From d58f67078860af1bc4dcecc64955c8f45b46d2a5 Mon Sep 17 00:00:00 2001
From: Florian Hahn <flo@fhahn.com>
Date: Wed, 21 Dec 2022 13:23:25 +0000
Subject: [PATCH] [LV] Add test for #59459.

---
 .../X86/epilog-vectorization-inductions.ll         | 131 +++++++++++++++++++++
 1 file changed, 131 insertions(+)
 create mode 100644 llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll

diff --git a/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll b/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
new file mode 100644
index 0000000..5aa30ea
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
@@ -0,0 +1,131 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -passes=loop-vectorize -mcpu=znver2 -S %s | FileCheck %s
+
+target triple = "x86_64-unknown-linux-gnu"
+
+define void @test_pr59459(i64 %iv.start, ptr %arr) {
+; CHECK-LABEL: @test_pr59459(
+; CHECK-NEXT:  iter.check:
+; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[IV_START:%.*]] to i32
+; CHECK-NEXT:    [[SMAX1:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP0]], i32 92)
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i32 [[SMAX1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP2:%.*]] = zext i32 [[TMP1]] to i64
+; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[TMP2]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP3]], 8
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
+; CHECK:       vector.scevcheck:
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[IV_START]] to i32
+; CHECK-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP4]], i32 92)
+; CHECK-NEXT:    [[TMP5:%.*]] = sub i32 [[SMAX]], [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add i32 [[TMP4]], -1
+; CHECK-NEXT:    [[TMP7:%.*]] = add i32 [[TMP6]], [[TMP5]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp ult i32 [[TMP7]], [[TMP6]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK:       vector.main.loop.iter.check:
+; CHECK-NEXT:    [[MIN_ITERS_CHECK2:%.*]] = icmp ult i64 [[TMP3]], 16
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK2]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP3]], 16
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP9:%.*]] = trunc i64 [[IV_START]] to i32
+; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <16 x i32> poison, i32 [[TMP9]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <16 x i32> [[DOTSPLATINSERT]], <16 x i32> poison, <16 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <16 x i32> [[DOTSPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15>
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <16 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[IV_START]], [[INDEX]]
+; CHECK-NEXT:    [[TMP10:%.*]] = trunc i64 [[OFFSET_IDX]] to i32
+; CHECK-NEXT:    [[TMP11:%.*]] = add i32 [[TMP10]], 0
+; CHECK-NEXT:    [[TMP12:%.*]] = add i32 [[TMP11]], -1
+; CHECK-NEXT:    [[TMP13:%.*]] = mul <16 x i32> [[VEC_IND]], <i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608>
+; CHECK-NEXT:    [[TMP14:%.*]] = lshr exact <16 x i32> [[TMP13]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; CHECK-NEXT:    [[TMP15:%.*]] = trunc <16 x i32> [[TMP14]] to <16 x i16>
+; CHECK-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP12]] to i64
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i16, ptr [[ARR:%.*]], i64 [[TMP16]]
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP17]], i32 0
+; CHECK-NEXT:    store <16 x i16> [[TMP15]], ptr [[TMP18]], align 2
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <16 x i32> [[VEC_IND]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK:       vec.epilog.iter.check:
+; CHECK-NEXT:    [[IND_END5:%.*]] = add i64 [[IV_START]], [[N_VEC]]
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[TMP3]], [[N_VEC]]
+; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8
+; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK:       vec.epilog.ph:
+; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[N_MOD_VF3:%.*]] = urem i64 [[TMP3]], 8
+; CHECK-NEXT:    [[N_VEC4:%.*]] = sub i64 [[TMP3]], [[N_MOD_VF3]]
+; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[IV_START]], [[N_VEC4]]
+; CHECK-NEXT:    [[TMP20:%.*]] = trunc i64 [[IV_START]] to i32
+; CHECK-NEXT:    [[DOTSPLATINSERT8:%.*]] = insertelement <8 x i32> poison, i32 [[TMP20]], i32 0
+; CHECK-NEXT:    [[DOTSPLAT9:%.*]] = shufflevector <8 x i32> [[DOTSPLATINSERT8]], <8 x i32> poison, <8 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION10:%.*]] = add <8 x i32> [[DOTSPLAT9]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
+; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK:       vec.epilog.vector.body:
+; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND11:%.*]] = phi <8 x i32> [ [[INDUCTION10]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT12:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX13:%.*]] = add i64 [[IV_START]], [[INDEX7]]
+; CHECK-NEXT:    [[TMP21:%.*]] = trunc i64 [[OFFSET_IDX13]] to i32
+; CHECK-NEXT:    [[TMP22:%.*]] = add i32 [[TMP21]], 0
+; CHECK-NEXT:    [[TMP23:%.*]] = add i32 [[TMP22]], -1
+; CHECK-NEXT:    [[TMP24:%.*]] = mul <8 x i32> [[VEC_IND11]], <i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608, i32 196608>
+; CHECK-NEXT:    [[TMP25:%.*]] = lshr exact <8 x i32> [[TMP24]], <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; CHECK-NEXT:    [[TMP26:%.*]] = trunc <8 x i32> [[TMP25]] to <8 x i16>
+; CHECK-NEXT:    [[TMP27:%.*]] = zext i32 [[TMP23]] to i64
+; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i16, ptr [[ARR]], i64 [[TMP27]]
+; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i16, ptr [[TMP28]], i32 0
+; CHECK-NEXT:    store <8 x i16> [[TMP26]], ptr [[TMP29]], align 2
+; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX7]], 8
+; CHECK-NEXT:    [[VEC_IND_NEXT12]] = add <8 x i32> [[VEC_IND11]], <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC4]]
+; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[CMP_N6:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC4]]
+; CHECK-NEXT:    br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK:       vec.epilog.scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END5]], [[VEC_EPILOG_ITER_CHECK]] ], [ [[IV_START]], [[VECTOR_SCEVCHECK]] ], [ [[IV_START]], [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP:%.*]]
+; CHECK:       loop:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV_TRUNC:%.*]] = trunc i64 [[IV]] to i32
+; CHECK-NEXT:    [[STORE_IDX:%.*]] = add i32 [[IV_TRUNC]], -1
+; CHECK-NEXT:    [[X:%.*]] = mul i32 [[IV_TRUNC]], 196608
+; CHECK-NEXT:    [[Y:%.*]] = lshr exact i32 [[X]], 16
+; CHECK-NEXT:    [[STORE_VAL:%.*]] = trunc i32 [[Y]] to i16
+; CHECK-NEXT:    [[STORE_IDX_WIDE:%.*]] = zext i32 [[STORE_IDX]] to i64
+; CHECK-NEXT:    [[ADDR:%.*]] = getelementptr i16, ptr [[ARR]], i64 [[STORE_IDX_WIDE]]
+; CHECK-NEXT:    store i16 [[STORE_VAL]], ptr [[ADDR]], align 2
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; CHECK-NEXT:    [[LOOP_COND:%.*]] = icmp sgt i32 [[IV_TRUNC]], 91
+; CHECK-NEXT:    br i1 [[LOOP_COND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop
+
+loop:
+  %iv = phi i64 [ %iv.start, %entry ], [ %iv.next, %loop ]
+  %iv.trunc = trunc i64 %iv to i32
+  %store.idx = add i32 %iv.trunc, -1
+  %x = mul i32 %iv.trunc, 196608
+  %y = lshr exact i32 %x, 16
+  %store.val = trunc i32 %y to i16
+  %store.idx.wide = zext i32 %store.idx to i64
+  %addr = getelementptr i16, ptr %arr, i64 %store.idx.wide
+  store i16 %store.val, ptr %addr
+  %iv.next = add nuw nsw i64 %iv, 1
+  %loop.cond = icmp sgt i32 %iv.trunc, 91
+  br i1 %loop.cond, label %exit, label %loop
+
+exit:
+  ret void
+}
+
-- 
2.7.4