235 lines · plain
1; RUN: opt -passes=loop-simplify,loop-distribute -enable-loop-distribute -S < %s 2>&1 \2; RUN: | FileCheck %s --check-prefix=ALWAYS --check-prefix=NO_REMARKS3; RUN: opt -passes=loop-simplify,loop-distribute -enable-loop-distribute -S \4; RUN: -pass-remarks-missed=loop-distribute < %s 2>&1 \5; RUN: | FileCheck %s --check-prefix=ALWAYS --check-prefix=MISSED_REMARKS6; RUN: opt -passes=loop-simplify,loop-distribute -enable-loop-distribute -S \7; RUN: -pass-remarks-analysis=loop-distribute < %s 2>&1 \8; RUN: | FileCheck %s --check-prefix=ALWAYS --check-prefix=ANALYSIS_REMARKS9; RUN: opt -passes=loop-simplify,loop-distribute -enable-loop-distribute -S \10; RUN: -pass-remarks=loop-distribute < %s 2>&1 \11; RUN: | FileCheck %s --check-prefix=ALWAYS --check-prefix=REMARKS12 13; This is the input program:14;15; 1 void forced (char *A, char *B, char *C, int N) {16; 2 #pragma clang loop distribute(enable)17; 3 for(int i = 0; i < N; i++) {18; 4 A[i] = B[i] * C[i];19; 5 }20; 6 }21; 722; 8 void not_forced (char *A, char *B, char *C, int N) {23; 9 for(int i = 0; i < N; i++) {24; 10 A[i] = B[i] * C[i];25; 11 }26; 12 }27; 1328; 14 void success (char *A, char *B, char *C, char *D, char *E, int N) {29; 15 for(int i = 0; i < N; i++) {30; 16 A[i + 1] = A[i] + B[i];31; 17 C[i] = D[i] * E[i];32; 18 }33; 19 }34 35; MISSED_REMARKS: remark: /tmp/t.c:3:3: loop not distributed: use -Rpass-analysis=loop-distribute for more info36; ALWAYS: remark: /tmp/t.c:3:3: loop not distributed: memory operations are safe for vectorization37; ALWAYS: warning: /tmp/t.c:3:3: loop not distributed: failed explicitly specified loop distribution38 39define void @forced(ptr %A, ptr %B, ptr %C, i32 %N) !dbg !7 {40entry:41 %cmp12 = icmp sgt i32 %N, 0, !dbg !942 br i1 %cmp12, label %ph, label %for.cond.cleanup, !dbg !1043 44ph:45 br label %for.body46 47for.body:48 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]49 %arrayidx = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !1250 %0 = load i8, ptr %arrayidx, align 1, !dbg !12, !tbaa !1351 %arrayidx2 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !1652 %1 = load i8, ptr %arrayidx2, align 1, !dbg !16, !tbaa !1353 %mul = mul i8 %1, %0, !dbg !1754 %arrayidx6 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !1855 store i8 %mul, ptr %arrayidx6, align 1, !dbg !19, !tbaa !1356 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !1057 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !1058 %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !1059 br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !10, !llvm.loop !2060 61for.cond.cleanup:62 ret void, !dbg !1163}64 65; NO_REMARKS-NOT: remark: /tmp/t.c:9:3: loop not distributed: memory operations are safe for vectorization66; MISSED_REMARKS: remark: /tmp/t.c:9:3: loop not distributed: use -Rpass-analysis=loop-distribute for more info67; ANALYSIS_REMARKS: remark: /tmp/t.c:9:3: loop not distributed: memory operations are safe for vectorization68; ALWAYS-NOT: warning: /tmp/t.c:9:3: loop not distributed: failed explicitly specified loop distribution69 70define void @not_forced(ptr %A, ptr %B, ptr %C, i32 %N) !dbg !22 {71entry:72 %cmp12 = icmp sgt i32 %N, 0, !dbg !2373 br i1 %cmp12, label %ph, label %for.cond.cleanup, !dbg !2474 75ph:76 br label %for.body77 78for.body:79 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]80 %arrayidx = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !2681 %0 = load i8, ptr %arrayidx, align 1, !dbg !26, !tbaa !1382 %arrayidx2 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !2783 %1 = load i8, ptr %arrayidx2, align 1, !dbg !27, !tbaa !1384 %mul = mul i8 %1, %0, !dbg !2885 %arrayidx6 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !2986 store i8 %mul, ptr %arrayidx6, align 1, !dbg !30, !tbaa !1387 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !2488 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !2489 %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !2490 br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !2491 92for.cond.cleanup:93 ret void, !dbg !2594}95 96; REMARKS: remark: /tmp/t.c:15:3: distributed loop97 98define void @success(ptr %A, ptr %B, ptr %C, ptr %D, ptr %E, i32 %N) !dbg !31 {99entry:100 %cmp28 = icmp sgt i32 %N, 0, !dbg !32101 br i1 %cmp28, label %ph, label %for.cond.cleanup, !dbg !33102 103ph:104 br label %for.body105 106for.body:107 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]108 %arrayidx = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !35109 %0 = load i8, ptr %arrayidx, align 1, !dbg !35, !tbaa !13110 %arrayidx2 = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !36111 %1 = load i8, ptr %arrayidx2, align 1, !dbg !36, !tbaa !13112 %add = add i8 %1, %0, !dbg !37113 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !33114 %arrayidx7 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv.next, !dbg !38115 store i8 %add, ptr %arrayidx7, align 1, !dbg !39, !tbaa !13116 %arrayidx9 = getelementptr inbounds i8, ptr %D, i64 %indvars.iv, !dbg !40117 %2 = load i8, ptr %arrayidx9, align 1, !dbg !40, !tbaa !13118 %arrayidx12 = getelementptr inbounds i8, ptr %E, i64 %indvars.iv, !dbg !41119 %3 = load i8, ptr %arrayidx12, align 1, !dbg !41, !tbaa !13120 %mul = mul i8 %3, %2, !dbg !42121 %arrayidx16 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !43122 store i8 %mul, ptr %arrayidx16, align 1, !dbg !44, !tbaa !13123 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !33124 %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !33125 br i1 %exitcond, label %for.cond.cleanup, label %for.body, !dbg !33126 127for.cond.cleanup:128 ret void, !dbg !34129}130 131; MISSED_REMARKS: /tmp/t.c:27:5: loop not distributed: use -Rpass-analysis=loop-distribute for more info132; ANALYSIS_REMARKS: /tmp/t.c:27:5: loop not distributed: may not insert runtime check with convergent operation133; ALWAYS: warning: /tmp/t.c:27:5: loop not distributed: failed explicitly specified loop distribution134define void @convergent(ptr %A, ptr %B, ptr %C, ptr %D, ptr %E, i32 %N) #1 !dbg !45 {135entry:136 %cmp28 = icmp sgt i32 %N, 0, !dbg !46137 br i1 %cmp28, label %ph, label %for.cond.cleanup, !dbg !47138 139ph:140 br label %for.body141 142for.body:143 %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %ph ]144 %arrayidx = getelementptr inbounds i8, ptr %A, i64 %indvars.iv, !dbg !49145 %0 = load i8, ptr %arrayidx, align 1, !dbg !49, !tbaa !13146 %arrayidx2 = getelementptr inbounds i8, ptr %B, i64 %indvars.iv, !dbg !50147 %1 = load i8, ptr %arrayidx2, align 1, !dbg !50, !tbaa !13148 %add = add i8 %1, %0, !dbg !51149 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !57150 %arrayidx7 = getelementptr inbounds i8, ptr %A, i64 %indvars.iv.next, !dbg !52151 store i8 %add, ptr %arrayidx7, align 1, !dbg !53, !tbaa !13152 %arrayidx9 = getelementptr inbounds i8, ptr %D, i64 %indvars.iv, !dbg !54153 %2 = load i8, ptr %arrayidx9, align 1, !dbg !54, !tbaa !13154 %arrayidx12 = getelementptr inbounds i8, ptr %E, i64 %indvars.iv, !dbg !55155 %3 = load i8, ptr %arrayidx12, align 1, !dbg !55, !tbaa !13156 %mul = mul i8 %3, %2, !dbg !56157 %arrayidx16 = getelementptr inbounds i8, ptr %C, i64 %indvars.iv, !dbg !57158 store i8 %mul, ptr %arrayidx16, align 1, !dbg !58, !tbaa !13159 call void @llvm.convergent()160 %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !57161 %exitcond = icmp eq i32 %lftr.wideiv, %N, !dbg !57162 br i1 %exitcond, label %for.cond.cleanup, label %for.body, !llvm.loop !20, !dbg !57163 164for.cond.cleanup:165 ret void, !dbg !58166}167 168 169declare void @llvm.convergent() #0170 171attributes #0 = { nounwind readnone convergent }172attributes #1 = { nounwind convergent }173 174 175!llvm.dbg.cu = !{!0}176!llvm.module.flags = !{!3, !4}177 178!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.9.0 (trunk 267633) (llvm/trunk 267675)", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2)179!1 = !DIFile(filename: "/tmp/t.c", directory: "/tmp")180!2 = !{}181!3 = !{i32 2, !"Dwarf Version", i32 2}182!4 = !{i32 2, !"Debug Info Version", i32 3}183!7 = distinct !DISubprogram(name: "forced", scope: !1, file: !1, line: 1, type: !8, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)184!8 = !DISubroutineType(types: !2)185!9 = !DILocation(line: 3, column: 20, scope: !7)186!10 = !DILocation(line: 3, column: 3, scope: !7)187!11 = !DILocation(line: 6, column: 1, scope: !7)188!12 = !DILocation(line: 4, column: 12, scope: !7)189!13 = !{!14, !14, i64 0}190!14 = !{!"omnipotent char", !15, i64 0}191!15 = !{!"Simple C/C++ TBAA"}192!16 = !DILocation(line: 4, column: 19, scope: !7)193!17 = !DILocation(line: 4, column: 17, scope: !7)194!18 = !DILocation(line: 4, column: 5, scope: !7)195!19 = !DILocation(line: 4, column: 10, scope: !7)196!20 = distinct !{!20, !21}197!21 = !{!"llvm.loop.distribute.enable", i1 true}198!22 = distinct !DISubprogram(name: "not_forced", scope: !1, file: !1, line: 8, type: !8, isLocal: false, isDefinition: true, scopeLine: 8, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)199!23 = !DILocation(line: 9, column: 20, scope: !22)200!24 = !DILocation(line: 9, column: 3, scope: !22)201!25 = !DILocation(line: 12, column: 1, scope: !22)202!26 = !DILocation(line: 10, column: 12, scope: !22)203!27 = !DILocation(line: 10, column: 19, scope: !22)204!28 = !DILocation(line: 10, column: 17, scope: !22)205!29 = !DILocation(line: 10, column: 5, scope: !22)206!30 = !DILocation(line: 10, column: 10, scope: !22)207!31 = distinct !DISubprogram(name: "success", scope: !1, file: !1, line: 14, type: !8, isLocal: false, isDefinition: true, scopeLine: 14, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)208!32 = !DILocation(line: 15, column: 20, scope: !31)209!33 = !DILocation(line: 15, column: 3, scope: !31)210!34 = !DILocation(line: 19, column: 1, scope: !31)211!35 = !DILocation(line: 16, column: 16, scope: !31)212!36 = !DILocation(line: 16, column: 23, scope: !31)213!37 = !DILocation(line: 16, column: 21, scope: !31)214!38 = !DILocation(line: 16, column: 5, scope: !31)215!39 = !DILocation(line: 16, column: 14, scope: !31)216!40 = !DILocation(line: 17, column: 12, scope: !31)217!41 = !DILocation(line: 17, column: 19, scope: !31)218!42 = !DILocation(line: 17, column: 17, scope: !31)219!43 = !DILocation(line: 17, column: 5, scope: !31)220!44 = !DILocation(line: 17, column: 10, scope: !31)221!45 = distinct !DISubprogram(name: "convergent", scope: !1, file: !1, line: 24, type: !8, isLocal: false, isDefinition: true, scopeLine: 24, flags: DIFlagPrototyped, isOptimized: true, unit: !0, retainedNodes: !2)222!46 = !DILocation(line: 25, column: 20, scope: !45)223!47 = !DILocation(line: 25, column: 3, scope: !45)224!48 = !DILocation(line: 29, column: 1, scope: !45)225!49 = !DILocation(line: 26, column: 16, scope: !45)226!50 = !DILocation(line: 26, column: 23, scope: !45)227!51 = !DILocation(line: 26, column: 21, scope: !45)228!52 = !DILocation(line: 26, column: 5, scope: !45)229!53 = !DILocation(line: 26, column: 14, scope: !45)230!54 = !DILocation(line: 27, column: 12, scope: !45)231!55 = !DILocation(line: 27, column: 19, scope: !45)232!56 = !DILocation(line: 27, column: 17, scope: !45)233!57 = !DILocation(line: 27, column: 5, scope: !45)234!58 = !DILocation(line: 27, column: 10, scope: !45)235