[DOCS] add missing docs

[khushi-411]

Hi,
While reading through the documentation, I noticed some missing information, like below:

The same is true for fdiv and umulhi. This PR addresses those gaps.

I'm looking forward to hearing from the team. :-)
Thank you!

[zhanglx13]

Is there an assumption that trans operates data in shared memory?

[Jokeren]

Is there an assumption that trans operates data in shared memory?

Yes, that's why it caused several bugs.

[Jokeren]

both tl.trans and tl.reshape (tl.view) are unstable at the current stage because they adjust the tensor layout without actually adjusting the position of elements.

[zhanglx13]

Yes, that's why it caused several bugs.

Is it because not "everyone" (the user and different parts of the lowering passes) knows that trans operates in shared memory?
At python level, triton exposes the concept of global memory, so load is needed. But shared memory concept is hidden.

[Jokeren]

Is it because not "everyone" (the user and different parts of the lowering passes) knows that trans operates in shared memory?

The users shouldn't be aware of that. And I think it's just yet another TODO item that needs to be improved.

[khushi-411]

Hi, @Jokeren, @zhanglx13!
Gentle ping! Could you please take a look at the PR whenever you are free? Please let me know if there's anything I need to add more. Thanks :-)