linking printf

Open
#2,157 0 comments 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
35/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Stale
Tech stack
cpp
Domain
testing-qa

Research direction

Start with the minimal Python reproduction in the issue and the two related C++ examples in OpenCL-CLHPP issue 316. Add coverage for one and two kernels containing printf, including an empty second source, checking compilation and linking and, where execution is available, the output. Done means the cases are represented in the conformance tests and their expected behavior is explicit.

Written by the indexing model from the issue text.

Description

clLinkProgram has issues with printf.

Consider this pyopencl example:

import pyopencl as cl

plattform = cl.get_platforms()[0]
device = plattform.get_devices()[0]

ctx = cl.Context([device])
queue = cl.CommandQueue(ctx)

src = r"""
kernel void tet() {
    const long m = get_global_id(0);
	printf("ID %ld\n", m);
}
"""

src2 = r"""
kernel void other() {
    const long n = get_global_id(0);
    printf("Do not print %ld\n", n);
}
"""

p = cl.Program(ctx, src).compile()
p2 = cl.Program(ctx, src2).compile()

program = cl.link_program(ctx, [p, p2])

program.tet(queue, (2,), None)

This should return

ID 0
ID 1

And using pocl with my CPU it does.
However, using Nvidia the linking segfaults.
Using AMD, I get the following:

ID 0
Do not print 4294967297
(null)

These problems are related to the presence of a printf in both kernels. If I get rid of one of them the issues disappear as well.
I have two related C++ examples here: https://github.com/KhronosGroup/OpenCL-CLHPP/issues/316
For some conditions linking “just” fails under Nvidia instead of a segfault. Nvidia can sometimes segfault with a printf in a single kernel as well though (see example in linked issue).

A test covering this (One kernel with a printf, two kernels with a printf) would be appreciated.
While part of the issue can be captured by testing if the kernels compile and can be linked, to capture what AMD is doing wrong, execution is required. And I do not know how to test that.

By the way, while testing this I removed the contents of src2. This is how I discovered, that passing an empty string (which should be completely valid, and can happen with generated code) also creates issues.

Dominant language
C++
Stars
234
Forks
240
Avg merge
6d 12h
Merged PRs (30d)
15

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from KhronosGroup/OpenCL-CTS

All issues in KhronosGroup/OpenCL-CTS

Similar issues

More C++ issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.