File: trace-through-global.cu

package info (click to toggle)
llvm-toolchain-9 1%3A9.0.1-16.1
  • links: PTS, VCS
  • area: main
  • in suites: bullseye
  • size: 882,388 kB
  • sloc: cpp: 4,167,636; ansic: 714,256; asm: 457,610; python: 155,927; objc: 65,094; sh: 42,856; lisp: 26,908; perl: 7,786; pascal: 7,722; makefile: 6,881; ml: 5,581; awk: 3,648; cs: 2,027; xml: 888; javascript: 381; ruby: 156
file content (48 lines) | stat: -rw-r--r-- 1,383 bytes parent folder | download | duplicates (8)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
// RUN: %clang_cc1 -fsyntax-only -verify %s

// Check that it's OK for kernels to call HD functions that call device-only
// functions.

#include "Inputs/cuda.h"

__device__ void device_fn(int) {}
// expected-note@-1 2 {{declared here}}

inline __host__ __device__ int hd1() {
  device_fn(0);  // expected-error {{reference to __device__ function}}
  return 0;
}

inline __host__ __device__ int hd2() {
  // No error here because hd2 is only referenced from a kernel.
  device_fn(0);
  return 0;
}

inline __host__ __device__ void hd3(int) {
  device_fn(0);  // expected-error {{reference to __device__ function 'device_fn'}}
}
inline __host__ __device__ void hd3(double) {}

inline __host__ __device__ void hd4(int) {}
inline __host__ __device__ void hd4(double) {
  device_fn(0);  // No error; this function is never called.
}

__global__ void kernel(int) { hd2(); }

template <typename T>
void launch_kernel() {
  kernel<<<0, 0>>>(T());

  // Notice that these two diagnostics are different: Because the call to hd1
  // is not dependent on T, the call to hd1 comes from 'launch_kernel', while
  // the call to hd3, being dependent, comes from 'launch_kernel<int>'.
  hd1(); // expected-note {{called by 'launch_kernel'}}
  hd3(T()); // expected-note {{called by 'launch_kernel<int>'}}
}

void host_fn() {
  launch_kernel<int>();
  // expected-note@-1 2 {{called by 'host_fn'}}
}