270 lines · plain
1; REQUIRES: asserts2; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=false < %s | FileCheck %s --check-prefixes=V0,V013; RUN: llvm-profdata merge --sample -profile-isfs -o %t0.afdo %S/Inputs/fsloader.afdo4; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=false -fs-profile-file=%t0.afdo -show-fs-branchprob -disable-ra-fsprofile-loader=false -disable-layout-fsprofile-loader=false < %s 2>&1 | FileCheck %s --check-prefixes=LOADERV0,LOADER5; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=true < %s | FileCheck %s --check-prefixes=V1,V016; RUN: llvm-profdata merge --sample -profile-isfs -o %t1.afdo %S/Inputs/fsloader_v1.afdo7; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=true -fs-profile-file=%t1.afdo -show-fs-branchprob -disable-ra-fsprofile-loader=false -disable-layout-fsprofile-loader=false < %s 2>&1 | FileCheck %s --check-prefixes=LOADERV1,LOADER8; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=true -fs-profile-file=%S/Inputs/fsloader_v1.afdo -profile-isfs -show-fs-branchprob -disable-ra-fsprofile-loader=false -disable-layout-fsprofile-loader=false < %s 2>&1 | FileCheck %s --check-prefixes=LOADERV1,LOADER9; RUN: llc -enable-fs-discriminator -improved-fs-discriminator=true -fs-profile-file=%S/Inputs/fsloader_v1.afdo -show-fs-branchprob -disable-ra-fsprofile-loader=false -disable-layout-fsprofile-loader=false < %s 2>&1 | FileCheck %s --check-prefixes=NOLOAD10;;11;; C source code for the test (compiler at -O3):12;; // A test case for loop unroll.13;;14;; __attribute__((noinline)) int bar(int i){15;; volatile int j;16;; j = i;17;; return j;18;; }19;;20;; unsigned sum;21;; __attribute__((noinline)) void work(int i){22;; if (sum % 7)23;; sum += i;24;; else25;; sum -= i;26;; }27;;28;; __attribute__((noinline)) void foo(){29;; int i, j;30;; for (j = 0; j < 48; j++)31;; for (i = 0; i < 4; i++) {32;; int ii = bar(i+j*48);33;; if (ii % 2)34;; work(ii*2);35;; if (ii % 4)36;; work(ii*3);37;; }38;; }39;;40;; int main() {41;; int i;42;; for (i = 0; i < 10000000; i++) {43;; foo();44;; }45;; }46;;47;; Check that fs-afdo discriminators are generated.48; V01: .loc 1 23 9 is_stmt 0 discriminator 1 # unroll.c:23:949; V0: .loc 1 23 9 is_stmt 0 discriminator 3585 # unroll.c:23:950; V0: .loc 1 23 9 is_stmt 0 discriminator 8705 # unroll.c:23:951; V0: .loc 1 23 9 is_stmt 0 discriminator 4097 # unroll.c:23:952; V1: .loc 1 23 9 is_stmt 0 discriminator 257 # unroll.c:23:953; V1: .loc 1 23 9 is_stmt 0 discriminator 513 # unroll.c:23:954; V1: .loc 1 23 9 is_stmt 0 discriminator 769 # unroll.c:23:955;;56;; Check that variable __llvm_fs_discriminator__ is generated.57; V01: .type __llvm_fs_discriminator__,@object # @__llvm_fs_discriminator__58; V01: .section .rodata,"a",@progbits59; V01: .weak __llvm_fs_discriminator__60; V01: __llvm_fs_discriminator__:61; V01: .byte 162; V01: .size __llvm_fs_discriminator__, 163 64;; Check that new branch probs are generated.65; LOADER: Set branch fs prob: MBB (1 -> 3): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x7aca7894 / 0x80000000 = 95.93%66; LOADER: Set branch fs prob: MBB (1 -> 2): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x0535876c / 0x80000000 = 4.07%67; LOADER: Set branch fs prob: MBB (3 -> 5): unroll.c:24:11-->unroll.c:22:11 W=283590 0x30000000 / 0x80000000 = 37.50% --> 0x7aca7894 / 0x80000000 = 95.93%68; LOADER: Set branch fs prob: MBB (3 -> 4): unroll.c:24:11 W=283590 0x50000000 / 0x80000000 = 62.50% --> 0x0535876c / 0x80000000 = 4.07%69; LOADER: Set branch fs prob: MBB (5 -> 8): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x021c112e / 0x80000000 = 1.65%70; LOADER: Set branch fs prob: MBB (5 -> 7): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x7de3eed2 / 0x80000000 = 98.35%71; LOADER: Set branch fs prob: MBB (8 -> 10): unroll.c:24:11-->unroll.c:22:11 W=283590 0x30000000 / 0x80000000 = 37.50% --> 0x00000000 / 0x80000000 = 0.00%72; LOADER: Set branch fs prob: MBB (8 -> 9): unroll.c:24:11 W=283590 0x50000000 / 0x80000000 = 62.50% --> 0x80000000 / 0x80000000 = 100.00%73; LOADERV0: Set branch fs prob: MBB (10 -> 12): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x7aca7894 / 0x80000000 = 95.93%74; LOADERV1: Set branch fs prob: MBB (10 -> 12): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x0a5856e1 / 0x80000000 = 8.08%75; LOADERV0: Set branch fs prob: MBB (10 -> 11): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x0535876c / 0x80000000 = 4.07%76; LOADERV1: Set branch fs prob: MBB (10 -> 11): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x75a7a91f / 0x80000000 = 91.92%77; LOADER: Set branch fs prob: MBB (12 -> 14): unroll.c:24:11-->unroll.c:22:11 W=283590 0x30000000 / 0x80000000 = 37.50% --> 0x02012507 / 0x80000000 = 1.57%78; LOADER: Set branch fs prob: MBB (12 -> 13): unroll.c:24:11 W=283590 0x50000000 / 0x80000000 = 62.50% --> 0x7dfedaf9 / 0x80000000 = 98.43%79; LOADERV0: Set branch fs prob: MBB (14 -> 16): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x0a5856e1 / 0x80000000 = 8.08%80; LOADERV1: Set branch fs prob: MBB (14 -> 16): unroll.c:22:11-->unroll.c:24:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x7aca7894 / 0x80000000 = 95.93%81; LOADERV0: Set branch fs prob: MBB (14 -> 15): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x75a7a91f / 0x80000000 = 91.92%82; LOADERV1: Set branch fs prob: MBB (14 -> 15): unroll.c:22:11 W=283590 0x40000000 / 0x80000000 = 50.00% --> 0x0535876c / 0x80000000 = 4.07%83; LOADER: Set branch fs prob: MBB (16 -> 18): unroll.c:24:11-->unroll.c:19:3 W=283590 0x30000000 / 0x80000000 = 37.50% --> 0x16588166 / 0x80000000 = 17.46%84; LOADER: Set branch fs prob: MBB (16 -> 17): unroll.c:24:11 W=283590 0x50000000 / 0x80000000 = 62.50% --> 0x69a77e9a / 0x80000000 = 82.54%85 86;; Check that the profile is not loaded since the reader doesn't know it is a FS profile.87; NOLOAD-NOT: Set branch fs prob88 89target triple = "x86_64-unknown-linux-gnu"90 91@sum = dso_local local_unnamed_addr global i32 0, align 492 93declare i32 @bar(i32 %i) #094declare void @work(i32 %i) #295 96define dso_local void @foo() #0 !dbg !29 {97entry:98 br label %for.cond1.preheader, !dbg !3099 100for.cond1.preheader:101 %j.012 = phi i32 [ 0, %entry ], [ %inc11, %if.end9.3 ]102 %mul = mul nuw nsw i32 %j.012, 48103 %call = tail call i32 @bar(i32 %mul), !dbg !32104 %0 = and i32 %call, 1, !dbg !33105 %tobool.not = icmp eq i32 %0, 0, !dbg !33106 br i1 %tobool.not, label %if.end, label %if.then, !dbg !35107 108if.then:109 %mul4 = shl nsw i32 %call, 1, !dbg !36110 tail call void @work(i32 %mul4), !dbg !37111 br label %if.end, !dbg !38112 113if.end:114 %1 = and i32 %call, 3, !dbg !39115 %tobool6.not = icmp eq i32 %1, 0, !dbg !39116 br i1 %tobool6.not, label %if.end9, label %if.then7, !dbg !40117 118if.then7:119 %mul8 = mul nsw i32 %call, 3, !dbg !41120 tail call void @work(i32 %mul8), !dbg !42121 br label %if.end9, !dbg !43122 123if.end9:124 %add.1 = or i32 %mul, 1, !dbg !44125 %call.1 = tail call i32 @bar(i32 %add.1), !dbg !32126 %2 = and i32 %call.1, 1, !dbg !33127 %tobool.not.1 = icmp eq i32 %2, 0, !dbg !33128 br i1 %tobool.not.1, label %if.end.1, label %if.then.1, !dbg !35129 130for.end12:131 ret void, !dbg !45132 133if.then.1:134 %mul4.1 = shl nsw i32 %call.1, 1, !dbg !36135 tail call void @work(i32 %mul4.1), !dbg !37136 br label %if.end.1, !dbg !38137 138if.end.1:139 %3 = and i32 %call.1, 3, !dbg !39140 %tobool6.not.1 = icmp eq i32 %3, 0, !dbg !39141 br i1 %tobool6.not.1, label %if.end9.1, label %if.then7.1, !dbg !40142 143if.then7.1:144 %mul8.1 = mul nsw i32 %call.1, 3, !dbg !41145 tail call void @work(i32 %mul8.1), !dbg !42146 br label %if.end9.1, !dbg !43147 148if.end9.1:149 %add.2 = or i32 %mul, 2, !dbg !44150 %call.2 = tail call i32 @bar(i32 %add.2), !dbg !32151 %4 = and i32 %call.2, 1, !dbg !33152 %tobool.not.2 = icmp eq i32 %4, 0, !dbg !33153 br i1 %tobool.not.2, label %if.end.2, label %if.then.2, !dbg !35154 155if.then.2:156 %mul4.2 = shl nsw i32 %call.2, 1, !dbg !36157 tail call void @work(i32 %mul4.2), !dbg !37158 br label %if.end.2, !dbg !38159 160if.end.2:161 %5 = and i32 %call.2, 3, !dbg !39162 %tobool6.not.2 = icmp eq i32 %5, 0, !dbg !39163 br i1 %tobool6.not.2, label %if.end9.2, label %if.then7.2, !dbg !40164 165if.then7.2:166 %mul8.2 = mul nsw i32 %call.2, 3, !dbg !41167 tail call void @work(i32 %mul8.2), !dbg !42168 br label %if.end9.2, !dbg !43169 170if.end9.2:171 %add.3 = or i32 %mul, 3, !dbg !44172 %call.3 = tail call i32 @bar(i32 %add.3), !dbg !32173 %6 = and i32 %call.3, 1, !dbg !33174 %tobool.not.3 = icmp eq i32 %6, 0, !dbg !33175 br i1 %tobool.not.3, label %if.end.3, label %if.then.3, !dbg !35176 177if.then.3:178 %mul4.3 = shl nsw i32 %call.3, 1, !dbg !36179 tail call void @work(i32 %mul4.3), !dbg !37180 br label %if.end.3, !dbg !38181 182if.end.3:183 %7 = and i32 %call.3, 3, !dbg !39184 %tobool6.not.3 = icmp eq i32 %7, 0, !dbg !39185 br i1 %tobool6.not.3, label %if.end9.3, label %if.then7.3, !dbg !40186 187if.then7.3:188 %mul8.3 = mul nsw i32 %call.3, 3, !dbg !41189 tail call void @work(i32 %mul8.3), !dbg !42190 br label %if.end9.3, !dbg !43191 192if.end9.3:193 %inc11 = add nuw nsw i32 %j.012, 1, !dbg !46194 %exitcond.not = icmp eq i32 %inc11, 48, !dbg !48195 br i1 %exitcond.not, label %for.end12, label %for.cond1.preheader, !dbg !30, !llvm.loop !49196}197 198 199attributes #0 = { noinline nounwind uwtable "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-soft-float"="false" }200attributes #1 = { argmemonly nounwind willreturn }201attributes #2 = { nofree noinline norecurse nounwind uwtable "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-soft-float"="false" }202attributes #3 = { nounwind uwtable "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "use-soft-float"="false" }203 204!llvm.dbg.cu = !{!0}205!llvm.module.flags = !{!3, !4, !5}206 207!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2, splitDebugInlining: false, debugInfoForProfiling: true, nameTableKind: None)208!1 = !DIFile(filename: "unroll.c", directory: "a/")209!2 = !{}210!3 = !{i32 7, !"Dwarf Version", i32 4}211!4 = !{i32 2, !"Debug Info Version", i32 3}212!5 = !{i32 1, !"wchar_size", i32 4}213!7 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 3, type: !8, scopeLine: 3, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)214!8 = !DISubroutineType(types: !2)215!9 = !DILocation(line: 4, column: 3, scope: !7)216!10 = !DILocation(line: 5, column: 5, scope: !7)217!11 = !{!12, !12, i64 0}218!12 = !{!"int", !13, i64 0}219!13 = !{!"omnipotent char", !14, i64 0}220!14 = !{!"Simple C/C++ TBAA"}221!15 = !DILocation(line: 6, column: 10, scope: !7)222!16 = !DILocation(line: 7, column: 1, scope: !7)223!17 = !DILocation(line: 6, column: 3, scope: !18)224!18 = !DILexicalBlockFile(scope: !7, file: !1, discriminator: 1)225!19 = distinct !DISubprogram(name: "work", scope: !1, file: !1, line: 10, type: !8, scopeLine: 10, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)226!20 = !DILocation(line: 11, column: 7, scope: !19)227!21 = !DILocation(line: 11, column: 11, scope: !22)228!22 = !DILexicalBlockFile(scope: !19, file: !1, discriminator: 1)229!23 = !DILocation(line: 11, column: 11, scope: !24)230!24 = !DILexicalBlockFile(scope: !19, file: !1, discriminator: 2)231!25 = !DILocation(line: 11, column: 7, scope: !26)232!26 = !DILexicalBlockFile(scope: !19, file: !1, discriminator: 3)233!27 = !DILocation(line: 0, scope: !22)234!28 = !DILocation(line: 15, column: 1, scope: !19)235!29 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 17, type: !8, scopeLine: 17, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)236!30 = !DILocation(line: 19, column: 3, scope: !31)237!31 = !DILexicalBlockFile(scope: !29, file: !1, discriminator: 2)238!32 = !DILocation(line: 21, column: 16, scope: !31)239!33 = !DILocation(line: 22, column: 14, scope: !34)240!34 = !DILexicalBlockFile(scope: !29, file: !1, discriminator: 1)241!35 = !DILocation(line: 22, column: 11, scope: !31)242!36 = !DILocation(line: 23, column: 16, scope: !29)243!37 = !DILocation(line: 23, column: 9, scope: !34)244!38 = !DILocation(line: 23, column: 9, scope: !31)245!39 = !DILocation(line: 24, column: 14, scope: !34)246!40 = !DILocation(line: 24, column: 11, scope: !31)247!41 = !DILocation(line: 25, column: 16, scope: !29)248!42 = !DILocation(line: 25, column: 9, scope: !34)249!43 = !DILocation(line: 25, column: 9, scope: !31)250!44 = !DILocation(line: 21, column: 21, scope: !34)251!45 = !DILocation(line: 27, column: 1, scope: !29)252!46 = !DILocation(line: 19, column: 24, scope: !47)253!47 = !DILexicalBlockFile(scope: !29, file: !1, discriminator: 3)254!48 = !DILocation(line: 19, column: 17, scope: !34)255!49 = distinct !{!49, !50, !51}256!50 = !DILocation(line: 19, column: 3, scope: !29)257!51 = !DILocation(line: 26, column: 3, scope: !29)258!52 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 29, type: !8, scopeLine: 29, flags: DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)259!53 = !DILocation(line: 31, column: 3, scope: !54)260!54 = !DILexicalBlockFile(scope: !52, file: !1, discriminator: 2)261!55 = !DILocation(line: 32, column: 5, scope: !52)262!56 = !DILocation(line: 31, column: 30, scope: !57)263!57 = !DILexicalBlockFile(scope: !52, file: !1, discriminator: 3)264!58 = !DILocation(line: 31, column: 17, scope: !59)265!59 = !DILexicalBlockFile(scope: !52, file: !1, discriminator: 1)266!60 = distinct !{!60, !61, !62}267!61 = !DILocation(line: 31, column: 3, scope: !52)268!62 = !DILocation(line: 33, column: 3, scope: !52)269!63 = !DILocation(line: 34, column: 1, scope: !52)270