1 ; RUN: llc -march=amdgcn -mcpu=verde -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
2 ; RUN: llc -march=amdgcn -mcpu=tonga -mattr=-flat-for-global -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
3 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=EG -check-prefix=FUNC %s
5 ; FUNC-LABEL: {{^}}udiv_i32:
8 define void @udiv_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
9 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
10 %a = load i32, i32 addrspace(1) * %in
11 %b = load i32, i32 addrspace(1) * %b_ptr
12 %result = udiv i32 %a, %b
13 store i32 %result, i32 addrspace(1)* %out
17 ; FUNC-LABEL: {{^}}s_udiv_i32:
19 define void @s_udiv_i32(i32 addrspace(1)* %out, i32 %a, i32 %b) {
20 %result = udiv i32 %a, %b
21 store i32 %result, i32 addrspace(1)* %out
26 ; The code generated by udiv is long and complex and may frequently
27 ; change. The goal of this test is to make sure the ISel doesn't fail
28 ; when it gets a v4i32 udiv
30 ; FUNC-LABEL: {{^}}udiv_v2i32:
34 define void @udiv_v2i32(<2 x i32> addrspace(1)* %out, <2 x i32> addrspace(1)* %in) {
35 %b_ptr = getelementptr <2 x i32>, <2 x i32> addrspace(1)* %in, i32 1
36 %a = load <2 x i32>, <2 x i32> addrspace(1) * %in
37 %b = load <2 x i32>, <2 x i32> addrspace(1) * %b_ptr
38 %result = udiv <2 x i32> %a, %b
39 store <2 x i32> %result, <2 x i32> addrspace(1)* %out
43 ; FUNC-LABEL: {{^}}udiv_v4i32:
46 define void @udiv_v4i32(<4 x i32> addrspace(1)* %out, <4 x i32> addrspace(1)* %in) {
47 %b_ptr = getelementptr <4 x i32>, <4 x i32> addrspace(1)* %in, i32 1
48 %a = load <4 x i32>, <4 x i32> addrspace(1) * %in
49 %b = load <4 x i32>, <4 x i32> addrspace(1) * %b_ptr
50 %result = udiv <4 x i32> %a, %b
51 store <4 x i32> %result, <4 x i32> addrspace(1)* %out
55 ; FUNC-LABEL: {{^}}udiv_i32_div_pow2:
56 ; SI: buffer_load_dword [[VAL:v[0-9]+]]
57 ; SI: v_lshrrev_b32_e32 [[RESULT:v[0-9]+]], 4, [[VAL]]
58 ; SI: buffer_store_dword [[RESULT]]
59 define void @udiv_i32_div_pow2(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
60 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
61 %a = load i32, i32 addrspace(1)* %in
62 %result = udiv i32 %a, 16
63 store i32 %result, i32 addrspace(1)* %out
67 ; FUNC-LABEL: {{^}}udiv_i32_div_k_even:
68 ; SI-DAG: buffer_load_dword [[VAL:v[0-9]+]]
69 ; SI-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0xfabbd9c1
70 ; SI: v_mul_hi_u32 [[MULHI:v[0-9]+]], [[K]], [[VAL]]
71 ; SI: v_lshrrev_b32_e32 [[RESULT:v[0-9]+]], 25, [[MULHI]]
72 ; SI: buffer_store_dword [[RESULT]]
73 define void @udiv_i32_div_k_even(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
74 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
75 %a = load i32, i32 addrspace(1)* %in
76 %result = udiv i32 %a, 34259182
77 store i32 %result, i32 addrspace(1)* %out
81 ; FUNC-LABEL: {{^}}udiv_i32_div_k_odd:
82 ; SI-DAG: buffer_load_dword [[VAL:v[0-9]+]]
83 ; SI-DAG: v_mov_b32_e32 [[K:v[0-9]+]], 0x7d5deca3
84 ; SI: v_mul_hi_u32 [[MULHI:v[0-9]+]], [[K]], [[VAL]]
85 ; SI: v_lshrrev_b32_e32 [[RESULT:v[0-9]+]], 24, [[MULHI]]
86 ; SI: buffer_store_dword [[RESULT]]
87 define void @udiv_i32_div_k_odd(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
88 %b_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
89 %a = load i32, i32 addrspace(1)* %in
90 %result = udiv i32 %a, 34259183
91 store i32 %result, i32 addrspace(1)* %out
95 ; FUNC-LABEL: {{^}}v_udiv_i8:
97 ; SI: v_and_b32_e32 [[TRUNC:v[0-9]+]], 0xff, v{{[0-9]+}}
98 ; SI: buffer_store_dword [[TRUNC]]
99 define void @v_udiv_i8(i32 addrspace(1)* %out, i8 addrspace(1)* %in) {
100 %den_ptr = getelementptr i8, i8 addrspace(1)* %in, i8 1
101 %num = load i8, i8 addrspace(1) * %in
102 %den = load i8, i8 addrspace(1) * %den_ptr
103 %result = udiv i8 %num, %den
104 %result.ext = zext i8 %result to i32
105 store i32 %result.ext, i32 addrspace(1)* %out
109 ; FUNC-LABEL: {{^}}v_udiv_i16:
111 ; SI: v_and_b32_e32 [[TRUNC:v[0-9]+]], 0xffff, v{{[0-9]+}}
112 ; SI: buffer_store_dword [[TRUNC]]
113 define void @v_udiv_i16(i32 addrspace(1)* %out, i16 addrspace(1)* %in) {
114 %den_ptr = getelementptr i16, i16 addrspace(1)* %in, i16 1
115 %num = load i16, i16 addrspace(1) * %in
116 %den = load i16, i16 addrspace(1) * %den_ptr
117 %result = udiv i16 %num, %den
118 %result.ext = zext i16 %result to i32
119 store i32 %result.ext, i32 addrspace(1)* %out
123 ; FUNC-LABEL: {{^}}v_udiv_i23:
125 ; SI: v_and_b32_e32 [[TRUNC:v[0-9]+]], 0x7fffff, v{{[0-9]+}}
126 ; SI: buffer_store_dword [[TRUNC]]
127 define void @v_udiv_i23(i32 addrspace(1)* %out, i23 addrspace(1)* %in) {
128 %den_ptr = getelementptr i23, i23 addrspace(1)* %in, i23 1
129 %num = load i23, i23 addrspace(1) * %in
130 %den = load i23, i23 addrspace(1) * %den_ptr
131 %result = udiv i23 %num, %den
132 %result.ext = zext i23 %result to i32
133 store i32 %result.ext, i32 addrspace(1)* %out
137 ; FUNC-LABEL: {{^}}v_udiv_i24:
139 define void @v_udiv_i24(i32 addrspace(1)* %out, i24 addrspace(1)* %in) {
140 %den_ptr = getelementptr i24, i24 addrspace(1)* %in, i24 1
141 %num = load i24, i24 addrspace(1) * %in
142 %den = load i24, i24 addrspace(1) * %den_ptr
143 %result = udiv i24 %num, %den
144 %result.ext = zext i24 %result to i32
145 store i32 %result.ext, i32 addrspace(1)* %out
149 ; FUNC-LABEL: @scalarize_mulhu_4xi32
155 define void @scalarize_mulhu_4xi32(<4 x i32> addrspace(1)* nocapture readonly %in, <4 x i32> addrspace(1)* nocapture %out) {
156 %1 = load <4 x i32>, <4 x i32> addrspace(1)* %in, align 16
157 %2 = udiv <4 x i32> %1, <i32 53668, i32 53668, i32 53668, i32 53668>
158 store <4 x i32> %2, <4 x i32> addrspace(1)* %out, align 16