; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -mcpu=verde < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=amdgcn -mcpu=tonga -mattr=-flat-for-global < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
; RUN:  llc -amdgpu-scalarize-global-loads=false  -mtriple=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s

; The code generated by urem is long and complex and may frequently
; change.  The goal of this test is to make sure the ISel doesn't fail
; when it gets a v2i32/v4i32 urem

; FUNC-LABEL: {{^}}test_urem_i32:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr i32, ptr addrspace(1) %in, i32 1
  %a = load i32, ptr addrspace(1) %in
  %b = load i32, ptr addrspace(1) %b_ptr
  %result = urem i32 %a, %b
  store i32 %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_i32_7:
; SI: s_mov_b32 [[MAGIC:s[0-9]+]], 0x24924925
; SI: v_mul_hi_u32 {{v[0-9]+}}, {{v[0-9]+}}, [[MAGIC]]
; SI: v_sub_{{[iu]}}32
; SI: v_mul_lo_u32
; SI: v_subrev_{{[iu]}}32
; SI: buffer_store_dword
; SI: s_endpgm
define amdgpu_kernel void @test_urem_i32_7(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %num = load i32, ptr addrspace(1) %in
  %result = urem i32 %num, 7
  store i32 %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_v2i32:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_v2i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr <2 x i32>, ptr addrspace(1) %in, i32 1
  %a = load <2 x i32>, ptr addrspace(1) %in
  %b = load <2 x i32>, ptr addrspace(1) %b_ptr
  %result = urem <2 x i32> %a, %b
  store <2 x i32> %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_v4i32:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_v4i32(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr <4 x i32>, ptr addrspace(1) %in, i32 1
  %a = load <4 x i32>, ptr addrspace(1) %in
  %b = load <4 x i32>, ptr addrspace(1) %b_ptr
  %result = urem <4 x i32> %a, %b
  store <4 x i32> %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_i64:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr i64, ptr addrspace(1) %in, i64 1
  %a = load i64, ptr addrspace(1) %in
  %b = load i64, ptr addrspace(1) %b_ptr
  %result = urem i64 %a, %b
  store i64 %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_v2i64:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_v2i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr <2 x i64>, ptr addrspace(1) %in, i64 1
  %a = load <2 x i64>, ptr addrspace(1) %in
  %b = load <2 x i64>, ptr addrspace(1) %b_ptr
  %result = urem <2 x i64> %a, %b
  store <2 x i64> %result, ptr addrspace(1) %out
  ret void
}

; FUNC-LABEL: {{^}}test_urem_v4i64:
; SI: s_endpgm
; EG: CF_END
define amdgpu_kernel void @test_urem_v4i64(ptr addrspace(1) %out, ptr addrspace(1) %in) {
  %b_ptr = getelementptr <4 x i64>, ptr addrspace(1) %in, i64 1
  %a = load <4 x i64>, ptr addrspace(1) %in
  %b = load <4 x i64>, ptr addrspace(1) %b_ptr
  %result = urem <4 x i64> %a, %b
  store <4 x i64> %result, ptr addrspace(1) %out
  ret void
}