[PATCH] D47672: [Headers] Add _Interlocked*_HLEAcquire/_HLERelease

Tue Jun 5 01:09:07 PDT 2018

ethanhs added a comment.

In https://reviews.llvm.org/D47672#1121181, @rnk wrote:

>

> They are a hint to the processor that this is a short critical section, and it is likely that the entire critical section can be entered, run, and committed to memory before another thread needs to use the memory used by the critical section. Your implementation doesn't add these prefixes, but the code will execute correctly. This functionality relies on TSX, which it looks like Intel added some time in 2014, so it's probably widely available.
> 
> Personally, I don't want to implement these intrinsics this way. We've already implemented intrinsics "our way" instead of doing whatever Visual C does, and it just leads to developer confusion when they discover that the compiler didn't emit the instruction they want. For example, the rotate intrinsics often don't work (https://llvm.org/pr37387) and the bittestandset (bts) intrinsics are just broken (http://llvm.org/pr33188).

I see. AIUI, the HLE versions just add optional hints, so there should not be any functional differences beyond using the hints or not (theses are optimization hints, yes?). Perhaps I misunderstood the documentation.

> are there any plans for representing HLE hints on atomic instructions in LLVM IR?

This would be quite nice to have, the main reason I implemented these functions this way is I don't have the bandwidth at the moment to correctly implement HLE in LLVM/Clang, nor some of the requisite knowledge I'm sure. I also am keen to have Python build with clang-cl on Windows :) But I definitely can understand wanting to do things correctly from the start.

I am going to the LLVM Bay Area monthly social Thursday so if someone wants to discuss this there I'd be more than happy to.

Repository:
  rC Clang

https://reviews.llvm.org/D47672