; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -hoist-runtime-checks=false -S < %s | FileCheck %s --check-prefix=CHECK
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -S < %s | FileCheck %s --check-prefix=CHECK-HOIST
; This is the test case from PR26314.
; When we were retrying dependence checking with memchecks only,
; the loop-invariant access in the inner loop was incorrectly determined to be wrapping
; because it was not strided in the inner loop.
; Improved wrapping detection allows vectorization in the following case.
; #define Z 32
; typedef struct s {
; int v1[Z];
; int v2[Z];
; int v3[Z][Z];
; } s;
;
; void slow_function (s* const obj, int z) {
; for (int j=0; j<Z; j++) {
; for (int k=0; k<z; k++) {
; int x = obj->v1[k] + obj->v2[j];
; obj->v3[j][k] += x;
; }
; }
; }
target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
%struct.s = type { [32 x i32], [32 x i32], [32 x [32 x i32]] }
define void @Test(ptr nocapture %obj, i64 %z) #0 {
; CHECK-LABEL: @Test(
; CHECK-NEXT: [[TMP1:%.*]] = shl i64 [[Z:%.*]], 2
; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], 256
; CHECK-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[OBJ:%.*]], i64 [[TMP1]]
; CHECK-NEXT: br label [[DOTOUTER_PREHEADER:%.*]]
; CHECK: .outer.preheader:
; CHECK-NEXT: [[I:%.*]] = phi i64 [ 0, [[TMP0:%.*]] ], [ [[I_NEXT:%.*]], [[DOTOUTER:%.*]] ]
; CHECK-NEXT: [[TMP3:%.*]] = shl nuw nsw i64 [[I]], 7
; CHECK-NEXT: [[TMP4:%.*]] = add i64 [[TMP3]], 256
; CHECK-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 [[TMP4]]
; CHECK-NEXT: [[TMP5:%.*]] = add i64 [[TMP2]], [[TMP3]]
; CHECK-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP5]]
; CHECK-NEXT: [[TMP6:%.*]] = shl nuw nsw i64 [[I]], 2
; CHECK-NEXT: [[TMP7:%.*]] = add i64 [[TMP6]], 128
; CHECK-NEXT: [[SCEVGEP3:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 [[TMP7]]
; CHECK-NEXT: [[TMP8:%.*]] = add i64 [[TMP6]], 132
; CHECK-NEXT: [[SCEVGEP4:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP8]]
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[OBJ]], i64 0, i32 1, i64 [[I]]
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[Z]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]
; CHECK-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[OBJ]], [[SCEVGEP1]]
; CHECK-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-NEXT: [[BOUND05:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP4]]
; CHECK-NEXT: [[BOUND16:%.*]] = icmp ult ptr [[SCEVGEP3]], [[SCEVGEP1]]
; CHECK-NEXT: [[FOUND_CONFLICT7:%.*]] = and i1 [[BOUND05]], [[BOUND16]]
; CHECK-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[FOUND_CONFLICT]], [[FOUND_CONFLICT7]]
; CHECK-NEXT: br i1 [[CONFLICT_RDX]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[Z]], 4
; CHECK-NEXT: [[N_VEC:%.*]] = sub i64 [[Z]], [[N_MOD_VF]]
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[TMP10]]
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4, !alias.scope [[META0:![0-9]+]]
; CHECK-NEXT: [[TMP13:%.*]] = load i32, ptr [[TMP9]], align 4, !alias.scope [[META3:![0-9]+]]
; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP13]], i64 0
; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-NEXT: [[TMP14:%.*]] = add nsw <4 x i32> [[BROADCAST_SPLAT]], [[WIDE_LOAD]]
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[TMP10]]
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
; CHECK-NEXT: [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP16]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
; CHECK-NEXT: [[TMP17:%.*]] = add nsw <4 x i32> [[TMP14]], [[WIDE_LOAD8]]
; CHECK-NEXT: store <4 x i32> [[TMP17]], ptr [[TMP16]], align 4, !alias.scope [[META5]], !noalias [[META7]]
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[Z]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[DOTOUTER]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTOUTER_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[DOTINNER:%.*]]
; CHECK: .exit:
; CHECK-NEXT: ret void
; CHECK: .outer:
; CHECK-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-NEXT: [[EXITCOND_OUTER:%.*]] = icmp eq i64 [[I_NEXT]], 32
; CHECK-NEXT: br i1 [[EXITCOND_OUTER]], label [[DOTEXIT:%.*]], label [[DOTOUTER_PREHEADER]]
; CHECK: .inner:
; CHECK-NEXT: [[J:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[J_NEXT:%.*]], [[DOTINNER]] ]
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[J]]
; CHECK-NEXT: [[TMP20:%.*]] = load i32, ptr [[TMP19]], align 4
; CHECK-NEXT: [[TMP21:%.*]] = load i32, ptr [[TMP9]], align 4
; CHECK-NEXT: [[TMP22:%.*]] = add nsw i32 [[TMP21]], [[TMP20]]
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[J]]
; CHECK-NEXT: [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
; CHECK-NEXT: [[TMP25:%.*]] = add nsw i32 [[TMP22]], [[TMP24]]
; CHECK-NEXT: store i32 [[TMP25]], ptr [[TMP23]], align 4
; CHECK-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 1
; CHECK-NEXT: [[EXITCOND_INNER:%.*]] = icmp eq i64 [[J_NEXT]], [[Z]]
; CHECK-NEXT: br i1 [[EXITCOND_INNER]], label [[DOTOUTER]], label [[DOTINNER]], !llvm.loop [[LOOP11:![0-9]+]]
;
; CHECK-HOIST-LABEL: @Test(
; CHECK-HOIST-NEXT: [[SCEVGEP:%.*]] = getelementptr nuw i8, ptr [[OBJ:%.*]], i64 256
; CHECK-HOIST-NEXT: [[TMP1:%.*]] = shl i64 [[Z:%.*]], 2
; CHECK-HOIST-NEXT: [[TMP2:%.*]] = add i64 [[TMP1]], 4224
; CHECK-HOIST-NEXT: [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP2]]
; CHECK-HOIST-NEXT: [[SCEVGEP2:%.*]] = getelementptr i8, ptr [[OBJ]], i64 [[TMP1]]
; CHECK-HOIST-NEXT: [[SCEVGEP3:%.*]] = getelementptr nuw i8, ptr [[OBJ]], i64 128
; CHECK-HOIST-NEXT: br label [[DOTOUTER_PREHEADER:%.*]]
; CHECK-HOIST: .outer.preheader:
; CHECK-HOIST-NEXT: [[I:%.*]] = phi i64 [ 0, [[TMP0:%.*]] ], [ [[I_NEXT:%.*]], [[DOTOUTER:%.*]] ]
; CHECK-HOIST-NEXT: [[TMP3:%.*]] = getelementptr inbounds [[STRUCT_S:%.*]], ptr [[OBJ]], i64 0, i32 1, i64 [[I]]
; CHECK-HOIST-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[Z]], 4
; CHECK-HOIST-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK-HOIST: vector.memcheck:
; CHECK-HOIST-NEXT: [[BOUND0:%.*]] = icmp ult ptr [[SCEVGEP]], [[SCEVGEP2]]
; CHECK-HOIST-NEXT: [[BOUND1:%.*]] = icmp ult ptr [[OBJ]], [[SCEVGEP1]]
; CHECK-HOIST-NEXT: [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
; CHECK-HOIST-NEXT: [[BOUND14:%.*]] = icmp ult ptr [[SCEVGEP3]], [[SCEVGEP1]]
; CHECK-HOIST-NEXT: br i1 [[FOUND_CONFLICT]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK-HOIST: vector.ph:
; CHECK-HOIST-NEXT: [[N_MOD_VF:%.*]] = urem i64 [[Z]], 4
; CHECK-HOIST-NEXT: [[N_VEC:%.*]] = sub i64 [[Z]], [[N_MOD_VF]]
; CHECK-HOIST-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK-HOIST: vector.body:
; CHECK-HOIST-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-HOIST-NEXT: [[TMP4:%.*]] = add i64 [[INDEX]], 0
; CHECK-HOIST-NEXT: [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[TMP4]]
; CHECK-HOIST-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
; CHECK-HOIST-NEXT: [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META0:![0-9]+]]
; CHECK-HOIST-NEXT: [[TMP7:%.*]] = load i32, ptr [[TMP3]], align 4, !alias.scope [[META3:![0-9]+]]
; CHECK-HOIST-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP7]], i64 0
; CHECK-HOIST-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
; CHECK-HOIST-NEXT: [[TMP8:%.*]] = add nsw <4 x i32> [[BROADCAST_SPLAT]], [[WIDE_LOAD]]
; CHECK-HOIST-NEXT: [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[TMP4]]
; CHECK-HOIST-NEXT: [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
; CHECK-HOIST-NEXT: [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
; CHECK-HOIST-NEXT: [[TMP11:%.*]] = add nsw <4 x i32> [[TMP8]], [[WIDE_LOAD5]]
; CHECK-HOIST-NEXT: store <4 x i32> [[TMP11]], ptr [[TMP10]], align 4, !alias.scope [[META5]], !noalias [[META7]]
; CHECK-HOIST-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-HOIST-NEXT: [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-HOIST-NEXT: br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
; CHECK-HOIST: middle.block:
; CHECK-HOIST-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[Z]], [[N_VEC]]
; CHECK-HOIST-NEXT: br i1 [[CMP_N]], label [[DOTOUTER]], label [[SCALAR_PH]]
; CHECK-HOIST: scalar.ph:
; CHECK-HOIST-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[DOTOUTER_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-HOIST-NEXT: br label [[DOTINNER:%.*]]
; CHECK-HOIST: .exit:
; CHECK-HOIST-NEXT: ret void
; CHECK-HOIST: .outer:
; CHECK-HOIST-NEXT: [[I_NEXT]] = add nuw nsw i64 [[I]], 1
; CHECK-HOIST-NEXT: [[EXITCOND_OUTER:%.*]] = icmp eq i64 [[I_NEXT]], 32
; CHECK-HOIST-NEXT: br i1 [[EXITCOND_OUTER]], label [[DOTEXIT:%.*]], label [[DOTOUTER_PREHEADER]]
; CHECK-HOIST: .inner:
; CHECK-HOIST-NEXT: [[J:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[J_NEXT:%.*]], [[DOTINNER]] ]
; CHECK-HOIST-NEXT: [[TMP13:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[J]]
; CHECK-HOIST-NEXT: [[TMP14:%.*]] = load i32, ptr [[TMP13]], align 4
; CHECK-HOIST-NEXT: [[TMP15:%.*]] = load i32, ptr [[TMP3]], align 4
; CHECK-HOIST-NEXT: [[TMP16:%.*]] = add nsw i32 [[TMP15]], [[TMP14]]
; CHECK-HOIST-NEXT: [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[J]]
; CHECK-HOIST-NEXT: [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
; CHECK-HOIST-NEXT: [[TMP19:%.*]] = add nsw i32 [[TMP16]], [[TMP18]]
; CHECK-HOIST-NEXT: store i32 [[TMP19]], ptr [[TMP17]], align 4
; CHECK-HOIST-NEXT: [[J_NEXT]] = add nuw nsw i64 [[J]], 1
; CHECK-HOIST-NEXT: [[EXITCOND_INNER:%.*]] = icmp eq i64 [[J_NEXT]], [[Z]]
; CHECK-HOIST-NEXT: br i1 [[EXITCOND_INNER]], label [[DOTOUTER]], label [[DOTINNER]], !llvm.loop [[LOOP11:![0-9]+]]
;
br label %.outer.preheader
.outer.preheader:
%i = phi i64 [ 0, %0 ], [ %i.next, %.outer ]
%1 = getelementptr inbounds %struct.s, ptr %obj, i64 0, i32 1, i64 %i
br label %.inner
.exit:
ret void
.outer:
%i.next = add nuw nsw i64 %i, 1
%exitcond.outer = icmp eq i64 %i.next, 32
br i1 %exitcond.outer, label %.exit, label %.outer.preheader
.inner:
%j = phi i64 [ 0, %.outer.preheader ], [ %j.next, %.inner ]
%2 = getelementptr inbounds %struct.s, ptr %obj, i64 0, i32 0, i64 %j
%3 = load i32, ptr %2
%4 = load i32, ptr %1
%5 = add nsw i32 %4, %3
%6 = getelementptr inbounds %struct.s, ptr %obj, i64 0, i32 2, i64 %i, i64 %j
%7 = load i32, ptr %6
%8 = add nsw i32 %5, %7
store i32 %8, ptr %6
%j.next = add nuw nsw i64 %j, 1
%exitcond.inner = icmp eq i64 %j.next, %z
br i1 %exitcond.inner, label %.outer, label %.inner
}
Codebase Browser
llvm