| 1 | // RUN: %clang_cc1 -triple riscv64 -emit-llvm %s -o - | FileCheck %s |
| 2 | |
| 3 | #include <stddef.h> |
| 4 | #include <stdint.h> |
| 5 | |
| 6 | // CHECK-LABEL: define void @f_void() |
| 7 | void f_void(void) {} |
| 8 | |
| 9 | // Scalar arguments and return values smaller than the word size are extended |
| 10 | // according to the sign of their type, up to 32 bits |
| 11 | |
| 12 | // CHECK-LABEL: define zeroext i1 @f_scalar_0(i1 zeroext %x) |
| 13 | _Bool f_scalar_0(_Bool x) { return x; } |
| 14 | |
| 15 | // CHECK-LABEL: define signext i8 @f_scalar_1(i8 signext %x) |
| 16 | int8_t f_scalar_1(int8_t x) { return x; } |
| 17 | |
| 18 | // CHECK-LABEL: define zeroext i8 @f_scalar_2(i8 zeroext %x) |
| 19 | uint8_t f_scalar_2(uint8_t x) { return x; } |
| 20 | |
| 21 | // CHECK-LABEL: define signext i32 @f_scalar_3(i32 signext %x) |
| 22 | uint32_t f_scalar_3(int32_t x) { return x; } |
| 23 | |
| 24 | // CHECK-LABEL: define i64 @f_scalar_4(i64 %x) |
| 25 | int64_t f_scalar_4(int64_t x) { return x; } |
| 26 | |
| 27 | // CHECK-LABEL: define float @f_fp_scalar_1(float %x) |
| 28 | float f_fp_scalar_1(float x) { return x; } |
| 29 | |
| 30 | // CHECK-LABEL: define double @f_fp_scalar_2(double %x) |
| 31 | double f_fp_scalar_2(double x) { return x; } |
| 32 | |
| 33 | // CHECK-LABEL: define fp128 @f_fp_scalar_3(fp128 %x) |
| 34 | long double f_fp_scalar_3(long double x) { return x; } |
| 35 | |
| 36 | // Empty structs or unions are ignored. |
| 37 | |
| 38 | struct empty_s {}; |
| 39 | |
| 40 | // CHECK-LABEL: define void @f_agg_empty_struct() |
| 41 | struct empty_s f_agg_empty_struct(struct empty_s x) { |
| 42 | return x; |
| 43 | } |
| 44 | |
| 45 | union empty_u {}; |
| 46 | |
| 47 | // CHECK-LABEL: define void @f_agg_empty_union() |
| 48 | union empty_u f_agg_empty_union(union empty_u x) { |
| 49 | return x; |
| 50 | } |
| 51 | |
| 52 | // Aggregates <= 2*xlen may be passed in registers, so will be coerced to |
| 53 | // integer arguments. The rules for return are the same. |
| 54 | |
| 55 | struct tiny { |
| 56 | uint16_t a, b, c, d; |
| 57 | }; |
| 58 | |
| 59 | // CHECK-LABEL: define void @f_agg_tiny(i64 %x.coerce) |
| 60 | void f_agg_tiny(struct tiny x) { |
| 61 | x.a += x.b; |
| 62 | x.c += x.d; |
| 63 | } |
| 64 | |
| 65 | // CHECK-LABEL: define i64 @f_agg_tiny_ret() |
| 66 | struct tiny f_agg_tiny_ret() { |
| 67 | return (struct tiny){1, 2, 3, 4}; |
| 68 | } |
| 69 | |
| 70 | typedef uint16_t v4i16 __attribute__((vector_size(8))); |
| 71 | typedef int64_t v1i64 __attribute__((vector_size(8))); |
| 72 | |
| 73 | // CHECK-LABEL: define void @f_vec_tiny_v4i16(i64 %x.coerce) |
| 74 | void f_vec_tiny_v4i16(v4i16 x) { |
| 75 | x[0] = x[1]; |
| 76 | x[2] = x[3]; |
| 77 | } |
| 78 | |
| 79 | // CHECK-LABEL: define i64 @f_vec_tiny_v4i16_ret() |
| 80 | v4i16 f_vec_tiny_v4i16_ret() { |
| 81 | return (v4i16){1, 2, 3, 4}; |
| 82 | } |
| 83 | |
| 84 | // CHECK-LABEL: define void @f_vec_tiny_v1i64(i64 %x.coerce) |
| 85 | void f_vec_tiny_v1i64(v1i64 x) { |
| 86 | x[0] = 114; |
| 87 | } |
| 88 | |
| 89 | // CHECK-LABEL: define i64 @f_vec_tiny_v1i64_ret() |
| 90 | v1i64 f_vec_tiny_v1i64_ret() { |
| 91 | return (v1i64){1}; |
| 92 | } |
| 93 | |
| 94 | struct small { |
| 95 | int64_t a, *b; |
| 96 | }; |
| 97 | |
| 98 | // CHECK-LABEL: define void @f_agg_small([2 x i64] %x.coerce) |
| 99 | void f_agg_small(struct small x) { |
| 100 | x.a += *x.b; |
| 101 | x.b = &x.a; |
| 102 | } |
| 103 | |
| 104 | // CHECK-LABEL: define [2 x i64] @f_agg_small_ret() |
| 105 | struct small f_agg_small_ret() { |
| 106 | return (struct small){1, 0}; |
| 107 | } |
| 108 | |
| 109 | typedef uint16_t v8i16 __attribute__((vector_size(16))); |
| 110 | typedef __int128_t v1i128 __attribute__((vector_size(16))); |
| 111 | |
| 112 | // CHECK-LABEL: define void @f_vec_small_v8i16(i128 %x.coerce) |
| 113 | void f_vec_small_v8i16(v8i16 x) { |
| 114 | x[0] = x[7]; |
| 115 | } |
| 116 | |
| 117 | // CHECK-LABEL: define i128 @f_vec_small_v8i16_ret() |
| 118 | v8i16 f_vec_small_v8i16_ret() { |
| 119 | return (v8i16){1, 2, 3, 4, 5, 6, 7, 8}; |
| 120 | } |
| 121 | |
| 122 | // CHECK-LABEL: define void @f_vec_small_v1i128(i128 %x.coerce) |
| 123 | void f_vec_small_v1i128(v1i128 x) { |
| 124 | x[0] = 114; |
| 125 | } |
| 126 | |
| 127 | // CHECK-LABEL: define i128 @f_vec_small_v1i128_ret() |
| 128 | v1i128 f_vec_small_v1i128_ret() { |
| 129 | return (v1i128){1}; |
| 130 | } |
| 131 | |
| 132 | // Aggregates of 2*xlen size and 2*xlen alignment should be coerced to a |
| 133 | // single 2*xlen-sized argument, to ensure that alignment can be maintained if |
| 134 | // passed on the stack. |
| 135 | |
| 136 | struct small_aligned { |
| 137 | __int128_t a; |
| 138 | }; |
| 139 | |
| 140 | // CHECK-LABEL: define void @f_agg_small_aligned(i128 %x.coerce) |
| 141 | void f_agg_small_aligned(struct small_aligned x) { |
| 142 | x.a += x.a; |
| 143 | } |
| 144 | |
| 145 | // CHECK-LABEL: define i128 @f_agg_small_aligned_ret(i128 %x.coerce) |
| 146 | struct small_aligned f_agg_small_aligned_ret(struct small_aligned x) { |
| 147 | return (struct small_aligned){10}; |
| 148 | } |
| 149 | |
| 150 | // Aggregates greater > 2*xlen will be passed and returned indirectly |
| 151 | struct large { |
| 152 | int64_t a, b, c, d; |
| 153 | }; |
| 154 | |
| 155 | // CHECK-LABEL: define void @f_agg_large(%struct.large* %x) |
| 156 | void f_agg_large(struct large x) { |
| 157 | x.a = x.b + x.c + x.d; |
| 158 | } |
| 159 | |
| 160 | // The address where the struct should be written to will be the first |
| 161 | // argument |
| 162 | // CHECK-LABEL: define void @f_agg_large_ret(%struct.large* noalias sret %agg.result, i32 signext %i, i8 signext %j) |
| 163 | struct large f_agg_large_ret(int32_t i, int8_t j) { |
| 164 | return (struct large){1, 2, 3, 4}; |
| 165 | } |
| 166 | |
| 167 | typedef unsigned char v32i8 __attribute__((vector_size(32))); |
| 168 | |
| 169 | // CHECK-LABEL: define void @f_vec_large_v32i8(<32 x i8>*) |
| 170 | void f_vec_large_v32i8(v32i8 x) { |
| 171 | x[0] = x[7]; |
| 172 | } |
| 173 | |
| 174 | // CHECK-LABEL: define void @f_vec_large_v32i8_ret(<32 x i8>* noalias sret %agg.result) |
| 175 | v32i8 f_vec_large_v32i8_ret() { |
| 176 | return (v32i8){1, 2, 3, 4, 5, 6, 7, 8}; |
| 177 | } |
| 178 | |
| 179 | // Scalars passed on the stack should have signext/zeroext attributes (they |
| 180 | // are anyext). |
| 181 | |
| 182 | // CHECK-LABEL: define signext i32 @f_scalar_stack_1(i64 %a.coerce, [2 x i64] %b.coerce, i128 %c.coerce, %struct.large* %d, i8 zeroext %e, i8 signext %f, i8 %g, i8 %h) |
| 183 | int f_scalar_stack_1(struct tiny a, struct small b, struct small_aligned c, |
| 184 | struct large d, uint8_t e, int8_t f, uint8_t g, int8_t h) { |
| 185 | return g + h; |
| 186 | } |
| 187 | |
| 188 | // CHECK-LABEL: define signext i32 @f_scalar_stack_2(i32 signext %a, i128 %b, float %c, fp128 %d, <32 x i8>*, i8 zeroext %f, i8 %g, i8 %h) |
| 189 | int f_scalar_stack_2(int32_t a, __int128_t b, float c, long double d, v32i8 e, |
| 190 | uint8_t f, int8_t g, uint8_t h) { |
| 191 | return g + h; |
| 192 | } |
| 193 | |
| 194 | // Ensure that scalars passed on the stack are still determined correctly in |
| 195 | // the presence of large return values that consume a register due to the need |
| 196 | // to pass a pointer. |
| 197 | |
| 198 | // CHECK-LABEL: define void @f_scalar_stack_3(%struct.large* noalias sret %agg.result, i32 signext %a, i128 %b, fp128 %c, <32 x i8>*, i8 zeroext %e, i8 %f, i8 %g) |
| 199 | struct large f_scalar_stack_3(uint32_t a, __int128_t b, long double c, v32i8 d, |
| 200 | uint8_t e, int8_t f, uint8_t g) { |
| 201 | return (struct large){a, e, f, g}; |
| 202 | } |
| 203 | |
| 204 | // Ensure that ABI lowering happens as expected for vararg calls. |
| 205 | // Specifically, ensure that signext is emitted for varargs that will be |
| 206 | // passed in registers but not on the stack. Ensure this takes into account |
| 207 | // the use of "aligned" register pairs for varargs with 2*xlen alignment. |
| 208 | |
| 209 | int f_va_callee(int, ...); |
| 210 | |
| 211 | // CHECK-LABEL: define void @f_va_caller() |
| 212 | void f_va_caller() { |
| 213 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i64 3, double 4.000000e+00, double 5.000000e+00, i64 {{%.*}}, [2 x i64] {{%.*}}, i128 {{%.*}}, %struct.large* {{%.*}}) |
| 214 | f_va_callee(1, 2, 3LL, 4.0f, 5.0, (struct tiny){6, 7, 8, 9}, |
| 215 | (struct small){10, NULL}, (struct small_aligned){11}, |
| 216 | (struct large){12, 13, 14, 15}); |
| 217 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, fp128 0xL00000000000000004001400000000000, i32 signext 6, i32 signext 7, i32 8, i32 9) |
| 218 | f_va_callee(1, 2, 3, 4, 5.0L, 6, 7, 8, 9); |
| 219 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i128 {{%.*}}, i32 signext 6, i32 signext 7, i32 8, i32 9) |
| 220 | f_va_callee(1, 2, 3, 4, (struct small_aligned){5}, 6, 7, 8, 9); |
| 221 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, [2 x i64] {{%.*}}, i32 signext 6, i32 signext 7, i32 8, i32 9) |
| 222 | f_va_callee(1, 2, 3, 4, (struct small){5, NULL}, 6, 7, 8, 9); |
| 223 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, fp128 0xL00000000000000004001800000000000, i32 7, i32 8, i32 9) |
| 224 | f_va_callee(1, 2, 3, 4, 5, 6.0L, 7, 8, 9); |
| 225 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, i128 {{%.*}}, i32 7, i32 8, i32 9) |
| 226 | f_va_callee(1, 2, 3, 4, 5, (struct small_aligned){6}, 7, 8, 9); |
| 227 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, [2 x i64] {{%.*}}, i32 signext 7, i32 8, i32 9) |
| 228 | f_va_callee(1, 2, 3, 4, 5, (struct small){6, NULL}, 7, 8, 9); |
| 229 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, i32 signext 6, fp128 0xL00000000000000004001C00000000000, i32 8, i32 9) |
| 230 | f_va_callee(1, 2, 3, 4, 5, 6, 7.0L, 8, 9); |
| 231 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, i32 signext 6, i128 {{%.*}}, i32 8, i32 9) |
| 232 | f_va_callee(1, 2, 3, 4, 5, 6, (struct small_aligned){7}, 8, 9); |
| 233 | // CHECK: call signext i32 (i32, ...) @f_va_callee(i32 signext 1, i32 signext 2, i32 signext 3, i32 signext 4, i32 signext 5, i32 signext 6, [2 x i64] {{.*}}, i32 8, i32 9) |
| 234 | f_va_callee(1, 2, 3, 4, 5, 6, (struct small){7, NULL}, 8, 9); |
| 235 | } |
| 236 | |
| 237 | // CHECK-LABEL: define signext i32 @f_va_1(i8* %fmt, ...) {{.*}} { |
| 238 | // CHECK: [[FMT_ADDR:%.*]] = alloca i8*, align 8 |
| 239 | // CHECK: [[VA:%.*]] = alloca i8*, align 8 |
| 240 | // CHECK: [[V:%.*]] = alloca i32, align 4 |
| 241 | // CHECK: store i8* %fmt, i8** [[FMT_ADDR]], align 8 |
| 242 | // CHECK: [[VA1:%.*]] = bitcast i8** [[VA]] to i8* |
| 243 | // CHECK: call void @llvm.va_start(i8* [[VA1]]) |
| 244 | // CHECK: [[ARGP_CUR:%.*]] = load i8*, i8** [[VA]], align 8 |
| 245 | // CHECK: [[ARGP_NEXT:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR]], i64 8 |
| 246 | // CHECK: store i8* [[ARGP_NEXT]], i8** [[VA]], align 8 |
| 247 | // CHECK: [[TMP0:%.*]] = bitcast i8* [[ARGP_CUR]] to i32* |
| 248 | // CHECK: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 8 |
| 249 | // CHECK: store i32 [[TMP1]], i32* [[V]], align 4 |
| 250 | // CHECK: [[VA2:%.*]] = bitcast i8** [[VA]] to i8* |
| 251 | // CHECK: call void @llvm.va_end(i8* [[VA2]]) |
| 252 | // CHECK: [[TMP2:%.*]] = load i32, i32* [[V]], align 4 |
| 253 | // CHECK: ret i32 [[TMP2]] |
| 254 | // CHECK: } |
| 255 | int f_va_1(char *fmt, ...) { |
| 256 | __builtin_va_list va; |
| 257 | |
| 258 | __builtin_va_start(va, fmt); |
| 259 | int v = __builtin_va_arg(va, int); |
| 260 | __builtin_va_end(va); |
| 261 | |
| 262 | return v; |
| 263 | } |
| 264 | |
| 265 | // An "aligned" register pair (where the first register is even-numbered) is |
| 266 | // used to pass varargs with 2x xlen alignment and 2x xlen size. Ensure the |
| 267 | // correct offsets are used. |
| 268 | |
| 269 | // CHECK-LABEL: @f_va_2( |
| 270 | // CHECK: [[FMT_ADDR:%.*]] = alloca i8*, align 8 |
| 271 | // CHECK-NEXT: [[VA:%.*]] = alloca i8*, align 8 |
| 272 | // CHECK-NEXT: [[V:%.*]] = alloca fp128, align 16 |
| 273 | // CHECK-NEXT: store i8* [[FMT:%.*]], i8** [[FMT_ADDR]], align 8 |
| 274 | // CHECK-NEXT: [[VA1:%.*]] = bitcast i8** [[VA]] to i8* |
| 275 | // CHECK-NEXT: call void @llvm.va_start(i8* [[VA1]]) |
| 276 | // CHECK-NEXT: [[ARGP_CUR:%.*]] = load i8*, i8** [[VA]], align 8 |
| 277 | // CHECK-NEXT: [[TMP0:%.*]] = ptrtoint i8* [[ARGP_CUR]] to i64 |
| 278 | // CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], 15 |
| 279 | // CHECK-NEXT: [[TMP2:%.*]] = and i64 [[TMP1]], -16 |
| 280 | // CHECK-NEXT: [[ARGP_CUR_ALIGNED:%.*]] = inttoptr i64 [[TMP2]] to i8* |
| 281 | // CHECK-NEXT: [[ARGP_NEXT:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR_ALIGNED]], i64 16 |
| 282 | // CHECK-NEXT: store i8* [[ARGP_NEXT]], i8** [[VA]], align 8 |
| 283 | // CHECK-NEXT: [[TMP3:%.*]] = bitcast i8* [[ARGP_CUR_ALIGNED]] to fp128* |
| 284 | // CHECK-NEXT: [[TMP4:%.*]] = load fp128, fp128* [[TMP3]], align 16 |
| 285 | // CHECK-NEXT: store fp128 [[TMP4]], fp128* [[V]], align 16 |
| 286 | // CHECK-NEXT: [[VA2:%.*]] = bitcast i8** [[VA]] to i8* |
| 287 | // CHECK-NEXT: call void @llvm.va_end(i8* [[VA2]]) |
| 288 | // CHECK-NEXT: [[TMP5:%.*]] = load fp128, fp128* [[V]], align 16 |
| 289 | // CHECK-NEXT: ret fp128 [[TMP5]] |
| 290 | long double f_va_2(char *fmt, ...) { |
| 291 | __builtin_va_list va; |
| 292 | |
| 293 | __builtin_va_start(va, fmt); |
| 294 | long double v = __builtin_va_arg(va, long double); |
| 295 | __builtin_va_end(va); |
| 296 | |
| 297 | return v; |
| 298 | } |
| 299 | |
| 300 | // Two "aligned" register pairs. |
| 301 | |
| 302 | // CHECK-LABEL: @f_va_3( |
| 303 | // CHECK: [[FMT_ADDR:%.*]] = alloca i8*, align 8 |
| 304 | // CHECK-NEXT: [[VA:%.*]] = alloca i8*, align 8 |
| 305 | // CHECK-NEXT: [[V:%.*]] = alloca fp128, align 16 |
| 306 | // CHECK-NEXT: [[W:%.*]] = alloca i32, align 4 |
| 307 | // CHECK-NEXT: [[X:%.*]] = alloca fp128, align 16 |
| 308 | // CHECK-NEXT: store i8* [[FMT:%.*]], i8** [[FMT_ADDR]], align 8 |
| 309 | // CHECK-NEXT: [[VA1:%.*]] = bitcast i8** [[VA]] to i8* |
| 310 | // CHECK-NEXT: call void @llvm.va_start(i8* [[VA1]]) |
| 311 | // CHECK-NEXT: [[ARGP_CUR:%.*]] = load i8*, i8** [[VA]], align 8 |
| 312 | // CHECK-NEXT: [[TMP0:%.*]] = ptrtoint i8* [[ARGP_CUR]] to i64 |
| 313 | // CHECK-NEXT: [[TMP1:%.*]] = add i64 [[TMP0]], 15 |
| 314 | // CHECK-NEXT: [[TMP2:%.*]] = and i64 [[TMP1]], -16 |
| 315 | // CHECK-NEXT: [[ARGP_CUR_ALIGNED:%.*]] = inttoptr i64 [[TMP2]] to i8* |
| 316 | // CHECK-NEXT: [[ARGP_NEXT:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR_ALIGNED]], i64 16 |
| 317 | // CHECK-NEXT: store i8* [[ARGP_NEXT]], i8** [[VA]], align 8 |
| 318 | // CHECK-NEXT: [[TMP3:%.*]] = bitcast i8* [[ARGP_CUR_ALIGNED]] to fp128* |
| 319 | // CHECK-NEXT: [[TMP4:%.*]] = load fp128, fp128* [[TMP3]], align 16 |
| 320 | // CHECK-NEXT: store fp128 [[TMP4]], fp128* [[V]], align 16 |
| 321 | // CHECK-NEXT: [[ARGP_CUR2:%.*]] = load i8*, i8** [[VA]], align 8 |
| 322 | // CHECK-NEXT: [[ARGP_NEXT3:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR2]], i64 8 |
| 323 | // CHECK-NEXT: store i8* [[ARGP_NEXT3]], i8** [[VA]], align 8 |
| 324 | // CHECK-NEXT: [[TMP5:%.*]] = bitcast i8* [[ARGP_CUR2]] to i32* |
| 325 | // CHECK-NEXT: [[TMP6:%.*]] = load i32, i32* [[TMP5]], align 8 |
| 326 | // CHECK-NEXT: store i32 [[TMP6]], i32* [[W]], align 4 |
| 327 | // CHECK-NEXT: [[ARGP_CUR4:%.*]] = load i8*, i8** [[VA]], align 8 |
| 328 | // CHECK-NEXT: [[TMP7:%.*]] = ptrtoint i8* [[ARGP_CUR4]] to i64 |
| 329 | // CHECK-NEXT: [[TMP8:%.*]] = add i64 [[TMP7]], 15 |
| 330 | // CHECK-NEXT: [[TMP9:%.*]] = and i64 [[TMP8]], -16 |
| 331 | // CHECK-NEXT: [[ARGP_CUR4_ALIGNED:%.*]] = inttoptr i64 [[TMP9]] to i8* |
| 332 | // CHECK-NEXT: [[ARGP_NEXT5:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR4_ALIGNED]], i64 16 |
| 333 | // CHECK-NEXT: store i8* [[ARGP_NEXT5]], i8** [[VA]], align 8 |
| 334 | // CHECK-NEXT: [[TMP10:%.*]] = bitcast i8* [[ARGP_CUR4_ALIGNED]] to fp128* |
| 335 | // CHECK-NEXT: [[TMP11:%.*]] = load fp128, fp128* [[TMP10]], align 16 |
| 336 | // CHECK-NEXT: store fp128 [[TMP11]], fp128* [[X]], align 16 |
| 337 | // CHECK-NEXT: [[VA6:%.*]] = bitcast i8** [[VA]] to i8* |
| 338 | // CHECK-NEXT: call void @llvm.va_end(i8* [[VA6]]) |
| 339 | // CHECK-NEXT: [[TMP12:%.*]] = load fp128, fp128* [[V]], align 16 |
| 340 | // CHECK-NEXT: [[TMP13:%.*]] = load fp128, fp128* [[X]], align 16 |
| 341 | // CHECK-NEXT: [[ADD:%.*]] = fadd fp128 [[TMP12]], [[TMP13]] |
| 342 | // CHECK-NEXT: ret fp128 [[ADD]] |
| 343 | long double f_va_3(char *fmt, ...) { |
| 344 | __builtin_va_list va; |
| 345 | |
| 346 | __builtin_va_start(va, fmt); |
| 347 | long double v = __builtin_va_arg(va, long double); |
| 348 | int w = __builtin_va_arg(va, int); |
| 349 | long double x = __builtin_va_arg(va, long double); |
| 350 | __builtin_va_end(va); |
| 351 | |
| 352 | return v + x; |
| 353 | } |
| 354 | |
| 355 | // CHECK-LABEL: @f_va_4( |
| 356 | // CHECK: [[FMT_ADDR:%.*]] = alloca i8*, align 8 |
| 357 | // CHECK-NEXT: [[VA:%.*]] = alloca i8*, align 8 |
| 358 | // CHECK-NEXT: [[V:%.*]] = alloca i32, align 4 |
| 359 | // CHECK-NEXT: [[TS:%.*]] = alloca [[STRUCT_TINY:%.*]], align 2 |
| 360 | // CHECK-NEXT: [[SS:%.*]] = alloca [[STRUCT_SMALL:%.*]], align 8 |
| 361 | // CHECK-NEXT: [[LS:%.*]] = alloca [[STRUCT_LARGE:%.*]], align 8 |
| 362 | // CHECK-NEXT: [[RET:%.*]] = alloca i32, align 4 |
| 363 | // CHECK-NEXT: store i8* [[FMT:%.*]], i8** [[FMT_ADDR]], align 8 |
| 364 | // CHECK-NEXT: [[VA1:%.*]] = bitcast i8** [[VA]] to i8* |
| 365 | // CHECK-NEXT: call void @llvm.va_start(i8* [[VA1]]) |
| 366 | // CHECK-NEXT: [[ARGP_CUR:%.*]] = load i8*, i8** [[VA]], align 8 |
| 367 | // CHECK-NEXT: [[ARGP_NEXT:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR]], i64 8 |
| 368 | // CHECK-NEXT: store i8* [[ARGP_NEXT]], i8** [[VA]], align 8 |
| 369 | // CHECK-NEXT: [[TMP0:%.*]] = bitcast i8* [[ARGP_CUR]] to i32* |
| 370 | // CHECK-NEXT: [[TMP1:%.*]] = load i32, i32* [[TMP0]], align 8 |
| 371 | // CHECK-NEXT: store i32 [[TMP1]], i32* [[V]], align 4 |
| 372 | // CHECK-NEXT: [[ARGP_CUR2:%.*]] = load i8*, i8** [[VA]], align 8 |
| 373 | // CHECK-NEXT: [[ARGP_NEXT3:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR2]], i64 8 |
| 374 | // CHECK-NEXT: store i8* [[ARGP_NEXT3]], i8** [[VA]], align 8 |
| 375 | // CHECK-NEXT: [[TMP2:%.*]] = bitcast i8* [[ARGP_CUR2]] to %struct.tiny* |
| 376 | // CHECK-NEXT: [[TMP3:%.*]] = bitcast %struct.tiny* [[TS]] to i8* |
| 377 | // CHECK-NEXT: [[TMP4:%.*]] = bitcast %struct.tiny* [[TMP2]] to i8* |
| 378 | // CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 2 [[TMP3]], i8* align 8 [[TMP4]], i64 8, i1 false) |
| 379 | // CHECK-NEXT: [[ARGP_CUR4:%.*]] = load i8*, i8** [[VA]], align 8 |
| 380 | // CHECK-NEXT: [[ARGP_NEXT5:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR4]], i64 16 |
| 381 | // CHECK-NEXT: store i8* [[ARGP_NEXT5]], i8** [[VA]], align 8 |
| 382 | // CHECK-NEXT: [[TMP5:%.*]] = bitcast i8* [[ARGP_CUR4]] to %struct.small* |
| 383 | // CHECK-NEXT: [[TMP6:%.*]] = bitcast %struct.small* [[SS]] to i8* |
| 384 | // CHECK-NEXT: [[TMP7:%.*]] = bitcast %struct.small* [[TMP5]] to i8* |
| 385 | // CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP6]], i8* align 8 [[TMP7]], i64 16, i1 false) |
| 386 | // CHECK-NEXT: [[ARGP_CUR6:%.*]] = load i8*, i8** [[VA]], align 8 |
| 387 | // CHECK-NEXT: [[ARGP_NEXT7:%.*]] = getelementptr inbounds i8, i8* [[ARGP_CUR6]], i64 8 |
| 388 | // CHECK-NEXT: store i8* [[ARGP_NEXT7]], i8** [[VA]], align 8 |
| 389 | // CHECK-NEXT: [[TMP8:%.*]] = bitcast i8* [[ARGP_CUR6]] to %struct.large** |
| 390 | // CHECK-NEXT: [[TMP9:%.*]] = load %struct.large*, %struct.large** [[TMP8]], align 8 |
| 391 | // CHECK-NEXT: [[TMP10:%.*]] = bitcast %struct.large* [[LS]] to i8* |
| 392 | // CHECK-NEXT: [[TMP11:%.*]] = bitcast %struct.large* [[TMP9]] to i8* |
| 393 | // CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 [[TMP10]], i8* align 8 [[TMP11]], i64 32, i1 false) |
| 394 | // CHECK-NEXT: [[VA8:%.*]] = bitcast i8** [[VA]] to i8* |
| 395 | // CHECK-NEXT: call void @llvm.va_end(i8* [[VA8]]) |
| 396 | // CHECK-NEXT: [[A:%.*]] = getelementptr inbounds [[STRUCT_TINY]], %struct.tiny* [[TS]], i32 0, i32 0 |
| 397 | // CHECK-NEXT: [[TMP12:%.*]] = load i16, i16* [[A]], align 2 |
| 398 | // CHECK-NEXT: [[CONV:%.*]] = zext i16 [[TMP12]] to i64 |
| 399 | // CHECK-NEXT: [[A9:%.*]] = getelementptr inbounds [[STRUCT_SMALL]], %struct.small* [[SS]], i32 0, i32 0 |
| 400 | // CHECK-NEXT: [[TMP13:%.*]] = load i64, i64* [[A9]], align 8 |
| 401 | // CHECK-NEXT: [[ADD:%.*]] = add nsw i64 [[CONV]], [[TMP13]] |
| 402 | // CHECK-NEXT: [[C:%.*]] = getelementptr inbounds [[STRUCT_LARGE]], %struct.large* [[LS]], i32 0, i32 2 |
| 403 | // CHECK-NEXT: [[TMP14:%.*]] = load i64, i64* [[C]], align 8 |
| 404 | // CHECK-NEXT: [[ADD10:%.*]] = add nsw i64 [[ADD]], [[TMP14]] |
| 405 | // CHECK-NEXT: [[CONV11:%.*]] = trunc i64 [[ADD10]] to i32 |
| 406 | // CHECK-NEXT: store i32 [[CONV11]], i32* [[RET]], align 4 |
| 407 | // CHECK-NEXT: [[TMP15:%.*]] = load i32, i32* [[RET]], align 4 |
| 408 | // CHECK-NEXT: ret i32 [[TMP15]] |
| 409 | int f_va_4(char *fmt, ...) { |
| 410 | __builtin_va_list va; |
| 411 | |
| 412 | __builtin_va_start(va, fmt); |
| 413 | int v = __builtin_va_arg(va, int); |
| 414 | struct tiny ts = __builtin_va_arg(va, struct tiny); |
| 415 | struct small ss = __builtin_va_arg(va, struct small); |
| 416 | struct large ls = __builtin_va_arg(va, struct large); |
| 417 | __builtin_va_end(va); |
| 418 | |
| 419 | int ret = ts.a + ss.a + ls.c; |
| 420 | |
| 421 | return ret; |
| 422 | } |
| 423 | |