CHECK: [[LOOP_HEADER]]: |
| ; CHECK-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ] |
| ; CHECK-NEXT: [[VAL:%.*]] = load i16, ptr [[PTR]], align 2 |
| ; CHECK-NEXT: [[FOUND:%.*]] = icmp eq i16 [[VAL]], 1 |
| ; CHECK-NEXT: br i1 [[FOUND]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_LATCH]] |
| ; CHECK: [[LOOP_LATCH]]: |
| ; CHECK-NEXT: [[PTR_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR]], i64 2 |
| ; CHECK-NEXT: [[DONE:%.*]] = icmp eq ptr [[PTR_NEXT]], [[END]] |
| ; CHECK-NEXT: br i1 [[DONE]], label %[[EXIT_LOOPEXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP13:![0-9]+]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RESULT_PH:%.*]] = phi ptr [ [[END_PTR]], %[[LOOP_LATCH]] ], [ [[PTR]], %[[LOOP_HEADER]] ], [ [[END_PTR]], %[[MIDDLE_BLOCK]] ], [ [[TMP14]], %[[VECTOR_EARLY_EXIT]] ] |
| ; CHECK-NEXT: [[DOTRESULT_INT:%.*]] = ptrtoint ptr [[RESULT_PH]] to i64 |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[RESULT_INT:%.*]] = phi i64 [ [[DOTRESULT_INT]], %[[EXIT_LOOPEXIT]] ], [ [[BEGIN_INT]], %[[ENTRY]] ] |
| ; CHECK-NEXT: ret i64 [[RESULT_INT]] |
| ; |
| entry: |
| %begin = load ptr, ptr %vec, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %begin, i64 2) ] |
| %begin.int = ptrtoint ptr %begin to i64 |
| %end.gep = getelementptr inbounds nuw i8, ptr %vec, i64 8 |
| %end = load ptr, ptr %end.gep, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 2) ] |
| %end.int = ptrtoint ptr %end to i64 |
| %distance = sub i64 %end.int, %begin.int |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %begin, i64 %distance) ] |
| %is.empty = icmp eq ptr %begin, %end |
| br i1 %is.empty, label %exit, label %loop.preheader |
| |
| loop.preheader: |
| %end.ptr = getelementptr i8, ptr %begin, i64 %distance |
| br label %loop.header |
| |
| loop.header: |
| %ptr = phi ptr [ %ptr.next, %loop.latch ], [ %begin, %loop.preheader ] |
| %val = load i16, ptr %ptr, align 2 |
| %found = icmp eq i16 %val, 1 |
| br i1 %found, label %exit.loopexit, label %loop.latch |
| |
| loop.latch: |
| %ptr.next = getelementptr inbounds nuw i8, ptr %ptr, i64 2 |
| %done = icmp eq ptr %ptr.next, %end |
| br i1 %done, label %exit.loopexit, label %loop.header |
| |
| exit.loopexit: |
| %result.ph = phi ptr [ %end.ptr, %loop.latch ], [ %ptr, %loop.header ] |
| %.result.int = ptrtoint ptr %result.ph to i64 |
| br label %exit |
| |
| exit: |
| %result.int = phi i64 [ %.result.int, %exit.loopexit ], [ %begin.int, %entry ] |
| ret i64 %result.int |
| } |
| |
| declare void @may_free() |
| |
| define i64 @find_if_pointer_distance_deref_via_assumption_mayfree_call(ptr %vec) nofree { |
| ; CHECK-LABEL: define i64 @find_if_pointer_distance_deref_via_assumption_mayfree_call( |
| ; CHECK-SAME: ptr [[VEC:%.*]]) #[[ATTR2:[0-9]+]] { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[BEGIN:%.*]] = load ptr, ptr [[VEC]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[BEGIN]], i64 2) ] |
| ; CHECK-NEXT: [[BEGIN_INT:%.*]] = ptrtoint ptr [[BEGIN]] to i64 |
| ; CHECK-NEXT: [[END_GEP:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8 |
| ; CHECK-NEXT: [[END:%.*]] = load ptr, ptr [[END_GEP]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 2) ] |
| ; CHECK-NEXT: [[END_INT:%.*]] = ptrtoint ptr [[END]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[END_INT]], [[BEGIN_INT]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[BEGIN]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[IS_EMPTY:%.*]] = icmp eq ptr [[BEGIN]], [[END]] |
| ; CHECK-NEXT: br i1 [[IS_EMPTY]], label %[[EXIT:.*]], label %[[LOOP_PREHEADER:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: call void @may_free() |
| ; CHECK-NEXT: [[END_PTR:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[DISTANCE]] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ [[BEGIN]], %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: [[VAL:%.*]] = load i16, ptr [[PTR]], align 2 |
| ; CHECK-NEXT: [[FOUND:%.*]] = icmp eq i16 [[VAL]], 1 |
| ; CHECK-NEXT: br i1 [[FOUND]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP_INC]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[PTR_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR]], i64 2 |
| ; CHECK-NEXT: [[DONE:%.*]] = icmp eq ptr [[PTR_NEXT]], [[END]] |
| ; CHECK-NEXT: br i1 [[DONE]], label %[[EXIT_LOOPEXIT]], label %[[LOOP]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RESULT_PH:%.*]] = phi ptr [ [[END_PTR]], %[[LOOP_INC]] ], [ [[PTR]], %[[LOOP]] ] |
| ; CHECK-NEXT: [[DOTRESULT_INT:%.*]] = ptrtoint ptr [[RESULT_PH]] to i64 |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[RESULT_INT:%.*]] = phi i64 [ [[DOTRESULT_INT]], %[[EXIT_LOOPEXIT]] ], [ [[BEGIN_INT]], %[[ENTRY]] ] |
| ; CHECK-NEXT: ret i64 [[RESULT_INT]] |
| ; |
| entry: |
| %begin = load ptr, ptr %vec, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %begin, i64 2) ] |
| %begin.int = ptrtoint ptr %begin to i64 |
| %end.gep = getelementptr inbounds nuw i8, ptr %vec, i64 8 |
| %end = load ptr, ptr %end.gep, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 2) ] |
| %end.int = ptrtoint ptr %end to i64 |
| %distance = sub i64 %end.int, %begin.int |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %begin, i64 %distance) ] |
| %is.empty = icmp eq ptr %begin, %end |
| br i1 %is.empty, label %exit, label %loop.preheader |
| |
| loop.preheader: |
| call void @may_free() |
| %end.ptr = getelementptr i8, ptr %begin, i64 %distance |
| br label %loop |
| |
| loop: |
| %ptr = phi ptr [ %ptr.next, %loop.inc ], [ %begin, %loop.preheader ] |
| %val = load i16, ptr %ptr, align 2 |
| %found = icmp eq i16 %val, 1 |
| br i1 %found, label %exit.loopexit, label %loop.inc |
| |
| loop.inc: |
| %ptr.next = getelementptr inbounds nuw i8, ptr %ptr, i64 2 |
| %done = icmp eq ptr %ptr.next, %end |
| br i1 %done, label %exit.loopexit, label %loop |
| |
| exit.loopexit: |
| %result.ph = phi ptr [ %end.ptr, %loop.inc ], [ %ptr, %loop ] |
| %.result.int = ptrtoint ptr %result.ph to i64 |
| br label %exit |
| |
| exit: |
| %result.int = phi i64 [ %.result.int, %exit.loopexit ], [ %begin.int, %entry ] |
| ret i64 %result.int |
| } |
| |
| ; Test with an invoke terminator between the block containing the assume and loop. |
| ; This should NOT vectorize because the invoke could free memory. |
| define i64 @find_if_deref_via_assumption_invoke_may_free_in_assume_block(ptr %vec) nofree nosync personality ptr @__gxx_personality_v0 { |
| ; CHECK-LABEL: define i64 @find_if_deref_via_assumption_invoke_may_free_in_assume_block( |
| ; CHECK-SAME: ptr [[VEC:%.*]]) #[[ATTR0]] personality ptr @__gxx_personality_v0 { |
| ; CHECK-NEXT: [[ENTRY:.*:]] |
| ; CHECK-NEXT: [[BEGIN:%.*]] = load ptr, ptr [[VEC]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[BEGIN]], i64 2) ] |
| ; CHECK-NEXT: [[BEGIN_INT:%.*]] = ptrtoint ptr [[BEGIN]] to i64 |
| ; CHECK-NEXT: [[END_GEP:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8 |
| ; CHECK-NEXT: [[END:%.*]] = load ptr, ptr [[END_GEP]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 2) ] |
| ; CHECK-NEXT: [[END_INT:%.*]] = ptrtoint ptr [[END]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[END_INT]], [[BEGIN_INT]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[BEGIN]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[NOT_EMPTY:%.*]] = icmp ne ptr [[BEGIN]], [[END]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 [[NOT_EMPTY]]) |
| ; CHECK-NEXT: invoke void @may_free() |
| ; CHECK-NEXT: to label %[[LOOP_PREHEADER:.*]] unwind label %[[LPAD:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: [[END_PTR:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[DISTANCE]] |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ [[BEGIN]], %[[LOOP_PREHEADER]] ] |
| ; CHECK-NEXT: [[VAL:%.*]] = load i16, ptr [[PTR]], align 2 |
| ; CHECK-NEXT: [[FOUND:%.*]] = icmp eq i16 [[VAL]], 1 |
| ; CHECK-NEXT: br i1 [[FOUND]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP_INC]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[PTR_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR]], i64 2 |
| ; CHECK-NEXT: [[DONE:%.*]] = icmp eq ptr [[PTR_NEXT]], [[END]] |
| ; CHECK-NEXT: br i1 [[DONE]], label %[[EXIT_LOOPEXIT]], label %[[LOOP]] |
| ; CHECK: [[LPAD]]: |
| ; CHECK-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } |
| ; CHECK-NEXT: cleanup |
| ; CHECK-NEXT: resume { ptr, i32 } [[LP]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RESULT_PH:%.*]] = phi ptr [ [[END_PTR]], %[[LOOP_INC]] ], [ [[PTR]], %[[LOOP]] ] |
| ; CHECK-NEXT: [[DOTRESULT_INT:%.*]] = ptrtoint ptr [[RESULT_PH]] to i64 |
| ; CHECK-NEXT: br label %[[EXIT:.*]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[RESULT_INT:%.*]] = phi i64 [ [[DOTRESULT_INT]], %[[EXIT_LOOPEXIT]] ] |
| ; CHECK-NEXT: ret i64 [[RESULT_INT]] |
| ; |
| entry: |
| %begin = load ptr, ptr %vec, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %begin, i64 2) ] |
| %begin.int = ptrtoint ptr %begin to i64 |
| %end.gep = getelementptr inbounds nuw i8, ptr %vec, i64 8 |
| %end = load ptr, ptr %end.gep, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 2) ] |
| %end.int = ptrtoint ptr %end to i64 |
| %distance = sub i64 %end.int, %begin.int |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %begin, i64 %distance) ] |
| %not.empty = icmp ne ptr %begin, %end |
| call void @llvm.assume(i1 %not.empty) |
| invoke void @may_free() to label %loop.preheader unwind label %lpad |
| |
| loop.preheader: |
| %end.ptr = getelementptr i8, ptr %begin, i64 %distance |
| br label %loop |
| |
| loop: |
| %ptr = phi ptr [ %ptr.next, %loop.inc ], [ %begin, %loop.preheader ] |
| %val = load i16, ptr %ptr, align 2 |
| %found = icmp eq i16 %val, 1 |
| br i1 %found, label %exit.loopexit, label %loop.inc |
| |
| loop.inc: |
| %ptr.next = getelementptr inbounds nuw i8, ptr %ptr, i64 2 |
| %done = icmp eq ptr %ptr.next, %end |
| br i1 %done, label %exit.loopexit, label %loop |
| |
| lpad: |
| %lp = landingpad { ptr, i32 } cleanup |
| resume { ptr, i32 } %lp |
| |
| exit.loopexit: |
| %result.ph = phi ptr [ %end.ptr, %loop.inc ], [ %ptr, %loop ] |
| %.result.int = ptrtoint ptr %result.ph to i64 |
| br label %exit |
| |
| exit: |
| %result.int = phi i64 [ %.result.int, %exit.loopexit ] |
| ret i64 %result.int |
| } |
| |
| ; Test with an invoke terminator between the block containing the assume and loop. |
| ; This should NOT vectorize because the invoke could free memory. |
| define i64 @find_if_deref_via_assumption_invoke_may_free_in_preheader(ptr %vec) nofree nosync personality ptr @__gxx_personality_v0 { |
| ; CHECK-LABEL: define i64 @find_if_deref_via_assumption_invoke_may_free_in_preheader( |
| ; CHECK-SAME: ptr [[VEC:%.*]]) #[[ATTR0]] personality ptr @__gxx_personality_v0 { |
| ; CHECK-NEXT: [[ENTRY:.*]]: |
| ; CHECK-NEXT: [[BEGIN:%.*]] = load ptr, ptr [[VEC]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[BEGIN]], i64 2) ] |
| ; CHECK-NEXT: [[BEGIN_INT:%.*]] = ptrtoint ptr [[BEGIN]] to i64 |
| ; CHECK-NEXT: [[END_GEP:%.*]] = getelementptr inbounds nuw i8, ptr [[VEC]], i64 8 |
| ; CHECK-NEXT: [[END:%.*]] = load ptr, ptr [[END_GEP]], align 8 |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "align"(ptr [[END]], i64 2) ] |
| ; CHECK-NEXT: [[END_INT:%.*]] = ptrtoint ptr [[END]] to i64 |
| ; CHECK-NEXT: [[DISTANCE:%.*]] = sub i64 [[END_INT]], [[BEGIN_INT]] |
| ; CHECK-NEXT: call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[BEGIN]], i64 [[DISTANCE]]) ] |
| ; CHECK-NEXT: [[IS_EMPTY:%.*]] = icmp eq ptr [[BEGIN]], [[END]] |
| ; CHECK-NEXT: br i1 [[IS_EMPTY]], label %[[EXIT:.*]], label %[[LOOP_PREHEADER:.*]] |
| ; CHECK: [[LOOP_PREHEADER]]: |
| ; CHECK-NEXT: [[END_PTR:%.*]] = getelementptr i8, ptr [[BEGIN]], i64 [[DISTANCE]] |
| ; CHECK-NEXT: invoke void @may_free() |
| ; CHECK-NEXT: to label %[[LOOP_PREHEADER1:.*]] unwind label %[[LPAD:.*]] |
| ; CHECK: [[LOOP_PREHEADER1]]: |
| ; CHECK-NEXT: br label %[[LOOP:.*]] |
| ; CHECK: [[LOOP]]: |
| ; CHECK-NEXT: [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], %[[LOOP_INC:.*]] ], [ [[BEGIN]], %[[LOOP_PREHEADER1]] ] |
| ; CHECK-NEXT: [[VAL:%.*]] = load i16, ptr [[PTR]], align 2 |
| ; CHECK-NEXT: [[FOUND:%.*]] = icmp eq i16 [[VAL]], 1 |
| ; CHECK-NEXT: br i1 [[FOUND]], label %[[EXIT_LOOPEXIT:.*]], label %[[LOOP_INC]] |
| ; CHECK: [[LOOP_INC]]: |
| ; CHECK-NEXT: [[PTR_NEXT]] = getelementptr inbounds nuw i8, ptr [[PTR]], i64 2 |
| ; CHECK-NEXT: [[DONE:%.*]] = icmp eq ptr [[PTR_NEXT]], [[END]] |
| ; CHECK-NEXT: br i1 [[DONE]], label %[[EXIT_LOOPEXIT]], label %[[LOOP]] |
| ; CHECK: [[LPAD]]: |
| ; CHECK-NEXT: [[LP:%.*]] = landingpad { ptr, i32 } |
| ; CHECK-NEXT: cleanup |
| ; CHECK-NEXT: resume { ptr, i32 } [[LP]] |
| ; CHECK: [[EXIT_LOOPEXIT]]: |
| ; CHECK-NEXT: [[RESULT_PH:%.*]] = phi ptr [ [[END_PTR]], %[[LOOP_INC]] ], [ [[PTR]], %[[LOOP]] ] |
| ; CHECK-NEXT: [[DOTRESULT_INT:%.*]] = ptrtoint ptr [[RESULT_PH]] to i64 |
| ; CHECK-NEXT: br label %[[EXIT]] |
| ; CHECK: [[EXIT]]: |
| ; CHECK-NEXT: [[RESULT_INT:%.*]] = phi i64 [ [[DOTRESULT_INT]], %[[EXIT_LOOPEXIT]] ], [ [[BEGIN_INT]], %[[ENTRY]] ] |
| ; CHECK-NEXT: ret i64 [[RESULT_INT]] |
| ; |
| entry: |
| %begin = load ptr, ptr %vec, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %begin, i64 2) ] |
| %begin.int = ptrtoint ptr %begin to i64 |
| %end.gep = getelementptr inbounds nuw i8, ptr %vec, i64 8 |
| %end = load ptr, ptr %end.gep, align 8 |
| call void @llvm.assume(i1 true) [ "align"(ptr %end, i64 2) ] |
| %end.int = ptrtoint ptr %end to i64 |
| %distance = sub i64 %end.int, %begin.int |
| call void @llvm.assume(i1 true) [ "dereferenceable"(ptr %begin, i64 %distance) ] |
| %is.empty = icmp eq ptr %begin, %end |
| br i1 %is.empty, label %exit, label %loop.preheader |
| |
| loop.preheader: |
| %end.ptr = getelementptr i8, ptr %begin, i64 %distance |
| invoke void @may_free() to label %loop unwind label %lpad |
| |
| loop: |
| %ptr = phi ptr [ %ptr.next, %loop.inc ], [ %begin, %loop.preheader ] |
| %val = load i16, ptr %ptr, align 2 |
| %found = icmp eq i16 %val, 1 |
| br i1 %found, label %exit.loopexit, label %loop.inc |
| |
| loop.inc: |
| %ptr.next = getelementptr inbounds nuw i8, ptr %ptr, i64 2 |
| %done = icmp eq ptr %ptr.next, %end |
| br i1 %done, label %exit.loopexit, label %loop |
| |
| lpad: |
| %lp = landingpad { ptr, i32 } cleanup |
| resume { ptr, i32 } %lp |
| |
| exit.loopexit: |
| %result.ph = phi ptr [ %end.ptr, %loop.inc ], [ %ptr, %loop ] |
| %.result.int = ptrtoint ptr %result.ph to i64 |
| br label %exit |
| |
| exit: |
| %result.int = phi i64 [ %.result.int, %exit.loopexit ], [ %begin.int, %entry ] |
| ret i64 %result.int |
| } |
| |
| declare i32 @__gxx_personality_v0(...) |
