llvm/test/Transforms/LoopVectorize/RISCV/strided-access-wide-stride.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
; RUN: opt < %s -passes=loop-vectorize -mtriple=riscv64 -mattr=+v,+zvl256b -S | FileCheck %s

; Test for https://github.com/llvm/llvm-project/issues/199640
; convertToStridedAccesses uses the canonical IV type (i32) for the stride
; constant, but the pointer stride exceeds INT32_MAX on rv64 with a narrow IV.

; TODO: Should form a strided load with SCEV predicates.
define void @stride_sext_exceeds_i32_max(ptr noalias readonly %src, ptr noalias %dst, i16 %start) {
; CHECK-LABEL: define void @stride_sext_exceeds_i32_max(
; CHECK-SAME: ptr noalias readonly [[SRC:%.*]], ptr noalias [[DST:%.*]], i16 [[START:%.*]]) #[[ATTR0:[0-9]+]] {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = sub i16 -1, [[START]]
; CHECK-NEXT:    [[TMP1:%.*]] = zext i16 [[TMP0]] to i32
; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[TMP1]], 1
; CHECK-NEXT:    br label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x ptr> poison, ptr [[DST]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x ptr> [[BROADCAST_SPLATINSERT]], <vscale x 16 x ptr> poison, <vscale x 16 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP3:%.*]] = call <vscale x 16 x i16> @llvm.stepvector.nxv16i16()
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i16> poison, i16 [[START]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i16> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer
; CHECK-NEXT:    [[INDUCTION:%.*]] = add <vscale x 16 x i16> [[BROADCAST_SPLAT2]], [[TMP3]]
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i16> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[AVL:%.*]] = phi i32 [ [[TMP2]], %[[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
; CHECK-NEXT:    [[TMP5:%.*]] = trunc i32 [[TMP4]] to i16
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <vscale x 16 x i16> poison, i16 [[TMP5]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <vscale x 16 x i16> [[BROADCAST_SPLATINSERT3]], <vscale x 16 x i16> poison, <vscale x 16 x i32> zeroinitializer
; CHECK-NEXT:    [[TMP6:%.*]] = sext <vscale x 16 x i16> [[VEC_IND]] to <vscale x 16 x i64>
; CHECK-NEXT:    [[TMP7:%.*]] = mul nsw <vscale x 16 x i64> [[TMP6]], splat (i64 3000000000)
; CHECK-NEXT:    [[WIDE_GEP:%.*]] = getelementptr i8, ptr [[SRC]], <vscale x 16 x i64> [[TMP7]]
; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[WIDE_GEP]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP4]])
; CHECK-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x ptr> align 1 [[BROADCAST_SPLAT]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP4]])
; CHECK-NEXT:    [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP4]]
; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i16> [[VEC_IND]], [[BROADCAST_SPLAT4]]
; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %loop

loop:
  %iv = phi i16 [ %start, %entry ], [ %iv.next, %loop ]
  %iv.ext = sext i16 %iv to i64
  %offset = mul nsw i64 %iv.ext, 3000000000
  %gep = getelementptr i8, ptr %src, i64 %offset
  %val = load i8, ptr %gep, align 1
  store i8 %val, ptr %dst, align 1
  %iv.next = add i16 %iv, 1
  %cmp = icmp ne i16 %iv.next, 0
  br i1 %cmp, label %loop, label %exit

exit:
  ret void
}

define void @stride_zext_exceeds_i32_max(ptr noalias readonly %src, ptr noalias %dst, i16 %start) {
; CHECK-LABEL: define void @stride_zext_exceeds_i32_max(
; CHECK-SAME: ptr noalias readonly [[SRC:%.*]], ptr noalias [[DST:%.*]], i16 [[START:%.*]]) #[[ATTR0]] {
; CHECK-NEXT:  [[ENTRY:.*:]]
; CHECK-NEXT:    [[TMP0:%.*]] = sub i16 -1, [[START]]
; CHECK-NEXT:    [[TMP1:%.*]] = zext i16 [[TMP0]] to i32
; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[TMP1]], 1
; CHECK-NEXT:    br label %[[VECTOR_PH:.*]]
; CHECK:       [[VECTOR_PH]]:
; CHECK-NEXT:    [[TMP3:%.*]] = zext i16 [[START]] to i64
; CHECK-NEXT:    [[TMP4:%.*]] = mul nuw nsw i64 [[TMP3]], 3000000000
; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP4]]
; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x ptr> poison, ptr [[DST]], i64 0
; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x ptr> [[BROADCAST_SPLATINSERT]], <vscale x 16 x ptr> poison, <vscale x 16 x i32> zeroinitializer
; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
; CHECK:       [[VECTOR_BODY]]:
; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[CURRENT_ITERATION_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[AVL:%.*]] = phi i32 [ [[TMP2]], %[[VECTOR_PH]] ], [ [[AVL_NEXT:%.*]], %[[VECTOR_BODY]] ]
; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
; CHECK-NEXT:    [[TMP7:%.*]] = zext i32 [[INDEX]] to i64
; CHECK-NEXT:    [[TMP8:%.*]] = mul i64 [[TMP7]], 3000000000
; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP5]], i64 [[TMP8]]
; CHECK-NEXT:    [[TMP10:%.*]] = call <vscale x 16 x i8> @llvm.experimental.vp.strided.load.nxv16i8.p0.i64(ptr align 1 [[TMP9]], i64 3000000000, <vscale x 16 x i1> splat (i1 true), i32 [[TMP6]])
; CHECK-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x ptr> align 1 [[BROADCAST_SPLAT]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP6]])
; CHECK-NEXT:    [[CURRENT_ITERATION_NEXT]] = add nuw i32 [[TMP6]], [[INDEX]]
; CHECK-NEXT:    [[AVL_NEXT]] = sub nuw i32 [[AVL]], [[TMP6]]
; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[AVL_NEXT]], 0
; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
; CHECK:       [[MIDDLE_BLOCK]]:
; CHECK-NEXT:    br label %[[EXIT:.*]]
; CHECK:       [[EXIT]]:
; CHECK-NEXT:    ret void
;
entry:
  br label %loop

loop:
  %iv = phi i16 [ %start, %entry ], [ %iv.next, %loop ]
  %iv.ext = zext i16 %iv to i64
  %offset = mul nsw i64 %iv.ext, 3000000000
  %gep = getelementptr i8, ptr %src, i64 %offset
  %val = load i8, ptr %gep, align 1
  store i8 %val, ptr %dst, align 1
  %iv.next = add i16 %iv, 1
  %cmp = icmp ne i16 %iv.next, 0
  br i1 %cmp, label %loop, label %exit

exit:
  ret void
}