; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -basic-aa -loop-distribute -enable-loop-distribute -S < %s | \
; RUN: FileCheck %s --check-prefix=DEFAULT
; RUN: opt -basic-aa -loop-distribute -enable-loop-distribute -S -enable-mem-access-versioning=0 < %s | \
; RUN: FileCheck %s --check-prefix=NO-VERSION
; If we don't speculate stride for 1 we can't distribute along the line
; because we could have a backward dependence:
;
; for (i = 0; i < n; i++) {
; A[i + 1] = A[i] * B[i];
; =======================
; C[i] = D[i] * A[stride * i];
; }
target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
target triple = "x86_64-apple-macosx10.10.0"
define void @f(i32* noalias %a,
;
;
; DEFAULT-LABEL: @f(
; DEFAULT-NEXT: entry:
; DEFAULT-NEXT: br label [[FOR_BODY_LVER_CHECK:%.*]]
; DEFAULT: for.body.lver.check:
; DEFAULT-NEXT: [[IDENT_CHECK:%.*]] = icmp ne i64 [[STRIDE:%.*]], 1
; DEFAULT-NEXT: br i1 [[IDENT_CHECK]], label [[FOR_BODY_PH_LVER_ORIG:%.*]], label [[FOR_BODY_PH_LDIST1:%.*]]
; DEFAULT: for.body.ph.lver.orig:
; DEFAULT-NEXT: br label [[FOR_BODY_LVER_ORIG:%.*]]
; DEFAULT: for.body.lver.orig:
; DEFAULT-NEXT: [[IND_LVER_ORIG:%.*]] = phi i64 [ 0, [[FOR_BODY_PH_LVER_ORIG]] ], [ [[ADD_LVER_ORIG:%.*]], [[FOR_BODY_LVER_ORIG]] ]
; DEFAULT-NEXT: [[ARRAYIDXA_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[IND_LVER_ORIG]]
; DEFAULT-NEXT: [[LOADA_LVER_ORIG:%.*]] = load i32, i32* [[ARRAYIDXA_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[ARRAYIDXB_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[IND_LVER_ORIG]]
; DEFAULT-NEXT: [[LOADB_LVER_ORIG:%.*]] = load i32, i32* [[ARRAYIDXB_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[MULA_LVER_ORIG:%.*]] = mul i32 [[LOADB_LVER_ORIG]], [[LOADA_LVER_ORIG]]
; DEFAULT-NEXT: [[ADD_LVER_ORIG]] = add nuw nsw i64 [[IND_LVER_ORIG]], 1
; DEFAULT-NEXT: [[ARRAYIDXA_PLUS_4_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[ADD_LVER_ORIG]]
; DEFAULT-NEXT: store i32 [[MULA_LVER_ORIG]], i32* [[ARRAYIDXA_PLUS_4_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[ARRAYIDXD_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[D:%.*]], i64 [[IND_LVER_ORIG]]
; DEFAULT-NEXT: [[LOADD_LVER_ORIG:%.*]] = load i32, i32* [[ARRAYIDXD_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[MUL_LVER_ORIG:%.*]] = mul i64 [[IND_LVER_ORIG]], [[STRIDE]]
; DEFAULT-NEXT: [[ARRAYIDXSTRIDEDA_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[MUL_LVER_ORIG]]
; DEFAULT-NEXT: [[LOADSTRIDEDA_LVER_ORIG:%.*]] = load i32, i32* [[ARRAYIDXSTRIDEDA_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[MULC_LVER_ORIG:%.*]] = mul i32 [[LOADD_LVER_ORIG]], [[LOADSTRIDEDA_LVER_ORIG]]
; DEFAULT-NEXT: [[ARRAYIDXC_LVER_ORIG:%.*]] = getelementptr inbounds i32, i32* [[C:%.*]], i64 [[IND_LVER_ORIG]]
; DEFAULT-NEXT: store i32 [[MULC_LVER_ORIG]], i32* [[ARRAYIDXC_LVER_ORIG]], align 4
; DEFAULT-NEXT: [[EXITCOND_LVER_ORIG:%.*]] = icmp eq i64 [[ADD_LVER_ORIG]], 20
; DEFAULT-NEXT: br i1 [[EXITCOND_LVER_ORIG]], label [[FOR_END_LOOPEXIT:%.*]], label [[FOR_BODY_LVER_ORIG]]
; DEFAULT: for.body.ph.ldist1:
; DEFAULT-NEXT: br label [[FOR_BODY_LDIST1:%.*]]
; DEFAULT: for.body.ldist1:
; DEFAULT-NEXT: [[IND_LDIST1:%.*]] = phi i64 [ 0, [[FOR_BODY_PH_LDIST1]] ], [ [[ADD_LDIST1:%.*]], [[FOR_BODY_LDIST1]] ]
; DEFAULT-NEXT: [[ARRAYIDXA_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[IND_LDIST1]]
; DEFAULT-NEXT: [[LOADA_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXA_LDIST1]], align 4
; DEFAULT-NEXT: [[ARRAYIDXB_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[B]], i64 [[IND_LDIST1]]
; DEFAULT-NEXT: [[LOADB_LDIST1:%.*]] = load i32, i32* [[ARRAYIDXB_LDIST1]], align 4
; DEFAULT-NEXT: [[MULA_LDIST1:%.*]] = mul i32 [[LOADB_LDIST1]], [[LOADA_LDIST1]]
; DEFAULT-NEXT: [[ADD_LDIST1]] = add nuw nsw i64 [[IND_LDIST1]], 1
; DEFAULT-NEXT: [[ARRAYIDXA_PLUS_4_LDIST1:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[ADD_LDIST1]]
; DEFAULT-NEXT: store i32 [[MULA_LDIST1]], i32* [[ARRAYIDXA_PLUS_4_LDIST1]], align 4
; DEFAULT-NEXT: [[EXITCOND_LDIST1:%.*]] = icmp eq i64 [[ADD_LDIST1]], 20
; DEFAULT-NEXT: br i1 [[EXITCOND_LDIST1]], label [[FOR_BODY_PH:%.*]], label [[FOR_BODY_LDIST1]]
; DEFAULT: for.body.ph:
; DEFAULT-NEXT: br label [[FOR_BODY:%.*]]
; DEFAULT: for.body:
; DEFAULT-NEXT: [[IND:%.*]] = phi i64 [ 0, [[FOR_BODY_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
; DEFAULT-NEXT: [[ADD]] = add nuw nsw i64 [[IND]], 1
; DEFAULT-NEXT: [[ARRAYIDXD:%.*]] = getelementptr inbounds i32, i32* [[D]], i64 [[IND]]
; DEFAULT-NEXT: [[LOADD:%.*]] = load i32, i32* [[ARRAYIDXD]], align 4
; DEFAULT-NEXT: [[MUL:%.*]] = mul i64 [[IND]], [[STRIDE]]
; DEFAULT-NEXT: [[ARRAYIDXSTRIDEDA:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[MUL]]
; DEFAULT-NEXT: [[LOADSTRIDEDA:%.*]] = load i32, i32* [[ARRAYIDXSTRIDEDA]], align 4
; DEFAULT-NEXT: [[MULC:%.*]] = mul i32 [[LOADD]], [[LOADSTRIDEDA]]
; DEFAULT-NEXT: [[ARRAYIDXC:%.*]] = getelementptr inbounds i32, i32* [[C]], i64 [[IND]]
; DEFAULT-NEXT: store i32 [[MULC]], i32* [[ARRAYIDXC]], align 4
; DEFAULT-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[ADD]], 20
; DEFAULT-NEXT: br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT1:%.*]], label [[FOR_BODY]]
; DEFAULT: for.end.loopexit:
; DEFAULT-NEXT: br label [[FOR_END:%.*]]
; DEFAULT: for.end.loopexit1:
; DEFAULT-NEXT: br label [[FOR_END]]
; DEFAULT: for.end:
; DEFAULT-NEXT: ret void
;
; NO-VERSION-LABEL: @f(
; NO-VERSION-NEXT: entry:
; NO-VERSION-NEXT: br label [[FOR_BODY:%.*]]
; NO-VERSION: for.body:
; NO-VERSION-NEXT: [[IND:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
; NO-VERSION-NEXT: [[ARRAYIDXA:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[IND]]
; NO-VERSION-NEXT: [[LOADA:%.*]] = load i32, i32* [[ARRAYIDXA]], align 4
; NO-VERSION-NEXT: [[ARRAYIDXB:%.*]] = getelementptr inbounds i32, i32* [[B:%.*]], i64 [[IND]]
; NO-VERSION-NEXT: [[LOADB:%.*]] = load i32, i32* [[ARRAYIDXB]], align 4
; NO-VERSION-NEXT: [[MULA:%.*]] = mul i32 [[LOADB]], [[LOADA]]
; NO-VERSION-NEXT: [[ADD]] = add nuw nsw i64 [[IND]], 1
; NO-VERSION-NEXT: [[ARRAYIDXA_PLUS_4:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[ADD]]
; NO-VERSION-NEXT: store i32 [[MULA]], i32* [[ARRAYIDXA_PLUS_4]], align 4
; NO-VERSION-NEXT: [[ARRAYIDXD:%.*]] = getelementptr inbounds i32, i32* [[D:%.*]], i64 [[IND]]
; NO-VERSION-NEXT: [[LOADD:%.*]] = load i32, i32* [[ARRAYIDXD]], align 4
; NO-VERSION-NEXT: [[MUL:%.*]] = mul i64 [[IND]], [[STRIDE:%.*]]
; NO-VERSION-NEXT: [[ARRAYIDXSTRIDEDA:%.*]] = getelementptr inbounds i32, i32* [[A]], i64 [[MUL]]
; NO-VERSION-NEXT: [[LOADSTRIDEDA:%.*]] = load i32, i32* [[ARRAYIDXSTRIDEDA]], align 4
; NO-VERSION-NEXT: [[MULC:%.*]] = mul i32 [[LOADD]], [[LOADSTRIDEDA]]
; NO-VERSION-NEXT: [[ARRAYIDXC:%.*]] = getelementptr inbounds i32, i32* [[C:%.*]], i64 [[IND]]
; NO-VERSION-NEXT: store i32 [[MULC]], i32* [[ARRAYIDXC]], align 4
; NO-VERSION-NEXT: [[EXITCOND:%.*]] = icmp eq i64 [[ADD]], 20
; NO-VERSION-NEXT: br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]]
; NO-VERSION: for.end:
; NO-VERSION-NEXT: ret void
;
i32* noalias %b,
i32* noalias %c,
i32* noalias %d,
i64 %stride) {
entry:
br label %for.body
for.body: ; preds = %for.body, %entry
%ind = phi i64 [ 0, %entry ], [ %add, %for.body ]
%arrayidxA = getelementptr inbounds i32, i32* %a, i64 %ind
%loadA = load i32, i32* %arrayidxA, align 4
%arrayidxB = getelementptr inbounds i32, i32* %b, i64 %ind
%loadB = load i32, i32* %arrayidxB, align 4
%mulA = mul i32 %loadB, %loadA
%add = add nuw nsw i64 %ind, 1
%arrayidxA_plus_4 = getelementptr inbounds i32, i32* %a, i64 %add
store i32 %mulA, i32* %arrayidxA_plus_4, align 4
%arrayidxD = getelementptr inbounds i32, i32* %d, i64 %ind
%loadD = load i32, i32* %arrayidxD, align 4
%mul = mul i64 %ind, %stride
%arrayidxStridedA = getelementptr inbounds i32, i32* %a, i64 %mul
%loadStridedA = load i32, i32* %arrayidxStridedA, align 4
%mulC = mul i32 %loadD, %loadStridedA
%arrayidxC = getelementptr inbounds i32, i32* %c, i64 %ind
store i32 %mulC, i32* %arrayidxC, align 4
%exitcond = icmp eq i64 %add, 20
br i1 %exitcond, label %for.end, label %for.body
for.end: ; preds = %for.body
ret void
}