| 1 | // RUN: %clang_cc1 -emit-llvm %s -o - | FileCheck %s |
| 2 | |
| 3 | typedef __attribute__(( ext_vector_type(4) )) float float4; |
| 4 | typedef __attribute__(( ext_vector_type(2) )) float float2; |
| 5 | typedef __attribute__(( ext_vector_type(4) )) int int4; |
| 6 | typedef __attribute__(( ext_vector_type(4) )) unsigned int uint4; |
| 7 | |
| 8 | // CHECK: @foo = {{(dso_local )?}}global <4 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00, float 4.000000e+00> |
| 9 | float4 foo = (float4){ 1.0, 2.0, 3.0, 4.0 }; |
| 10 | |
| 11 | // CHECK: @bar = {{(dso_local )?}}constant <4 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00, float 0x7FF0000000000000> |
| 12 | const float4 bar = (float4){ 1.0, 2.0, 3.0, __builtin_inff() }; |
| 13 | |
| 14 | // CHECK: @test1 |
| 15 | // CHECK: fadd <4 x float> |
| 16 | float4 test1(float4 V) { |
| 17 | return V.wzyx+V; |
| 18 | } |
| 19 | |
| 20 | float2 vec2, vec2_2; |
| 21 | float4 vec4, vec4_2; |
| 22 | float f; |
| 23 | |
| 24 | // CHECK: @test2 |
| 25 | // CHECK: shufflevector {{.*}} <i32 0, i32 1> |
| 26 | // CHECK: extractelement |
| 27 | // CHECK: shufflevector {{.*}} <i32 1, i32 1, i32 1, i32 1> |
| 28 | // CHECK: insertelement |
| 29 | // CHECK: shufflevector {{.*}} <i32 1, i32 0> |
| 30 | void test2() { |
| 31 | vec2 = vec4.xy; // shorten |
| 32 | f = vec2.x; // extract elt |
| 33 | vec4 = vec4.yyyy; // splat |
| 34 | |
| 35 | vec2.x = f; // insert one. |
| 36 | vec2.yx = vec2; // reverse |
| 37 | } |
| 38 | |
| 39 | // CHECK: @test3 |
| 40 | // CHECK: store <4 x float> <float 1.000000e+00, float 2.000000e+00, float 3.000000e+00, float 4.000000e+00> |
| 41 | void test3(float4 *out) { |
| 42 | *out = ((float4) {1.0f, 2.0f, 3.0f, 4.0f }); |
| 43 | } |
| 44 | |
| 45 | // CHECK: @test4 |
| 46 | // CHECK: store <4 x float> |
| 47 | // CHECK: store <4 x float> |
| 48 | void test4(float4 *out) { |
| 49 | float a = 1.0f; |
| 50 | float b = 2.0f; |
| 51 | float c = 3.0f; |
| 52 | float d = 4.0f; |
| 53 | *out = ((float4) {a,b,c,d}); |
| 54 | } |
| 55 | |
| 56 | // CHECK: @test5 |
| 57 | // CHECK: shufflevector {{.*}} <4 x i32> zeroinitializer |
| 58 | // CHECK: fmul <4 x float> |
| 59 | // CHECK: fmul <4 x float> |
| 60 | // CHECK: shufflevector {{.*}} <4 x i32> zeroinitializer |
| 61 | // CHECK: fmul <4 x float> |
| 62 | void test5(float4 *out) { |
| 63 | float a; |
| 64 | float4 b; |
| 65 | |
| 66 | a = 1.0f; |
| 67 | b = a; |
| 68 | b = b * 5.0f; |
| 69 | b = 5.0f * b; |
| 70 | b *= a; |
| 71 | |
| 72 | *out = b; |
| 73 | } |
| 74 | |
| 75 | // CHECK: @test6 |
| 76 | void test6(float4 *ap, float4 *bp, float c) { |
| 77 | float4 a = *ap; |
| 78 | float4 b = *bp; |
| 79 | |
| 80 | // CHECK: fadd <4 x float> |
| 81 | // CHECK: fsub <4 x float> |
| 82 | // CHECK: fmul <4 x float> |
| 83 | // CHECK: fdiv <4 x float> |
| 84 | a = a + b; |
| 85 | a = a - b; |
| 86 | a = a * b; |
| 87 | a = a / b; |
| 88 | |
| 89 | // CHECK: fadd <4 x float> |
| 90 | // CHECK: fsub <4 x float> |
| 91 | // CHECK: fmul <4 x float> |
| 92 | // CHECK: fdiv <4 x float> |
| 93 | a = a + c; |
| 94 | a = a - c; |
| 95 | a = a * c; |
| 96 | a = a / c; |
| 97 | |
| 98 | // CHECK: fadd <4 x float> |
| 99 | // CHECK: fsub <4 x float> |
| 100 | // CHECK: fmul <4 x float> |
| 101 | // CHECK: fdiv <4 x float> |
| 102 | a += b; |
| 103 | a -= b; |
| 104 | a *= b; |
| 105 | a /= b; |
| 106 | |
| 107 | // CHECK: fadd <4 x float> |
| 108 | // CHECK: fsub <4 x float> |
| 109 | // CHECK: fmul <4 x float> |
| 110 | // CHECK: fdiv <4 x float> |
| 111 | a += c; |
| 112 | a -= c; |
| 113 | a *= c; |
| 114 | a /= c; |
| 115 | |
| 116 | // Vector comparisons can sometimes crash the x86 backend: rdar://6326239, |
| 117 | // reject them until the implementation is stable. |
| 118 | #if 0 |
| 119 | int4 cmp; |
| 120 | cmp = a < b; |
| 121 | cmp = a <= b; |
| 122 | cmp = a < b; |
| 123 | cmp = a >= b; |
| 124 | cmp = a == b; |
| 125 | cmp = a != b; |
| 126 | #endif |
| 127 | } |
| 128 | |
| 129 | // CHECK: @test7 |
| 130 | void test7(int4 *ap, int4 *bp, int c) { |
| 131 | int4 a = *ap; |
| 132 | int4 b = *bp; |
| 133 | |
| 134 | // CHECK: add <4 x i32> |
| 135 | // CHECK: sub <4 x i32> |
| 136 | // CHECK: mul <4 x i32> |
| 137 | // CHECK: sdiv <4 x i32> |
| 138 | // CHECK: srem <4 x i32> |
| 139 | a = a + b; |
| 140 | a = a - b; |
| 141 | a = a * b; |
| 142 | a = a / b; |
| 143 | a = a % b; |
| 144 | |
| 145 | // CHECK: add <4 x i32> |
| 146 | // CHECK: sub <4 x i32> |
| 147 | // CHECK: mul <4 x i32> |
| 148 | // CHECK: sdiv <4 x i32> |
| 149 | // CHECK: srem <4 x i32> |
| 150 | a = a + c; |
| 151 | a = a - c; |
| 152 | a = a * c; |
| 153 | a = a / c; |
| 154 | a = a % c; |
| 155 | |
| 156 | // CHECK: add <4 x i32> |
| 157 | // CHECK: sub <4 x i32> |
| 158 | // CHECK: mul <4 x i32> |
| 159 | // CHECK: sdiv <4 x i32> |
| 160 | // CHECK: srem <4 x i32> |
| 161 | a += b; |
| 162 | a -= b; |
| 163 | a *= b; |
| 164 | a /= b; |
| 165 | a %= b; |
| 166 | |
| 167 | // CHECK: add <4 x i32> |
| 168 | // CHECK: sub <4 x i32> |
| 169 | // CHECK: mul <4 x i32> |
| 170 | // CHECK: sdiv <4 x i32> |
| 171 | // CHECK: srem <4 x i32> |
| 172 | a += c; |
| 173 | a -= c; |
| 174 | a *= c; |
| 175 | a /= c; |
| 176 | a %= c; |
| 177 | |
| 178 | |
| 179 | // Vector comparisons. |
| 180 | // CHECK: icmp slt |
| 181 | // CHECK: icmp sle |
| 182 | // CHECK: icmp sgt |
| 183 | // CHECK: icmp sge |
| 184 | // CHECK: icmp eq |
| 185 | // CHECK: icmp ne |
| 186 | int4 cmp; |
| 187 | cmp = a < b; |
| 188 | cmp = a <= b; |
| 189 | cmp = a > b; |
| 190 | cmp = a >= b; |
| 191 | cmp = a == b; |
| 192 | cmp = a != b; |
| 193 | } |
| 194 | |
| 195 | // CHECK: @test8 |
| 196 | void test8(float4 *ap, float4 *bp, int c) { |
| 197 | float4 a = *ap; |
| 198 | float4 b = *bp; |
| 199 | |
| 200 | // Vector comparisons. |
| 201 | // CHECK: fcmp olt |
| 202 | // CHECK: fcmp ole |
| 203 | // CHECK: fcmp ogt |
| 204 | // CHECK: fcmp oge |
| 205 | // CHECK: fcmp oeq |
| 206 | // CHECK: fcmp une |
| 207 | int4 cmp; |
| 208 | cmp = a < b; |
| 209 | cmp = a <= b; |
| 210 | cmp = a > b; |
| 211 | cmp = a >= b; |
| 212 | cmp = a == b; |
| 213 | cmp = a != b; |
| 214 | } |
| 215 | |
| 216 | // CHECK: @test9 |
| 217 | // CHECK: extractelement <4 x i32> |
| 218 | int test9(int4 V) { |
| 219 | return V.xy.x; |
| 220 | } |
| 221 | |
| 222 | // CHECK: @test10 |
| 223 | // CHECK: add <4 x i32> |
| 224 | // CHECK: extractelement <4 x i32> |
| 225 | int test10(int4 V) { |
| 226 | return (V+V).x; |
| 227 | } |
| 228 | |
| 229 | // CHECK: @test11 |
| 230 | // CHECK: extractelement <4 x i32> |
| 231 | int4 test11a(); |
| 232 | int test11() { |
| 233 | return test11a().x; |
| 234 | } |
| 235 | |
| 236 | // CHECK: @test12 |
| 237 | // CHECK: shufflevector {{.*}} <i32 2, i32 1, i32 0> |
| 238 | // CHECK: shufflevector {{.*}} <i32 0, i32 1, i32 2, i32 undef> |
| 239 | // CHECK: shufflevector {{.*}} <i32 4, i32 5, i32 6, i32 3> |
| 240 | int4 test12(int4 V) { |
| 241 | V.xyz = V.zyx; |
| 242 | return V; |
| 243 | } |
| 244 | |
| 245 | // CHECK: @test13 |
| 246 | // CHECK: shufflevector {{.*}} <i32 2, i32 1, i32 0, i32 3> |
| 247 | int4 test13(int4 *V) { |
| 248 | return V->zyxw; |
| 249 | } |
| 250 | |
| 251 | // CHECK: @test14 |
| 252 | void test14(uint4 *ap, uint4 *bp, unsigned c) { |
| 253 | uint4 a = *ap; |
| 254 | uint4 b = *bp; |
| 255 | int4 d; |
| 256 | |
| 257 | // CHECK: udiv <4 x i32> |
| 258 | // CHECK: urem <4 x i32> |
| 259 | a = a / b; |
| 260 | a = a % b; |
| 261 | |
| 262 | // CHECK: udiv <4 x i32> |
| 263 | // CHECK: urem <4 x i32> |
| 264 | a = a / c; |
| 265 | a = a % c; |
| 266 | |
| 267 | // CHECK: icmp ult |
| 268 | // CHECK: icmp ule |
| 269 | // CHECK: icmp ugt |
| 270 | // CHECK: icmp uge |
| 271 | // CHECK: icmp eq |
| 272 | // CHECK: icmp ne |
| 273 | d = a < b; |
| 274 | d = a <= b; |
| 275 | d = a > b; |
| 276 | d = a >= b; |
| 277 | d = a == b; |
| 278 | d = a != b; |
| 279 | } |
| 280 | |
| 281 | // CHECK: @test15 |
| 282 | int4 test15(uint4 V0) { |
| 283 | // CHECK: icmp eq <4 x i32> |
| 284 | int4 V = !V0; |
| 285 | V = V && V; |
| 286 | V = V || V; |
| 287 | return V; |
| 288 | } |
| 289 | |
| 290 | // CHECK: @test16 |
| 291 | void test16(float2 a, float2 b) { |
| 292 | float2 t0 = (a + b) / 2; |
| 293 | } |
| 294 | |
| 295 | typedef char char16 __attribute__((ext_vector_type(16))); |
| 296 | |
| 297 | // CHECK: @test17 |
| 298 | void test17(void) { |
| 299 | char16 valA; |
| 300 | char valB; |
| 301 | char valC; |
| 302 | char16 destVal = valC ? valA : valB; |
| 303 | } |
| 304 | |
| 305 | typedef __attribute__(( ext_vector_type(16) )) float float16; |
| 306 | |
| 307 | float16 vec16, vec16_2; |
| 308 | |
| 309 | // CHECK: @test_rgba |
| 310 | void test_rgba() { |
| 311 | // CHECK: fadd <4 x float> |
| 312 | vec4_2 = vec4.abgr + vec4; |
| 313 | |
| 314 | // CHECK: shufflevector {{.*}} <i32 0, i32 1> |
| 315 | vec2 = vec4.rg; |
| 316 | // CHECK: shufflevector {{.*}} <i32 2, i32 3> |
| 317 | vec2_2 = vec4.ba; |
| 318 | // CHECK: extractelement {{.*}} 2 |
| 319 | f = vec4.b; |
| 320 | // CHECK: shufflevector {{.*}} <i32 2, i32 2, i32 2, i32 2> |
| 321 | vec4_2 = vec4_2.bbbb; |
| 322 | |
| 323 | // CHECK: insertelement {{.*}} 0 |
| 324 | vec2.r = f; |
| 325 | // CHECK: shufflevector {{.*}} <i32 1, i32 0> |
| 326 | vec2.gr = vec2; |
| 327 | |
| 328 | // CHECK: extractelement {{.*}} 0 |
| 329 | f = vec4_2.rg.r; |
| 330 | // CHECK: shufflevector {{.*}} <i32 2, i32 1, i32 0> |
| 331 | // CHECK: shufflevector {{.*}} <i32 0, i32 1, i32 2, i32 undef> |
| 332 | // CHECK: shufflevector {{.*}} <i32 4, i32 5, i32 6, i32 3> |
| 333 | vec4.rgb = vec4.bgr; |
| 334 | |
| 335 | // CHECK: extractelement {{.*}} 11 |
| 336 | // CHECK: insertelement {{.*}} 2 |
| 337 | vec4.b = vec16.sb; |
| 338 | // CHECK: shufflevector {{.*}} <i32 10, i32 11, i32 12, i32 13> |
| 339 | vec4_2 = vec16.sabcd; |
| 340 | } |
| 341 | |