Updated tests and clamping number of elements to 2

Updated data_sharing_stack_init_common

Only places where it was accessed are here and here. Jon's observation is correct. The maximum number of threads on both amdgcn and nvptx is 1024. However, on amdgcn, wave size is 64 and so maximum number of waves can be 16 and on nvptx, the warp size is 32 and maximum number of warps is 32.

@arsenm , let me know if it is good to land.

Updated review comments.

Added support for vector types.

Review comments

removed unneeded changes

Added runRUN lines with OpenMP 5.0 version option

@phosek Let me know if this is good to land.

In D81991#2127205, @MaskRay wrote:

Hi, your git commit contains extra Phabricator tags. You can drop Reviewers: Subscribers: Tags: and the text Summary: from the git commit with the following script:

arcfilter () {
        arc amend
        git log -1 --pretty=%B | awk '/Reviewers:|Subscribers:/{p=1} /Reviewed By:|Differential Revision:/{p=0} !p && !/^Summary:$/ {sub(/^Summary: /,"");print}' | git commit --amend --date=now -F -
}

Reviewed By: is considered important by some people. Please keep the tag. (--date=now is my personal preference (author dates are usually not useful. Using committer dates can make log almost monotonic in time))

llvm/utils/git/pre-push.py can validate the message does not include unneeded tags.

In D79400#2121685, @vsapsai wrote:

Seems like this change causes ninja clean to fail with the error

ninja: error: remove(include/llvm/Support): Directory not empty

Full repro steps are

<cmake configuration>
ninja install
ninja clean

Simpler steps are

<cmake configuration>
ninja include/llvm/Support/all
ninja clean

@arsenm Please let me know if more information is required.

My understanding is that clang already emits amdgpu-flat-work-group-size in range of [1, 256] as default value and HIP defaults to [1,1024]. And logic to calculate min-waves-per-eu is dependent on the maximum value of work-group size not minimum.

Added lit test case

3. Another way is to gracefully handle the file write error, for which I don't think there is a portable way. (which @scott.linder also suggested)

I have found the portable way to do this. So cmake provides a command-line tool touch (doc) which can be used with [[ https://cmake.org/cmake/help/v3.0/command/execute_process.html | execute_process ]]. execute_process provides a way to suppress the failure using ERROR_QUIET option. Coming back to touch, it behaves similar to file(WRITE ..) command however, additionally the exit code can be used to check if file was created successfully. This retains the old behaviour while still solving the original problem. I will update the patch.

Sorry for late reply.

I understand that .git/logs/HEAD acts as a dependency for vcs_revision_h target. However, problem here is that cmake fails when it tries to create .git/logs/HEAD in read-only filesystem.

Removing find_first_existing_vc_file makes a lot of sense, as since llvm has moved from svn to git, there is no need to have logic for svn dependency. Even the generation script is dependent on git executable only. I will keep my patch ready just in case other reviewers also feel the same.

Added more context to diff.

In D79400#2021255, @scott.linder wrote:

Can you provide more context in the diff? Looking at the current master locally it seems like this patch changes the behavior of the llvm_vcsrevision_h target when logs/HEAD isn't found, such that it only depends on the generation script. I.e. if you configure CMake without the HEAD file present, and then you do something which changes the working directory and creates HEAD the target will not be out of date and won't be rebuilt. Before this patch it would have noticed the change to HEAD and forced regenerating the version header.

If you look at the generation script https://github.com/llvm/llvm-project/blob/master/llvm/cmake/modules/GenerateVersionFromVCS.cmake and https://github.com/llvm/llvm-project/blob/master/llvm/cmake/modules/VersionFromVCS.cmake, these are not directly dependent on .git/logs/HEAD instead are dependent on .git/HEAD. The LLVM_REVISION is generated using following command

# https://github.com/llvm/llvm-project/blob/master/llvm/cmake/modules/VersionFromVCS.cmake#L17
git rev-parse HEAD

which does not seem to use .git/logs/HEAD (instead uses .git/HEAD as revealed by strace)