; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -o - -S %s | FileCheck %s
; Test case with a large number of pointer groups to check for memory
; conflicts, but with many redundant checks that can be simplified.
define void @test_large_number_of_group(ptr %dst, i64 %off, i64 %N) {
; CHECK-LABEL: @test_large_number_of_group(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[OFF_MUL_2:%.*]] = shl i64 [[OFF:%.*]], 1
; CHECK-NEXT: [[OFF_MUL_3:%.*]] = mul i64 [[OFF]], 3
; CHECK-NEXT: [[OFF_MUL_4:%.*]] = shl i64 [[OFF]], 2
; CHECK-NEXT: [[OFF_MUL_5:%.*]] = mul i64 [[OFF]], 5
; CHECK-NEXT: [[OFF_MUL_6:%.*]] = mul i64 [[OFF]], 6
; CHECK-NEXT: [[OFF_MUL_7:%.*]] = mul i64 [[OFF]], 7
; CHECK-NEXT: [[OFF_MUL_8:%.*]] = shl i64 [[OFF]], 3
; CHECK-NEXT: [[OFF_MUL_9:%.*]] = mul i64 [[OFF]], 9
; CHECK-NEXT: [[OFF_MUL_10:%.*]] = mul i64 [[OFF]], 10
; CHECK-NEXT: [[OFF_MUL_11:%.*]] = mul i64 [[OFF]], 11
; CHECK-NEXT: [[OFF_MUL_12:%.*]] = mul i64 [[OFF]], 12
; CHECK-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 4
; CHECK-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[OFF_MUL_8]], 32
; CHECK-NEXT: [[TMP0:%.*]] = shl i64 [[OFF]], 4
; CHECK-NEXT: [[DIFF_CHECK1:%.*]] = icmp ult i64 [[TMP0]], 32
; CHECK-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK1]]
; CHECK-NEXT: [[TMP1:%.*]] = mul i64 [[OFF]], 24
; CHECK-NEXT: [[DIFF_CHECK2:%.*]] = icmp ult i64 [[TMP1]], 32
; CHECK-NEXT: [[CONFLICT_RDX3:%.*]] = or i1 [[CONFLICT_RDX]], [[DIFF_CHECK2]]
; CHECK-NEXT: [[TMP2:%.*]] = shl i64 [[OFF]], 5
; CHECK-NEXT: [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP2]], 32
; CHECK-NEXT: [[CONFLICT_RDX5:%.*]] = or i1 [[CONFLICT_RDX3]], [[DIFF_CHECK4]]
; CHECK-NEXT: [[TMP3:%.*]] = mul i64 [[OFF]], 40
; CHECK-NEXT: [[DIFF_CHECK6:%.*]] = icmp ult i64 [[TMP3]], 32
; CHECK-NEXT: [[CONFLICT_RDX7:%.*]] = or i1 [[CONFLICT_RDX5]], [[DIFF_CHECK6]]
; CHECK-NEXT: [[TMP4:%.*]] = mul i64 [[OFF]], 48
; CHECK-NEXT: [[DIFF_CHECK8:%.*]] = icmp ult i64 [[TMP4]], 32
; CHECK-NEXT: [[CONFLICT_RDX9:%.*]] = or i1 [[CONFLICT_RDX7]], [[DIFF_CHECK8]]
; CHECK-NEXT: [[TMP5:%.*]] = mul i64 [[OFF]], 56
; CHECK-NEXT: [[DIFF_CHECK10:%.*]] = icmp ult i64 [[TMP5]], 32
; CHECK-NEXT: [[CONFLICT_RDX11:%.*]] = or i1 [[CONFLICT_RDX9]], [[DIFF_CHECK10]]
; CHECK-NEXT: [[TMP6:%.*]] = shl i64 [[OFF]], 6
; CHECK-NEXT: [[DIFF_CHECK12:%.*]] = icmp ult i64 [[TMP6]], 32
; CHECK-NEXT: [[CONFLICT_RDX13:%.*]] = or i1 [[CONFLICT_RDX11]], [[DIFF_CHECK12]]
; CHECK-NEXT: [[TMP7:%.*]] = mul i64 [[OFF]], 72
; CHECK-NEXT: [[DIFF_CHECK14:%.*]] = icmp ult i64 [[TMP7]], 32
; CHECK-NEXT: [[CONFLICT_RDX15:%.*]] = or i1 [[CONFLICT_RDX13]], [[DIFF_CHECK14]]
; CHECK-NEXT: [[TMP8:%.*]] = mul i64 [[OFF]], 80
; CHECK-NEXT: [[DIFF_CHECK16:%.*]] = icmp ult i64 [[TMP8]], 32
; CHECK-NEXT: [[CONFLICT_RDX17:%.*]] = or i1 [[CONFLICT_RDX15]], [[DIFF_CHECK16]]
; CHECK-NEXT: [[TMP9:%.*]] = mul i64 [[OFF]], 88
; CHECK-NEXT: [[DIFF_CHECK18:%.*]] = icmp ult i64 [[TMP9]], 32
; CHECK-NEXT: [[CONFLICT_RDX19:%.*]] = or i1 [[CONFLICT_RDX17]], [[DIFF_CHECK18]]
; CHECK-NEXT: br i1 [[CONFLICT_RDX19]], 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: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP10:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP11:%.*]] = add nsw i64 [[TMP10]], -5
; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[TMP11]], [[OFF]]
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i64, ptr [[DST:%.*]], i64 [[TMP12]]
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr double, ptr [[TMP13]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP14]], align 8
; CHECK-NEXT: [[TMP15:%.*]] = add i64 [[TMP11]], [[OFF_MUL_2]]
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP15]]
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr double, ptr [[TMP16]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP17]], align 8
; CHECK-NEXT: [[TMP18:%.*]] = add i64 [[TMP11]], [[OFF_MUL_3]]
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP18]]
; CHECK-NEXT: [[TMP20:%.*]] = getelementptr double, ptr [[TMP19]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP20]], align 8
; CHECK-NEXT: [[TMP21:%.*]] = add i64 [[TMP11]], [[OFF_MUL_4]]
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP21]]
; CHECK-NEXT: [[TMP23:%.*]] = getelementptr double, ptr [[TMP22]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP23]], align 8
; CHECK-NEXT: [[TMP24:%.*]] = add i64 [[TMP11]], [[OFF_MUL_5]]
; CHECK-NEXT: [[TMP25:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP24]]
; CHECK-NEXT: [[TMP26:%.*]] = getelementptr double, ptr [[TMP25]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP26]], align 8
; CHECK-NEXT: [[TMP27:%.*]] = add i64 [[TMP11]], [[OFF_MUL_6]]
; CHECK-NEXT: [[TMP28:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP27]]
; CHECK-NEXT: [[TMP29:%.*]] = getelementptr double, ptr [[TMP28]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP29]], align 8
; CHECK-NEXT: [[TMP30:%.*]] = add i64 [[TMP11]], [[OFF_MUL_7]]
; CHECK-NEXT: [[TMP31:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP30]]
; CHECK-NEXT: [[TMP32:%.*]] = getelementptr double, ptr [[TMP31]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP32]], align 8
; CHECK-NEXT: [[TMP33:%.*]] = add i64 [[TMP11]], [[OFF_MUL_8]]
; CHECK-NEXT: [[TMP34:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP33]]
; CHECK-NEXT: [[TMP35:%.*]] = getelementptr double, ptr [[TMP34]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP35]], align 8
; CHECK-NEXT: [[TMP36:%.*]] = add i64 [[TMP11]], [[OFF_MUL_9]]
; CHECK-NEXT: [[TMP37:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP36]]
; CHECK-NEXT: [[TMP38:%.*]] = getelementptr double, ptr [[TMP37]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP38]], align 8
; CHECK-NEXT: [[TMP39:%.*]] = add i64 [[TMP11]], [[OFF_MUL_10]]
; CHECK-NEXT: [[TMP40:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP39]]
; CHECK-NEXT: [[TMP41:%.*]] = getelementptr double, ptr [[TMP40]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP41]], align 8
; CHECK-NEXT: [[TMP42:%.*]] = add i64 [[TMP11]], [[OFF_MUL_11]]
; CHECK-NEXT: [[TMP43:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP42]]
; CHECK-NEXT: [[TMP44:%.*]] = getelementptr double, ptr [[TMP43]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP44]], align 8
; CHECK-NEXT: [[TMP45:%.*]] = add i64 [[TMP11]], [[OFF_MUL_12]]
; CHECK-NEXT: [[TMP46:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP45]]
; CHECK-NEXT: [[TMP47:%.*]] = getelementptr double, ptr [[TMP46]], i32 0
; CHECK-NEXT: store <4 x double> zeroinitializer, ptr [[TMP47]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP48:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
; CHECK-NEXT: br i1 [[TMP48]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
; CHECK-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[IV_SUB_5:%.*]] = add nsw i64 [[IV]], -5
; CHECK-NEXT: [[IDX_1:%.*]] = add i64 [[IV_SUB_5]], [[OFF]]
; CHECK-NEXT: [[GEP_1:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_1]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_1]], align 8
; CHECK-NEXT: [[IDX_2:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_2]]
; CHECK-NEXT: [[GEP_2:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_2]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_2]], align 8
; CHECK-NEXT: [[IDX_3:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_3]]
; CHECK-NEXT: [[GEP_3:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_3]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_3]], align 8
; CHECK-NEXT: [[IDX_4:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_4]]
; CHECK-NEXT: [[GEP_4:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_4]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_4]], align 8
; CHECK-NEXT: [[IDX_5:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_5]]
; CHECK-NEXT: [[GEP_5:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_5]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_5]], align 8
; CHECK-NEXT: [[IDX_6:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_6]]
; CHECK-NEXT: [[GEP_6:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_6]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_6]], align 8
; CHECK-NEXT: [[IDX_7:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_7]]
; CHECK-NEXT: [[GEP_7:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_7]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_7]], align 8
; CHECK-NEXT: [[IDX_8:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_8]]
; CHECK-NEXT: [[GEP_8:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_8]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_8]], align 8
; CHECK-NEXT: [[IDX_9:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_9]]
; CHECK-NEXT: [[GEP_9:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_9]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_9]], align 8
; CHECK-NEXT: [[IDX_10:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_10]]
; CHECK-NEXT: [[GEP_10:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_10]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_10]], align 8
; CHECK-NEXT: [[IDX_11:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_11]]
; CHECK-NEXT: [[GEP_11:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_11]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_11]], align 8
; CHECK-NEXT: [[IDX_12:%.*]] = add i64 [[IV_SUB_5]], [[OFF_MUL_12]]
; CHECK-NEXT: [[GEP_12:%.*]] = getelementptr i64, ptr [[DST]], i64 [[IDX_12]]
; CHECK-NEXT: store double 0.000000e+00, ptr [[GEP_12]], align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
; CHECK-NEXT: br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%off.mul.2 = shl i64 %off, 1
%off.mul.3 = mul i64 %off, 3
%off.mul.4 = shl i64 %off, 2
%off.mul.5 = mul i64 %off, 5
%off.mul.6 = mul i64 %off, 6
%off.mul.7 = mul i64 %off, 7
%off.mul.8 = shl i64 %off, 3
%off.mul.9 = mul i64 %off, 9
%off.mul.10 = mul i64 %off, 10
%off.mul.11 = mul i64 %off, 11
%off.mul.12 = mul i64 %off, 12
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%iv.sub.5 = add nsw i64 %iv, -5
%idx.1 = add i64 %iv.sub.5, %off
%gep.1 = getelementptr i64, ptr %dst, i64 %idx.1
store double 0.000000e+00, ptr %gep.1, align 8
%idx.2 = add i64 %iv.sub.5, %off.mul.2
%gep.2 = getelementptr i64, ptr %dst, i64 %idx.2
store double 0.000000e+00, ptr %gep.2, align 8
%idx.3 = add i64 %iv.sub.5, %off.mul.3
%gep.3 = getelementptr i64, ptr %dst, i64 %idx.3
store double 0.000000e+00, ptr %gep.3, align 8
%idx.4 = add i64 %iv.sub.5, %off.mul.4
%gep.4 = getelementptr i64, ptr %dst, i64 %idx.4
store double 0.000000e+00, ptr %gep.4, align 8
%idx.5 = add i64 %iv.sub.5, %off.mul.5
%gep.5 = getelementptr i64, ptr %dst, i64 %idx.5
store double 0.000000e+00, ptr %gep.5, align 8
%idx.6 = add i64 %iv.sub.5, %off.mul.6
%gep.6 = getelementptr i64, ptr %dst, i64 %idx.6
store double 0.000000e+00, ptr %gep.6, align 8
%idx.7 = add i64 %iv.sub.5, %off.mul.7
%gep.7 = getelementptr i64, ptr %dst, i64 %idx.7
store double 0.000000e+00, ptr %gep.7, align 8
%idx.8 = add i64 %iv.sub.5, %off.mul.8
%gep.8 = getelementptr i64, ptr %dst, i64 %idx.8
store double 0.000000e+00, ptr %gep.8, align 8
%idx.9 = add i64 %iv.sub.5, %off.mul.9
%gep.9 = getelementptr i64, ptr %dst, i64 %idx.9
store double 0.000000e+00, ptr %gep.9, align 8
%idx.10 = add i64 %iv.sub.5, %off.mul.10
%gep.10 = getelementptr i64, ptr %dst, i64 %idx.10
store double 0.000000e+00, ptr %gep.10, align 8
%idx.11 = add i64 %iv.sub.5, %off.mul.11
%gep.11 = getelementptr i64, ptr %dst, i64 %idx.11
store double 0.000000e+00, ptr %gep.11, align 8
%idx.12 = add i64 %iv.sub.5, %off.mul.12
%gep.12 = getelementptr i64, ptr %dst, i64 %idx.12
store double 0.000000e+00, ptr %gep.12, align 8
%iv.next = add nuw nsw i64 %iv, 1
%ec = icmp eq i64 %iv.next, %N
br i1 %ec, label %exit, label %loop
exit:
ret void
}
define void @check_creation_order(ptr %a, ptr %b, i32 %m) {
; CHECK-LABEL: @check_creation_order(
; CHECK-NEXT: entry:
; CHECK-NEXT: [[B2:%.*]] = ptrtoint ptr [[B:%.*]] to i64
; CHECK-NEXT: [[A1:%.*]] = ptrtoint ptr [[A:%.*]] to i64
; CHECK-NEXT: [[M_EXT:%.*]] = sext i32 [[M:%.*]] to i64
; CHECK-NEXT: [[INVARIANT_GEP:%.*]] = getelementptr double, ptr [[A]], i64 [[M_EXT]]
; CHECK-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_MEMCHECK:%.*]]
; CHECK: vector.memcheck:
; CHECK-NEXT: [[TMP0:%.*]] = mul nsw i64 [[M_EXT]], -8
; CHECK-NEXT: [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP0]], 32
; CHECK-NEXT: [[TMP1:%.*]] = sub i64 [[A1]], [[B2]]
; CHECK-NEXT: [[DIFF_CHECK3:%.*]] = icmp ult i64 [[TMP1]], 32
; CHECK-NEXT: [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK3]]
; CHECK-NEXT: br i1 [[CONFLICT_RDX]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
; CHECK: vector.ph:
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
; CHECK: vector.body:
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
; CHECK-NEXT: [[TMP2:%.*]] = add i64 [[INDEX]], 0
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr double, ptr [[INVARIANT_GEP]], i64 [[TMP2]]
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr double, ptr [[TMP3]], i32 0
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP4]], align 8
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[TMP2]]
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 0
; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x double>, ptr [[TMP6]], align 8
; CHECK-NEXT: [[TMP7:%.*]] = fadd <4 x double> [[WIDE_LOAD]], [[WIDE_LOAD4]]
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[TMP2]]
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds double, ptr [[TMP8]], i32 0
; CHECK-NEXT: store <4 x double> [[TMP7]], ptr [[TMP9]], align 8
; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
; CHECK-NEXT: [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 31996
; CHECK-NEXT: br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
; CHECK: middle.block:
; CHECK-NEXT: br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
; CHECK: scalar.ph:
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 31996, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
; CHECK-NEXT: br label [[LOOP:%.*]]
; CHECK: loop:
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
; CHECK-NEXT: [[GEP:%.*]] = getelementptr double, ptr [[INVARIANT_GEP]], i64 [[IV]]
; CHECK-NEXT: [[L_0:%.*]] = load double, ptr [[GEP]], align 8
; CHECK-NEXT: [[GEP_B:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]]
; CHECK-NEXT: [[L_1:%.*]] = load double, ptr [[GEP_B]], align 8
; CHECK-NEXT: [[ADD3:%.*]] = fadd double [[L_0]], [[L_1]]
; CHECK-NEXT: [[GEP_A:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
; CHECK-NEXT: store double [[ADD3]], ptr [[GEP_A]], align 8
; CHECK-NEXT: [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
; CHECK-NEXT: [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 31999
; CHECK-NEXT: br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
; CHECK: exit:
; CHECK-NEXT: ret void
;
entry:
%m.ext = sext i32 %m to i64
%invariant.gep = getelementptr double, ptr %a, i64 %m.ext
br label %loop
loop:
%iv = phi i64 [ 0, %entry ], [ %iv.next, %loop ]
%gep = getelementptr double, ptr %invariant.gep, i64 %iv
%l.0 = load double, ptr %gep, align 8
%gep.b = getelementptr inbounds double, ptr %b, i64 %iv
%l.1 = load double, ptr %gep.b, align 8
%add3 = fadd double %l.0, %l.1
%gep.a = getelementptr inbounds double, ptr %a, i64 %iv
store double %add3, ptr %gep.a, align 8
%iv.next = add nuw nsw i64 %iv, 1
%exitcond.not = icmp eq i64 %iv.next, 31999
br i1 %exitcond.not, label %exit, label %loop
exit:
ret void
}
Codebase Browser
llvm