test/Transforms/InstCombine/gep-vector-indices.ll

; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt -passes=instcombine %s -S | FileCheck %s

define ptr @vector_splat_indices_v2i64_ext0(ptr %a) {
; CHECK-LABEL: @vector_splat_indices_v2i64_ext0(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 4
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, ptr %a, <2 x i64> <i64 4, i64 4>
  %res = extractelement <2 x ptr> %gep, i32 0
  ret ptr %res
}

define ptr @vector_splat_indices_nxv2i64_ext0(ptr %a) {
; CHECK-LABEL: @vector_splat_indices_nxv2i64_ext0(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 4
; CHECK-NEXT:    ret ptr [[RES]]
;
  %tmp = insertelement <vscale x 2 x i64> poison, i64 4, i32 0
  %splatof4 = shufflevector <vscale x 2 x i64> %tmp, <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
  %gep = getelementptr inbounds i32, ptr %a, <vscale x 2 x i64> %splatof4
  %res = extractelement <vscale x 2 x ptr> %gep, i32 0
  ret ptr %res
}

define ptr @vector_indices_v2i64_ext0(ptr %a, <2 x i64> %indices) {
; CHECK-LABEL: @vector_indices_v2i64_ext0(
; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <2 x i64> [[INDICES:%.*]], i64 0
; CHECK-NEXT:    [[RES:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[TMP1]]
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, ptr %a, <2 x i64> %indices
  %res = extractelement <2 x ptr> %gep, i32 0
  ret ptr %res
}

define ptr @vector_indices_nxv1i64_ext0(ptr %a, <vscale x 1 x i64> %indices) {
; CHECK-LABEL: @vector_indices_nxv1i64_ext0(
; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <vscale x 1 x i64> [[INDICES:%.*]], i64 0
; CHECK-NEXT:    [[RES:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[TMP1]]
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, ptr %a, <vscale x 1 x i64> %indices
  %res = extractelement <vscale x 1 x ptr> %gep, i32 0
  ret ptr %res
}


define ptr @vector_splat_ptrs_v2i64_ext0(ptr %a, i64 %index) {
; CHECK-LABEL: @vector_splat_ptrs_v2i64_ext0(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX:%.*]]
; CHECK-NEXT:    ret ptr [[RES]]
;
  %tmp = insertelement <2 x ptr> poison, ptr %a, i32 0
  %splatofa = shufflevector <2 x ptr> %tmp, <2 x ptr> poison, <2 x i32> zeroinitializer
  %gep = getelementptr i32, <2 x ptr> %splatofa, i64 %index
  %res = extractelement <2 x ptr> %gep, i32 0
  ret ptr %res
}


define ptr @vector_splat_ptrs_nxv2i64_ext0(ptr %a, i64 %index) {
; CHECK-LABEL: @vector_splat_ptrs_nxv2i64_ext0(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX:%.*]]
; CHECK-NEXT:    ret ptr [[RES]]
;
  %tmp = insertelement <vscale x 2 x ptr> poison, ptr %a, i32 0
  %splatofa = shufflevector <vscale x 2 x ptr> %tmp, <vscale x 2 x ptr> poison, <vscale x 2 x i32> zeroinitializer
  %gep = getelementptr i32, <vscale x 2 x ptr> %splatofa, i64 %index
  %res = extractelement <vscale x 2 x ptr> %gep, i32 0
  ret ptr %res
}


define ptr @vector_struct1_splat_indices_v4i64_ext1(ptr %a) {
; CHECK-LABEL: @vector_struct1_splat_indices_v4i64_ext1(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr { float, float }, ptr [[A:%.*]], i64 4, i32 0
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr {float, float}, ptr %a, <4 x i32> <i32 4, i32 4, i32 4, i32 4>, i32 0
  %res = extractelement <4 x ptr> %gep, i32 1
  ret ptr %res
}


define ptr @vector_struct2_splat_indices_v4i64_ext1(ptr %a) {
; CHECK-LABEL: @vector_struct2_splat_indices_v4i64_ext1(
; CHECK-NEXT:    [[RES:%.*]] = getelementptr { float, [8 x float] }, ptr [[A:%.*]], i64 2, i32 1, i64 4
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr {float, [8 x float]}, ptr %a, i32 2, i32 1, <4 x i32> <i32 4, i32 4, i32 4, i32 4>
  %res = extractelement <4 x ptr> %gep, i32 1
  ret ptr %res
}


; Negative tests

define ptr @vector_indices_nxv2i64_ext3(ptr %a, <vscale x 2 x i64> %indices) {
; CHECK-LABEL: @vector_indices_nxv2i64_ext3(
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[A:%.*]], <vscale x 2 x i64> [[INDICES:%.*]]
; CHECK-NEXT:    [[RES:%.*]] = extractelement <vscale x 2 x ptr> [[GEP]], i64 3
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, ptr %a, <vscale x 2 x i64> %indices
  %res = extractelement <vscale x 2 x ptr> %gep, i32 3
  ret ptr %res
}

define ptr @vector_indices_nxv2i64_extN(ptr %a, <vscale x 2 x i64> %indices, i32 %N) {
; CHECK-LABEL: @vector_indices_nxv2i64_extN(
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[A:%.*]], <vscale x 2 x i64> [[INDICES:%.*]]
; CHECK-NEXT:    [[RES:%.*]] = extractelement <vscale x 2 x ptr> [[GEP]], i32 [[N:%.*]]
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, ptr %a, <vscale x 2 x i64> %indices
  %res = extractelement <vscale x 2 x ptr> %gep, i32 %N
  ret ptr %res
}

define void @vector_indices_nxv2i64_mulitple_use(ptr %a, <vscale x 2 x i64> %indices, ptr %b, ptr %c) {
; CHECK-LABEL: @vector_indices_nxv2i64_mulitple_use(
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[A:%.*]], <vscale x 2 x i64> [[INDICES:%.*]]
; CHECK-NEXT:    [[LANE0:%.*]] = extractelement <vscale x 2 x ptr> [[GEP]], i64 0
; CHECK-NEXT:    [[LANE1:%.*]] = extractelement <vscale x 2 x ptr> [[GEP]], i64 1
; CHECK-NEXT:    store ptr [[LANE0]], ptr [[B:%.*]], align 8
; CHECK-NEXT:    store ptr [[LANE1]], ptr [[C:%.*]], align 8
; CHECK-NEXT:    ret void
;
  %gep = getelementptr i32, ptr %a, <vscale x 2 x i64> %indices
  %lane0 = extractelement <vscale x 2 x ptr> %gep, i32 0
  %lane1 = extractelement <vscale x 2 x ptr> %gep, i32 1
  store ptr %lane0, ptr %b, align 8
  store ptr %lane1, ptr %c, align 8
  ret void
}

define ptr @vector_ptrs_and_indices_ext0(<vscale x 4 x ptr> %a, <vscale x 4 x i64> %indices) {
; CHECK-LABEL: @vector_ptrs_and_indices_ext0(
; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, <vscale x 4 x ptr> [[A:%.*]], <vscale x 4 x i64> [[INDICES:%.*]]
; CHECK-NEXT:    [[RES:%.*]] = extractelement <vscale x 4 x ptr> [[GEP]], i64 0
; CHECK-NEXT:    ret ptr [[RES]]
;
  %gep = getelementptr i32, <vscale x 4 x ptr> %a, <vscale x 4 x i64> %indices
  %res = extractelement <vscale x 4 x ptr> %gep, i32 0
  ret ptr %res
}