137 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 42; RUN: opt -S < %s -passes=loop-vectorize -mtriple aarch64-linux-gnu -mattr=+sve 2>&1 | FileCheck %s3 4; TODO: Remove single-iteration vector loop.5define void @clamped_tc_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range(1,16) {6; CHECK-LABEL: define void @clamped_tc_8(7; CHECK-SAME: ptr captures(none) [[DST:%.*]], i32 [[N:%.*]], i64 [[VAL:%.*]]) #[[ATTR0:[0-9]+]] {8; CHECK-NEXT: entry:9; CHECK-NEXT: br label [[VECTOR_PH:%.*]]10; CHECK: vector.ph:11; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()12; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 813; CHECK-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 8)14; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i64> poison, i64 [[VAL]], i64 015; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i64> poison, <vscale x 8 x i32> zeroinitializer16; CHECK-NEXT: [[TMP8:%.*]] = call <vscale x 8 x i64> @llvm.stepvector.nxv8i64()17; CHECK-NEXT: [[TMP3:%.*]] = mul <vscale x 8 x i64> [[TMP8]], splat (i64 1)18; CHECK-NEXT: [[INDUCTION:%.*]] = add <vscale x 8 x i64> zeroinitializer, [[TMP3]]19; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i64> poison, i64 [[TMP1]], i64 020; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i64> [[DOTSPLATINSERT]], <vscale x 8 x i64> poison, <vscale x 8 x i32> zeroinitializer21; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]22; CHECK: vector.body:23; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]24; CHECK-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]25; CHECK-NEXT: [[VEC_IND:%.*]] = phi <vscale x 8 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]26; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]27; CHECK-NEXT: [[TMP10:%.*]] = shl nuw nsw <vscale x 8 x i64> [[VEC_IND]], splat (i64 3)28; CHECK-NEXT: [[TMP11:%.*]] = lshr <vscale x 8 x i64> [[BROADCAST_SPLAT]], [[TMP10]]29; CHECK-NEXT: [[TMP14:%.*]] = trunc <vscale x 8 x i64> [[TMP11]] to <vscale x 8 x i8>30; CHECK-NEXT: call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr align 1 [[NEXT_GEP]], <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])31; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]32; CHECK-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 8)33; CHECK-NEXT: [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]34; CHECK-NEXT: br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]35; CHECK: middle.block:36; CHECK-NEXT: br label [[FOR_BODY:%.*]]37; CHECK: for.cond.cleanup:38; CHECK-NEXT: ret void39;40entry:41 br label %for.body42 43for.body: ; preds = %entry, %for.body44 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]45 %p_out_tail.09 = phi ptr [ %dst, %entry ], [ %incdec.ptr, %for.body ]46 %0 = shl nuw nsw i64 %indvars.iv, 347 %shr3 = lshr i64 %val, %048 %conv4 = trunc i64 %shr3 to i849 store i8 %conv4, ptr %p_out_tail.09, align 150 %incdec.ptr = getelementptr inbounds i8, ptr %p_out_tail.09, i64 151 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 152 %exitcond.not = icmp eq i64 %indvars.iv.next, 853 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body54 55for.cond.cleanup: ; preds = %for.body56 ret void57}58 59; TODO: Remove single-iteration vector loop.60define void @clamped_tc_max_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range(1,16) {61; CHECK-LABEL: define void @clamped_tc_max_8(62; CHECK-SAME: ptr captures(none) [[DST:%.*]], i32 [[N:%.*]], i64 [[VAL:%.*]]) #[[ATTR0]] {63; CHECK-NEXT: entry:64; CHECK-NEXT: [[REM:%.*]] = and i32 [[N]], 6365; CHECK-NEXT: [[CMP8_NOT:%.*]] = icmp eq i32 [[REM]], 066; CHECK-NEXT: br i1 [[CMP8_NOT]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY_PREHEADER:%.*]]67; CHECK: for.body.preheader:68; CHECK-NEXT: [[ADD:%.*]] = add nuw nsw i32 [[REM]], 769; CHECK-NEXT: [[SHR:%.*]] = lshr i32 [[ADD]], 370; CHECK-NEXT: [[WIDE_TRIP_COUNT:%.*]] = zext i32 [[SHR]] to i6471; CHECK-NEXT: br label [[VECTOR_PH:%.*]]72; CHECK: vector.ph:73; CHECK-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()74; CHECK-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 875; CHECK-NEXT: [[ACTIVE_LANE_MASK_ENTRY:%.*]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 0, i64 [[WIDE_TRIP_COUNT]])76; CHECK-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i64> poison, i64 [[VAL]], i64 077; CHECK-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i64> poison, <vscale x 8 x i32> zeroinitializer78; CHECK-NEXT: [[TMP8:%.*]] = call <vscale x 8 x i64> @llvm.stepvector.nxv8i64()79; CHECK-NEXT: [[TMP3:%.*]] = mul <vscale x 8 x i64> [[TMP8]], splat (i64 1)80; CHECK-NEXT: [[INDUCTION:%.*]] = add <vscale x 8 x i64> zeroinitializer, [[TMP3]]81; CHECK-NEXT: [[DOTSPLATINSERT:%.*]] = insertelement <vscale x 8 x i64> poison, i64 [[TMP1]], i64 082; CHECK-NEXT: [[DOTSPLAT:%.*]] = shufflevector <vscale x 8 x i64> [[DOTSPLATINSERT]], <vscale x 8 x i64> poison, <vscale x 8 x i32> zeroinitializer83; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]84; CHECK: vector.body:85; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]86; CHECK-NEXT: [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]87; CHECK-NEXT: [[VEC_IND:%.*]] = phi <vscale x 8 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]88; CHECK-NEXT: [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]89; CHECK-NEXT: [[TMP10:%.*]] = shl nuw nsw <vscale x 8 x i64> [[VEC_IND]], splat (i64 3)90; CHECK-NEXT: [[TMP11:%.*]] = lshr <vscale x 8 x i64> [[BROADCAST_SPLAT]], [[TMP10]]91; CHECK-NEXT: [[TMP14:%.*]] = trunc <vscale x 8 x i64> [[TMP11]] to <vscale x 8 x i8>92; CHECK-NEXT: call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr align 1 [[NEXT_GEP]], <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])93; CHECK-NEXT: [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP1]]94; CHECK-NEXT: [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])95; CHECK-NEXT: [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]96; CHECK-NEXT: br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]97; CHECK: middle.block:98; CHECK-NEXT: br label [[FOR_BODY:%.*]]99; CHECK: for.cond.cleanup.loopexit:100; CHECK-NEXT: br label [[FOR_COND_CLEANUP]]101; CHECK: for.cond.cleanup:102; CHECK-NEXT: ret void103;104 105entry:106 %rem = and i32 %n, 63107 %cmp8.not = icmp eq i32 %rem, 0108 br i1 %cmp8.not, label %for.cond.cleanup, label %for.body.preheader109 110for.body.preheader: ; preds = %entry111 %add = add nuw nsw i32 %rem, 7112 %shr = lshr i32 %add, 3113 %wide.trip.count = zext i32 %shr to i64114 br label %for.body115 116for.body: ; preds = %for.body.preheader, %for.body117 %indvars.iv = phi i64 [ 0, %for.body.preheader ], [ %indvars.iv.next, %for.body ]118 %p_out_tail.09 = phi ptr [ %dst, %for.body.preheader ], [ %incdec.ptr, %for.body ]119 %0 = shl nuw nsw i64 %indvars.iv, 3120 %shr3 = lshr i64 %val, %0121 %conv4 = trunc i64 %shr3 to i8122 store i8 %conv4, ptr %p_out_tail.09, align 1123 %incdec.ptr = getelementptr inbounds i8, ptr %p_out_tail.09, i64 1124 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1125 %exitcond.not = icmp eq i64 %indvars.iv.next, %wide.trip.count126 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body127 128for.cond.cleanup: ; preds = %for.body129 ret void130}131;.132; CHECK: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}133; CHECK: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}134; CHECK: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}135; CHECK: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]], [[META2]]}136;.137