# RUN: llc -O0 -run-pass=legalizer -global-isel %s -o - | FileCheck %s

--- |
  target datalayout = "e-m:o-i64:64-i128:128-n32:64-S128"
  target triple = "aarch64--"
  define void @test_inserts_1() { ret void }
  define void @test_inserts_2() { ret void }
  define void @test_inserts_3() { ret void }
  define void @test_inserts_4() { ret void }
  define void @test_inserts_5() { ret void }
  define void @test_inserts_6() { ret void }
...

---
name:            test_inserts_1
body: |
  bb.0:
    liveins: %w0

      ; Low part of insertion wipes out the old register entirely, so %0 gets
      ; forwarded to the G_STORE. Hi part is unchanged so (split) G_LOAD gets
      ; forwarded.
    ; CHECK-LABEL: name: test_inserts_1
    ; CHECK: [[LO:%[0-9]+]](s64) = G_LOAD
    ; CHECK: [[HI:%[0-9]+]](s64) = G_LOAD
    ; CHECK: G_STORE %0(s64)
    ; CHECK: G_STORE [[HI]]
    %0:_(s64) = COPY %x0
    %1:_(s32) = COPY %w1
    %2:_(p0) = COPY %x2
    %3:_(s128) = G_LOAD %2(p0) :: (load 16)
    %4:_(s128) = G_INSERT %3(s128), %0(s64), 0
    G_STORE %4(s128), %2(p0) :: (store 16)
    RET_ReallyLR
...

---
name:            test_inserts_2
body: |
  bb.0:
    liveins: %w0

      ; Low insertion wipes out the old register entirely, so %0 gets forwarded
      ; to the G_STORE again. Second insertion is real.
    ; CHECK-LABEL: name: test_inserts_2
    ; CHECK: [[LO:%[0-9]+]](s64) = G_LOAD
    ; CHECK: [[HI:%[0-9]+]](s64) = G_LOAD
    ; CHECK: [[NEWHI:%[0-9]+]](s64) = G_INSERT [[HI]], %1(s32), 0
    ; CHECK: G_STORE %0(s64)
    ; CHECK: G_STORE [[NEWHI]]
    %0:_(s64) = COPY %x0
    %1:_(s32) = COPY %w1
    %2:_(p0) = COPY %x2
    %3:_(s128) = G_LOAD %2(p0) :: (load 16)
    %4:_(s128) = G_INSERT %3(s128), %0(s64), 0
    %5:_(s128) = G_INSERT %4(s128), %1(s32), 64
    G_STORE %5(s128), %2(p0) :: (store 16)
    RET_ReallyLR
...

---
name:            test_inserts_3
body: |
  bb.0:
    liveins: %w0

      ; I'm not entirely convinced inserting a p0 into an s64 is valid, but it's
      ; certainly better than the alternative of directly forwarding the value
      ; which would cause a nasty type mismatch.
    ; CHECK-LABEL: name: test_inserts_3
    ; CHECK: [[LO:%[0-9]+]](s64) = G_LOAD
    ; CHECK: [[HI:%[0-9]+]](s64) = G_LOAD
    ; CHECK: [[NEWLO:%[0-9]+]](s64) = G_PTRTOINT %0(p0)
    ; CHECK: G_STORE [[NEWLO]](s64)
    ; CHECK: G_STORE [[HI]]
    %0:_(p0) = COPY %x0
    %1:_(s32) = COPY %w1
    %2:_(p0) = COPY %x2
    %3:_(s128) = G_LOAD %2(p0) :: (load 16)
    %4:_(s128) = G_INSERT %3(s128), %0(p0), 0
    G_STORE %4(s128), %2(p0) :: (store 16)
    RET_ReallyLR
...

---
name:            test_inserts_4
body: |
  bb.0:
    liveins: %w0

      ; A narrow insert gets surrounded by a G_ANYEXT/G_TRUNC pair.
    ; CHECK-LABEL: name: test_inserts_4
    ; CHECK: [[VALEXT:%[0-9]+]](s32) = G_ANYEXT %1(s8)
    ; CHECK: [[VAL:%[0-9]+]](s32) = G_INSERT [[VALEXT]], %0(s1), 0
    ; CHECK: %3(s8) = G_TRUNC [[VAL]](s32)
    %0:_(s1) = COPY %w0
    %1:_(s8) = COPY %w1
    %2:_(p0) = COPY %x2
    %3:_(s8) = G_INSERT %1(s8), %0(s1), 0
    G_STORE %3(s8), %2(p0) :: (store 1)
    RET_ReallyLR
...

---
name:            test_inserts_5
body: |
  bb.0:
    liveins: %x0, %x1, %x2


    ; CHECK-LABEL: name: test_inserts_5
    ; CHECK: [[INS_LO:%[0-9]+]](s32) = G_EXTRACT %2(s64), 0
    ; CHECK: [[VAL_LO:%[0-9]+]](s64) = G_INSERT %0, [[INS_LO]](s32), 32
    ; CHECK: [[INS_HI:%[0-9]+]](s32) = G_EXTRACT %2(s64), 32
    ; CHECK: [[VAL_HI:%[0-9]+]](s64) = G_INSERT %1, [[INS_HI]](s32), 0
    ; CHECK: %4(s128) = G_MERGE_VALUES [[VAL_LO]](s64), [[VAL_HI]](s64)
    %0:_(s64) = COPY %x0
    %1:_(s64) = COPY %x1
    %2:_(s64) = COPY %x2
    %3:_(s128) = G_MERGE_VALUES %0, %1
    %4:_(s128) = G_INSERT %3, %2, 32
    RET_ReallyLR
...

---
name:            test_inserts_6
body: |
  bb.0:
    liveins: %x0, %x1, %x2


    ; CHECK-LABEL: name: test_inserts_6
    ; CHECK: [[VAL_LO:%[0-9]+]](s64) = G_INSERT %0, %2(s32), 32
    ; CHECK: %4(s128) = G_MERGE_VALUES [[VAL_LO]](s64), %1(s64)
    %0:_(s64) = COPY %x0
    %1:_(s64) = COPY %x1
    %2:_(s32) = COPY %w2
    %3:_(s128) = G_MERGE_VALUES %0, %1
    %4:_(s128) = G_INSERT %3, %2, 32
    RET_ReallyLR
...