143 lines · plain
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py2; RUN: opt %s -S -riscv-gather-scatter-lowering -mtriple=riscv64 -mattr=+m,+v,+zvl256b | FileCheck %s3 4; This contains negative tests for the strided load/store recognition in5; RISCVGatherScatterLowering.cpp6 7; Negative test for treating OR as ADD.8define void @gather_bad_or(ptr noalias nocapture %A, ptr noalias nocapture readonly %B) {9; CHECK-LABEL: @gather_bad_or(10; CHECK-NEXT: entry:11; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]12; CHECK: vector.body:13; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]14; CHECK-NEXT: [[VEC_IND:%.*]] = phi <32 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15, i64 16, i64 17, i64 18, i64 19, i64 20, i64 21, i64 22, i64 23, i64 24, i64 25, i64 26, i64 27, i64 28, i64 29, i64 30, i64 31>, [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]15; CHECK-NEXT: [[I:%.*]] = mul nuw nsw <32 x i64> [[VEC_IND]], splat (i64 5)16; CHECK-NEXT: [[OR:%.*]] = or <32 x i64> [[I]], splat (i64 1)17; CHECK-NEXT: [[I1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], <32 x i64> [[OR]]18; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> align 1 [[I1]], <32 x i1> splat (i1 true), <32 x i8> poison)19; CHECK-NEXT: [[I2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]20; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[I2]], align 121; CHECK-NEXT: [[I4:%.*]] = add <32 x i8> [[WIDE_LOAD]], [[WIDE_MASKED_GATHER]]22; CHECK-NEXT: store <32 x i8> [[I4]], ptr [[I2]], align 123; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 3224; CHECK-NEXT: [[VEC_IND_NEXT]] = add <32 x i64> [[VEC_IND]], splat (i64 32)25; CHECK-NEXT: [[I6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 102426; CHECK-NEXT: br i1 [[I6]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]]27; CHECK: for.cond.cleanup:28; CHECK-NEXT: ret void29;30entry:31 br label %vector.body32 33vector.body: ; preds = %vector.body, %entry34 %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]35 %vec.ind = phi <32 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15, i64 16, i64 17, i64 18, i64 19, i64 20, i64 21, i64 22, i64 23, i64 24, i64 25, i64 26, i64 27, i64 28, i64 29, i64 30, i64 31>, %entry ], [ %vec.ind.next, %vector.body ]36 %i = mul nuw nsw <32 x i64> %vec.ind, <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>37 %or = or <32 x i64> %i, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>38 %i1 = getelementptr inbounds i8, ptr %B, <32 x i64> %or39 %wide.masked.gather = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> %i1, i32 1, <32 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <32 x i8> poison)40 %i2 = getelementptr inbounds i8, ptr %A, i64 %index41 %wide.load = load <32 x i8>, ptr %i2, align 142 %i4 = add <32 x i8> %wide.load, %wide.masked.gather43 store <32 x i8> %i4, ptr %i2, align 144 %index.next = add nuw i64 %index, 3245 %vec.ind.next = add <32 x i64> %vec.ind, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>46 %i6 = icmp eq i64 %index.next, 102447 br i1 %i6, label %for.cond.cleanup, label %vector.body48 49for.cond.cleanup: ; preds = %vector.body50 ret void51}52 53; Don't transform since we might not handle wrap correctly with narrow indices.54define void @gather_narrow_index(ptr noalias nocapture %A, ptr noalias nocapture readonly %B) {55; CHECK-LABEL: @gather_narrow_index(56; CHECK-NEXT: entry:57; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]58; CHECK: vector.body:59; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]60; CHECK-NEXT: [[VEC_IND:%.*]] = phi <32 x i32> [ <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>, [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]61; CHECK-NEXT: [[TMP0:%.*]] = mul nuw nsw <32 x i32> [[VEC_IND]], splat (i32 5)62; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], <32 x i32> [[TMP0]]63; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> align 1 [[TMP1]], <32 x i1> splat (i1 true), <32 x i8> poison)64; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]65; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[TMP2]], align 166; CHECK-NEXT: [[TMP4:%.*]] = add <32 x i8> [[WIDE_LOAD]], [[WIDE_MASKED_GATHER]]67; CHECK-NEXT: store <32 x i8> [[TMP4]], ptr [[TMP2]], align 168; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 3269; CHECK-NEXT: [[VEC_IND_NEXT]] = add <32 x i32> [[VEC_IND]], splat (i32 32)70; CHECK-NEXT: [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 102471; CHECK-NEXT: br i1 [[TMP6]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]]72; CHECK: for.cond.cleanup:73; CHECK-NEXT: ret void74;75entry:76 br label %vector.body77 78vector.body: ; preds = %vector.body, %entry79 %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]80 %vec.ind = phi <32 x i32> [ <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>, %entry ], [ %vec.ind.next, %vector.body ]81 %0 = mul nuw nsw <32 x i32> %vec.ind, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>82 %1 = getelementptr inbounds i8, ptr %B, <32 x i32> %083 %wide.masked.gather = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> %1, i32 1, <32 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <32 x i8> poison)84 %2 = getelementptr inbounds i8, ptr %A, i64 %index85 %wide.load = load <32 x i8>, ptr %2, align 186 %3 = add <32 x i8> %wide.load, %wide.masked.gather87 store <32 x i8> %3, ptr %2, align 188 %index.next = add nuw i64 %index, 3289 %vec.ind.next = add <32 x i32> %vec.ind, <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32>90 %4 = icmp eq i64 %index.next, 102491 br i1 %4, label %for.cond.cleanup, label %vector.body92 93for.cond.cleanup: ; preds = %vector.body94 ret void95}96 97; The last element of the start value of the phi has the wrong stride.98define void @gather_broken_stride(ptr noalias nocapture %A, ptr noalias nocapture readonly %B) {99; CHECK-LABEL: @gather_broken_stride(100; CHECK-NEXT: entry:101; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]102; CHECK: vector.body:103; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]104; CHECK-NEXT: [[VEC_IND:%.*]] = phi <32 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15, i64 16, i64 17, i64 18, i64 19, i64 20, i64 21, i64 22, i64 23, i64 24, i64 25, i64 26, i64 27, i64 28, i64 29, i64 30, i64 32>, [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]105; CHECK-NEXT: [[I:%.*]] = mul nuw nsw <32 x i64> [[VEC_IND]], splat (i64 5)106; CHECK-NEXT: [[OR:%.*]] = or <32 x i64> [[I]], splat (i64 1)107; CHECK-NEXT: [[I1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], <32 x i64> [[OR]]108; CHECK-NEXT: [[WIDE_MASKED_GATHER:%.*]] = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> align 1 [[I1]], <32 x i1> splat (i1 true), <32 x i8> poison)109; CHECK-NEXT: [[I2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]110; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[I2]], align 1111; CHECK-NEXT: [[I4:%.*]] = add <32 x i8> [[WIDE_LOAD]], [[WIDE_MASKED_GATHER]]112; CHECK-NEXT: store <32 x i8> [[I4]], ptr [[I2]], align 1113; CHECK-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32114; CHECK-NEXT: [[VEC_IND_NEXT]] = add <32 x i64> [[VEC_IND]], splat (i64 32)115; CHECK-NEXT: [[I6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024116; CHECK-NEXT: br i1 [[I6]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]]117; CHECK: for.cond.cleanup:118; CHECK-NEXT: ret void119;120entry:121 br label %vector.body122 123vector.body: ; preds = %vector.body, %entry124 %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]125 %vec.ind = phi <32 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15, i64 16, i64 17, i64 18, i64 19, i64 20, i64 21, i64 22, i64 23, i64 24, i64 25, i64 26, i64 27, i64 28, i64 29, i64 30, i64 32>, %entry ], [ %vec.ind.next, %vector.body ]126 %i = mul nuw nsw <32 x i64> %vec.ind, <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>127 %or = or <32 x i64> %i, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>128 %i1 = getelementptr inbounds i8, ptr %B, <32 x i64> %or129 %wide.masked.gather = call <32 x i8> @llvm.masked.gather.v32i8.v32p0(<32 x ptr> %i1, i32 1, <32 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <32 x i8> poison)130 %i2 = getelementptr inbounds i8, ptr %A, i64 %index131 %wide.load = load <32 x i8>, ptr %i2, align 1132 %i4 = add <32 x i8> %wide.load, %wide.masked.gather133 store <32 x i8> %i4, ptr %i2, align 1134 %index.next = add nuw i64 %index, 32135 %vec.ind.next = add <32 x i64> %vec.ind, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>136 %i6 = icmp eq i64 %index.next, 1024137 br i1 %i6, label %for.cond.cleanup, label %vector.body138 139for.cond.cleanup: ; preds = %vector.body140 ret void141}142 143