Added WaveOps tests for QuadReadAcrossDiagonal

[kcloudy0717]

PR for QuadReadAcrossDiagonal tests, resolves #881.

Note: After following some conversation in other PRs, I added convergence suffix to QuadReadAcross* control flow tests in the PR as well.

[kcloudy0717]

Pinging @bob80905 and @farzonl for review.

[bob80905]

This is great, thanks.
You might consider at some point updating issue #986
so that it mentions it's DXC specific, if the failure is not found on clang also.
Your XFAIL condition is specific to DXC, so that edit to the issue would be good, or maybe remove the DXC condition from the XFAIL if it isn't necessary. I know it's hard to tell right now with the implementation in clang not ready yet.

[kcloudy0717]

This is great, thanks. You might consider at some point updating issue #986 so that it mentions it's DXC specific, if the failure is not found on clang also. Your XFAIL condition is specific to DXC, so that edit to the issue would be good, or maybe remove the DXC condition from the XFAIL if it isn't necessary. I know it's hard to tell right now with the implementation in clang not ready yet.

Sounds good, once the PR for QuadReadAcrossY is merged to main, I'll open up the PR for QuadReadAcrossDiagonal intrinsic support.

[bob80905]

This is great, thanks. You might consider at some point updating issue #986 so that it mentions it's DXC specific, if the failure is not found on clang also. Your XFAIL condition is specific to DXC, so that edit to the issue would be good, or maybe remove the DXC condition from the XFAIL if it isn't necessary. I know it's hard to tell right now with the implementation in clang not ready yet.

Sounds good, once the PR for QuadReadAcrossY is merged to main, I'll open up the PR for QuadReadAcrossDiagonal intrinsic support.

Also on the topic of QuadReadAcrossY, the implementation should be in now, so if you'd like, please open a PR to remove the XFAIL for Clang on those tests since it's now supported 🚀

[kcloudy0717]

This is great, thanks. You might consider at some point updating issue #986 so that it mentions it's DXC specific, if the failure is not found on clang also. Your XFAIL condition is specific to DXC, so that edit to the issue would be good, or maybe remove the DXC condition from the XFAIL if it isn't necessary. I know it's hard to tell right now with the implementation in clang not ready yet.

I've updated the issue to include DXC as part of the title. AFAIK, this does only trigger on DXC, since previous PRs for QuadReadAcross[X|Y] also have the same issue, so I did update the title for include diagonal as well.

Looking at the checks, it looks like failing test is unrelated to this PR, so this PR is good to go for me. If we want to get another pair of eyes for review then merge, that would be appreciated. I'm working on some other intrinsic support in the meantime.