; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck -check-prefixes=GFX950,GFX950-SDAG %s
; RUN: llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck -check-prefixes=GFX950,GFX950-GISEL %s

; RUN: not --crash llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 -filetype=null < %s 2>&1 | FileCheck -check-prefix=ERR-SDAG %s
; RUN: not llc -global-isel=1 -new-reg-bank-select -mtriple=amdgcn -mcpu=gfx942 -filetype=null < %s 2>&1 | FileCheck -check-prefix=ERR-GISEL %s

; ERR-SDAG: LLVM ERROR: Cannot select: intrinsic %llvm.amdgcn.global.load.lds

; ERR-GISEL: LLVM ERROR: cannot select: G_INTRINSIC_W_SIDE_EFFECTS intrinsic(@llvm.amdgcn.global.load.lds),


declare void @llvm.amdgcn.global.load.lds(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr, i32 %size, i32 %offset, i32 %aux)

;---------------------------------------------------------------------y
; dwordx3
;---------------------------------------------------------------------

define amdgpu_ps void @global_load_lds_dwordx3_vaddr(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr) {
; GFX950-LABEL: global_load_lds_dwordx3_vaddr:
; GFX950:       ; %bb.0:
; GFX950-NEXT:    v_readfirstlane_b32 s0, v2
; GFX950-NEXT:    s_mov_b32 m0, s0
; GFX950-NEXT:    s_nop 0
; GFX950-NEXT:    global_load_lds_dwordx3 v[0:1], off offset:16 sc0
; GFX950-NEXT:    s_endpgm
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 12, i32 16, i32 1)
  ret void
}

define amdgpu_ps void @global_load_lds_dwordx3_saddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr) {
; GFX950-SDAG-LABEL: global_load_lds_dwordx3_saddr:
; GFX950-SDAG:       ; %bb.0:
; GFX950-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 0
; GFX950-SDAG-NEXT:    s_mov_b32 m0, s2
; GFX950-SDAG-NEXT:    s_nop 0
; GFX950-SDAG-NEXT:    global_load_lds_dwordx3 v1, s[0:1] offset:32 nt
; GFX950-SDAG-NEXT:    s_endpgm
;
; GFX950-GISEL-LABEL: global_load_lds_dwordx3_saddr:
; GFX950-GISEL:       ; %bb.0:
; GFX950-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
; GFX950-GISEL-NEXT:    s_mov_b32 m0, s2
; GFX950-GISEL-NEXT:    s_nop 0
; GFX950-GISEL-NEXT:    global_load_lds_dwordx3 v0, s[0:1] offset:32 nt
; GFX950-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 12, i32 32, i32 2)
  ret void
}

define amdgpu_ps void @global_load_lds_dwordx3_saddr_and_vaddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr, i32 %voffset) {
; GFX950-LABEL: global_load_lds_dwordx3_saddr_and_vaddr:
; GFX950:       ; %bb.0:
; GFX950-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-NEXT:    s_mov_b32 m0, s2
; GFX950-NEXT:    s_nop 0
; GFX950-NEXT:    global_load_lds_dwordx3 v1, s[0:1] offset:48 sc1
; GFX950-NEXT:    s_endpgm
  %voffset.64 = zext i32 %voffset to i64
  %gep = getelementptr i8, ptr addrspace(1) %gptr, i64 %voffset.64
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gep, ptr addrspace(3) %lptr, i32 12, i32 48, i32 16)
  ret void
}

;---------------------------------------------------------------------
; dwordx4
;---------------------------------------------------------------------

define amdgpu_ps void @global_load_lds_dwordx4_vaddr(ptr addrspace(1) nocapture %gptr, ptr addrspace(3) nocapture %lptr) {
; GFX950-LABEL: global_load_lds_dwordx4_vaddr:
; GFX950:       ; %bb.0:
; GFX950-NEXT:    v_readfirstlane_b32 s0, v2
; GFX950-NEXT:    s_mov_b32 m0, s0
; GFX950-NEXT:    s_nop 0
; GFX950-NEXT:    global_load_lds_dwordx4 v[0:1], off offset:16 sc0
; GFX950-NEXT:    s_endpgm
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 16, i32 16, i32 1)
  ret void
}

define amdgpu_ps void @global_load_lds_dwordx4_saddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr) {
; GFX950-SDAG-LABEL: global_load_lds_dwordx4_saddr:
; GFX950-SDAG:       ; %bb.0:
; GFX950-SDAG-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 0
; GFX950-SDAG-NEXT:    s_mov_b32 m0, s2
; GFX950-SDAG-NEXT:    s_nop 0
; GFX950-SDAG-NEXT:    global_load_lds_dwordx4 v1, s[0:1] offset:32 nt
; GFX950-SDAG-NEXT:    s_endpgm
;
; GFX950-GISEL-LABEL: global_load_lds_dwordx4_saddr:
; GFX950-GISEL:       ; %bb.0:
; GFX950-GISEL-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
; GFX950-GISEL-NEXT:    s_mov_b32 m0, s2
; GFX950-GISEL-NEXT:    s_nop 0
; GFX950-GISEL-NEXT:    global_load_lds_dwordx4 v0, s[0:1] offset:32 nt
; GFX950-GISEL-NEXT:    s_endpgm
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gptr, ptr addrspace(3) %lptr, i32 16, i32 32, i32 2)
  ret void
}

define amdgpu_ps void @global_load_lds_dwordx4_saddr_and_vaddr(ptr addrspace(1) nocapture inreg %gptr, ptr addrspace(3) nocapture %lptr, i32 %voffset) {
; GFX950-LABEL: global_load_lds_dwordx4_saddr_and_vaddr:
; GFX950:       ; %bb.0:
; GFX950-NEXT:    v_readfirstlane_b32 s2, v0
; GFX950-NEXT:    s_mov_b32 m0, s2
; GFX950-NEXT:    s_nop 0
; GFX950-NEXT:    global_load_lds_dwordx4 v1, s[0:1] offset:48 sc1
; GFX950-NEXT:    s_endpgm
  %voffset.64 = zext i32 %voffset to i64
  %gep = getelementptr i8, ptr addrspace(1) %gptr, i64 %voffset.64
  call void @llvm.amdgcn.global.load.lds(ptr addrspace(1) %gep, ptr addrspace(3) %lptr, i32 16, i32 48, i32 16)
  ret void
}