; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn < %s | FileCheck -enable-var-scope -check-prefixes=SI %s
; RUN: llc -amdgpu-scalarize-global-loads=false -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -enable-var-scope -check-prefixes=VI %s

define amdgpu_kernel void @trunc_i64_bitcast_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i64_bitcast_v2i32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i64_bitcast_v2i32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <2 x i32>, ptr addrspace(1) %in
  %bc = bitcast <2 x i32> %ld to i64
  %trunc = trunc i64 %bc to i32
  store i32 %trunc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @trunc_i96_bitcast_v3i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i96_bitcast_v3i32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i96_bitcast_v3i32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <3 x i32>, ptr addrspace(1) %in
  %bc = bitcast <3 x i32> %ld to i96
  %trunc = trunc i96 %bc to i32
  store i32 %trunc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @trunc_i128_bitcast_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i128_bitcast_v4i32:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i128_bitcast_v4i32:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_dword v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <4 x i32>, ptr addrspace(1) %in
  %bc = bitcast <4 x i32> %ld to i128
  %trunc = trunc i128 %bc to i32
  store i32 %trunc, ptr addrspace(1) %out
  ret void
}

; Don't want load width reduced in this case.
define amdgpu_kernel void @trunc_i16_bitcast_v2i16(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i16_bitcast_v2i16:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i16_bitcast_v2i16:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <2 x i16>, ptr addrspace(1) %in
  %bc = bitcast <2 x i16> %ld to i32
  %trunc = trunc i32 %bc to i16
  store i16 %trunc, ptr addrspace(1) %out
  ret void
}

; FIXME We need to teach the dagcombiner to reduce load width for:
;   t21: v2i32,ch = load<LD8[%in(addrspace=1)]> t12, t10, undef:i64
;        t23: i64 = bitcast t21
;      t30: i16 = truncate t23
define amdgpu_kernel void @trunc_i16_bitcast_v4i16(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i16_bitcast_v4i16:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_short v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i16_bitcast_v4i16:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_short v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <4 x i16>, ptr addrspace(1) %in
  %bc = bitcast <4 x i16> %ld to i64
  %trunc = trunc i64 %bc to i16
  store i16 %trunc, ptr addrspace(1) %out
  ret void
}

; FIXME: Consistently shrink or not here
define amdgpu_kernel void @trunc_i8_bitcast_v2i8(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i8_bitcast_v2i8:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_ubyte v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i8_bitcast_v2i8:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_ushort v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <2 x i8>, ptr addrspace(1) %in
  %bc = bitcast <2 x i8> %ld to i16
  %trunc = trunc i16 %bc to i8
  store i8 %trunc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @trunc_i32_bitcast_v4i8(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i32_bitcast_v4i8:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i32_bitcast_v4i8:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <4 x i8>, ptr addrspace(1) %in
  %bc = bitcast <4 x i8> %ld to i32
  %trunc = trunc i32 %bc to i8
  store i8 %trunc, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @trunc_i24_bitcast_v3i8(ptr addrspace(1) %out, ptr addrspace(1) %in) {
; SI-LABEL: trunc_i24_bitcast_v3i8:
; SI:       ; %bb.0:
; SI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x9
; SI-NEXT:    s_mov_b32 s7, 0xf000
; SI-NEXT:    s_mov_b32 s6, -1
; SI-NEXT:    s_mov_b32 s10, s6
; SI-NEXT:    s_mov_b32 s11, s7
; SI-NEXT:    s_waitcnt lgkmcnt(0)
; SI-NEXT:    s_mov_b32 s8, s2
; SI-NEXT:    s_mov_b32 s9, s3
; SI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; SI-NEXT:    s_mov_b32 s4, s0
; SI-NEXT:    s_mov_b32 s5, s1
; SI-NEXT:    s_waitcnt vmcnt(0)
; SI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; SI-NEXT:    s_endpgm
;
; VI-LABEL: trunc_i24_bitcast_v3i8:
; VI:       ; %bb.0:
; VI-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
; VI-NEXT:    s_mov_b32 s7, 0xf000
; VI-NEXT:    s_mov_b32 s6, -1
; VI-NEXT:    s_mov_b32 s10, s6
; VI-NEXT:    s_mov_b32 s11, s7
; VI-NEXT:    s_waitcnt lgkmcnt(0)
; VI-NEXT:    s_mov_b32 s8, s2
; VI-NEXT:    s_mov_b32 s9, s3
; VI-NEXT:    buffer_load_dword v0, off, s[8:11], 0
; VI-NEXT:    s_mov_b32 s4, s0
; VI-NEXT:    s_mov_b32 s5, s1
; VI-NEXT:    s_waitcnt vmcnt(0)
; VI-NEXT:    buffer_store_byte v0, off, s[4:7], 0
; VI-NEXT:    s_endpgm
  %ld = load <3 x i8>, ptr addrspace(1) %in
  %bc = bitcast <3 x i8> %ld to i24
  %trunc = trunc i24 %bc to i8
  store i8 %trunc, ptr addrspace(1) %out
  ret void
}