109 lines · plain
1# RUN: llc -mtriple=hexagon -run-pass pipeliner -debug-only=pipeliner %s -o /dev/null 2>&1 -pipeliner-experimental-cg=true | FileCheck %s2# REQUIRES: asserts3 4# Test that loop-carried memory dependencies are added correctly.5# The original code is as follows.6#7# ```8# void f(int *a, int n) {9# for (int i = 0; i < n-1; i++) {10# a[i] += a[i];11# a[i+1] += i;12# }13# }14# ```15# 16# Loop-carried dependencies exist from store for a[i+1] to load/store for a[i], but not vice versa.17 18# CHECK: ===== Loop Carried Edges Begin =====19# CHECK-NEXT: Loop carried edges from SU(6)20# CHECK-NEXT: Order21# CHECK-NEXT: SU(4)22# CHECK-NEXT: Loop carried edges from SU(8)23# CHECK-NEXT: Order24# CHECK-NEXT: SU(4)25# CHECK-NEXT: ===== Loop Carried Edges End =====26 27--- |28 define dso_local void @f(ptr nocapture noundef %a, i32 noundef %n) local_unnamed_addr {29 entry:30 %cmp12 = icmp sgt i32 %n, 131 br i1 %cmp12, label %for.body.preheader, label %for.cond.cleanup32 33 for.body.preheader:34 %.pre = load i32, ptr %a, align 4, !tbaa !535 %0 = add i32 %n, -136 %cgep = getelementptr i8, ptr %a, i32 437 br label %for.body38 39 for.cond.cleanup:40 ret void41 42 for.body:43 %lsr.iv14 = phi ptr [ %cgep, %for.body.preheader ], [ %cgep18, %for.body ]44 %lsr.iv = phi i32 [ %0, %for.body.preheader ], [ %lsr.iv.next, %for.body ]45 %1 = phi i32 [ %add4, %for.body ], [ %.pre, %for.body.preheader ]46 %i.013 = phi i32 [ %add2, %for.body ], [ 0, %for.body.preheader ]47 %add = shl nsw i32 %1, 148 %cgep17 = getelementptr i8, ptr %lsr.iv14, i32 -449 store i32 %add, ptr %cgep17, align 4, !tbaa !550 %add2 = add nuw nsw i32 %i.013, 151 %2 = load i32, ptr %lsr.iv14, align 4, !tbaa !552 %add4 = add nsw i32 %2, %i.01353 %3 = add i32 %i.013, %254 store i32 %3, ptr %lsr.iv14, align 4, !tbaa !555 %lsr.iv.next = add i32 %lsr.iv, -156 %exitcond.not = icmp eq i32 %lsr.iv.next, 057 %cgep18 = getelementptr i8, ptr %lsr.iv14, i32 458 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body59 }60 61 !5 = !{!6, !6, i64 0}62 !6 = !{!"int", !7, i64 0}63 !7 = !{!"omnipotent char", !8, i64 0}64 !8 = !{!"Simple C/C++ TBAA"}65 66...67---68name: f69tracksRegLiveness: true70machineFunctionInfo: {}71body: |72 bb.0.entry:73 successors: %bb.1, %bb.274 liveins: $r0, $r175 76 %12:intregs = COPY $r177 %11:intregs = COPY $r078 %13:predregs = C2_cmpgti %12, 179 J2_jumpf %13, %bb.2, implicit-def dead $pc80 J2_jump %bb.1, implicit-def dead $pc81 82 bb.1.for.body.preheader:83 %0:intregs, %2:intregs = L2_loadri_pi %11, 4 :: (load (s32) from %ir.a, !tbaa !5)84 %1:intregs = A2_addi %12, -185 %15:intregs = A2_tfrsi 086 %19:intregs = COPY %187 J2_loop0r %bb.3, %19, implicit-def $lc0, implicit-def $sa0, implicit-def $usr88 J2_jump %bb.3, implicit-def dead $pc89 90 bb.2.for.cond.cleanup:91 PS_jmpret $r31, implicit-def dead $pc92 93 bb.3.for.body:94 successors: %bb.2, %bb.395 96 %3:intregs = PHI %2, %bb.1, %10, %bb.397 %5:intregs = PHI %0, %bb.1, %8, %bb.398 %6:intregs = PHI %15, %bb.1, %7, %bb.399 %16:intregs = nsw S2_asl_i_r %5, 1100 S2_storeri_io %3, -4, killed %16 :: (store (s32) into %ir.cgep17, !tbaa !5)101 %7:intregs = nuw nsw A2_addi %6, 1102 %17:intregs = L2_loadri_io %3, 0 :: (load (s32) from %ir.lsr.iv14, !tbaa !5)103 %8:intregs = A2_add killed %17, %6104 S2_storeri_io %3, 0, %8 :: (store (s32) into %ir.lsr.iv14, !tbaa !5)105 %10:intregs = A2_addi %3, 4106 ENDLOOP0 %bb.3, implicit-def $pc, implicit-def $lc0, implicit $sa0, implicit $lc0107 J2_jump %bb.2, implicit-def $pc108...109