1; RUN: opt -S -passes=loop-unroll --debug-only=loop-unroll < %s 2>&1 | FileCheck %s -check-prefix=LOOP-UNROLL 2; RUN: opt -S -passes='require<opt-remark-emit>,loop(loop-unroll-full)' --debug-only=loop-unroll < %s 2>&1 | FileCheck %s -check-prefix=LOOP-UNROLL-FULL 3 4; REQUIRES: asserts 5 6%struct.HIP_vector_type = type { %union.anon } 7%union.anon = type { <2 x float> } 8 9 10; LOOP-UNROLL-LABEL: Loop Unroll: F[pragma_unroll] Loop %for.body 11; LOOP-UNROLL-NEXT: Loop Size = 9 12; LOOP-UNROLL-NEXT: runtime unrolling with count: 8 13; LOOP-UNROLL-NEXT: Exiting block %for.body: TripCount=0, TripMultiple=1, BreakoutTrip=1 14; LOOP-UNROLL-NEXT: Trying runtime unrolling on Loop: 15; LOOP-UNROLL-NEXT: Loop at depth 1 containing: %for.body<header><latch><exiting> 16; LOOP-UNROLL-NEXT: Using epilog remainder. 17; LOOP-UNROLL-NEXT: UNROLLING loop %for.body by 8 with run-time trip count! 18 19; LOOP-UNROLL-FULL-LABEL: Loop Unroll: F[pragma_unroll] Loop %for.body 20; LOOP-UNROLL-FULL-NEXT: Loop Size = 9 21; LOOP-UNROLL-FULL-NEXT: runtime unrolling with count: 8 22; LOOP-UNROLL-FULL-NEXT: Not attempting partial/runtime unroll in FullLoopUnroll 23define void @pragma_unroll(ptr %queue, i32 %num_elements) { 24entry: 25 %cmp5 = icmp sgt i32 %num_elements, 0 26 br i1 %cmp5, label %for.body.preheader, label %for.cond.cleanup 27 28for.body.preheader: ; preds = %entry 29 br label %for.body 30 31for.cond.cleanup.loopexit: ; preds = %for.body 32 br label %for.cond.cleanup 33 34for.cond.cleanup: ; preds = %for.cond.cleanup.loopexit, %entry 35 ret void 36 37for.body: ; preds = %for.body.preheader, %for.body 38 %i.06 = phi i32 [ %add, %for.body ], [ 0, %for.body.preheader ] 39 %add = add nuw nsw i32 %i.06, 1 40 %idxprom = zext i32 %add to i64 41 %arrayidx = getelementptr inbounds %struct.HIP_vector_type, ptr %queue, i64 %idxprom 42 %idxprom1 = zext i32 %i.06 to i64 43 %arrayidx2 = getelementptr inbounds %struct.HIP_vector_type, ptr %queue, i64 %idxprom1 44 %0 = load i64, ptr %arrayidx, align 8 45 store i64 %0, ptr %arrayidx2, align 8 46 %exitcond = icmp ne i32 %add, %num_elements 47 br i1 %exitcond, label %for.body, label %for.cond.cleanup.loopexit, !llvm.loop !1 48} 49 50; LOOP-UNROLL-LABEL: Loop Unroll: F[pragma_unroll_count1] Loop %for.body 51; LOOP-UNROLL-NEXT: Loop Size = 9 52; LOOP-UNROLL-NEXT: Exiting block %for.body: TripCount=0, TripMultiple=1, BreakoutTrip=1 53; LOOP-UNROLL-NEXT: Trying runtime unrolling on Loop: 54; LOOP-UNROLL-NEXT: Loop at depth 1 containing: %for.body<header><latch><exiting> 55; LOOP-UNROLL-NEXT: Using epilog remainder. 56; LOOP-UNROLL-NEXT: UNROLLING loop %for.body by 5 with run-time trip count! 57 58; LOOP-UNROLL-FULL-LABEL: Loop Unroll: F[pragma_unroll_count1] Loop %for.body 59; LOOP-UNROLL-FULL-NEXT: Loop Size = 9 60; LOOP-UNROLL-FULL-NEXT: Not attempting partial/runtime unroll in FullLoopUnroll 61define void @pragma_unroll_count1(ptr %queue, i32 %num_elements) { 62entry: 63 %cmp5 = icmp sgt i32 %num_elements, 0 64 br i1 %cmp5, label %for.body.preheader, label %for.cond.cleanup 65 66for.body.preheader: ; preds = %entry 67 br label %for.body 68 69for.cond.cleanup.loopexit: ; preds = %for.body 70 br label %for.cond.cleanup 71 72for.cond.cleanup: ; preds = %for.cond.cleanup.loopexit, %entry 73 ret void 74 75for.body: ; preds = %for.body.preheader, %for.body 76 %i.06 = phi i32 [ %add, %for.body ], [ 0, %for.body.preheader ] 77 %add = add nuw nsw i32 %i.06, 1 78 %idxprom = zext i32 %add to i64 79 %arrayidx = getelementptr inbounds %struct.HIP_vector_type, ptr %queue, i64 %idxprom 80 %idxprom1 = zext i32 %i.06 to i64 81 %arrayidx2 = getelementptr inbounds %struct.HIP_vector_type, ptr %queue, i64 %idxprom1 82 %0 = load i64, ptr %arrayidx, align 8 83 store i64 %0, ptr %arrayidx2, align 8 84 %exitcond = icmp ne i32 %add, %num_elements 85 br i1 %exitcond, label %for.body, label %for.cond.cleanup.loopexit, !llvm.loop !3 86} 87 88; LOOP-UNROLL: llvm.loop.unroll.disable 89; LOOP-UNROLL-FULL: llvm.loop.unroll.enable 90!0 = !{!"llvm.loop.unroll.enable"} 91!1 = distinct !{!1, !0} 92 93!2 = !{!"llvm.loop.unroll.count", i32 5} 94!3 = distinct !{!3, !2} 95