1 | // REQUIRES: x86-registered-target |
2 | // REQUIRES: nvptx-registered-target |
3 | |
4 | // RUN: %clang_cc1 -triple nvptx64-nvidia-cuda -fcuda-is-device -emit-llvm \ |
5 | // RUN: -o - %s | FileCheck %s |
6 | |
7 | #include "Inputs/cuda.h" |
8 | |
9 | extern "C" __device__ int vprintf(const char*, const char*); |
10 | |
11 | // Check a simple call to printf end-to-end. |
12 | // CHECK: [[SIMPLE_PRINTF_TY:%[a-zA-Z0-9_]+]] = type { i32, i64, double } |
13 | __device__ int CheckSimple() { |
14 | // CHECK: [[BUF:%[a-zA-Z0-9_]+]] = alloca [[SIMPLE_PRINTF_TY]] |
15 | // CHECK: [[FMT:%[0-9]+]] = load{{.*}}%fmt |
16 | const char* fmt = "%d %lld %f"; |
17 | // CHECK: [[PTR0:%[0-9]+]] = getelementptr inbounds [[SIMPLE_PRINTF_TY]], [[SIMPLE_PRINTF_TY]]* [[BUF]], i32 0, i32 0 |
18 | // CHECK: store i32 1, i32* [[PTR0]], align 4 |
19 | // CHECK: [[PTR1:%[0-9]+]] = getelementptr inbounds [[SIMPLE_PRINTF_TY]], [[SIMPLE_PRINTF_TY]]* [[BUF]], i32 0, i32 1 |
20 | // CHECK: store i64 2, i64* [[PTR1]], align 8 |
21 | // CHECK: [[PTR2:%[0-9]+]] = getelementptr inbounds [[SIMPLE_PRINTF_TY]], [[SIMPLE_PRINTF_TY]]* [[BUF]], i32 0, i32 2 |
22 | // CHECK: store double 3.0{{[^,]*}}, double* [[PTR2]], align 8 |
23 | // CHECK: [[BUF_CAST:%[0-9]+]] = bitcast [[SIMPLE_PRINTF_TY]]* [[BUF]] to i8* |
24 | // CHECK: [[RET:%[0-9]+]] = call i32 @vprintf(i8* [[FMT]], i8* [[BUF_CAST]]) |
25 | // CHECK: ret i32 [[RET]] |
26 | return printf(fmt, 1, 2ll, 3.0); |
27 | } |
28 | |
29 | __device__ void CheckNoArgs() { |
30 | // CHECK: call i32 @vprintf({{.*}}, i8* null){{$}} |
31 | printf("hello, world!"); |
32 | } |
33 | |
34 | // Check that printf's alloca happens in the entry block, not inside the if |
35 | // statement. |
36 | __device__ bool foo(); |
37 | __device__ void CheckAllocaIsInEntryBlock() { |
38 | // CHECK: alloca %printf_args |
39 | // CHECK: call {{.*}} @_Z3foov() |
40 | if (foo()) { |
41 | printf("%d", 42); |
42 | } |
43 | } |
44 | |