; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 4
; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx942 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX942 %s

; RUN: llc -mtriple=amdgcn--amdhsa -mcpu=gfx90a -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX90a %s

define amdgpu_kernel void @ptr1_i8(ptr addrspace(1) inreg %out, i8 inreg %arg0) #0 {
; GFX942-LABEL: ptr1_i8:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB0_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB0_0:
; GFX942-NEXT:    s_and_b32 s0, s4, 0xff
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i8:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB0_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB0_0:
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xff
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  %ext = zext i8 %arg0 to i32
  store i32 %ext, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @ptr1_i8_zext_arg(ptr addrspace(1) inreg %out, i8 zeroext inreg %arg0) #0 {
; GFX942-LABEL: ptr1_i8_zext_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB1_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB1_0:
; GFX942-NEXT:    s_and_b32 s0, s4, 0xff
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i8_zext_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB1_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB1_0:
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xff
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  %ext = zext i8 %arg0 to i32
  store i32 %ext, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @ptr1_i16_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %arg0) #0 {
; GFX942-LABEL: ptr1_i16_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB2_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB2_0:
; GFX942-NEXT:    s_and_b32 s0, s4, 0xffff
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i16_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB2_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB2_0:
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xffff
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  %ext = zext i16 %arg0 to i32
  store i32 %ext, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @ptr1_i32_preload_arg(ptr addrspace(1) inreg %out, i32 inreg %arg0) #0 {
; GFX942-LABEL: ptr1_i32_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB3_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB3_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i32_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB3_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB3_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store i32 %arg0, ptr addrspace(1) %out
  ret void
}


define amdgpu_kernel void @i32_ptr1_i32_preload_arg(i32 inreg %arg0, ptr addrspace(1) inreg %out, i32 inreg %arg1) #0 {
; GFX942-LABEL: i32_ptr1_i32_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_load_dword s6, s[0:1], 0x10
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB4_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB4_0:
; GFX942-NEXT:    s_add_i32 s0, s2, s6
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[4:5]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i32_ptr1_i32_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dword s10, s[4:5], 0x10
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB4_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB4_0:
; GFX90a-NEXT:    s_add_i32 s0, s6, s10
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[8:9]
; GFX90a-NEXT:    s_endpgm
  %add = add i32 %arg0, %arg1
  store i32 %add, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @ptr1_i16_i16_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %arg0, i16 inreg %arg1) #0 {
; GFX942-LABEL: ptr1_i16_i16_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB5_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB5_0:
; GFX942-NEXT:    s_lshr_b32 s0, s4, 16
; GFX942-NEXT:    s_and_b32 s1, s4, 0xffff
; GFX942-NEXT:    s_add_i32 s0, s1, s0
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i16_i16_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB5_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB5_0:
; GFX90a-NEXT:    s_lshr_b32 s0, s8, 16
; GFX90a-NEXT:    s_and_b32 s1, s8, 0xffff
; GFX90a-NEXT:    s_add_i32 s0, s1, s0
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  %ext = zext i16 %arg0 to i32
  %ext1 = zext i16 %arg1 to i32
  %add = add i32 %ext, %ext1
  store i32 %add, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @ptr1_v2i8_preload_arg(ptr addrspace(1) inreg %out, <2 x i8> inreg %in) #0 {
; GFX942-LABEL: ptr1_v2i8_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB6_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB6_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_v2i8_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB6_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB6_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <2 x i8> %in, ptr addrspace(1) %out
  ret void
}


define amdgpu_kernel void @byref_preload_arg(ptr addrspace(1) inreg %out, ptr addrspace(4) byref(i32) align(256) %in.byref, i32 %after.offset) #0 {
; GFX942-LABEL: byref_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB7_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB7_0:
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x100
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    v_mov_b32_e32 v2, s5
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
; GFX942-NEXT:    s_waitcnt vmcnt(0)
; GFX942-NEXT:    global_store_dword v0, v2, s[2:3] sc0 sc1
; GFX942-NEXT:    s_waitcnt vmcnt(0)
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: byref_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB7_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB7_0:
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x100
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    v_mov_b32_e32 v2, s1
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_waitcnt vmcnt(0)
; GFX90a-NEXT:    global_store_dword v0, v2, s[6:7]
; GFX90a-NEXT:    s_waitcnt vmcnt(0)
; GFX90a-NEXT:    s_endpgm
  %in = load i32, ptr addrspace(4) %in.byref
  store volatile i32 %in, ptr addrspace(1) %out, align 4
  store volatile i32 %after.offset, ptr addrspace(1) %out, align 4
  ret void
}

; The second argument is not expected to be preloaded with the current behavior.

define amdgpu_kernel void @byref_staggered_preload_arg(ptr addrspace(1) inreg %out, ptr addrspace(4) byref(i32) align(256) %in.byref, i32 inreg %after.offset) #0 {
; GFX942-LABEL: byref_staggered_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB8_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB8_0:
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x100
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    v_mov_b32_e32 v2, s5
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
; GFX942-NEXT:    s_waitcnt vmcnt(0)
; GFX942-NEXT:    global_store_dword v0, v2, s[2:3] sc0 sc1
; GFX942-NEXT:    s_waitcnt vmcnt(0)
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: byref_staggered_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB8_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB8_0:
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x100
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    v_mov_b32_e32 v2, s1
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_waitcnt vmcnt(0)
; GFX90a-NEXT:    global_store_dword v0, v2, s[6:7]
; GFX90a-NEXT:    s_waitcnt vmcnt(0)
; GFX90a-NEXT:    s_endpgm
  %in = load i32, ptr addrspace(4) %in.byref
  store volatile i32 %in, ptr addrspace(1) %out, align 4
  store volatile i32 %after.offset, ptr addrspace(1) %out, align 4
  ret void
}


define amdgpu_kernel void @v8i32_arg(ptr addrspace(1) nocapture inreg %out, <8 x i32> inreg %in) #0 {
; GFX942-LABEL: v8i32_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB9_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB9_0:
; GFX942-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x20
; GFX942-NEXT:    v_mov_b32_e32 v4, 0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    v_mov_b32_e32 v0, s8
; GFX942-NEXT:    v_mov_b32_e32 v1, s9
; GFX942-NEXT:    v_mov_b32_e32 v2, s10
; GFX942-NEXT:    v_mov_b32_e32 v3, s11
; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3] offset:16
; GFX942-NEXT:    s_nop 1
; GFX942-NEXT:    v_mov_b32_e32 v0, s4
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    v_mov_b32_e32 v2, s6
; GFX942-NEXT:    v_mov_b32_e32 v3, s7
; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v8i32_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB9_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB9_0:
; GFX90a-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x20
; GFX90a-NEXT:    v_mov_b32_e32 v4, 0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    v_mov_b32_e32 v0, s12
; GFX90a-NEXT:    v_mov_b32_e32 v1, s13
; GFX90a-NEXT:    v_mov_b32_e32 v2, s14
; GFX90a-NEXT:    v_mov_b32_e32 v3, s15
; GFX90a-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7] offset:16
; GFX90a-NEXT:    s_nop 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s8
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    v_mov_b32_e32 v2, s10
; GFX90a-NEXT:    v_mov_b32_e32 v3, s11
; GFX90a-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <8 x i32> %in, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @v3i16_preload_arg(ptr addrspace(1) nocapture inreg %out, <3 x i16> inreg %in) #0 {
; GFX942-LABEL: v3i16_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB10_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB10_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    global_store_short v0, v1, s[2:3] offset:4
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v3i16_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB10_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB10_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7] offset:4
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <3 x i16> %in, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @v3i32_preload_arg(ptr addrspace(1) nocapture inreg %out, <3 x i32> inreg %in) #0 {
; GFX942-LABEL: v3i32_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x18
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB11_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB11_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v3i32_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x18
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB11_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB11_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <3 x i32> %in, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @v3f32_preload_arg(ptr addrspace(1) nocapture inreg %out, <3 x float> inreg %in) #0 {
; GFX942-LABEL: v3f32_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x18
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB12_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB12_0:
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v3f32_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x18
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB12_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB12_0:
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <3 x float> %in, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @v5i8_preload_arg(ptr addrspace(1) nocapture inreg %out, <5 x i8> inreg %in) #0 {
; GFX942-LABEL: v5i8_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB13_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB13_0:
; GFX942-NEXT:    s_lshr_b32 s1, s4, 24
; GFX942-NEXT:    s_and_b32 s0, s4, 0xffff
; GFX942-NEXT:    s_lshl_b32 s1, s1, 8
; GFX942-NEXT:    s_bfe_u32 s4, s4, 0x80010
; GFX942-NEXT:    s_or_b32 s1, s4, s1
; GFX942-NEXT:    s_lshl_b32 s1, s1, 16
; GFX942-NEXT:    s_or_b32 s0, s0, s1
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    global_store_byte v0, v1, s[2:3] offset:4
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v5i8_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB13_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB13_0:
; GFX90a-NEXT:    s_lshr_b32 s1, s8, 24
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 8
; GFX90a-NEXT:    s_bfe_u32 s2, s8, 0x80010
; GFX90a-NEXT:    s_or_b32 s1, s2, s1
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xffff
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 16
; GFX90a-NEXT:    s_or_b32 s0, s0, s1
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    global_store_byte v0, v1, s[6:7] offset:4
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <5 x i8> %in, ptr addrspace(1) %out, align 4
  ret void
}

define amdgpu_kernel void @v5f64_arg(ptr addrspace(1) nocapture inreg %out, <5 x double> inreg %in) #0 {
; GFX942-LABEL: v5f64_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB14_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB14_0:
; GFX942-NEXT:    s_load_dwordx2 s[12:13], s[0:1], 0x60
; GFX942-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x40
; GFX942-NEXT:    v_mov_b32_e32 v4, 0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[12:13]
; GFX942-NEXT:    v_mov_b32_e32 v0, s8
; GFX942-NEXT:    global_store_dwordx2 v4, v[2:3], s[2:3] offset:32
; GFX942-NEXT:    v_mov_b32_e32 v1, s9
; GFX942-NEXT:    v_mov_b32_e32 v2, s10
; GFX942-NEXT:    v_mov_b32_e32 v3, s11
; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3] offset:16
; GFX942-NEXT:    s_nop 1
; GFX942-NEXT:    v_mov_b32_e32 v0, s4
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    v_mov_b32_e32 v2, s6
; GFX942-NEXT:    v_mov_b32_e32 v3, s7
; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v5f64_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB14_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB14_0:
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x60
; GFX90a-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x40
; GFX90a-NEXT:    v_mov_b32_e32 v4, 0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    v_pk_mov_b32 v[2:3], s[0:1], s[0:1] op_sel:[0,1]
; GFX90a-NEXT:    v_mov_b32_e32 v0, s12
; GFX90a-NEXT:    global_store_dwordx2 v4, v[2:3], s[6:7] offset:32
; GFX90a-NEXT:    v_mov_b32_e32 v1, s13
; GFX90a-NEXT:    v_mov_b32_e32 v2, s14
; GFX90a-NEXT:    v_mov_b32_e32 v3, s15
; GFX90a-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7] offset:16
; GFX90a-NEXT:    s_nop 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s8
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    v_mov_b32_e32 v2, s10
; GFX90a-NEXT:    v_mov_b32_e32 v3, s11
; GFX90a-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <5 x double> %in, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @v8i8_preload_arg(ptr addrspace(1) inreg %out, <8 x i8> inreg %in) #0 {
; GFX942-LABEL: v8i8_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB15_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB15_0:
; GFX942-NEXT:    s_lshr_b32 s1, s5, 24
; GFX942-NEXT:    s_and_b32 s0, s5, 0xffff
; GFX942-NEXT:    s_lshl_b32 s1, s1, 8
; GFX942-NEXT:    s_bfe_u32 s5, s5, 0x80010
; GFX942-NEXT:    s_or_b32 s1, s5, s1
; GFX942-NEXT:    s_lshl_b32 s1, s1, 16
; GFX942-NEXT:    s_lshr_b32 s5, s4, 24
; GFX942-NEXT:    s_or_b32 s0, s0, s1
; GFX942-NEXT:    s_and_b32 s1, s4, 0xffff
; GFX942-NEXT:    s_lshl_b32 s5, s5, 8
; GFX942-NEXT:    s_bfe_u32 s4, s4, 0x80010
; GFX942-NEXT:    s_or_b32 s4, s4, s5
; GFX942-NEXT:    s_lshl_b32 s4, s4, 16
; GFX942-NEXT:    s_or_b32 s1, s1, s4
; GFX942-NEXT:    v_mov_b32_e32 v0, s1
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    v_mov_b32_e32 v2, 0
; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v8i8_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB15_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB15_0:
; GFX90a-NEXT:    s_lshr_b32 s1, s9, 24
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 8
; GFX90a-NEXT:    s_bfe_u32 s2, s9, 0x80010
; GFX90a-NEXT:    s_or_b32 s1, s2, s1
; GFX90a-NEXT:    s_lshr_b32 s2, s8, 24
; GFX90a-NEXT:    s_lshl_b32 s2, s2, 8
; GFX90a-NEXT:    s_bfe_u32 s3, s8, 0x80010
; GFX90a-NEXT:    s_and_b32 s0, s9, 0xffff
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 16
; GFX90a-NEXT:    s_or_b32 s2, s3, s2
; GFX90a-NEXT:    s_or_b32 s0, s0, s1
; GFX90a-NEXT:    s_and_b32 s1, s8, 0xffff
; GFX90a-NEXT:    s_lshl_b32 s2, s2, 16
; GFX90a-NEXT:    s_or_b32 s1, s1, s2
; GFX90a-NEXT:    v_mov_b32_e32 v0, s1
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    v_mov_b32_e32 v2, 0
; GFX90a-NEXT:    global_store_dwordx2 v2, v[0:1], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <8 x i8> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @i64_kernel_preload_arg(ptr addrspace(1) inreg %out, i64 inreg %a) #0 {
; GFX942-LABEL: i64_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB16_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB16_0:
; GFX942-NEXT:    v_mov_b32_e32 v2, 0
; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i64_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB16_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB16_0:
; GFX90a-NEXT:    v_mov_b32_e32 v2, 0
; GFX90a-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
; GFX90a-NEXT:    global_store_dwordx2 v2, v[0:1], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store i64 %a, ptr addrspace(1) %out, align 8
  ret void
}

define amdgpu_kernel void @f64_kernel_preload_arg(ptr addrspace(1) inreg %out, double inreg %in) #0 {
; GFX942-LABEL: f64_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB17_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB17_0:
; GFX942-NEXT:    v_mov_b32_e32 v2, 0
; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: f64_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB17_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB17_0:
; GFX90a-NEXT:    v_mov_b32_e32 v2, 0
; GFX90a-NEXT:    v_pk_mov_b32 v[0:1], s[8:9], s[8:9] op_sel:[0,1]
; GFX90a-NEXT:    global_store_dwordx2 v2, v[0:1], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store double %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @half_kernel_preload_arg(ptr addrspace(1) inreg %out, half inreg %in) #0 {
; GFX942-LABEL: half_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB18_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB18_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: half_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB18_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB18_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store half %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @bfloat_kernel_preload_arg(ptr addrspace(1) inreg %out, bfloat inreg %in) #0 {
; GFX942-LABEL: bfloat_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB19_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB19_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: bfloat_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB19_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB19_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store bfloat %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @v2bfloat_kernel_preload_arg(ptr addrspace(1) inreg %out, <2 x bfloat> inreg %in) #0 {
; GFX942-LABEL: v2bfloat_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB20_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB20_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v2bfloat_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB20_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB20_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <2 x bfloat> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @v3bfloat_kernel_preload_arg(ptr addrspace(1) inreg %out, <3 x bfloat> inreg %in) #0 {
; GFX942-LABEL: v3bfloat_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB21_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB21_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    global_store_short v0, v1, s[2:3] offset:4
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v3bfloat_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB21_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB21_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7] offset:4
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <3 x bfloat> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @v6bfloat_kernel_preload_arg(ptr addrspace(1) inreg %out, <6 x bfloat> inreg %in) #0 {
; GFX942-LABEL: v6bfloat_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x18
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB22_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB22_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v6bfloat_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x18
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB22_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB22_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <6 x bfloat> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @half_v7bfloat_kernel_preload_arg(ptr addrspace(1) inreg %out, half inreg %in, <7 x bfloat> inreg %in2, ptr addrspace(1) inreg %out2) #0 {
; GFX942-LABEL: half_v7bfloat_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB23_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB23_0:
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    v_mov_b32_e32 v0, s4
; GFX942-NEXT:    global_store_short v3, v0, s[2:3]
; GFX942-NEXT:    v_mov_b32_e32 v0, s9
; GFX942-NEXT:    global_store_short v3, v0, s[10:11] offset:12
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[10:11]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: half_v7bfloat_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB23_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB23_0:
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x20
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s8
; GFX90a-NEXT:    global_store_short v3, v0, s[6:7]
; GFX90a-NEXT:    v_mov_b32_e32 v0, s13
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    global_store_short v3, v0, s[0:1] offset:12
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[0:1]
; GFX90a-NEXT:    s_endpgm
  store half %in, ptr addrspace(1) %out
  store <7 x bfloat> %in2, ptr addrspace(1) %out2
  ret void
}

define amdgpu_kernel void @i1_kernel_preload_arg(ptr addrspace(1) inreg %out, i1 inreg %in) #0 {
; GFX942-LABEL: i1_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dword s4, s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB24_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB24_0:
; GFX942-NEXT:    s_and_b32 s0, s4, 1
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_byte v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i1_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dword s8, s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB24_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB24_0:
; GFX90a-NEXT:    s_and_b32 s0, s8, 1
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_byte v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store i1 %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @fp128_kernel_preload_arg(ptr addrspace(1) inreg %out, fp128 inreg %in) #0 {
; GFX942-LABEL: fp128_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x18
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB25_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB25_0:
; GFX942-NEXT:    v_mov_b32_e32 v4, 0
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    v_mov_b32_e32 v3, s9
; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: fp128_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x18
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB25_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB25_0:
; GFX90a-NEXT:    v_mov_b32_e32 v4, 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    v_mov_b32_e32 v3, s13
; GFX90a-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store fp128 %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @v7i8_kernel_preload_arg(ptr addrspace(1) inreg %out, <7 x i8> inreg %in) #0 {
; GFX942-LABEL: v7i8_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB26_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB26_0:
; GFX942-NEXT:    s_lshr_b32 s1, s4, 24
; GFX942-NEXT:    s_and_b32 s0, s4, 0xffff
; GFX942-NEXT:    s_lshl_b32 s1, s1, 8
; GFX942-NEXT:    s_bfe_u32 s4, s4, 0x80010
; GFX942-NEXT:    s_or_b32 s1, s4, s1
; GFX942-NEXT:    s_lshl_b32 s1, s1, 16
; GFX942-NEXT:    s_or_b32 s0, s0, s1
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    global_store_byte_d16_hi v0, v1, s[2:3] offset:6
; GFX942-NEXT:    global_store_short v0, v1, s[2:3] offset:4
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v7i8_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB26_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB26_0:
; GFX90a-NEXT:    s_lshr_b32 s1, s8, 24
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 8
; GFX90a-NEXT:    s_bfe_u32 s2, s8, 0x80010
; GFX90a-NEXT:    s_or_b32 s1, s2, s1
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xffff
; GFX90a-NEXT:    s_lshl_b32 s1, s1, 16
; GFX90a-NEXT:    s_or_b32 s0, s0, s1
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    global_store_byte_d16_hi v0, v1, s[6:7] offset:6
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7] offset:4
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <7 x i8> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @v7half_kernel_preload_arg(ptr addrspace(1) inreg %out, <7 x half> inreg %in) #0 {
; GFX942-LABEL: v7half_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[0:1], 0x18
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB27_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB27_0:
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    v_mov_b32_e32 v0, s9
; GFX942-NEXT:    global_store_short v3, v0, s[2:3] offset:12
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: v7half_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x18
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB27_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB27_0:
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    v_mov_b32_e32 v0, s13
; GFX90a-NEXT:    global_store_short v3, v0, s[6:7] offset:12
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[6:7]
; GFX90a-NEXT:    s_endpgm
  store <7 x half> %in, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @i16_i32_kernel_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %in, i32 inreg %in2, ptr addrspace(1) inreg %out2) #0 {
; GFX942-LABEL: i16_i32_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB28_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB28_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    v_mov_b32_e32 v1, s5
; GFX942-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i16_i32_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB28_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB28_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    v_mov_b32_e32 v1, s9
; GFX90a-NEXT:    global_store_dword v0, v1, s[10:11]
; GFX90a-NEXT:    s_endpgm
  store i16 %in, ptr addrspace(1) %out
  store i32 %in2, ptr addrspace(1) %out2
  ret void
}

define amdgpu_kernel void @i16_v3i32_kernel_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %in, <3 x i32> inreg %in2, ptr addrspace(1) inreg %out2) #0 {
; GFX942-LABEL: i16_v3i32_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx8 s[4:11], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB29_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB29_0:
; GFX942-NEXT:    v_mov_b32_e32 v3, 0
; GFX942-NEXT:    v_mov_b32_e32 v4, s4
; GFX942-NEXT:    v_mov_b32_e32 v0, s6
; GFX942-NEXT:    v_mov_b32_e32 v1, s7
; GFX942-NEXT:    v_mov_b32_e32 v2, s8
; GFX942-NEXT:    global_store_short v3, v4, s[2:3]
; GFX942-NEXT:    global_store_dwordx3 v3, v[0:2], s[10:11]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i16_v3i32_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB29_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB29_0:
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x20
; GFX90a-NEXT:    v_mov_b32_e32 v3, 0
; GFX90a-NEXT:    v_mov_b32_e32 v4, s8
; GFX90a-NEXT:    v_mov_b32_e32 v0, s10
; GFX90a-NEXT:    v_mov_b32_e32 v1, s11
; GFX90a-NEXT:    v_mov_b32_e32 v2, s12
; GFX90a-NEXT:    global_store_short v3, v4, s[6:7]
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    global_store_dwordx3 v3, v[0:2], s[0:1]
; GFX90a-NEXT:    s_endpgm
  store i16 %in, ptr addrspace(1) %out
  store <3 x i32> %in2, ptr addrspace(1) %out2
  ret void
}

define amdgpu_kernel void @i16_i16_kernel_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %in, i16 inreg %in2, ptr addrspace(1) inreg %out2) #0 {
; GFX942-LABEL: i16_i16_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB30_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB30_0:
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    global_store_short_d16_hi v0, v1, s[6:7]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i16_i16_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB30_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB30_0:
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    global_store_short_d16_hi v0, v1, s[10:11]
; GFX90a-NEXT:    s_endpgm
  store i16 %in, ptr addrspace(1) %out
  store i16 %in2, ptr addrspace(1) %out2
  ret void
}

define amdgpu_kernel void @i16_v2i8_kernel_preload_arg(ptr addrspace(1) inreg %out, i16 inreg %in, <2 x i8> inreg %in2, ptr addrspace(1) inreg %out2) #0 {
; GFX942-LABEL: i16_v2i8_kernel_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB31_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB31_0:
; GFX942-NEXT:    s_lshr_b32 s0, s4, 24
; GFX942-NEXT:    s_lshl_b32 s0, s0, 8
; GFX942-NEXT:    s_bfe_u32 s1, s4, 0x80010
; GFX942-NEXT:    s_or_b32 s0, s1, s0
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s4
; GFX942-NEXT:    global_store_short v0, v1, s[2:3]
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_short v0, v1, s[6:7]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i16_v2i8_kernel_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB31_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB31_0:
; GFX90a-NEXT:    s_lshr_b32 s0, s8, 24
; GFX90a-NEXT:    s_lshl_b32 s0, s0, 8
; GFX90a-NEXT:    s_bfe_u32 s1, s8, 0x80010
; GFX90a-NEXT:    s_or_b32 s0, s1, s0
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s8
; GFX90a-NEXT:    global_store_short v0, v1, s[6:7]
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_short v0, v1, s[10:11]
; GFX90a-NEXT:    s_endpgm
  store i16 %in, ptr addrspace(1) %out
  store <2 x i8> %in2, ptr addrspace(1) %out2
  ret void
}

; The second argument is not expected to be preloaded with the current behavior.

define amdgpu_kernel void @i32_ptr1_i32_staggered_preload_arg(i32 inreg %arg0, ptr addrspace(1) %out, i32 inreg %arg1) #0 {
; GFX942-LABEL: i32_ptr1_i32_staggered_preload_arg:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dword s2, s[0:1], 0x0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB32_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB32_0:
; GFX942-NEXT:    s_load_dword s3, s[0:1], 0x10
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_add_i32 s0, s2, s3
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[4:5]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: i32_ptr1_i32_staggered_preload_arg:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dword s6, s[4:5], 0x0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB32_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB32_0:
; GFX90a-NEXT:    s_load_dword s2, s[4:5], 0x10
; GFX90a-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x8
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_add_i32 s2, s6, s2
; GFX90a-NEXT:    v_mov_b32_e32 v1, s2
; GFX90a-NEXT:    global_store_dword v0, v1, s[0:1]
; GFX90a-NEXT:    s_endpgm
  %add = add i32 %arg0, %arg1
  store i32 %add, ptr addrspace(1) %out
  ret void
}

define amdgpu_kernel void @ptr1_i8_trailing_unused(ptr addrspace(1) inreg %out, i8 inreg %arg0, i32 inreg %unused) #0 {
; GFX942-LABEL: ptr1_i8_trailing_unused:
; GFX942:       ; %bb.1:
; GFX942-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
; GFX942-NEXT:    s_load_dwordx2 s[4:5], s[0:1], 0x8
; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
; GFX942-NEXT:    s_branch .LBB33_0
; GFX942-NEXT:    .p2align 8
; GFX942-NEXT:  ; %bb.2:
; GFX942-NEXT:  .LBB33_0:
; GFX942-NEXT:    s_and_b32 s0, s4, 0xff
; GFX942-NEXT:    v_mov_b32_e32 v0, 0
; GFX942-NEXT:    v_mov_b32_e32 v1, s0
; GFX942-NEXT:    global_store_dword v0, v1, s[2:3]
; GFX942-NEXT:    s_endpgm
;
; GFX90a-LABEL: ptr1_i8_trailing_unused:
; GFX90a:       ; %bb.1:
; GFX90a-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x0
; GFX90a-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x8
; GFX90a-NEXT:    s_waitcnt lgkmcnt(0)
; GFX90a-NEXT:    s_branch .LBB33_0
; GFX90a-NEXT:    .p2align 8
; GFX90a-NEXT:  ; %bb.2:
; GFX90a-NEXT:  .LBB33_0:
; GFX90a-NEXT:    s_and_b32 s0, s8, 0xff
; GFX90a-NEXT:    v_mov_b32_e32 v0, 0
; GFX90a-NEXT:    v_mov_b32_e32 v1, s0
; GFX90a-NEXT:    global_store_dword v0, v1, s[6:7]
; GFX90a-NEXT:    s_endpgm
  %ext = zext i8 %arg0 to i32
  store i32 %ext, ptr addrspace(1) %out
  ret void
}

attributes #0 = { "amdgpu-no-agpr" "amdgpu-no-completion-action" "amdgpu-no-default-queue" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "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" "uniform-work-group-size"="false" }