test/CodeGen/AMDGPU/load-select-ptr.ll

; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=fiji < %s | FileCheck -check-prefix=GCN %s

; Combine on select c, (load x), (load y) -> load (select c, x, y)
; drops MachinePointerInfo, so it can't be relied on for correctness.

define amdgpu_kernel void @select_ptr_crash_i64_flat(i32 %tmp, [8 x i32], ptr %ptr0, [8 x i32], ptr %ptr1, [8 x i32], ptr addrspace(1) %ptr2) {
; GCN-LABEL: select_ptr_crash_i64_flat:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_add_i32 s12, s12, s17
; GCN-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT:    s_load_dword s2, s[8:9], 0x0
; GCN-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x78
; GCN-NEXT:    s_add_u32 s4, s8, 40
; GCN-NEXT:    s_addc_u32 s3, s9, 0
; GCN-NEXT:    s_add_u32 s5, s8, 0x50
; GCN-NEXT:    s_addc_u32 s6, s9, 0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_cmp_eq_u32 s2, 0
; GCN-NEXT:    s_cselect_b32 s3, s3, s6
; GCN-NEXT:    s_cselect_b32 s2, s4, s5
; GCN-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GCN-NEXT:    s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    v_mov_b32_e32 v0, s2
; GCN-NEXT:    v_mov_b32_e32 v1, s3
; GCN-NEXT:    s_add_u32 s2, s2, 4
; GCN-NEXT:    flat_load_dword v0, v[0:1]
; GCN-NEXT:    s_addc_u32 s3, s3, 0
; GCN-NEXT:    v_mov_b32_e32 v1, s2
; GCN-NEXT:    v_mov_b32_e32 v2, s3
; GCN-NEXT:    flat_load_dword v1, v[1:2]
; GCN-NEXT:    v_mov_b32_e32 v3, s1
; GCN-NEXT:    v_mov_b32_e32 v2, s0
; GCN-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GCN-NEXT:    s_endpgm
  %tmp2 = icmp eq i32 %tmp, 0
  %tmp3 = load i64, ptr %ptr0, align 8
  %tmp4 = load i64, ptr %ptr1, align 8
  %tmp5 = select i1 %tmp2, i64 %tmp3, i64 %tmp4
  store i64 %tmp5, ptr addrspace(1) %ptr2, align 8
  ret void
}

; The transform currently doesn't happen for non-addrspace 0, but it
; should.

define amdgpu_kernel void @select_ptr_crash_i64_global(i32 %tmp, [8 x i32], ptr addrspace(1) %ptr0, [8 x i32], ptr addrspace(1) %ptr1, [8 x i32], ptr addrspace(1) %ptr2) {
; GCN-LABEL: select_ptr_crash_i64_global:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_add_i32 s12, s12, s17
; GCN-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT:    s_load_dword s2, s[8:9], 0x0
; GCN-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x78
; GCN-NEXT:    s_add_u32 s4, s8, 40
; GCN-NEXT:    s_addc_u32 s3, s9, 0
; GCN-NEXT:    s_add_u32 s5, s8, 0x50
; GCN-NEXT:    s_addc_u32 s6, s9, 0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_cmp_eq_u32 s2, 0
; GCN-NEXT:    s_cselect_b32 s3, s3, s6
; GCN-NEXT:    s_cselect_b32 s2, s4, s5
; GCN-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GCN-NEXT:    v_mov_b32_e32 v0, s0
; GCN-NEXT:    s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT:    v_mov_b32_e32 v1, s1
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_load_dwordx2 s[2:3], s[2:3], 0x0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    v_mov_b32_e32 v2, s2
; GCN-NEXT:    v_mov_b32_e32 v3, s3
; GCN-NEXT:    flat_store_dwordx2 v[0:1], v[2:3]
; GCN-NEXT:    s_endpgm
  %tmp2 = icmp eq i32 %tmp, 0
  %tmp3 = load i64, ptr addrspace(1) %ptr0, align 8
  %tmp4 = load i64, ptr addrspace(1) %ptr1, align 8
  %tmp5 = select i1 %tmp2, i64 %tmp3, i64 %tmp4
  store i64 %tmp5, ptr addrspace(1) %ptr2, align 8
  ret void
}

define amdgpu_kernel void @select_ptr_crash_i64_local(i32 %tmp, ptr addrspace(3) %ptr0, ptr addrspace(3) %ptr1, ptr addrspace(1) %ptr2) {
; GCN-LABEL: select_ptr_crash_i64_local:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
; GCN-NEXT:    s_load_dwordx2 s[4:5], s[8:9], 0x10
; GCN-NEXT:    s_add_i32 s12, s12, s17
; GCN-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT:    s_mov_b32 m0, -1
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_cmp_eq_u32 s0, 0
; GCN-NEXT:    s_cselect_b32 s0, s1, s2
; GCN-NEXT:    v_mov_b32_e32 v0, s0
; GCN-NEXT:    ds_read_b64 v[0:1], v0
; GCN-NEXT:    v_mov_b32_e32 v2, s4
; GCN-NEXT:    s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT:    v_mov_b32_e32 v3, s5
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GCN-NEXT:    s_endpgm
  %tmp2 = icmp eq i32 %tmp, 0
  %tmp3 = load i64, ptr addrspace(3) %ptr0, align 8
  %tmp4 = load i64, ptr addrspace(3) %ptr1, align 8
  %tmp5 = select i1 %tmp2, i64 %tmp3, i64 %tmp4
  store i64 %tmp5, ptr addrspace(1) %ptr2, align 8
  ret void
}

; The transform will break addressing mode matching, so unclear it
; would be good to do

define amdgpu_kernel void @select_ptr_crash_i64_local_offsets(i32 %tmp, ptr addrspace(3) %ptr0, ptr addrspace(3) %ptr1, ptr addrspace(1) %ptr2) {
; GCN-LABEL: select_ptr_crash_i64_local_offsets:
; GCN:       ; %bb.0:
; GCN-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
; GCN-NEXT:    s_load_dwordx2 s[4:5], s[8:9], 0x10
; GCN-NEXT:    s_add_i32 s12, s12, s17
; GCN-NEXT:    s_lshr_b32 flat_scratch_hi, s12, 8
; GCN-NEXT:    s_mov_b32 m0, -1
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_addk_i32 s1, 0x80
; GCN-NEXT:    s_addk_i32 s2, 0x200
; GCN-NEXT:    s_cmp_eq_u32 s0, 0
; GCN-NEXT:    s_cselect_b32 s0, s1, s2
; GCN-NEXT:    v_mov_b32_e32 v0, s0
; GCN-NEXT:    ds_read_b64 v[0:1], v0
; GCN-NEXT:    v_mov_b32_e32 v2, s4
; GCN-NEXT:    s_mov_b32 flat_scratch_lo, s13
; GCN-NEXT:    v_mov_b32_e32 v3, s5
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GCN-NEXT:    s_endpgm
  %tmp2 = icmp eq i32 %tmp, 0
  %gep0 = getelementptr inbounds i64, ptr addrspace(3) %ptr0, i64 16
  %gep1 = getelementptr inbounds i64, ptr addrspace(3) %ptr1, i64 64
  %tmp3 = load i64, ptr addrspace(3) %gep0, align 8
  %tmp4 = load i64, ptr addrspace(3) %gep1, align 8
  %tmp5 = select i1 %tmp2, i64 %tmp3, i64 %tmp4
  store i64 %tmp5, ptr addrspace(1) %ptr2, align 8
  ret void
}

; The resultant load cannot be treated as uniform
define amdgpu_kernel void @sample_test(ptr addrspace(1) %dest, ptr addrspace(1) %sourceA, ptr addrspace(1) %sourceB, i1 %tobool.not.i) #0 {
; GCN-LABEL: sample_test:
; GCN:       ; %bb.0: ; %entry
; GCN-NEXT:    s_load_dword s6, s[4:5], 0x18
; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; GCN-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
; GCN-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_bitcmp1_b32 s6, 0
; GCN-NEXT:    v_mov_b32_e32 v0, s3
; GCN-NEXT:    v_add_u32_e32 v3, vcc, s2, v2
; GCN-NEXT:    v_addc_u32_e32 v0, vcc, 0, v0, vcc
; GCN-NEXT:    v_mov_b32_e32 v1, s5
; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
; GCN-NEXT:    v_cndmask_b32_e32 v1, v1, v0, vcc
; GCN-NEXT:    v_mov_b32_e32 v0, s4
; GCN-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; GCN-NEXT:    v_mov_b32_e32 v3, s1
; GCN-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
; GCN-NEXT:    s_waitcnt vmcnt(0)
; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GCN-NEXT:    s_endpgm
entry:
  %0 = tail call i32 @llvm.amdgcn.workitem.id.x()
  %conv2.i.i.i1 = zext i32 %0 to i64
  %arrayidx.i = getelementptr i64, ptr addrspace(1) %sourceA, i64 %conv2.i.i.i1
  %dest.gep = getelementptr i64, ptr addrspace(1) %dest, i64 %conv2.i.i.i1
  %ld0 = load i64, ptr addrspace(1) %arrayidx.i, align 8, !amdgpu.noclobber !0
  %ld1 = load i64, ptr addrspace(1) %sourceB, align 8
  %cond.i = select i1 %tobool.not.i, i64 %ld0, i64 %ld1
  store i64 %cond.i, ptr addrspace(1) %dest.gep, align 8
  ret void
}

; The resultant load cannot be treated as uniform
define amdgpu_kernel void @constant_is_not_uniform(ptr addrspace(1) %dest, ptr addrspace(4) %sourceA, ptr addrspace(4) %sourceB, i1 %tobool.not.i) #0 {
; GCN-LABEL: constant_is_not_uniform:
; GCN:       ; %bb.0: ; %entry
; GCN-NEXT:    s_load_dword s6, s[4:5], 0x18
; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
; GCN-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x10
; GCN-NEXT:    v_lshlrev_b32_e32 v2, 3, v0
; GCN-NEXT:    s_waitcnt lgkmcnt(0)
; GCN-NEXT:    s_bitcmp1_b32 s6, 0
; GCN-NEXT:    v_mov_b32_e32 v0, s3
; GCN-NEXT:    v_add_u32_e32 v3, vcc, s2, v2
; GCN-NEXT:    v_addc_u32_e32 v0, vcc, 0, v0, vcc
; GCN-NEXT:    v_mov_b32_e32 v1, s5
; GCN-NEXT:    s_cselect_b64 vcc, -1, 0
; GCN-NEXT:    v_cndmask_b32_e32 v1, v1, v0, vcc
; GCN-NEXT:    v_mov_b32_e32 v0, s4
; GCN-NEXT:    v_cndmask_b32_e32 v0, v0, v3, vcc
; GCN-NEXT:    flat_load_dwordx2 v[0:1], v[0:1]
; GCN-NEXT:    v_mov_b32_e32 v3, s1
; GCN-NEXT:    v_add_u32_e32 v2, vcc, s0, v2
; GCN-NEXT:    v_addc_u32_e32 v3, vcc, 0, v3, vcc
; GCN-NEXT:    s_waitcnt vmcnt(0)
; GCN-NEXT:    flat_store_dwordx2 v[2:3], v[0:1]
; GCN-NEXT:    s_endpgm
entry:
  %0 = tail call i32 @llvm.amdgcn.workitem.id.x()
  %conv2.i.i.i1 = zext i32 %0 to i64
  %arrayidx.i = getelementptr i64, ptr addrspace(4) %sourceA, i64 %conv2.i.i.i1
  %dest.gep = getelementptr i64, ptr addrspace(1) %dest, i64 %conv2.i.i.i1
  %ld0 = load i64, ptr addrspace(4) %arrayidx.i, align 8
  %ld1 = load i64, ptr addrspace(4) %sourceB, align 8
  %cond.i = select i1 %tobool.not.i, i64 %ld0, i64 %ld1
  store i64 %cond.i, ptr addrspace(1) %dest.gep, align 8
  ret void
}

attributes #0 = { nounwind "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-flat-scratch-init" "amdgpu-no-heap-ptr" "amdgpu-no-hostcall-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-multigrid-sync-arg" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-x" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }

!0 = !{}