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
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
|
// Copyright (c) 2021 PoCL developers
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
/* See https://github.com/pocl/pocl/issues/893.
*
* This test will test that PoCL will not crash on LLVM 11 onward. The reason
* was a jump threading optimizations which will duplicate basic blocks and
* cause PoCL to form illegal parallel regions which does not follow
* single-entry single-exit rule. This test doesn't care about the kernel result
* or the arguments given to it. Point is to see that the kernel compiles
* without any problems.
*
* More specifically the issue was the jump threading will cause kernel for loop
* latch block to be duplicated when OpenCL is compiled to IR. Then later when
* standard -O3 is ran as part of the PoCL passes the loop backedge is
* duplicated, causing the loop duplication. When PoCL tries to form parallel
* regions from explicit barrier to loop latch barrier the parallel region is
* not form correctly. It contains incoming edges from outside of the parallel
* region which will cause assert to fail. Those incoming edges are coming from
* duplicated latch blocks.
*/
#include "pocl_opencl.h"
#define CL_HPP_ENABLE_EXCEPTIONS
#define CL_HPP_MINIMUM_OPENCL_VERSION 120
#define CL_HPP_TARGET_OPENCL_VERSION 120
#define CL_HPP_CL_1_2_DEFAULT_BUILD
#include <CL/opencl.hpp>
#include <iostream>
const char *SOURCE = R"RAW(
#define lid(N) ((int) get_local_id(N))
#define gid(N) ((int) get_group_id(N))
__kernel void loopy_kernel(__global float *__restrict__ out, int const im_h, int const im_w)
{
float acc_f_x_f_y_icolor;
if (-1 + -1 * /* im_y_inner */ lid(1) + -16 * /* im_y_outer */ gid(1) + im_h >= 0 && -1 + -1 * /* im_x_inner */ lid(0) + -16 * /* im_x_outer */ gid(0) + im_w >= 0)
acc_f_x_f_y_icolor = 0.0f;
for (int icolor = 0; icolor <= 2; ++icolor)
{
barrier(CLK_LOCAL_MEM_FENCE) /* for img_fetch (insn_f_x_f_y_icolor_update depends on img_fetch_rule) */;
if (-1 + -1 * /* im_y_inner */ lid(1) + -16 * /* im_y_outer */ gid(1) + im_h >= 0 && -1 + -1 * /* im_x_inner */ lid(0) + -16 * /* im_x_outer */ gid(0) + im_w >= 0)
for (int f_x = -3; f_x <= 3; ++f_x)
acc_f_x_f_y_icolor = acc_f_x_f_y_icolor + 10 * 3;
}
if (-1 + -1 * /* im_y_inner */ lid(1) + -16 * /* im_y_outer */ gid(1) + im_h >= 0 && -1 + -1 * /* im_x_inner */ lid(0) + -16 * /* im_x_outer */ gid(0) + im_w >= 0)
out[im_h * im_w * /* ifeat */ gid(2) + im_h * (16 * /* im_x_outer */ gid(0) + /* im_x_inner */ lid(0)) + 16 * /* im_y_outer */ gid(1) + /* im_y_inner */ lid(1)] = acc_f_x_f_y_icolor;
}
)RAW";
int main() {
int n = 8;
cl::Platform platform = cl::Platform::getDefault();
cl::Device device = cl::Device::getDefault();
try {
cl::CommandQueue queue = cl::CommandQueue::getDefault();
cl::Program program(SOURCE, true);
cl::Buffer buffer(CL_MEM_WRITE_ONLY, sizeof(float) * 256);
cl::Kernel kernel(program, "loopy_kernel");
kernel.setArg(0, buffer);
kernel.setArg(1, n);
kernel.setArg(2, n);
queue.enqueueNDRangeKernel(kernel, cl::NullRange, cl::NDRange(n),
cl::NDRange(n));
queue.finish();
} catch (cl::Error &err) {
std::cout << "FAIL with OpenCL error = " << err.err() << std::endl;
return EXIT_FAILURE;
}
platform.unloadCompiler();
std::cout << "OK" << std::endl;
return EXIT_SUCCESS;
}
|
