110 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 dependences are computed correctly.5# The original code is as follows.6#7# ```8# void f(int *a, int n) {9# for (int i = 0; i < n-2; i++) {10# a[i] += a[i+10];11# a[i+2] += i;12# }13# }14# ```15#16# Here is what each instruction does.17# SU(2): Load a[i+10]18# SU(3): Store it to a[i]19# SU(4): Load a[i+2], add i, then store it20#21# FIXME: Currently the following dependencies are missed.22#23 24# CHECK: ===== Loop Carried Edges Begin =====25# CHECK-NEXT: Loop carried edges from SU(2)26# CHECK-NEXT: Order27# CHECK-NEXT: SU(3)28# CHECK-NEXT: SU(4)29# CHECK-NEXT: Loop carried edges from SU(4)30# CHECK-NEXT: Order31# CHECK-NEXT: SU(3)32# CHECK-NEXT: ===== Loop Carried Edges End =====33 34--- |35 define dso_local void @f(ptr nocapture noundef %a, i32 noundef %n) {36 entry:37 %cmp13 = icmp sgt i32 %n, 238 br i1 %cmp13, label %for.body.preheader, label %for.cond.cleanup39 40 for.body.preheader:41 %0 = add i32 %n, -242 br label %for.body43 44 for.cond.cleanup:45 ret void46 47 for.body:48 %lsr.iv15 = phi ptr [ %a, %for.body.preheader ], [ %cgep19, %for.body ]49 %lsr.iv = phi i32 [ %0, %for.body.preheader ], [ %lsr.iv.next, %for.body ]50 %i.014 = phi i32 [ %inc, %for.body ], [ 0, %for.body.preheader ]51 %cgep = getelementptr i8, ptr %lsr.iv15, i32 4052 %1 = load i32, ptr %cgep, align 4, !tbaa !553 %2 = load i32, ptr %lsr.iv15, align 4, !tbaa !554 %add2 = add nsw i32 %2, %155 store i32 %add2, ptr %lsr.iv15, align 4, !tbaa !556 %cgep18 = getelementptr i8, ptr %lsr.iv15, i32 857 %3 = load i32, ptr %cgep18, align 4, !tbaa !558 %4 = add i32 %i.014, %359 store i32 %4, ptr %cgep18, align 4, !tbaa !560 %inc = add nuw nsw i32 %i.014, 161 %lsr.iv.next = add i32 %lsr.iv, -162 %exitcond.not = icmp eq i32 %lsr.iv.next, 063 %cgep19 = getelementptr i8, ptr %lsr.iv15, i32 464 br i1 %exitcond.not, label %for.cond.cleanup, label %for.body65 }66 67 !5 = !{!6, !6, i64 0}68 !6 = !{!"int", !7, i64 0}69 !7 = !{!"omnipotent char", !8, i64 0}70 !8 = !{!"Simple C/C++ TBAA"}71 72...73---74name: f75tracksRegLiveness: true76body: |77 bb.0.entry:78 successors: %bb.1, %bb.279 liveins: $r0, $r180 81 %8:intregs = COPY $r182 %7:intregs = COPY $r083 %9:predregs = C2_cmpgti %8, 284 J2_jumpf %9, %bb.2, implicit-def dead $pc85 J2_jump %bb.1, implicit-def dead $pc86 87 bb.1.for.body.preheader:88 %0:intregs = A2_addi %8, -289 %11:intregs = A2_tfrsi 090 %14:intregs = COPY %091 J2_loop0r %bb.3, %14, implicit-def $lc0, implicit-def $sa0, implicit-def $usr92 J2_jump %bb.3, implicit-def dead $pc93 94 bb.2.for.cond.cleanup:95 PS_jmpret $r31, implicit-def dead $pc96 97 bb.3.for.body:98 successors: %bb.2, %bb.399 100 %1:intregs = PHI %7, %bb.1, %6, %bb.3101 %3:intregs = PHI %11, %bb.1, %4, %bb.3102 %12:intregs = L2_loadri_io %1, 40 :: (load (s32) from %ir.cgep, !tbaa !5)103 L4_add_memopw_io %1, 0, killed %12 :: (store (s32) into %ir.lsr.iv15, !tbaa !5), (load (s32) from %ir.lsr.iv15, !tbaa !5)104 L4_add_memopw_io %1, 8, %3 :: (store (s32) into %ir.cgep18, !tbaa !5), (load (s32) from %ir.cgep18, !tbaa !5)105 %4:intregs = nuw nsw A2_addi %3, 1106 %6:intregs = A2_addi %1, 4107 ENDLOOP0 %bb.3, implicit-def $pc, implicit-def $lc0, implicit $sa0, implicit $lc0108 J2_jump %bb.2, implicit-def $pc109...110